DaCHS Reference Documentation ¶

Author:	Markus Demleitner
Email:	gavo@ari.uni-heidelberg.de
Date:	2023-09-01
Copyright:	Waived under CC-0

Contents

Resource Descriptor Element Reference ¶

The following (XML) elements are defined for resource descriptors. Some elements are polymorous (Grammars, Cores). See below for a reference on the respective real elements known to the software.

Each element description gives a general introduction to the element’s use (complain if it’s too technical; it’s not unlikely that it is since these texts are actually the defining classes’ docstrings).

Within RDs, element properties that can (but need not) be written in XML attributes, i.e., as a single string, are called “atomic”. Their types are given in parentheses after the attribute name along with a default value.

In general, items defaulted to Undefined are mandatory. Failing to give a value will result in an error at RD parse time.

Within RD XML documents, you can (almost always) give atomic children either as XML attribute (att="abc") or as child elements (<att>abc</abc>). Some of the “atomic” attributes actually contain lists of items. For those, you should normally write multiple child elements (<att>val1</att><att>val2</att>), although sometimes it’s allowed to mash together the individual list items using a variety of separators.

Here are some short words about the types you may encounter, together with valid literals:

boolean – these allow quite a number of literals; use True and False or yes and no and stick to your choice.
unicode string – there may be additional syntactical limitations on those. See the explanation
integer – only decimal integer literals are allowed
id reference – these are references to items within XML documents; all elements within RDs can have an id attribute, which can then be used as an id reference. Additionally, you can reference elements in different RDs using <rd-id>#<id>. Note that DaCHS does not support forward references (i.e., references to items lexically behind the referencing element).
list of id references – Lists of id references. The values could be mashed together with commas, but prefer multiple child elements.

There are also “Dict-like” attributes. These are built from XML like:

<d key="ab">val1</d>
<d key="cd">val2</d>

In addition to key, other (possibly more descriptive) attributes for the key within these mappings may also be allowed. In special circumstances (in particular with properties) it may be useful to add to a value:

<property key="brokencols">ab,cd</property>
<property key="brokencols" cumulative="True">,x</property>

will leave ab,cd,x in brokencols.

Many elements can also have “structure children”. These correspond to compound things with attributes and possibly children of their own. The name given at the start of each description is irrelevant to the pure user; it’s the attribute name you’d use when you have the corresponding python objects. For authoring XML, you use the name in the following link; thus, the phrase “colRefs (contains Element columnRef…” means you’d write <columnRef...>.

Here are some guidelines as to the naming of the attributes:

Attributes giving keys into dictionaries or similar (e.g., column names) should always be named key
Attributes giving references to some source of events or data should always be named source, never “src” or similar
Attributes referencing generic things should always be called ref; of course, references to specific things like tables or services should indicate in their names what they are supposed to reference.

Also note that examples for the usage of almost everything mentioned here can be found in in the GAVO datacenter element reference.

Element apply¶

A code fragment to manipulate the result row (and possibly more).

Apply elements allow embedding python code in rowmakers.

The current input fields from the grammar (including the rowmaker’s vars) are available in the vars dictionary and can be changed there. You can also add new keys.

You can add new keys for shipping out in the result dictionary.

The active rowmaker is available as parent. It is also used to expand macros.

The table that the rowmaker feeds to can be accessed as targetTable. You probably only want to change meta information here (e.g., warnings or infos).

As always in procApps, you can get the embedding RD as rd; this is useful to, e.g., resolve references using rd.getByRD, and specify resdir-relative file names using rd.getAbsPath.

May occur in Element rowmaker.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element bind¶

A binding of a procedure definition parameter to a concrete value.

The value to set is contained in the binding body in the form of a python expression. The body must not be empty.

May occur in Element phraseMaker, Element apply, Element job, Element processEarly, Element processLate, Element regTest, Element rowfilter, Element sourceFields, Element iterator, Element pargetter, Element makeQuery, Element dataFormatter, Element dataFunction, Element descriptorGenerator, Element metaMaker, Element coreProc.

Atomic Children¶

alias (unicode string; defaults to None) – A deprecated name for the parameter
Character content of the element (defaulting to <Not given/empty>) – The default for the parameter. The special value __NULL__ indicates a NULL (python None) as usual. An empty content means a non-preset parameter, which must be filled in applications. The magic value __EMPTY__ allows presetting an empty string.
description (whitespace normalized unicode string; defaults to None) – Some human-readable description of what the parameter is about
key (unicode string; defaults to <Undefined>) – The name of the parameter
late (boolean; defaults to ‘False’) – Bind the name not at setup time but at applying time. In rowmaker procedures, for example, this allows you to refer to variables like vars or rowIter in the bindings.

Element column¶

A database column.

Columns contain almost all metadata to describe a column in a database table or a VOTable (the exceptions are for column properties that may span several columns, most notably indices).

Note that the type system adopted by the DC software is a subset of postgres’ type system. Thus when defining types, you have to specify basically SQL types. Types for other type systems (like VOTable, XSD, or the software-internal representation in python values) are inferred from them.

Columns can have delimited identifiers as names. Don’t do this, it’s no end of trouble. For this reason, however, you should not use name but rather key to programmatially obtain field’s values from rows.

Properties evaluated:

std – set to 1 to tell the tap schema importer to have the column’s std column in TAP_SCHEMA 1 (it’s 0 otherwise).
statisticsTarget – an integer to be set as this column’s statistics-gathering target. Set this to something between 100 and 10000 on postgres if you have large tables and columns with strongly non-uniform distributions. Set to -1 to revert to the system default. gavo imp -m will apply changes here; you’ll manually have to run analyze <tablename> after that.
statistics – set this to “no” to keep DaCHS from using this column in dachs limits. Set this to “enumerate” to make DaCHS collect the discrete values allowed (currently only supported for strings).
targetType – for a column containing a URL, the media type of the resource pointed at. This is for producing extra annotation for Aladin and friends as per http://mail.ivoa.net/pipermail/dal/2018-May/008017.html
targetTitle – if you give targetType, use this to set the link title (defaults to “Link”).

May occur in Element table.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Columns admit data content but ignore it. This is exclusively a convenience for building columns from params and should not be used for anything else.
description (whitespace normalized unicode string; defaults to ‘’) – A short (one-line) description of the values in this column.
displayHint (Display hint; defaults to ‘’) – Suggested presentation; the format is <kw>=<value>{,<kw>=<value>}, where what is interpreted depends on the output format. See, e.g., documentation on HTML renderers and the formatter child of outputFields.
fixup (unicode string; defaults to None) – A python expression the value of which will replace this column’s value on database reads. Write a ___ to access the original value. You can use macros for the embedding table. This is for, e.g., simple URL generation (fixup=“‘internallink{/this/svc}’+___”). It will only kick in when tuples are deserialized from the database, i.e., not for values taken from tables in memory.
hidden (boolean; defaults to ‘False’) – Hide the column from most of the user interface (specifically, you can’t use it in TAP queries or results, and it won’t be in TAP_SCHEMA). You typically want this for internal, administrative columns.
name (a column name within an SQL table. These have to match the SQL regular_identifier production. In a desperate pinch, you can generate delimited identifiers (that can contain anything) by prefixing the name with ‘quoted/’; defaults to <Undefined>) – Name of the column
note (unicode string; defaults to None) – Reference to a note meta on this table explaining more about this column
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
required (boolean; defaults to ‘False’) – Record becomes invalid when this column is NULL
tablehead (unicode string; defaults to None) – Terse phrase to put into table headers for this column
type (a type name; the internal type system is similar to SQL’s with some restrictions and extensions. The known atomic types include: smallint, integer, bigint, real, boolean, double precision, text, char, unicode, date, timestamp, time, spoint, scircle, spoly, sbox, smoc, bytea, raw, file, box, vexpr-mjd, vexpr-string, vexpr- float, vexpr-date, pql-string, pql-float, pql-int, pql-date, pql- upload, int4range; defaults to ‘real’) – datatype for the column (SQL-like type system)
ucd (unicode string; defaults to ‘’) – UCD of the column
unit (unicode string; defaults to ‘’) – Unit of the values. Use VOUnits syntax and use single quotes when you use custom units (you should avoid that).
utype (unicode string; defaults to None) – utype for this column
verbLevel (integer; defaults to ‘20’) – Minimal verbosity level at which to include this column
xtype (unicode string; defaults to None) – VOTable xtype giving the serialization form; you usually do not want to set this, as the xtypes actually used are computed from database type. DaCHS xtypes are only used for a few unsavoury, hopefully temporary, hacks

Structure Children¶

values (contains Element values) – Specification of legal values

Other Children¶

dmRoles (read-only list of roles played by this column in DMs; defaults to []) – Roles played by this column; cannot be assigned to.
meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.
stc (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to unless instructed to do so)
stcUtype (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to)

Element columnRef¶

A reference from a group to a column within a table.

ColumnReferences do not support qualified references, i.e., you can only give simple names.

May occur in Element group.

Atomic Children¶

key (unicode string; defaults to <Undefined>) – The key (i.e., name) of the referenced column or param.
ucd (unicode string; defaults to None) – The UCD of the group
utype (unicode string; defaults to None) – A utype for the group

Element condDesc¶

A query specification for cores talking to the database.

CondDescs define inputs as a sequence of InputKeys (see Element InputKey). Internally, the values in the InputKeys can be translated to SQL.

May occur in Element resource, Element dbCore, Element fancyQueryCore, Element productCore, Element scsCore, Element siapCutoutCore, Element ssapCore.

Atomic Children¶

buildFrom (id reference; defaults to None) – A reference to a column or an InputKey to define this CondDesc
combining (boolean; defaults to ‘False’) – Allow some input keys to be missing when others are given? (you want this for pseudo- condDescs just collecting random input keys)
fixedSQL (unicode string; defaults to None) – Always insert this SQL statement into the query. Deprecated.
inputOptional (boolean; defaults to ‘False’) – Call the phrase maker even if no input to our input keys is there (useful to provide built-in defaults).
joiner (unicode string; defaults to ‘OR’) – When yielding multiple fragments, join them using this operator (probably the only thing besides OR is AND).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
required (boolean; defaults to ‘False’) – Reject queries not filling the InputKeys of this CondDesc
silent (boolean; defaults to ‘False’) – Do not produce SQL from this CondDesc. This can be used to convey meta information to the core. However, in general, a service is a more appropriate place to deal with such information, and thus you should prefer service InputKeys to silent CondDescs.

Structure Children¶

group (contains Element group) – Group child input keys in the input table (primarily interesting for web forms, where this grouping is shown graphically; Set the style property to compact to have a one-line group there)
inputKeys (contains Element inputKey and may be repeated zero or more times) – One or more InputKeys defining the condition’s input.
phraseMaker (contains Element phraseMaker) – Code to generate custom SQL from the input keys

Element coverage¶

The coverage of a resource.

For now, this is attached to the complete resource rather than the table, since this is where it sits in VOResource. DaCHS could be a bit more flexible, allowing different coverages per publish element. It is not right now, though.

Note: Technically, this will introduce or amend the coverage meta element. The information given here will be masked if you define a coverage meta on the service or table level. Just do not do that.

May occur in Element resource.

Atomic Children¶

spatial (unicode string; defaults to <Not given/empty>) – A MOC in ASCII representation giving the ICRS coverage of the resource
spectral (A sequence of intervals (a space-separated pair of floats; defaults to ‘[]’) – Interval(s) of spectral coverage, in Joules of BARYCENTER vacuum messenger particle energy.
temporal (A sequence of intervals (a space-separated pair of floats; defaults to ‘[]’) – Interval(s) of temporal coverage, in MJD (for TT BARYCENTER).

Structure Children¶

updater (contains Element updater) – Rules for automatic computation or updating of coverage information.

Element customDF¶

A custom data function for a service.

Custom data functions can be used to expose certain aspects of a service to Nevow templates. Thus, their definition usually only makes sense with custom templates, though you could, in principle, override built-in render functions.

In the data functions, you have the names ctx for a context and data for the “current data” (i.e., what’s last been set using n:data). In ctx, only use ctx.tag (the tag on which the n:render attribute sits) and, if necessary ctx.request (the t.w request object).

Also, the active renderer is visible as self; the one thing you might want to see from there is self.queryMeta, which contains, for instance, the input parameters.

You can access the embedding service as service, the embedding RD as service.rd.

You can return arbitrary python objects – whatever the render functions can deal with. You could, e.g., write:

<customDF name="now">
  return datetime.datetime.utcnow()
</customDF>

You can use the request to fetch request parameters. Within DaCHS, in addition to the clumsy request.args (mapping bytes to bytes), there is also request.strargs (mapping strings to strings). So, access a query parameter order like this:

sortOrder = ctx.request.strargs.get("order", ["authors"])

May occur in Element service.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Function body of the renderer or data function; the arguments are named request and tag, but you can also use legacy ctx.tag and ctx.request. If data has been set on a tag, you will see it as data.
name (unicode string; defaults to <Undefined>) – Name of the render or data function (use this in the n:render or n:data attribute in custom templates).

Element customRF¶

A custom render function for a service.

Custom render functions can be used to expose certain aspects of a service to Nevow templates. Thus, their definition usually only makes sense with custom templates, though you could, in principle, override built-in render functions.

In the render functions, you have the names ctx for a context and data for the “current data” (i.e., what’s last been set using n:data). In ctx, only use ctx.tag (the tag on which the n:render attribute sits) and, if necessary ctx.request (the t.w request object).

Also, the active renderer is visible as self; the one thing you might want to see from there is self.queryMeta, which contains, for instance, the input parameters (but be careful: the inputTable will be None when input errors are rendered, so better to code using it like this:

if self.queryMeta["inputTable"] and self.queryMeta["inputTable"]...:

You can return anything that can be in a stan DOM. Usually, this will be a string. To return HTML, use the stan DOM available under the T namespace.

As an example, the following code returns the current data as a link:

return ctx.tag[T.a(href=data)[data]]

You can access the embedding service as service, the embedding RD as service.rd.

May occur in Element service.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Function body of the renderer or data function; the arguments are named request and tag, but you can also use legacy ctx.tag and ctx.request. If data has been set on a tag, you will see it as data.
name (unicode string; defaults to <Undefined>) – Name of the render or data function (use this in the n:render or n:data attribute in custom templates).

Element data¶

A description of how to process data from a given set of sources.

Data descriptors bring together a grammar, a source specification and “makes”, each giving a table and a rowmaker to feed the table from the grammar output.

They are the “executable” parts of a resource descriptor. Their ids are used as arguments to gavoimp for partial imports.

May occur in Element resource.

Atomic Children¶

auto (boolean; defaults to ‘True’) – Import this data set if not explicitly mentioned on the command line?
dependents (Zero or more unicode string-typed recreateAfter elements; defaults to ‘[]’) – A data ID to recreate when this resource is remade; use # syntax to reference in other RDs.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
remakeOnDataChange (boolean; defaults to ‘False’) – If this data is made as a dependent of some other, remake it even if the other just has data updates? You generally want to keep this at the default for data that just makes views and the like. You’d want this true when the data produces some statistics or the like.
updating (boolean; defaults to ‘False’) – Keep existing tables on import? You usually want this False unless you have some kind of sources management, e.g., via a sources ignore specification.

Structure Children¶

grammar (contains one of binaryGrammar, cdfHeaderGrammar, columnGrammar, contextGrammar, customGrammar, dictlistGrammar, directGrammar, embeddedGrammar, fitsProdGrammar, freeREGrammar, hdf5Grammar, keyValueGrammar, mySQLDumpGrammar, nullGrammar, transparentGrammar, odbcGrammar, pdsGrammar, reGrammar, rowsetGrammar, unionGrammar, voTableGrammar, csvGrammar, fitsTableGrammar, xmlGrammar) – Grammar used to parse this data set.
makes (contains Element make and may be repeated zero or more times) – Specification of a target table and the rowmaker to feed them.
params (contains Element param and may be repeated zero or more times) – Param (“global columns”) for this data (mostly for VOTable serialization).
registration (contains Element publish (data)) – A registration (to the VO registry) of this table or data collection.
rowmakers (contains Element rowmaker and may be repeated zero or more times) – Embedded build rules (preferably put rowmakers directly into make elements)
sources (contains Element sources) – Specification of sources that should be fed to the grammar.
tables (contains Element table and may be repeated zero or more times) – Embedded table definitions (usually, tables are defined toplevel)

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element DEFAULTS¶

Defaults for macros.

In STREAMs and NXSTREAMs, DEFAULTS let you specify values filled into macros when a FEED doesn’t given them. Macro names are attribute names (or element names, if you insist), defaults are their values.

May occur in Element EDIT, Element events, Element lateEvents, Element NXSTREAM, Element STREAM.

Element dm¶

an annotation of a table in terms of data models.

The content of this element is a Simple Instance Language clause.

May occur in Element table, Element outputTable.

Atomic Children¶

Character content of the element (defaulting to ‘’) – SIL (simple instance language) annotation.

Element EDIT¶

an event stream targeted at editing other structures.

When replaying a stream in the presence of EDITs, the elements are are continually checked against ref. If an element matches, the children of edit will be played back into it.

May occur in Element mixinDef, Element FEED, Element LFEED, Element LOOP.

Atomic Children¶

doc (unicode string; defaults to None) – A description of this stream (should be restructured text).
passivate (unicode string; defaults to None) – If set to True, do not expand active elements immediately in the body of these events (as in an NXSTREAM)
ref (unicode string; defaults to <Undefined>) – Destination of the edits, in the form elementName[<name or id>]

Structure Children¶

DEFAULTS (contains Element DEFAULTS) – A mapping giving defaults for macros expanded in this stream. Macros not defaulted will fail when not given in a FEED’s attributes.

Element events¶

An event stream as a child of another element.

May occur in Element mixinDef, Element FEED, Element LFEED, Element LOOP.

Atomic Children¶

doc (unicode string; defaults to None) – A description of this stream (should be restructured text).
passivate (unicode string; defaults to None) – If set to True, do not expand active elements immediately in the body of these events (as in an NXSTREAM)

Structure Children¶

DEFAULTS (contains Element DEFAULTS) – A mapping giving defaults for macros expanded in this stream. Macros not defaulted will fail when not given in a FEED’s attributes.

Element execute¶

a container for calling code.

This is a cron-like functionality. The jobs are run in separate threads, so they need to be thread-safe with respect to the rest of DaCHS. DaCHS serializes calls, though, so that your code should never run twice at the same time.

At least on CPython, you must make sure your code does not block with the GIL held; this is still in the server process. If you do daring things, fork off (note that you must not use any database connections you may have after forking, which means you can’t safely use the RD passed in). See the docs on Element job.

Then testing/debugging such code, use gavo admin execute rd#id to immediately run the jobs.

May occur in Element resource.

Atomic Children¶

at (Comma-separated list of strings; defaults to <Not given/empty>) – One or more hour:minute pairs at which to run the code each day. This conflicts with every. Optionally, you can prefix each time by one of m<dom> or w<dow> for jobs only to be executed at some day of the month or week, both counted from 1. So, ‘m22 7:30, w3 15:02’ would execute on the 22nd of each month at 7:30 UTC and on every wednesday at 15:02.
debug (boolean; defaults to ‘False’) – If true, on execution of external processes (span or spawnPython), the output will be accumulated and mailed to the administrator. Note that output of the actual cron job itself is not caught (it might turn up in serverStderr). You could use execDef.outputAccum.append(<stuff>) to have information from within the code included.
every (integer; defaults to <Not given/empty>) – Run the job roughly every this many seconds. This conflicts with at. Note that the first execution of such a job is after every/10 seconds, and that the timers start anew at every server restart. So, if you restart often, these jobs may run much more frequently or not at all if the interval is large. If every is smaller than zero, the job will be executed immediately when the RD is being loaded and is then run every abs(every) seconds
title (unicode string; defaults to <Undefined>) – Some descriptive title for the job; this is used in diagnostics.

Structure Children¶

job (contains Element job) – The code to run.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element foreignKey¶

A description of a foreign key relation between this table and another one.

May occur in Element table, Element outputTable.

Atomic Children¶

dest (unicode string; defaults to <Not given/empty>) – Comma- separated list of columns in the target table belonging to its key. No checks for their existence, uniqueness, etc. are done here. If not given, defaults to source.
inTable (id reference; defaults to <Undefined>) – Reference to the table the foreign key points to.
metaOnly (boolean; defaults to ‘False’) – Do not tell the database to actually create the foreign key, just declare it in the metadata. This is for when you want to document a relationship but don’t want the DB to actually enforce this. This is typically a wise thing to do when you have, say a gigarecord of flux/density pairs and only several thousand metadata records – you may want to update the latter without having to tear down the former.
source (unicode string; defaults to <Undefined>) – Comma- separated list of local columns corresponding to the foreign key. No sanity checks are performed here.

Element group¶

A group is a collection of columns, parameters and other groups with a dash of metadata.

Within a group, you can refer to columns or params of the enclosing table by their names. Nothing outside of the enclosing table can be part of a group.

Rather than referring to params, you can also embed them into a group; they will then not be present in the embedding table.

Groups may contain groups.

One application for this is grouping input keys for the form renderer. For such groups, you probably want to give the label property (and possibly cssClass).

May occur in Element condDesc, Element table, Element outputTable, Element inputTable.

Atomic Children¶

description (whitespace normalized unicode string; defaults to None) – A short (one-line) description of the group
name (A name for a table or service parameter. These have to match [A-Za-z_][A-Za-z0-9_]*$.; defaults to None) – Name of the column (must be SQL-valid for onDisk tables)
ucd (unicode string; defaults to None) – The UCD of the group
utype (unicode string; defaults to None) – A utype for the group

Structure Children¶

columnRefs (contains Element columnRef and may be repeated zero or more times) – References to table columns belonging to this group
groups (contains an instance of the embedding element and may be repeated zero or more times) – Sub-groups of this group (names are still referenced from the enclosing table)
paramRefs (contains Element paramRef and may be repeated zero or more times) – Names of table parameters belonging to this group
params (contains Element param and may be repeated zero or more times) – Immediate param elements for this group (use paramref to reference params defined in the parent table)

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element httpUpload¶

An upload going with a URL.

May occur in Element url.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Inline data to be uploaded (conflicts with source)
fileName (unicode string; defaults to ‘upload.dat’) – Remote file name for the uploaded file.
name (unicode string; defaults to <Undefined>) – Name of the upload parameter
source (unicode string; defaults to <Not given/empty>) – Path to a file containing the data to be uploaded.

Element ignoreOn¶

A condition on a row that, if true, causes the row to be dropped.

Here, you can set bail to abort an import when the condition is met rather than just dropping the row.

May occur in Element rowmaker, Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar.

Atomic Children¶

bail (boolean; defaults to ‘False’) – Abort when condition is met?
name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Structure Children¶

triggers (contains any of keyPresent,keyMissing,keyNull,keyIs,not,and and may be repeated zero or more times) – One or more conditions joined by an implicit logical or. See Triggers for information on what can stand here.

Element ignoreSources¶

A specification of sources to ignore.

Sources mentioned here are compared against the inputsDir-relative path of sources generated by sources (cf. Element sources). If there is a match, the corresponding source will not be processed.

You can get ignored files from various sources. If you give more than one source, the set of ignored files is the union of the the individual sets.

fromdbUpdating is a bit special in that the query must return UTC timestamps of the file’s mtime during the last ingest in addition to the accrefs (see the reference documentation for an example).

Macros are expanded in the RD.

May occur in Element sources.

Atomic Children¶

fromdb (unicode string; defaults to None) – A DB query to obtain a set of sources to ignore; the select clause must select exactly one column containing the source key. See also Using fromdb on ignoreSources
fromdbUpdating (unicode string; defaults to None) – A DB query to obtain a set of sources to ignore unless they the timestamp on disk is newer than what’s returned. The query given must return pairs of accrefs and UTC timestamps of the last ingest. See also Using fromdbUpdating on ignoreSources
fromfile (unicode string; defaults to None) – A name of a file containing blacklisted source paths, one per line. Empty lines and lines beginning with a hash are ignored. If you have foolishly used non-ASCII in your filenames: The names in this file are interpreted as utf-8.
patterns (Zero or more unicode string-typed pattern elements; defaults to ‘[]’) – Shell patterns to ignore. Slashes are treated like any other character, i.e., patterns do not know about paths.

Element index¶

A description of an index in the database.

In real databases, indices may be fairly complex things; still, the most common usage here will be to just index a single column:

<index columns="my_col"/>

To index over functions, use the character content; you will have to put parentheses when using expressions. An explicit specification of the index expression is also necessary to allow RE pattern matches using indices in character columns (outside of the C locale). That would be:

<index columns="uri">uri text_pattern_ops</index>

(you still want to give columns so the metadata engine is aware of the index). See section “Operator Classes and Operator Families” in the Postgres documentation for details.

For pgsphere-valued columns, you at the time of writing need to specify the method:

<index columns="coverage" method="GIST"/>

To define q3c indices, use the //scs#q3cindex mixin; if you’re devious enough to require something more flexible, have a look at that mixin’s definition.

If indexed columns take part in a DaCHS-defined view, DaCHS will not notice. You should still declare the indices so users will see them in the metadata; writing:

<index columns="col1, col2, col3"/>

is sufficient for that.

May occur in Element table, Element outputTable.

Atomic Children¶

cluster (boolean; defaults to ‘False’) – Cluster the table according to this index?
columns (Comma-separated list of strings; defaults to ‘’) – Table columns taking part in the index (must be given even if there is an expression building the index and mention all columns taking part in the index generated by it
Character content of the element (defaulting to ‘’) – Raw SQL specifying an expression the table should be indexed for. If not given, the expression will be generated from columns (which is what you usually want).
kind (unicode string; defaults to ‘straight’) – A tag on the index; this is used by the ADQL translation engine in some situations. Consider it a DaCHS implementation detail and ignore it for now.
metaOnly (unicode string; defaults to False) – Do not tell the database to actually create the index, just declare it in the metadata. This is for when you want to tell users of the ADQL engine that columns in, say, a view are indexed when the database cannot actually create the index but the underlying tables provide one.
method (unicode string; defaults to None) – The indexing method, like an index type. In the 8.x, series of postgres, you need to set method=GIST for indices over pgsphere columns; otherwise, you should not need to worry about this.
name (unicode string; defaults to <Undefined>) – Name of the index. Defaults to something computed from columns; the name of the parent table will be prepended in the DB. The default will not work if you have multiple indices on one set of columns.
options (Zero or more unicode string-typed option elements; defaults to ‘[]’) – Index modifiers. For DaCHS, this is free text, except that DaCHS will order INCLUDE, WITH, TABLESPACE, and WHERE clauses it recognises to yield a syntactically correct postgres statement.

Element inputKey¶

A description of a piece of input.

Think of inputKeys as abstractions for input fields in forms, though they are used for services not actually exposing HTML forms as well.

Some of the DDL-type attributes (e.g., references) only make sense here if columns are being defined from the InputKey.

Properties evaluated:

defaultForForm – a value entered into form fields by default (be stingy with those; while it’s nice to not have to set things presumably right for almost everyone, having to delete stuff you don’t want over and over is really annoying).
adaptToRenderer – a true boolean literal here causes the param to be adapted for the renderer (e.g., float could become vexpr-float). You’ll usually not want this, because the expressions are generally evaluated by the database, and the condDescs do the adaptation themselves. This is mainly for rare situations like file uploads in custom cores.
notForRenderer – a renderer name for which this inputKey is suppressed
onlyForRenderer – a renderer name for which this inputKey will be preserved; it will be dropped for all others.

May occur in Element condDesc, Element service, Element contextGrammar, Element inputTable, Element datalinkCore.

Atomic Children¶

Character content of the element (defaulting to <Not given/empty>) – The value of parameter. It is parsed according to the param’s type using the default parser for the type VOTable tabledata.
description (whitespace normalized unicode string; defaults to ‘’) – A short (one-line) description of the values in this column.
displayHint (Display hint; defaults to ‘’) – Suggested presentation; the format is <kw>=<value>{,<kw>=<value>}, where what is interpreted depends on the output format. See, e.g., documentation on HTML renderers and the formatter child of outputFields.
fixup (unicode string; defaults to None) – A python expression the value of which will replace this column’s value on database reads. Write a ___ to access the original value. You can use macros for the embedding table. This is for, e.g., simple URL generation (fixup=“‘internallink{/this/svc}’+___”). It will only kick in when tuples are deserialized from the database, i.e., not for values taken from tables in memory.
inputUnit (unicode string; defaults to None) – Override unit of the table column with this.
multiplicity (unicode string; defaults to None) – Set this to single to have an atomic value (chosen at random if multiple input values are given), forced-single to have an atomic value and raise an exception if multiple values come in, or multiple to receive lists. On the form renderer, this is ignored, and the values are what gavo.formal passes in. If not given, it is single unless there is a values element with options, in which case it’s multiple.
name (A name for a table or service parameter. These have to match [A-Za-z_][A-Za-z0-9_]*$.; defaults to <Undefined>) – Name of the param
note (unicode string; defaults to None) – Reference to a note meta on this table explaining more about this column
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
preparse (unicode string; defaults to <Not given/empty>) – Function body morphing the input string before the type parser sees it. This must return a string or raise an Exception (that will be turned into a ValidationError). The input string is available as input. Note that need to be able to handle incoming None-s gracefully (e.g., input and input.lower()).
required (boolean; defaults to ‘False’) – Record becomes invalid when this column is NULL
showItems (integer; defaults to ‘3’) – Number of items to show at one time on selection widgets.
std (boolean; defaults to ‘False’) – Is this input key part of a standard interface for registry purposes?
tablehead (unicode string; defaults to None) – Terse phrase to put into table headers for this column
type (a type name; the internal type system is similar to SQL’s with some restrictions and extensions. The known atomic types include: smallint, integer, bigint, real, boolean, double precision, text, char, unicode, date, timestamp, time, spoint, scircle, spoly, sbox, smoc, bytea, raw, file, box, vexpr-mjd, vexpr-string, vexpr- float, vexpr-date, pql-string, pql-float, pql-int, pql-date, pql- upload, int4range; defaults to ‘real’) – datatype for the column (SQL-like type system)
ucd (unicode string; defaults to ‘’) – UCD of the column
unit (unicode string; defaults to ‘’) – Unit of the values. Use VOUnits syntax and use single quotes when you use custom units (you should avoid that).
utype (unicode string; defaults to None) – utype for this column
verbLevel (integer; defaults to ‘20’) – Minimal verbosity level at which to include this column
widgetFactory (unicode string; defaults to None) – A python expression for a custom widget factory for this input, e.g., ‘Hidden’ or ‘widgetFactory(TextArea, rows=15, cols=30)’
xtype (unicode string; defaults to None) – VOTable xtype giving the serialization form; you usually do not want to set this, as the xtypes actually used are computed from database type. DaCHS xtypes are only used for a few unsavoury, hopefully temporary, hacks

Structure Children¶

values (contains Element values) – Specification of legal values

Other Children¶

dmRoles (read-only list of roles played by this column in DMs; defaults to []) – Roles played by this column; cannot be assigned to.
meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.
stc (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to unless instructed to do so)
stcUtype (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to)

Element job¶

Python code for use within execute.

The resource descriptor this runs at is available as rd, the execute definition (having such attributes as title, job, plus any properties given in the RD) as execDef.

Note that no I/O capturing takes place (that’s impossible since in general the jobs run within the server). To have actual cron jobs, use execDef.spawn(["cmd", "arg1"...]). This will send a mail on failed execution and also raise a ReportableError in that case.

In the frequent use case of a resdir-relative python program, you can use the execDef.spawnPython(modulePath) function.

If you must stay within the server process, you can do something like:

mod, _ = utils.loadPythonModule(rd.getAbsPath("bin/coverageplot"))
mod.makePlot()

– in that way, your code can sit safely within the resource directory and you still don’t have to manipulate the module path.

May occur in Element execute.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element lateEvents¶

An event stream played back by a mixin when the substrate is being finalised (but before the early processing).

May occur in Element mixinDef.

Atomic Children¶

doc (unicode string; defaults to None) – A description of this stream (should be restructured text).
passivate (unicode string; defaults to None) – If set to True, do not expand active elements immediately in the body of these events (as in an NXSTREAM)

Structure Children¶

DEFAULTS (contains Element DEFAULTS) – A mapping giving defaults for macros expanded in this stream. Macros not defaulted will fail when not given in a FEED’s attributes.

Element macDef¶

A macro definition within an RD.

The macro defined is available on the parent; macros are expanded within the parent (behaviour is undefined if you try a recursive expansion).

May occur in Element resource.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Replacement text of the macro
name (unicode string; defaults to <Undefined>) – Name the macro will be available as

Element make¶

A build recipe for tables belonging to a data descriptor.

All makes belonging to a DD will be processed in the order in which they appear in the file.

May occur in Element data.

Atomic Children¶

parmaker (id reference; defaults to <Not given/empty>) – The parmaker (i.e., mapping rules from grammar parameters to table parameters) for the table being made. You will usually not give a parmaker.
role (unicode string; defaults to None) – The role of the embedded table within the data set
rowSource (One of: parameters, rows; defaults to ‘rows’) – Source for the raw rows processed by this rowmaker.
rowmaker (id reference; defaults to <Not given/empty>) – The rowmaker (i.e., mapping rules from grammar keys to table columns) for the table being made.
table (id reference; defaults to <Undefined>) – Reference to the table to be embedded

Structure Children¶

scripts (contains Element script and may be repeated zero or more times) – Code snippets attached to this object. See Scripting .

Element map¶

A mapping rule.

To specify the source of a mapping, you can either

grab a value from what’s emitted by the grammar or defined using var via the source attribute. The value given for source is converted to a python value and stored.
or give a python expression in the body. In that case, no further type conversion will be attempted.

If neither source or a body is given, map uses the key attribute as its source attribute.

The map rule generates a key/value pair in the result record.

May occur in Element rowmaker.

Atomic Children¶

Character content of the element (defaulting to ‘’) – A python expression giving the value for key.
key (a column name within an SQL table. These have to match the SQL regular_identifier production. In a desperate pinch, you can generate delimited identifiers (that can contain anything) by prefixing the name with ‘quoted/’; defaults to <Undefined>) – Name of the column the value is to end up in.
nullExcs (unicode string; defaults to <Not given/empty>) – Exceptions that should be caught and cause the value to be NULL, separated by commas.
nullExpr (unicode string; defaults to <Not given/empty>) – A python expression for a value that is mapped to NULL (None). Equality is checked after building the value, so this expression has to be of the column type. Use map with the parseWithNull function to catch null values before type conversion.
source (unicode string; defaults to None) – Source key name to convert to column value (either a grammar key or a var).

Element mixinDef¶

A definition for a resource mixin.

Resource mixins are resource descriptor fragments typically rooted in tables (though it’s conceivable that other structures could grow mixin attributes as well).

They are used to define and implement certain behaviours components of the DC software want to see:

products want to be added into their table, and certain fields are required within tables describing products
tables containing positions need some basic machinery to support scs.
siap needs quite a bunch of fields

Mixins consist of events that are played back on the structure mixing in before anything else happens (much like original) and two procedure definitions, viz, processEarly and processLate. These can access the structure that has the mixin as substrate.

processEarly is called at the lexical location of the mixin. processLate is executed just before the parser exits. This is the place to fix up anything that uses the table mixed in. Note, however, that you should be as conservative as possible here – you should think of DC structures as immutable as long as possible.

Programmatically, you can check if a certain table mixes in something by calling its mixesIn method.

Recursive application of mixins, even to separate objects, will deadlock.

May occur in Element resource.

Atomic Children¶

doc (unicode string; defaults to None) – Documentation for this mixin
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
reexpand (boolean; defaults to ‘False’) – Force re-expansion of macros; usually, when replaying, each string is only expanded once, mainly to avoid overly long backslash-fences. Set this to true to force further expansion.
source (id reference; defaults to None) – id of a stream to replay

Structure Children¶

edits (contains Element EDIT and may be repeated zero or more times) – Changes to be performed on the events played back.
events (contains Element events) – Events to be played back into the structure mixing this in at mixin time.
lateEvents (contains Element lateEvents) – Events to be played back into the structure mixing this in at completion time.
pars (contains Element mixinPar and may be repeated zero or more times) – Parameters available for this mixin.
processEarly (contains Element processEarly) – Code executed at element fixup.
processLate (contains Element processLate) – Code executed resource fixup.
prunes (contains Element PRUNE and may be repeated zero or more times) – Conditions for removing items from the playback stream.

Macros predefined here: Macro RSTservicelink, Macro RSTtable, Macro decapitalize, Macro getConfig, Macro internallink, Macro lower, Macro magicEmpty, Macro metaSeq, Macro metaString, Macro quote, Macro rdId, Macro rdIdDotted, Macro reSub, Macro resdir, Macro schema, Macro sql_standardPubDID, Macro sqlquote, Macro test, Macro today, Macro upper, Macro urlquote

Element mixinPar¶

A parameter definition for mixins.

The (optional) body provides a default for the parameter.

May occur in Element mixinDef.

Atomic Children¶

alias (unicode string; defaults to None) – A deprecated name for the parameter
Character content of the element (defaulting to <Not given/empty>) – The default for the parameter. A __NULL__ here does not directly mean None/NULL, but since the content will frequently end up in attributes, it will usually work as presetting None. An empty content means a non-preset parameter, which must be filled in applications. The magic value __EMPTY__ allows presetting an empty string.
description (whitespace normalized unicode string; defaults to None) – Some human-readable description of what the parameter is about
key (unicode string; defaults to <Undefined>) – The name of the parameter
late (boolean; defaults to ‘False’) – Bind the name not at setup time but at applying time. In rowmaker procedures, for example, this allows you to refer to variables like vars or rowIter in the bindings.

Element option¶

A value for enumerated columns.

For presentation purposes, an option can have a title, defaulting to the option’s value.

May occur in Element values.

Atomic Children¶

Character content of the element (defaulting to ‘’) – The value of the option; this is what is used in, e.g., queries and the like.
title (unicode string; defaults to <Not given/empty>) – A Label for presentation purposes; defaults to val.

Element outputField¶

A column for defining the output of a service.

It adds some attributes useful for rendering results, plus functionality specific to certain cores.

The optional formatter overrides the standard formatting code in HTML (which is based on units, ucds, and displayHints). You receive the item from the database as data and must return a string or t.w.template stan. In addition to the standard Functions available for row makers you have queryMeta and t.w.template’s tags in T.

Here’s an example for generating a link to another service using this facility:

<outputField name="more"
    select="array[centerAlpha,centerDelta]" tablehead="More"
    description="More exposures near the center of this plate">
  <formatter><![CDATA[
    return T.a(href=base.makeSitePath("/lswscans/res/positions/q/form?"
        "POS=%s,%s&SIZE=1&INTERSECT=OVERLAPS&cutoutSize=0.5"
        "&__nevow_form__=genForm"%tuple(data)
        ))["More"] ]]>
  </formatter>
</outputField>

Within the code, in addition to data, you see rd and queryMeta.

May occur in Element outputTable.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Columns admit data content but ignore it. This is exclusively a convenience for building columns from params and should not be used for anything else.
description (whitespace normalized unicode string; defaults to ‘’) – A short (one-line) description of the values in this column.
displayHint (Display hint; defaults to ‘’) – Suggested presentation; the format is <kw>=<value>{,<kw>=<value>}, where what is interpreted depends on the output format. See, e.g., documentation on HTML renderers and the formatter child of outputFields.
fixup (unicode string; defaults to None) – A python expression the value of which will replace this column’s value on database reads. Write a ___ to access the original value. You can use macros for the embedding table. This is for, e.g., simple URL generation (fixup=“‘internallink{/this/svc}’+___”). It will only kick in when tuples are deserialized from the database, i.e., not for values taken from tables in memory.
formatter (unicode string; defaults to None) – Function body to render this item to HTML.
hidden (boolean; defaults to ‘False’) – Hide the column from most of the user interface (specifically, you can’t use it in TAP queries or results, and it won’t be in TAP_SCHEMA). You typically want this for internal, administrative columns.
name (a column name within an SQL table. These have to match the SQL regular_identifier production. In a desperate pinch, you can generate delimited identifiers (that can contain anything) by prefixing the name with ‘quoted/’; defaults to <Undefined>) – Name of the column
note (unicode string; defaults to None) – Reference to a note meta on this table explaining more about this column
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
required (boolean; defaults to ‘False’) – Record becomes invalid when this column is NULL
select (unicode string; defaults to <Undefined>) – Use this SQL fragment rather than field name in the select list of a DB based core.
sets (Comma-separated list of strings; defaults to ‘’) – Output sets this field should be included in; ALL includes the field in all output sets.
tablehead (unicode string; defaults to None) – Terse phrase to put into table headers for this column
type (a type name; the internal type system is similar to SQL’s with some restrictions and extensions. The known atomic types include: smallint, integer, bigint, real, boolean, double precision, text, char, unicode, date, timestamp, time, spoint, scircle, spoly, sbox, smoc, bytea, raw, file, box, vexpr-mjd, vexpr-string, vexpr- float, vexpr-date, pql-string, pql-float, pql-int, pql-date, pql- upload, int4range; defaults to ‘real’) – datatype for the column (SQL-like type system)
ucd (unicode string; defaults to ‘’) – UCD of the column
unit (unicode string; defaults to ‘’) – Unit of the values. Use VOUnits syntax and use single quotes when you use custom units (you should avoid that).
utype (unicode string; defaults to None) – utype for this column
verbLevel (integer; defaults to ‘20’) – Minimal verbosity level at which to include this column
wantsRow (boolean; defaults to None) – Does formatter expect the entire row rather than the column value only?
xtype (unicode string; defaults to None) – VOTable xtype giving the serialization form; you usually do not want to set this, as the xtypes actually used are computed from database type. DaCHS xtypes are only used for a few unsavoury, hopefully temporary, hacks

Structure Children¶

values (contains Element values) – Specification of legal values

Other Children¶

dmRoles (read-only list of roles played by this column in DMs; defaults to []) – Roles played by this column; cannot be assigned to.
meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.
stc (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to unless instructed to do so)
stcUtype (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to)

Element outputTable¶

A table that has outputFields for columns.

Cores always have one of these, but they are implicitly defined by the underlying database tables in case of dbCores and such.

Services may define output tables to modify what is coming back fromt the core. Note that this usually only affects the output to web browsers. To use the output table also through VO protocols (and when producing VOTables, FITS files, and the like), you need to set the service’s votableRespectsOutputTable property to True.

May occur in Element service, Element resource.

Atomic Children¶

adql (boolean or ‘hidden’; defaults to ‘False’) – Should this table be available for ADQL queries? In addition to True/False, this can also be ‘hidden’ for tables readable from the TAP machinery but not published in the metadata; this is useful for, e.g., tables contributing to a published view. Warning: adql=hidden is incompatible with setting readProfiles manually.
allProfiles (Comma separated list of profile names, defaults to [db]maintainers from gavorc) – A (comma separated) list of profile names through which the object can be written or administred.
autoCols (Comma-separated list of strings; defaults to ‘’) – Column names obtained from fromTable; you can use shell patterns into the output table’s parent table (in a table core, that’s the queried table; in a service, it’s the core’s output table) here.
dupePolicy (One of: check, drop, dropOld, overwrite; defaults to None) – Handle duplicate rows with identical primary keys manually by raising an error if existing and new rows are not identical (check), dropping the new one (drop), updating the old one (overwrite), or dropping the old one and inserting the new one (dropOld)? Note that if you change this, you will have to re-create the table to make it take effect. The default is to have no special handling (which, if a primary key is there at all, is like check, except an error will be raised even if new and old row are identical).
forceUnique (boolean; defaults to ‘False’) – Ignored legacy attribute.
A mixin reference, typically to support certain protocol. See Mixins.
namePath (id reference; defaults to None) – Reference to an element tried to satisfy requests for names in id references of this element’s children.
nrows (integer; defaults to None) – Approximate number of rows in this table. While you can hard-code this here, running dachs limits will put an estimate into the database.
onDisk (boolean; defaults to ‘False’) – Table in the database rather than in memory?
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
primary (Comma-separated list of strings; defaults to ‘’) – Comma separated names of columns making up the primary key.
readProfiles (Comma separated list of profile names, defaults to [db]queryProfiles from gavorc) – A (comma separated) list of profile names through which the object can be read.
system (boolean; defaults to ‘False’) – Is this a system table? If it is, it will not be dropped on normal imports, and accesses to it will not be logged.
temporary (boolean; defaults to ‘False’) – If this is an onDisk table, make it temporary? This is mostly useful for custom cores and such.
verbLevel (integer; defaults to None) – Copy over columns from fromTable not more verbose than this.
viewStatement (unicode string; defaults to None) – A single SQL statement to create a view. Setting this makes this table a view. The statement will typically be something like CREATE VIEW \qName AS (SELECT \colNames FROM…).

Structure Children¶

columns (contains Element outputField and may be repeated zero or more times) – Output fields for this table.
dm (contains Element dm and may be repeated zero or more times) – Annotations for data models.
foreignKeys (contains Element foreignKey and may be repeated zero or more times) – Foreign keys used in this table
groups (contains Element group and may be repeated zero or more times) – Groups for columns and params of this table
indices (contains Element index and may be repeated zero or more times) – Indices defined on this table
params (contains Element param and may be repeated zero or more times) – Param (“global columns”) for this table.
registration (contains Element publish (data)) – A registration (to the VO registry) of this table or data collection.
scripts (contains Element script and may be repeated zero or more times) – Code snippets attached to this object. See Scripting .
stc (contains Element stc and may be repeated zero or more times) – STC-S definitions of coordinate systems.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Macros predefined here: Macro RSTservicelink, Macro RSTtable, Macro colNames, Macro curtable, Macro decapitalize, Macro getConfig, Macro getParam, Macro internallink, Macro lower, Macro magicEmpty, Macro metaSeq, Macro metaString, Macro nameForUCD, Macro nameForUCDs, Macro qName, Macro quote, Macro rdId, Macro rdIdDotted, Macro reSub, Macro resdir, Macro schema, Macro sql_standardPubDID, Macro sqlquote, Macro tablename, Macro test, Macro today, Macro upper, Macro urlquote

Element par¶

A parameter of a procedure definition.

Bodies of ProcPars are interpreted as python expressions, in which macros are expanded in the context of the procedure application’s parent. If a body is empty, the parameter has no default and has to be filled by the procedure application.

May occur in Element setup.

Atomic Children¶

alias (unicode string; defaults to None) – A deprecated name for the parameter
Character content of the element (defaulting to <Not given/empty>) – The default for the parameter. The special value __NULL__ indicates a NULL (python None) as usual. An empty content means a non-preset parameter, which must be filled in applications. The magic value __EMPTY__ allows presetting an empty string.
description (whitespace normalized unicode string; defaults to None) – Some human-readable description of what the parameter is about
key (unicode string; defaults to <Undefined>) – The name of the parameter
late (boolean; defaults to ‘False’) – Bind the name not at setup time but at applying time. In rowmaker procedures, for example, this allows you to refer to variables like vars or rowIter in the bindings.

Element param¶

A table parameter.

This is like a column, except that it conceptually applies to all rows in the table. In VOTables, params will be rendered as PARAMs.

While we validate the values passed using the DaCHS default parsers, at least the VOTable params will be literal copies of the string passed in.

You can obtain a parsed value from the value attribute.

Null value handling is a bit tricky with params. An empty param (like <param name="x"/>) is always NULL (None in python). In order to allow setting NULL even where syntactially something has to stand, we also turn any __NULL__ to None.

For floats, NaN will also yield NULLs. For integers, you can also use

<param name=”x” type=”integer”><values nullLiteral=”-1”/>-1</params>

For arrays, floats, and strings, the interpretation of values is undefined. Following VOTable practice, we do not tell empty strings and NULLs apart; for internal usage, there is a little hack: __EMPTY__ as literal does set an empty string. This is to allow defaulting of empty strings – in VOTables, these cannot be distinguished from “true” NULLs.

May occur in Element group, Element table, Element data, Element outputTable.

Atomic Children¶

Character content of the element (defaulting to <Not given/empty>) – The value of parameter. It is parsed according to the param’s type using the default parser for the type VOTable tabledata.
description (whitespace normalized unicode string; defaults to ‘’) – A short (one-line) description of the values in this column.
displayHint (Display hint; defaults to ‘’) – Suggested presentation; the format is <kw>=<value>{,<kw>=<value>}, where what is interpreted depends on the output format. See, e.g., documentation on HTML renderers and the formatter child of outputFields.
fixup (unicode string; defaults to None) – A python expression the value of which will replace this column’s value on database reads. Write a ___ to access the original value. You can use macros for the embedding table. This is for, e.g., simple URL generation (fixup=“‘internallink{/this/svc}’+___”). It will only kick in when tuples are deserialized from the database, i.e., not for values taken from tables in memory.
hidden (boolean; defaults to ‘False’) – Ignored on params, just present for constructor compatibility with column
name (A name for a table or service parameter. These have to match [A-Za-z_][A-Za-z0-9_]*$.; defaults to <Undefined>) – Name of the param
note (unicode string; defaults to None) – Reference to a note meta on this table explaining more about this column
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
required (boolean; defaults to ‘False’) – Record becomes invalid when this column is NULL
tablehead (unicode string; defaults to None) – Terse phrase to put into table headers for this column
type (a type name; the internal type system is similar to SQL’s with some restrictions and extensions. The known atomic types include: smallint, integer, bigint, real, boolean, double precision, text, char, unicode, date, timestamp, time, spoint, scircle, spoly, sbox, smoc, bytea, raw, file, box, vexpr-mjd, vexpr-string, vexpr- float, vexpr-date, pql-string, pql-float, pql-int, pql-date, pql- upload, int4range; defaults to ‘real’) – datatype for the column (SQL-like type system)
ucd (unicode string; defaults to ‘’) – UCD of the column
unit (unicode string; defaults to ‘’) – Unit of the values. Use VOUnits syntax and use single quotes when you use custom units (you should avoid that).
utype (unicode string; defaults to None) – utype for this column
verbLevel (integer; defaults to ‘20’) – Minimal verbosity level at which to include this column
xtype (unicode string; defaults to None) – VOTable xtype giving the serialization form; you usually do not want to set this, as the xtypes actually used are computed from database type. DaCHS xtypes are only used for a few unsavoury, hopefully temporary, hacks

Structure Children¶

values (contains Element values) – Specification of legal values

Other Children¶

dmRoles (read-only list of roles played by this column in DMs; defaults to []) – Roles played by this column; cannot be assigned to.
meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.
stc (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to unless instructed to do so)
stcUtype (non-settable internally used value; defaults to None) – Internally used STC information for this column (do not assign to)

Element paramRef¶

A reference from a group to a parameter within a table.

ParamReferences do not support qualified references, i.e., you can only give simple names.

Also note that programmatically, you usually want to resolve ParamReferences within the Table instance, not the table definition.

May occur in Element group.

Atomic Children¶

key (unicode string; defaults to <Undefined>) – The key (i.e., name) of the referenced column or param.
ucd (unicode string; defaults to None) – The UCD of the group
utype (unicode string; defaults to None) – A utype for the group

Element phraseMaker¶

A procedure application for generating SQL expressions from input keys.

PhraseMaker code must yield SQL fragments that can occur in WHERE clauses, i.e., boolean expressions (thus, they must be generator bodies). The clauses yielded by a single condDesc are combined with the joiner set in the containing CondDesc (default=OR).

The following names are available to them:

inputKeys – the list of input keys for the parent CondDesc

inPars – a dictionary mapping inputKey names to the values provided by the user

outPars – a dictionary that is later used as the parameter dictionary to the query.

core – the core to which this phrase maker’s condDesc belongs

To get the standard SQL a single key would generate, say:

yield base.getSQLForField(inputKeys[0], inPars, outPars)

To insert some value into outPars, do not simply use some key into outParse, since, e.g., the condDesc might be used multiple times. Instead, use getSQLKey, maybe like this:

ik = inputKeys[0]
yield "%s BETWEEN %%(%s)s AND %%(%s)s"%(ik.name,
  base.getSQLKey(ik.name, inPars[ik.name]-10, outPars),
  base.getSQLKey(ik.name, inPars[ik.name]+10, outPars))

getSQLKey will make sure unique names in outPars are chosen and enters the values there.

May occur in Element condDesc.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element procDef¶

An embedded procedure.

Embedded procedures are python code fragments with some interface defined by their type. They can occur at various places (which is called procedure application generically), e.g., as row generators in grammars, as apply-s in rowmakers, or as SQL phrase makers in condDescs.

They consist of the actual actual code and, optionally, definitions like the namespace setup, configuration parameters, or a documentation.

The procedure applications compile into python functions with special global namespaces. The signatures of the functions are determined by the type attribute.

ProcDefs are referred to by procedure applications using their id.

May occur in Element resource.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element processEarly¶

A code fragment run by the mixin machinery when the structure being worked on is being finished.

Within processEarly, you can access:

the structure the mixin is applied to as “substrate”
the mixin parameters as “mixinPars”
the parse context as “context”

(the context is particularly handy for context.resolveId)

May occur in Element mixinDef.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element processLate¶

A code fragment run by the mixin machinery when the parser parsing everything exits.

Within processLate, you can access:

the structure mixed in as “substrate”,
the root structure of the whole parse tree as root,
the parse context as “context”,
and the mixin parameters (a dictionary) as “mixinPars”.

May occur in Element mixinDef.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element PRUNE¶

An active tag that lets you selectively delete children of the current object.

You give it regular expression-valued attributes; on the replay of the stream, matching items and their children will not be replayed.

If you give more than one attribute, the result will be a conjunction of the specified conditions.

This only works if the items to be matched are true XML attributes (i.e., not written as children).

For instance, the following will filter out all elements with a name of VERB from the stream:

<PRUNE name="VERB"/>

May occur in Element mixinDef, Element FEED, Element LFEED, Element LOOP.

Element publish (data)¶

A request for registration of a data or table item.

This is much like publish for services, just for data and tables; since they have no renderers, you can only have one register element per such element.

Data registrations may refer to published services that make their data available.

May occur in Element table, Element data, Element outputTable.

Atomic Children¶

services (list of id references (comma separated or in distinct elements); defaults to []) – A DC-internal reference to a service that lets users query that within the data collection; tables with adql=True are automatically declared as isServiceFor the TAP service.
sets (Comma-separated list of strings; defaults to ‘ivo_managed’) – A comma-separated list of sets this data will be published in. To publish data to the VO registry, just say ivo_managed here. Other sets probably don’t make much sense right now. ivo_managed also is the default.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.

Element publish¶

A specification of how a service should be published.

This contains most of the metadata for what is an interface in registry speak.

May occur in Element resRec, Element service.

Atomic Children¶

auxiliary (boolean; defaults to ‘False’) – Auxiliary publications are for capabilities not intended to be picked up for all-VO queries, typically because they are already registered with other services. This is mostly used internally; you probably have no reason to touch it.
render (unicode string; defaults to <Undefined>) – The renderer the publication will point at.
service (id reference; defaults to <Not given/empty>) – Reference for a service actually implementing the capability corresponding to this publication. This is mainly when there is a vs:WebBrowser service accompanying a VO protocol service, and this other service should be published in the same resource record; you would say something like <publish render="form" sets="..." service="web"/>.
sets (Comma-separated list of strings; defaults to ‘’) – Comma- separated list of sets this service will be published in. Predefined are: local=publish on front page, ivo_managed=register with the VO registry. If you leave it empty, ‘local’ publication is assumed.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.

Element regSuite¶

A suite of regression tests.

May occur in Element resource.

Atomic Children¶

sequential (boolean; defaults to ‘False’) – Set to true if the individual tests need to be run in sequence.
title (whitespace normalized unicode string; defaults to None) – A short, human-readable phrase describing what this suite is about.

Structure Children¶

tests (contains Element regTest and may be repeated zero or more times) – Tests making up this suite

Element regTest¶

A regression test.

Tests are defined through url and code elements. See Regression Testing for more information.

May occur in Element regSuite.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
tags (Comma-separated list of strings; defaults to ‘’) – A list of (free-form) tags for this test. Tagged tests are only run when the runner is constructed with at least one of the tags given. This is mainly for restricting tags to production or development servers.
title (whitespace normalized unicode string; defaults to <Undefined>) – A short, human-readable phrase describing what this test is exercising.
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in
url (contains Element url) – The source from which to fetch the test data.

Element resource¶

A resource descriptor.

RDs collect all information about how to parse a particular source (like a collection of FITS images, a catalogue, or whatever), about the database tables the data ends up in, and the services used to access them.

In DaCHS’ RD XML serialisation, they correspond to the root element.

Atomic Children¶

allProfiles (Comma separated list of profile names, defaults to [db]maintainers from gavorc) – A (comma separated) list of profile names through which the object can be written or administred.
readProfiles (Comma separated list of profile names, defaults to [db]queryProfiles from gavorc) – A (comma separated) list of profile names through which the object can be read.
require (unicode string; defaults to None) – Import the named gavo module (for when you need something registered)
resdir (unicode string; defaults to None) – Base directory for source files and everything else belonging to the resource. Use a single dot (.) to say ‘the directory the RD resides in’, which is recommended in modern DaCHS.
schema (unicode string; defaults to <Undefined>) – Database schema for tables defined here. Follow the rule ‘one schema, one RD’ if at all possible. If two RDs share the same schema, the must generate exactly the same permissions for that schema; this means, in particular, that if one has an ADQL-published table, so must the other. In a nutshell: one schema, one RD.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Global condition descriptors for later reference
cores (contains any of adqlCore,customCore,datalinkCore,dbCore,debug Core,fancyQueryCore,fixedQueryCore,nullCore,productCore,pythonCore,r egistryCore,scsCore,siapCutoutCore,ssapCore,tapCore,uploadCore and may be repeated zero or more times) – Cores available in this resource.
coverage (contains Element coverage) – STC coverage of this resource.
dds (contains Element data and may be repeated zero or more times) – Descriptors for the data generated and/or published within this resource.
jobs (contains Element execute and may be repeated zero or more times) – Jobs to be run while this RD is active.
macDefs (contains Element macDef and may be repeated zero or more times) – User-defined macros available on this RD
mixdefs (contains Element mixinDef and may be repeated zero or more times) – Mixin definitions (usually not for users)
outputTables (contains Element outputTable and may be repeated zero or more times) – Canned output tables for later reference.
procDefs (contains Element procDef and may be repeated zero or more times) – Procedure definintions (rowgens, rowmaker apply-s)
resRecs (contains Element resRec and may be repeated zero or more times) – Non-service resources for the IVOA registry. They will be published when gavo publish is run on the RD.
rowmakers (contains Element rowmaker and may be repeated zero or more times) – Transformations for going from grammars to tables. If specified in the RD, they must be referenced from make elements to become active.
scripts (contains Element script and may be repeated zero or more times) – Code snippets attached to this object. See Scripting .
services (contains Element service and may be repeated zero or more times) – Services exposing data from this resource.
tables (contains Element table and may be repeated zero or more times) – A table used or created by this resource
tests (contains Element regSuite and may be repeated zero or more times) – Suites of regression tests connected to this RD.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Macros predefined here: Macro RSTcc0, Macro RSTccby, Macro RSTccbysa, Macro RSTservicelink, Macro RSTtable, Macro decapitalize, Macro getConfig, Macro internallink, Macro lower, Macro magicEmpty, Macro metaSeq, Macro metaString, Macro quote, Macro rdId, Macro rdIdDotted, Macro reSub, Macro resdir, Macro schema, Macro sql_standardPubDID, Macro sqlquote, Macro test, Macro today, Macro upper, Macro urlquote

Element resRec¶

A resource for pure registration purposes.

A Resource without DaCHS defined behaviour. This can be Organizations or Instruments, but possibly also external services

All resources must either have an id (which is used in the construction of their IVOID), or you must give an identifier meta item.

You must further set the following meta items:

resType specifying the kind of resource record. You should not use this element to build resource records for services or tables (use the normal elements, even if the actual resources are external to DaCHS). resType can be registry, organization, authority, deleted, or anything else for which registry.builders has a handling class.

title

subject(s)

description

referenceURL

creationDate

Additional meta keys (e.g., accessURL for a registry) may be required depending on resType. See the registry section in the operator’s guide.

ResRecs can also have publication children. These will be turned into the appropriate capabilities depending on the value of the render attribute.

May occur in Element resource.

Structure Children¶

publications (contains Element publish and may be repeated zero or more times) – Capabilities the record should have (this is empty for standards, organisations, instruments, etc.)

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.

Element rowmaker¶

A definition of the mapping between grammar input and finished rows ready for shipout.

Rowmakers consist of variables, procedures and mappings. They result in a python callable doing the mapping. In python code within rowmaker elements, you can use a large number of functions. See Functions available for row makers in the reference documentation.

RowmakerDefs double as macro packages for the expansion of various macros. The standard macros will need to be quoted, the rowmaker macros above yield python expressions.

Within map and var bodies as well as late apply pars and apply bodies, you can refer to the grammar input as vars[“name”] or, shorter @name.

To add output keys, use map or, in apply bodies, add keys to the result dictionary.

May occur in Element data, Element resource.

Atomic Children¶

idmaps (Comma-separated list of strings; defaults to ‘’) – List of column names that are just “mapped through” (like map with key only); you can use shell patterns to select multiple columns at once.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
simplemaps (Comma-separated list of <identifier>:<identifier> pairs; defaults to None) – Abbreviated notation for <map source>; each pair is destination:source

Structure Children¶

apps (contains Element apply and may be repeated zero or more times) – Procedure applications.
ignoreOn (contains Element ignoreOn) – Conditions on the input record coming from the grammar to cause the input record to be dropped by the rowmaker, i.e., for this specific table. If you need to drop a row for all tables being fed, use a trigger on the grammar.
maps (contains Element map and may be repeated zero or more times) – Mapping rules.
vars (contains Element var and may be repeated zero or more times) – Definitions of intermediate variables.

Macros predefined here: Macro RSTservicelink, Macro RSTtable, Macro decapitalize, Macro dlMetaURI, Macro docField, Macro fullPath, Macro getConfig, Macro inputRelativePath, Macro inputSize, Macro internallink, Macro lastSourceElements, Macro lower, Macro magicEmpty, Macro metaSeq, Macro metaString, Macro property, Macro qName, Macro quote, Macro rdId, Macro rdIdDotted, Macro reSub, Macro resdir, Macro rootlessPath, Macro rowsMade, Macro rowsProcessed, Macro schema, Macro sourceCDate, Macro sourceDate, Macro sql_standardPubDID, Macro sqlquote, Macro srcstem, Macro standardPubDID, Macro test, Macro today, Macro upper, Macro urlquote

Element script¶

A script, i.e., some executable item within a resource descriptor.

The content of scripts is given by their type – usually, they are either python scripts or SQL with special rules for breaking the script into individual statements (which are basically like python’s).

The special language AC_SQL is like SQL, but execution errors are ignored. This is not what you want for most data RDs (it’s intended for housekeeping scripts).

See Scripting.

May occur in Element make, Element table, Element outputTable, Element resource.

Atomic Children¶

Character content of the element (defaulting to ‘’) – The script body.
lang (One of: AC_SQL, SQL, python; defaults to <Undefined>) – Language of the script.
name (unicode string; defaults to <Not given/empty>) – A human- consumable designation of the script.
notify (boolean; defaults to ‘True’) – Send out a notification when running this script.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
type (One of: afterMeta, beforeDrop, newSource, postCreation, preCreation, preImport, preIndex, sourceDone; defaults to <Undefined>) – Point of time at which script is to run (not all script types are allowed on all elements).

Element service¶

A service definition.

A service is a combination of a core and one or more renderers. They can be published, and they carry the metadata published into the VO.

You can set the defaultSort property on the service to a name of an output column to preselect a sort order. Note again that this will slow down responses for all but the smallest tables unless there is an index on the corresponding column.

Properties evaluated:

votableRespectsOutputTable – usually, VOTable output puts in all columns from the underlying database table with low enough verbLevel (essentially). When this property is “True” (case-sensitive), that’s not done and only the service’s output table is evaluated.
fixedFormat – if this is set to “True”, the form renderer will not produce an output format selector (and we shouldn’t produce DALI RESPONSEFORMAT metadata).

May occur in Element resource.

Atomic Children¶

allowed (Comma-separated list of strings; defaults to ‘’) – Names of renderers allowed on this service; leave empty to allow the form renderer only.
core (id reference; defaults to <Undefined>) – The core that does the computations for this service. Instead of a reference, you can use an immediate element of some registered core.
customPage (unicode string; defaults to None) – resdir-relative path to custom page code. It is used by the ‘custom’ renderer
defaultRenderer (unicode string; defaults to None) – A name of a renderer used when none is provided in the URL (lets you have shorter URLs).
limitTo (unicode string; defaults to None) – Limit access to the group given; the empty default disables access control.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

customDFs (contains Element customDF and may be repeated zero or more times) – Custom data functions for use in custom templates.
customRFs (contains Element customRF and may be repeated zero or more times) – Custom render functions for use in custom templates.
outputTable (contains Element outputTable) – The output fields of this service.
publications (contains Element publish and may be repeated zero or more times) – Sets and renderers this service is published with.
serviceKeys (contains Element inputKey and may be repeated zero or more times) – Input widgets for processing by the service, e.g. output sets.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.
template (mapping; the value is the element content, the key is in the ‘key’ (or, equivalently, key) attribute) – Custom nevow templates for this service; use key “form” to replace the Form renderer’s standard template; qp uses resulttable and resultline depending on whether there’s many result lines or just one. Start the path with two slashes to access system templates.

Element setup¶

Prescriptions for setting up a namespace for a procedure application.

You can add names to this namespace you using par(ameter)s. If a parameter has no default and an procedure application does not provide them, an error is raised.

You can also add names by providing a code attribute containing a python function body in code. Within, the parameters are available. The procedure application’s parent can be accessed as parent. All names you define in the code are available as globals to the procedure body.

Caution: Macros are expanded within the code; this means you need double backslashes if you want a single backslash in python code.

May occur in Element phraseMaker, Element apply, Element job, Element processEarly, Element processLate, Element procDef, Element regTest, Element rowfilter, Element sourceFields, Element iterator, Element pargetter, Element makeQuery, Element dataFormatter, Element dataFunction, Element descriptorGenerator, Element metaMaker, Element coreProc.

Atomic Children¶

codeFrags (Zero or more unicode string-typed code elements; defaults to ‘[]’) – Python function bodies setting globals for the function application. Macros are expanded in the context of the procedure’s parent.
imports (unicode string; defaults to <Not given/empty>) – A list of comma-separated imports to put into the code’s namespace. Dottesd specs like a.b.c are converted to from a.b import c
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

pars (contains Element par and may be repeated zero or more times) – Names to add to the procedure’s global namespace.

Element sources¶

A Specification of a data descriptor’s inputs.

This will typcially be files taken from a file system. If so, DaCHS will, in each directory, process the files in alphabetical order. No guarantees are made as to the sequence directories are processed in.

Multiple patterns are processed in the order given in the RD.

May occur in Element data.

Atomic Children¶

Character content of the element (defaulting to ‘’) – A single file name (this is for convenience)
items (Zero or more unicode string-typed item elements; defaults to ‘[]’) – String literals to pass to grammars. In contrast to patterns, they are not interpreted as file names but passed to the grammar verbatim. Normal grammars do not like this. It is mainly intended for use with custom or null grammars.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
patterns (Zero or more unicode string-typed pattern elements; defaults to ‘[]’) – Paths to the source files. You can use shell patterns here.
recurse (boolean; defaults to ‘False’) – Search for pattern(s) recursively in their directory part(s)?

Structure Children¶

ignoredSources (contains Element ignoreSources) – Specification of sources that should not be processed although they match patterns. Typically used in update-type data descriptors.

Element stc¶

A definition of a space-time coordinate system using STC-S.

May occur in Element table, Element outputTable.

Atomic Children¶

Character content of the element (defaulting to ‘’) – An STC-S string with column references (using quote syntax) instead of values

Element table¶

A definition of a table, both on-disk and internal.

Some attributes are ignored for in-memory tables, e.g., roles or adql.

Properties for tables:

supportsModel – a short name of a data model supported through this table (for TAPRegExt dataModel); you can give multiple names separated by commas.
supportsModelURI – a URI of a data model supported through this table. You can give multiple URIs separated by blanks.
forceStats – if present (with any value), dachs limits on the embedding RD will obtain statistics of this even if it is a view.

Somewhat inconsistently, to set a table’s utype if you have to, set its utype meta.

If you give multiple data model names or URIs, the sequences of names and URIs must be identical (in particular, each name needs a URI).

May occur in Element data, Element resource.

Atomic Children¶

adql (boolean or ‘hidden’; defaults to ‘False’) – Should this table be available for ADQL queries? In addition to True/False, this can also be ‘hidden’ for tables readable from the TAP machinery but not published in the metadata; this is useful for, e.g., tables contributing to a published view. Warning: adql=hidden is incompatible with setting readProfiles manually.
allProfiles (Comma separated list of profile names, defaults to [db]maintainers from gavorc) – A (comma separated) list of profile names through which the object can be written or administred.
dupePolicy (One of: check, drop, dropOld, overwrite; defaults to None) – Handle duplicate rows with identical primary keys manually by raising an error if existing and new rows are not identical (check), dropping the new one (drop), updating the old one (overwrite), or dropping the old one and inserting the new one (dropOld)? Note that if you change this, you will have to re-create the table to make it take effect. The default is to have no special handling (which, if a primary key is there at all, is like check, except an error will be raised even if new and old row are identical).
forceUnique (boolean; defaults to ‘False’) – Ignored legacy attribute.
A mixin reference, typically to support certain protocol. See Mixins.
namePath (id reference; defaults to None) – Reference to an element tried to satisfy requests for names in id references of this element’s children.
nrows (integer; defaults to None) – Approximate number of rows in this table. While you can hard-code this here, running dachs limits will put an estimate into the database.
onDisk (boolean; defaults to ‘False’) – Table in the database rather than in memory?
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
primary (Comma-separated list of strings; defaults to ‘’) – Comma separated names of columns making up the primary key.
readProfiles (Comma separated list of profile names, defaults to [db]queryProfiles from gavorc) – A (comma separated) list of profile names through which the object can be read.
system (boolean; defaults to ‘False’) – Is this a system table? If it is, it will not be dropped on normal imports, and accesses to it will not be logged.
temporary (boolean; defaults to ‘False’) – If this is an onDisk table, make it temporary? This is mostly useful for custom cores and such.
viewStatement (unicode string; defaults to None) – A single SQL statement to create a view. Setting this makes this table a view. The statement will typically be something like CREATE VIEW \qName AS (SELECT \colNames FROM…).

Structure Children¶

columns (contains Element column and may be repeated zero or more times) – Columns making up this table.
dm (contains Element dm and may be repeated zero or more times) – Annotations for data models.
foreignKeys (contains Element foreignKey and may be repeated zero or more times) – Foreign keys used in this table
groups (contains Element group and may be repeated zero or more times) – Groups for columns and params of this table
indices (contains Element index and may be repeated zero or more times) – Indices defined on this table
params (contains Element param and may be repeated zero or more times) – Param (“global columns”) for this table.
registration (contains Element publish (data)) – A registration (to the VO registry) of this table or data collection.
scripts (contains Element script and may be repeated zero or more times) – Code snippets attached to this object. See Scripting .
stc (contains Element stc and may be repeated zero or more times) – STC-S definitions of coordinate systems.

Other Children¶

meta – a piece of meta information, giving at least a name and some content. See Metadata on what is permitted here.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element updater¶

Information on where and how to update a piece of coverage information.

May occur in Element coverage.

Atomic Children¶

mocOrder (integer; defaults to ‘6’) – Maximal HEALpix order to use in coverage MOCs (6 is about a degree resolution, each additional point doubles resolution).
sourceTable (id reference; defaults to <Not given/empty>) – A table from which to compute coverage by default.
spaceTable (id reference; defaults to <Not given/empty>) – A table from which to compute spatial coverage (overrides sourceTable).
spectralTable (id reference; defaults to <Not given/empty>) – A table from which to compute spectral coverage (overrides sourceTable)
timeTable (id reference; defaults to <Not given/empty>) – A table from which to compute temporal coverage (overrides sourceTable)

Element url¶

A source document for a regression test.

As string URLs, they specify where to get data from, but the additionally let you specify uploads, authentication, headers and http methods, while at the same time saving you manual escaping of parameters.

The bodies is the path to run the test against. This is interpreted as relative to the RD if there’s no leading slash, relative to the server if there’s a leading slash, and absolute if there’s a scheme.

The attributes are translated to parameters, except for a few pre-defined names. If you actually need those as URL parameters, should at us and we’ll provide some way of escaping these.

We don’t actually parse the URLs coming in here. GET parameters are appended with a & if there’s a ? in the existing URL, with a ? if not. Again, shout if this is too dumb for you (but urlparse really isn’t all that robust either…)

May occur in Element regTest.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Base for URL generation; embedded whitespace will be removed, so you’re free to break those wherever you like.
httpAuthKey (unicode string; defaults to <Not given/empty>) – A key into ~/.gavo/test.creds to find a user/password pair for this request.
httpChunkSize (integer; defaults to None) – If there are uploads, upload them in chunks of this many bytes using chunked encoding.
httpHonorRedirects (boolean; defaults to ‘False’) – Follow 30x redirects instead of just using status, headers, and payload of the initial request.
httpMethod (unicode string; defaults to ‘GET’) – Request method; usually one of GET or POST
parSet (One of: TAP, form; defaults to <Not given/empty>) – Preselect a default parameter set; form gives what our framework adds to form queries.
postPayload (unicode string; defaults to <Not given/empty>) – Path to a file containing material that should go with a POST request (conflicts with additional parameters).

Structure Children¶

uploads (contains Element httpUpload and may be repeated zero or more times) – HTTP uploads to add to request (must have httpMethod=”POST”)

Other Children¶

value (mapping; the value is the element content, the key is in the ‘key’ (or, equivalently, key) attribute) – Additional HTTP headers to pass.
(ignore)

Element values¶

Information on a column’s values, in particular its domain.

This is quite like the values element in a VOTable. In particular, to accommodate VOTable usage, we require nullLiteral to be a valid literal for the parent’s type.

Note that DaCHS does not validate for constraints from values on table import. This is mainly because before dachs limits has run, values may not represent the new dataset in semiautomatic values.

With HTTP parameters, values validation does take place (but again, that’s mostly not too helpful because there are query languages sitting in between most of the time).

Hence, the main utility of values is metadata declaration, both in the form renderer (where they become placeholders) and in datalink (where they are communicated as VOTable values).

May occur in Element param, Element inputKey, Element column, Element outputField.

Atomic Children¶

caseless (boolean; defaults to ‘False’) – When validating, ignore the case of string values. For non-string types, behaviour is undefined (i.e., DaCHS is going to spit on you).
default (unicode string; defaults to None) – A default value (currently only used for options).
fillFactor (unicode string; defaults to None) – Ratio of non- NULL values to the number of rows in the embedding table.
fromdb (unicode string; defaults to None) – A query fragment returning just one column to fill options from (will add to options if some are given). Do not write SELECT or anything, just the column name and the where clause. Do not do this for large tables even if there are reasonably few values, because there is no good way to speed up this kind of query using indices.
max (unicode string; defaults to None) – Maximum acceptable value as a datatype literal
median (unicode string; defaults to None) – Median of the distribution of this column.
min (unicode string; defaults to None) – Minimum acceptable value as a datatype literal
multiOk (boolean; defaults to ‘False’) – Deprecated, use multiplicity=multiple on input keys instead.
nullLiteral (unicode string; defaults to None) – An appropriate value representing a NULL for this column in VOTables and similar places. You usually should only set it for integer types and chars. Note that rowmakers make no use of this nullLiteral, i.e., you can and should choose null values independently of your source. Again, for reals, floats and (mostly) text you probably do not want to do this.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
percentile03 (unicode string; defaults to None) – Value at the 3rd percentile of the distribution of this column.
percentile97 (unicode string; defaults to None) – Value at the 97rd percentile of the distribution of this column.

Structure Children¶

options (contains Element option and may be repeated zero or more times) – List of acceptable values (if set)

Element var¶

A definition of a rowmaker variable.

It consists of a name and a python expression, including function calls. The variables are entered into the input row coming from the grammar.

var elements are evaluated before apply elements, in the sequence they are in the RD. You can refer to keys defined by vars already evaluated in the usual @key manner.

May occur in Element rowmaker.

Atomic Children¶

Character content of the element (defaulting to ‘’) – A python expression giving the value for key.
key (a column name within an SQL table. These have to match the SQL regular_identifier production. In a desperate pinch, you can generate delimited identifiers (that can contain anything) by prefixing the name with ‘quoted/’; defaults to <Undefined>) – Name of the column the value is to end up in.
nullExcs (unicode string; defaults to <Not given/empty>) – Exceptions that should be caught and cause the value to be NULL, separated by commas.
nullExpr (unicode string; defaults to <Not given/empty>) – A python expression for a value that is mapped to NULL (None). Equality is checked after building the value, so this expression has to be of the column type. Use map with the parseWithNull function to catch null values before type conversion.
source (unicode string; defaults to None) – Source key name to convert to column value (either a grammar key or a var).

Active Tags ¶

The following tags are “active”, which means that they do not directly contribute to the RD parsed. Instead they define, replay, or edit streams of elements.

Element FEED¶

An active tag that takes an event stream and replays the events, possibly filling variables.

This element supports arbitrary attributes with unicode values. These values are available as macros for replayed values.

Atomic Children¶

reexpand (boolean; defaults to ‘False’) – Force re-expansion of macros; usually, when replaying, each string is only expanded once, mainly to avoid overly long backslash-fences. Set this to true to force further expansion.
source (id reference; defaults to None) – id of a stream to replay

Structure Children¶

edits (contains Element EDIT and may be repeated zero or more times) – Changes to be performed on the events played back.
events (contains Element events) – Alternatively to source, an XML fragment to be replayed
prunes (contains Element PRUNE and may be repeated zero or more times) – Conditions for removing items from the playback stream.

Element LFEED¶

A ReplayedEventStream that does not expand active tag macros.

You only want this when embedding a stream into another stream that could want to expand the embedded macros.

Atomic Children¶

reexpand (boolean; defaults to ‘False’) – Force re-expansion of macros; usually, when replaying, each string is only expanded once, mainly to avoid overly long backslash-fences. Set this to true to force further expansion.
source (id reference; defaults to None) – id of a stream to replay

Structure Children¶

edits (contains Element EDIT and may be repeated zero or more times) – Changes to be performed on the events played back.
events (contains Element events) – Alternatively to source, an XML fragment to be replayed
prunes (contains Element PRUNE and may be repeated zero or more times) – Conditions for removing items from the playback stream.

Element LOOP¶

An active tag that replays a feed several times, each time with different values.

Atomic Children¶

codeItems (unicode string; defaults to None) – A python generator body that yields dictionaries that are then used as loop items. You can access the parse context as the context variable in these code snippets. context.replayTarget is the struct the loop is feeding to. You also get the utils and base namespaces, though using them it a bit dangerous (we don’t guarantee their stability).
csvItems (unicode string; defaults to None) – The items to loop over, in CSV-with-labels format.
listItems (unicode string; defaults to None) – The items to loop over, as space-separated single items. Each item will show up once, as ‘item’ macro.
reexpand (boolean; defaults to ‘False’) – Force re-expansion of macros; usually, when replaying, each string is only expanded once, mainly to avoid overly long backslash-fences. Set this to true to force further expansion.
source (id reference; defaults to None) – id of a stream to replay

Structure Children¶

edits (contains Element EDIT and may be repeated zero or more times) – Changes to be performed on the events played back.
events (contains Element events) – Alternatively to source, an XML fragment to be replayed
prunes (contains Element PRUNE and may be repeated zero or more times) – Conditions for removing items from the playback stream.

Element NXSTREAM¶

An event stream that records events, not expanding active tags.

Normal event streams expand embedded active tags in place. This is frequently what you want, but it means that you cannot, e.g., fill in loop variables through stream macros.

With non-expanded streams, you can do that:

<NXSTREAM id="cols">
  <LOOP listItems="\stuff">
    <events>
      <column name="\item"/>
    </events>
  </LOOP>
</NXSTREAM>
<table id="foo">
  <FEED source="cols" stuff="x y"/>
</table>

Note that the normal innermost-only rule for macro expansions within active tags does not apply for NXSTREAMS. Macros expanded by a replayed NXSTREAM will be re-expanded by the next active tag that sees them (this is allow embedded active tags to use macros; you need to double-escape macros for them, of course).

Atomic Children¶

doc (unicode string; defaults to None) – A description of this stream (should be restructured text).

Structure Children¶

DEFAULTS (contains Element DEFAULTS) – A mapping giving defaults for macros expanded in this stream. Macros not defaulted will fail when not given in a FEED’s attributes.

Element STREAM¶

An active tag that records events as they come in.

Their only direct effect is to leave a trace in the parser’s id map. The resulting event stream can be played back later.

Atomic Children¶

doc (unicode string; defaults to None) – A description of this stream (should be restructured text).

Structure Children¶

DEFAULTS (contains Element DEFAULTS) – A mapping giving defaults for macros expanded in this stream. Macros not defaulted will fail when not given in a FEED’s attributes.

Grammars Available ¶

The following elements are all grammar related. All grammar elements can occur in data descriptors.

Element binaryGrammar¶

A grammar that builds rowdicts from binary data.

The grammar expects the input to be in fixed-length records. the actual specification of the fields is done via a binaryRecordDef element.

Atomic Children¶

armor (One of: fortran; defaults to None) – Record armoring; by default it’s None meaning the data was dumped to the file sequentially. Set it to fortran for fortran unformatted files (4 byte length before and after the payload).
enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
skipBytes (integer; defaults to ‘0’) – Number of bytes to skip before parsing records.

Structure Children¶

fieldDefs (contains Element binaryRecordDef) – Definition of the record.
ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Macros predefined here: Macro RSTservicelink, Macro RSTtable, Macro colNames, Macro decapitalize, Macro dlMetaURI, Macro fullDLURL, Macro getConfig, Macro inputRelativePath, Macro inputSize, Macro internallink, Macro lower, Macro magicEmpty, Macro metaSeq, Macro metaString, Macro property, Macro quote, Macro rdId, Macro rdIdDotted, Macro reSub, Macro resdir, Macro rootlessPath, Macro schema, Macro sourceDate, Macro splitPreviewPath, Macro sql_standardPubDID, Macro sqlquote, Macro srcstem, Macro standardPreviewPath, Macro test, Macro today, Macro upper, Macro urlquote

Element binaryRecordDef¶

A definition of a binary record.

A binary records consists of a number of binary fields, each of which is defined by a name and a format code. The format codes supported here are a subset of what python’s struct module supports. The widths given below are for big, little, and packed binfmts. For native (which is the default), it depends on your platform.

<number>s – <number> characters making up a string
b,B – signed and unsigned byte (8 bit)
h,H – signed and unsigned short (16 bit)
i,I – signed and unsigned int (32 bit)
q,Q – signed and unsigned long (64 bit)
f,d – float and double.

The content of this element gives the record structure in the format <name>(<code>){<whitespace><name>(<code>)} where <name> is a c-style identifier.

May occur in Element binaryGrammar.

Atomic Children¶

binfmt (One of: big, little, native, packed; defaults to ‘native’) – Binary format of the input data; big and little stand for msb first and lsb first, and packed is like native except no alignment takes place.
Character content of the element (defaulting to ‘’) – The enumeration of the record fields.

Element cdfHeaderGrammar¶

A grammar that returns the header dictionary of a CDF file (global attributes).

This grammar yields a single dictionary per file, which corresponds to the global attributes. The values in this dictionary may have complex structure; in particular, sequences are returned as lists.

To use this grammar, additional software is required that (by 2014) is not packaged for Debian. See https://pythonhosted.org/SpacePy/install.html for installation instructions. Note that you must install the CDF library itself as described further down on that page; the default installation instructions do not install the library in a public place, so if you use these, you’ll have to set CDF_LIB to the right value, too, before running dachs imp.

Atomic Children¶

autoAtomize (boolean; defaults to ‘False’) – Unpack 1-element lists to their first value.
enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
mapKeys (contains Element mapKeys) – Prescription for how to map labels keys to grammar dictionary keys
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element columnGrammar¶

A grammar that builds rowdicts out of character index ranges.

This works by using the colRanges attribute like <col key=”mag”>12-16</col>, which will take the characters 12 through 16 inclusive from each input line to build the input column mag.

As a shortcut, you can also use the colDefs attribute; it contains a string of the form {<key>:<range>}, i.e., a whitespace-separated list of colon-separated items of key and range as accepted by cols, e.g.:

<colDefs>
  a: 3-4
  _u: 7
</colDefs>

Atomic Children¶

colDefs (unicode string; defaults to None) – Shortcut way of defining cols
commentIntroducer (unicode string; defaults to <Not given/empty>) – A character sequence that, when found at the beginning of a line makes this line ignored
enc (unicode string; defaults to None) – Encoding of the source file(s).
gunzip (boolean; defaults to ‘False’) – Unzip sources while reading? (Deprecated, use preFilter=’zcat’)
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
preFilter (unicode string; defaults to None) – Shell command to pipe the input through before passing it on to the grammar. Classical examples include zcat or bzcat, but you can commit arbitrary shell atrocities here.
strip (boolean; defaults to ‘True’) – Strip all parsed strings?
topIgnoredLines (integer; defaults to ‘0’) – Skip this many lines at the top of each source file.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

col (mapping; the value is the element content, the key is in the ‘key’ (or, equivalently, key) attribute) – Mapping of source keys to column ranges.
property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element contextGrammar¶

A grammar for web inputs.

The source tokens for context grammars are dictionaries; these are either typed dictionaries from gavo.formal, where the values usually are atomic, or, preferably, the dictionaries of lists from request.strargs.

ContextGrammars never yield rows, so they’re probably fairly useless in normal circumstances.

In normal usage, they just yield a single parameter row, corresponding to the source dictionary possibly completed with defaults, where non-requried input keys get None defaults where not given. Missing required parameters yield errors.

This parameter row honors the multiplicity specification, i.e., single or forced-single are just values, multiple are lists. The content are parsed values (using the InputKeys’ parsers).

Since most VO protocols require case-insensitive matching of parameter names, matching of input key names and the keys of the input dictionary is attempted first literally, then disregarding case.

Since ContextGrammars can be parents of inputKeys and thus are a bit table-like, they can also carry metadata.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
inputTD (id reference; defaults to <Not given/empty>) – The input table from which to take the input keys
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
inputKeys (contains Element inputKey and may be repeated zero or more times) – Extra input keys not defined in the inputTD. This is used when services want extra input processed by them rather than their core.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element csvGrammar¶

A grammar that uses python’s csv module to parse files.

Note that these grammars by default interpret the first line of the input file as the column names. When your files don’t follow that convention, you must give names (as in names='raj2000, dej2000, magV'), or you’ll lose the first line and have silly column names.

If data is left after filling the defined keys, it is available under the NOTASSIGNED key.

Atomic Children¶

delimiter (unicode string; defaults to ‘,’) – CSV delimiter
enc (unicode string; defaults to None) – Encoding of the source file(s).
gunzip (boolean; defaults to ‘False’) – Unzip sources while reading? (Deprecated, use preFilter=’zcat’)
names (Comma-separated list of strings; defaults to None) – Names for the parsed fields, in sequence of the comma separated values. The default is to read the field names from the first line of the csv file. You can use macros here, e.g., \colNames{someTable}.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
preFilter (unicode string; defaults to None) – Shell command to pipe the input through before passing it on to the grammar. Classical examples include zcat or bzcat, but you can commit arbitrary shell atrocities here.
strip (boolean; defaults to ‘False’) – If True, whitespace immediately following a delimiter is ignored.
topIgnoredLines (integer; defaults to ‘0’) – Skip this many lines at the top of each source file.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
mapKeys (contains Element mapKeys) – Prescription for how to map header keys to grammar dictionary keys
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element customGrammar¶

A Grammar with a user-defined row iterator taken from a module.

See the Writing Custom Grammars (in the reference manual) for details.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
isDispatching (boolean; defaults to ‘False’) – Is this a dispatching grammar (i.e., does the row iterator return pairs of role, row rather than only rows)?
module (unicode string; defaults to <Undefined>) – Path to module containing your row iterator.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element dictlistGrammar¶

A grammar that “parses” from lists of dicts.

Actually, it will just return the dicts as they are passed. This is mostly useful internally, though it might come in handy in custom code.

Atomic Children¶

asPars (boolean; defaults to ‘False’) – Just return the first item of the list as parameters row and exit?
enc (unicode string; defaults to None) – Encoding of the source file(s).
notify (boolean; defaults to ‘False’) – Enable notification of begin/end of processing (as for other grammars; dictlistGrammars are usually internal, where notification is unwelcome). Note that the -M option of gavo imp will only work if you set it to true.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element directGrammar¶

A user-defined external grammar.

See the booster.html on user-defined code on more on direct grammars.

You will almost always use these in connection with C code generated by dachs mkboost.

Properties:

dataset: For HDF5 boosters, the HDF path to the table to be imported (in vaex, not including “columns”). Certain, hopefully sensible, defaults apply depending on the HDF5 dialect.
chunkSize: In HDF5 boosters, now many rows to pull into memory at one time (defaults to 5000).

Atomic Children¶

autoNull (unicode string; defaults to None) – Use this string as general NULL value (when reading from plain text).
cBooster (unicode string; defaults to <Undefined>) – resdir- relative path to the booster C source.
customFlags (unicode string; defaults to ‘’) – Pass these flags to the C compiler when building the booster.
extension (integer; defaults to ‘1’) – For FITS table boosters, get the table from this extension.
gzippedInput (boolean; defaults to ‘False’) – Pipe gzip before booster? (will not work for FITS)
ignoreBadRecords (boolean; defaults to ‘False’) – Let booster ignore invalid records?
preFilter (unicode string; defaults to None) – Pipe input through this program before handing it to the booster; this string is shell-expanded (will not work for FITS).
recordSize (integer; defaults to ‘4000’) – For bin boosters, read this many bytes to make up a record; for line-based boosters, this is the maximum length of an input line.
splitChar (unicode string; defaults to ‘|’) – For split boosters, use this as the separator.
type (One of: bin, col, fits, hdf5rootarrs, hdf5vaex, split; defaults to ‘col’) – Make code for a booster parsing by column indices (col), by splitting along separators (split), by reading fixed-length binary records (bin), for from FITS binary tables (fits).

Structure Children¶

mapKeys (contains Element mapKeys) – For certain input formats (FITS, HDF5), map input column names to database column name. For instance, to generate code for a FITS column flx to end up in the DB column flux, say flux:flx).

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element embeddedGrammar¶

A Grammar defined by a code application.

To define this grammar, write a ProcApp iterator leading to code yielding row dictionaries. The grammar input is available as self.sourceToken; for normal grammars within data elements, that would be a fully qualified file name.

Grammars can also return one “parameter” dictionary per source (the input to a make’s parmaker). In an embedded grammar, you can define a pargetter to do that. It works like the iterator, except that it returns a single dictionary rather than yielding several of them.

This could look like this, when the grammar input is some iterable:

<embeddedGrammar>
  <iterator>
    <setup>
      <code>
        testData = "a"*1024
      </code>
    </setup>
    <code>
      for i in self.sourceToken:
        yield {'index': i, 'data': testData}
    </code>
  </iterator>
</embeddedGrammar>

If you need to raise an error from within an embeddedGrammar, use a SourceParseError, somewhat like this:

raise utils.SourceParseError(
  "Bad line",
  offending=inputLine,
  location=str(lineNumber+1),
  source=inputFile.name)

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
isDispatching (boolean; defaults to ‘False’) – Is this a dispatching grammar (i.e., does the row iterator return pairs of role, row rather than only rows)?
notify (boolean; defaults to ‘False’) – Enable notification of begin/end of processing (as for other grammars; embedded grammars often have odd source tokens for which you don’t want that). Note that the -M option of gavo imp will only work if you set it to true.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
iterator (contains Element iterator) – Code yielding row dictionaries
pargetter (contains Element pargetter) – Code returning a parameter dictionary
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element fitsProdGrammar¶

A grammar that returns FITS-headers as dictionaries.

This is the grammar you want when one FITS file corresponds to one row in the destination table.

The keywords of the grammar record are the cards in the primary header (or some other hdu using the same-named attribute). “-” in keywords is replaced with an underscore for easier @-referencing. You can use a mapKeys element to effect further name cosmetics.

This grammar should handle compressed FITS images transparently if set qnd=”False”. This means that you will essentially get the headers from the second extension for those even if you left hdu=”0”.

The original header is preserved as the value of the header_ key. This is mainly intended for use WCS use, as in wcs.WCS(@header_).

If you have more complex structures in your FITS files, you can get access to the pyfits HDU using the hdusField attribute. With hdusField="_H", you could say things like @_H[1].data[10][0] to get the first data item in the tenth row in the second HDU.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
hdu (integer; defaults to ‘0’) – Take the header from this HDU. You must say qnd=’False’ for this to take effect.
hdusField (unicode string; defaults to None) – If set, the complete pyfits HDU list for the FITS file is returned in this grammar field.
maxHeaderBlocks (integer; defaults to ‘40’) – Stop looking for FITS END cards and raise an error after this many blocks. You may need to raise this for people dumping obscene amounts of data or history into headers.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
qnd (boolean; defaults to ‘True’) – Use a hack to read the FITS header more quickly. This only works for the primary HDU

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
mapKeys (contains Element mapKeys) – Prescription for how to map header keys to grammar dictionary keys
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element fitsTableGrammar¶

A grammar parsing from FITS tables.

fitsTableGrammar result in typed records, i.e., values normally come in the types they are supposed to have. Of course, that won’t work for datetimes, STC-S regions, and the like.

The keys of the result dictionaries are simpily the names given in the FITS.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
hdu (integer; defaults to ‘1’) – Take the data from this extension (primary=0). Tabular data typically resides in the first extension.
lowerKeys (boolean; defaults to ‘False’) – Convert column names to lower case when building the rawdict.
nanIsNULL (boolean; defaults to ‘False’) – Map NaN floats to NULL when building the rawdict.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element freeREGrammar¶

A grammar allowing “free” regular expressions to parse a document.

Basically, you give a rowProduction to match individual records in the document. All matches of rowProduction will then be matched with parseRE, which in turn must have named groups. The dictionary from named groups to their matches makes up the input row.

For writing the parseRE, we recommend writing an element, using a CDATA construct, and taking advantage of python’s “verbose” regular expressions. Here’s an example:

<parseRE><![CDATA[(?xsm)^name::(?P<name>.*)
  ^query::(?P<query>.*)
  ^description::(?P<description>.*)\.\.
]]></parseRE>

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
ignoreJunk (boolean; defaults to ‘False’) – Ignore everything outside of the row production
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
parseRE (unicode string; defaults to <Undefined>) – RE containing named groups matching a record
rowProduction (unicode string; defaults to ‘(?m)^.+$\n’) – RE matching a complete record.
stripTokens (boolean; defaults to ‘False’) – Strip whitespace from result tokens?

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element handles¶

A declaration of what grammar to use within a UnionGrammar.

Each handler has a (full, python) regular expression defining what file names the grammar is responsible in the filePattern attribute; note that the pattern is matched against the full file name using search so you can match path parts, but you must take care not to overmatch. The other child is a normal DaCHS grammar.

May occur in Element unionGrammar.

Atomic Children¶

pattern (unicode string; defaults to <Undefined>) – Undocumented

Structure Children¶

grammar (contains one of binaryGrammar, cdfHeaderGrammar, columnGrammar, contextGrammar, customGrammar, dictlistGrammar, directGrammar, embeddedGrammar, fitsProdGrammar, freeREGrammar, hdf5Grammar, keyValueGrammar, mySQLDumpGrammar, nullGrammar, transparentGrammar, odbcGrammar, pdsGrammar, reGrammar, rowsetGrammar, unionGrammar, voTableGrammar, csvGrammar, fitsTableGrammar, xmlGrammar) – Grammar used to handle these kinds of files

Element hdf5Grammar¶

a grammar for parsing single tables from HDF5 files.

These result in typed records, i.e., values normally come in the types they are supposed to have. The keys in the rows are the column names as given in the HDF file.

Regrettably, there are about as many conventions to serialise tables in HDF5 as there are programmes writing HDF5. This grammar supports a few styles; ask to have more included.

Styles currently implemented:

astropy:	The table comes as a record array. The grammar is aware of the astropy convention of using adding mask columns as name+”.mask” and will turn masked values to Nones.
vaex:	The table comes as a group with the columns as individual arrays in the group member’s data dataset. Put the parent of the columns group into the dataset attribute here.

This class is not intended for ingesting large HDF5 files, as it will only process a few thousand rows per second on usual hardware. Use Element directgrammar for large files.

Atomic Children¶

dataset (unicode string; defaults to <Undefined>) – The name of the HDF5 dataset/group containing the table. At this point, only datasets that are children of root are supported.
enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
style (One of: astropy, vaex; defaults to ‘astropy’) – Style of the table serialisation.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element iterator¶

A definition of an iterator of a grammar.

The code defined here becomes the _iterRows method of a grammar.common.RowIterator class. This means that you can access self.grammar (the parent grammar; you can use this to transmit properties from the RD to your function) and self.sourceToken (whatever gets passed to parse()).

May occur in Element embeddedGrammar.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element keyValueGrammar¶

A grammar to parse key-value pairs from files.

The default assumes one pair per line, with # comments and = as separating character.

yieldPairs makes the grammar return an empty docdict and {“key”:, “value”:} rowdicts.

Whitespace around key and value is ignored.

Atomic Children¶

commentPattern (unicode string; defaults to ‘(?m)#.*’) – A regular expression describing comments.
enc (unicode string; defaults to None) – Encoding of the source file(s).
kvSeparators (unicode string; defaults to ‘:=’) – Characters accepted as separators between key and value
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
pairSeparators (unicode string; defaults to ‘n’) – Characters accepted as separators between pairs
yieldPairs (boolean; defaults to ‘False’) – Yield key-value pairs instead of complete records?

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
mapKeys (contains Element mapKeys) – Mappings to rename the keys coming from the source files. Use this, in particular, if the keys are not valid python identifiers.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element makeQuery¶

A generator of ODBC queries.

This is is mainly useful when doing ODBC queries to incrementally havest some external resource.

The current ODBC iterator will be available as self.

The procedures also see a function escapeSQL(val) that returns val as a SQL literal (actually, it’s psycopg2’s adapt at the moment).

This is intended to be used somewhat like this with a monotonously increasing column insertion_time:

<makeQuery>
  <code>
    # find to until when we have data locally
    try:
      with base.getTableConn() as conn:
        localMax = next(conn.query(
          "SELECT MAX(insertion_time) FROM \schema.main"))[0]

      fragment = " WHERE insertion_time>{}".format(
        sqlEscape(localMax))
    except base.DBError as msg:
      base.ui.notifyWarning(f"{msg} while harvesting: full re-harvest")
      fragment = ""

    return f"SELECT * FROM remote_table{fragment}"
  </code>
</makeQuery>

May occur in Element odbcGrammar.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element mapKeys¶

Mapping of names, specified in long or short forms.

mapKeys is necessary in grammars like keyValueGrammar or fitsProdGrammar. In these, the source files themselves give key names. Within the GAVO DC, keys are required to be valid python identifiers (i.e., match [A-Za-z\_][A-Za-z\_0-9]*). If keys coming in do not have this form, mapping can force proper names.

mapKeys could also be used to make incoming names more suitable for matching with shell patterns (like in rowmaker idmaps).

May occur in Element cdfHeaderGrammar, Element csvGrammar, Element directGrammar, Element fitsProdGrammar, Element keyValueGrammar, Element pdsGrammar.

Atomic Children¶

Character content of the element (defaulting to ‘’) – Simple mappings in the form<dest>:<src>{,<dest>:<src>}

Other Children¶

map (mapping; the key is the element content, the value is in the ‘key’ (or, equivalently, dest) attribute) – Map source names given in content to the name given in dest.

Element mySQLDumpGrammar¶

A grammar pulling information from MySQL dump files.

WARNING: This is a quick hack. If you want/need it, please contact the authors.

At this point this is nothing but an ugly RE mess with lots of assumptions about the dump file that’s easily fooled. Also, the entire dump file will be pulled into memory.

Since grammar semantics cannot do anything else, this will always only iterate over a single table. This currently is fixed to the first, but it’s conceivable to make that selectable.

Database NULLs are already translated into Nones.

In other words: It might do for simple cases. If you have something else, improve this or complain to the authors.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
gunzip (boolean; defaults to ‘False’) – Unzip sources while reading? (Deprecated, use preFilter=’zcat’)
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
preFilter (unicode string; defaults to None) – Shell command to pipe the input through before passing it on to the grammar. Classical examples include zcat or bzcat, but you can commit arbitrary shell atrocities here.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element nullGrammar¶

A grammar that never returns any rows.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element odbcGrammar¶

A grammar that feeds from a remote database.

This works as a sort of poor man’s foreign data wrapper: you pull data from a remote database now and then, mogrifying it into whatever format you want locally.

This expects files containing pyodbc connection strings as sources, so you’ll normally just have one source. Having the credentials externally helps keeping RDs using this safe for public version control.

An example for an ODBC connection string:

DRIVER={SQL Server};SERVER=localhost;DATABASE=testdb;UID=me;PWD=pass

This will only work if pyodbc (debian: python3-pyodbc) is installed. Additionally, you will have to install the odbc driver corresponding to your source database (e.g., odbc-postgresql).

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
query (unicode string; defaults to <Not given/empty>) – The query to run on the remote server. The keys of the grammar will be the names of the result columns.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
makeQuery (contains Element makeQuery) – Code returning the query to execute on the remote system.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element pargetter¶

A definition of the parameter getter of an embedded grammar.

The code defined here becomes the getParameters method of the generated row iterator. This means that the dictionary returned here becomes the input to a parmaker.

If you don’t define it, the parameter dict will be empty.

Like the iterators, pargetters see the current source token as self.sourceToken, and the grammar as self.grammar.

It is guaranteed that the pargetter is called exactly once before the iterator runs. Hence, if you have expensive initialisation to do, do it in the pargetter and pass the result to the iterator, typically somehow in the sourceToken.

May occur in Element embeddedGrammar.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element pdsGrammar¶

A grammar that returns labels of PDS documents as rowdicts.

PDS is the file format of the Planetary Data System; the labels are quite like, but not quite like FITS headers.

Extra care needs to be taken here since the values in the rawdicts can be complex objects (e.g., other labels). It’s likely that you will need constructs like @IMAGE["KEY"].

Current versions of PyPDS also don’t parse the values. This is particularly insiduous because general strings are marked with ” in PDS. When mapping those, you’ll probably want a @KEY.strip(‘”’).

You’ll need PyPDS to use this; there’s no Debian package for that yet, so you’ll have to do a source install from git://github.com/RyanBalfanz/PyPDS.git

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
mapKeys (contains Element mapKeys) – Prescription for how to map labels keys to grammar dictionary keys
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element reGrammar¶

A grammar that builds rowdicts from records and fields specified via REs separating them.

There is also a simple facility for “cleaning up” records. This can be used to remove standard shell-like comments; use recordCleaner="(?:#.*)?(.*)".

Atomic Children¶

commentPat (unicode string; defaults to None) – RE inter-record material to be ignored (note: make this match the entire comment, or you’ll get random mess from partly-matched comments. Use ‘(?m)^#.*$’ for beginning-of-line hash-comments.
enc (unicode string; defaults to None) – Encoding of the source file(s).
fieldSep (unicode string; defaults to ‘\s+’) – RE for separating two fields in a record.
gunzip (boolean; defaults to ‘False’) – Unzip sources while reading? (Deprecated, use preFilter=’zcat’)
lax (boolean; defaults to ‘False’) – allow more or less fields in source records than there are names
names (Comma-separated list of strings; defaults to ‘’) – Names for the parsed fields, in matching sequence. You can use macros here, e.g., \colNames{someTable}.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
preFilter (unicode string; defaults to None) – Shell command to pipe the input through before passing it on to the grammar. Classical examples include zcat or bzcat, but you can commit arbitrary shell atrocities here.
recordCleaner (unicode string; defaults to None) – A regular expression matched against each record. The matched groups in this RE are joined by blanks and used as the new pattern. This can be used for simple cleaning jobs; However, records not matching recordCleaner are rejected.
recordSep (unicode string; defaults to ‘n’) – RE for separating two records in the source.
stopPat (unicode string; defaults to None) – Stop parsing when a record matches this RE (this is for skipping non-data footers
topIgnoredLines (integer; defaults to ‘0’) – Skip this many lines at the top of each source file.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element rowfilter¶

A generator for rows coming from a grammar.

Rowfilters receive rows (i.e., dictionaries) as yielded by a grammar under the name row. Additionally, the embedding row iterator is available under the name rowIter.

Macros are expanded within the embedding grammar.

The procedure definition must result in a generator, i.e., there must be at least one yield; in general, this will typically be a yield row, but a rowfilter may swallow or create as many rows as desired.

If you forget to have a yield in the rowfilter source, you’ll get a “NoneType is not iterable” error that’s a bit hard to understand.

Here, you can only access whatever comes from the grammar. You can access grammar keys in late parameters as row[key] or, if key is like an identifier, as @key.

May occur in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element rowsetGrammar¶

A grammar handling sequences of tuples.

To add semantics to the field, it must know the “schema” of the data. This is defined via the table it is supposed to get the input from.

This grammar probably is only useful for internal purposes.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
fieldsFrom (id reference; defaults to <Undefined>) – the table defining the columns in the tuples.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element sourceFields¶

A procedure application that returns a dictionary added to all incoming rows.

Use this to programmatically provide information that can be computed once but that is then added to all rows coming from a single source, usually a file. This could be useful to add information on the source of a record or the like.

The code must return a dictionary. The source that is about to be parsed is passed in as sourceToken. When parsing from files, this simply is the file name. The data the rows will be delivered to is available as “data”, which is useful for adding or retrieving meta information.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element transparentGrammar¶

A grammar that returns its sourceToken as the row iterator.

This only makes sense in extreme situations and never without custom code. If you’re not sure you need this, you don’t want to know about it.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element unionGrammar¶

A grammar using one of a sequence of grammars to parse its sources.

(since version 2.7.2)

Use this if you have differing input formats eventually processible by the same row maker (of course, you can make the row maker flexible enough to cope with different grammar outputs). To do that, use two or more handles definitions, each giving a regular expression against the full file name (but matched with re.search) and a grammar to use for such files.

Handles definitions will be tried in sequence; you can hence have special cases early and catch-alls later.

The basic idea is that you write something like:

<unionGrammar>
  <handles pattern=".*\.txt$">
    <reGrammar...>
  </handles>
  <handles pattern=".*\.csv$">
    <csvGrammar...>
  </handles>
</unionGrammar>

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

handles (contains Element handles and may be repeated zero or more times) – Recipe for what grammar to use for what sort of file.
ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element voTableGrammar¶

A grammar parsing from VOTables.

Currently, the PARAM fields are ignored, only the data rows are returned.

voTableGrammars result in typed records, i.e., values normally come in the types they are supposed to have.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
gunzip (boolean; defaults to ‘False’) – Unzip sources while reading?
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element xmlGrammar¶

A grammar parsing from generic XML files.

Use this grammar to parse from generic XML files. For now, one rawdict per document is returned (later extensions might let you define elements that will yield rows).

The keys are xpaths (e.g., root/element or root/element/@attr), the values the (joined) text nodes that are immediate children or the element.

When elements are repeated within an element, [ct] is appended to the path element (e.g., root/element([0]).

For now, this grammar ignores namespaces.

Because most of the keys are not valid python identifiers, you cannot use the @key syntax when mapping this. Use vars[key] instead (or <map key=”dest” source=”path”/>).

Do not use this for VOTables; use VOTableGrammar instead.

Atomic Children¶

enc (unicode string; defaults to None) – Encoding of the source file(s).
normalizeWhitespace (boolean; defaults to ‘True’) – By default, the parser will return whitespace-only content as None and will turn internal whitespace to a single blank. Set this to False to preserve whitespace as present in the document.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

ignoreOn (contains Element ignoreOn) – Conditions for ignoring certain input records. These triggers drop an input record entirely. In modern RDs, prefer rowfilters raising SkipThis.
rowfilters (contains Element rowfilter and may be repeated zero or more times) – Row filters for this grammar.
sourceFields (contains Element sourceFields) – Code returning a dictionary of values added to all returned rows.

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Cores Available ¶

The following elements are related to cores. All cores can only occur toplevel, i.e. as direct children of resource descriptors. Cores are only useful with an id to make them referencable from services using that core.

Element adqlCore¶

A core taking an ADQL query from its query argument and returning the result of that query in a standard table.

Since the columns returned depend on the query, the outputTable of an ADQL core must not be defined.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element coreProc¶

A definition of a pythonCore’s functionalty.

This is a procApp complete with setup and code; you could inherit between these.

coreProcs see the embedding service, the input table passed, and the query metadata as service, inputTable, and queryMeta, respectively.

The core itself is available as self.

May occur in Element pythonCore.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element customCore¶

A wrapper around a core defined in a module.

This core lets you write your own cores in modules.

The module must define a class Core. When the custom core is encountered, this class will be instantiated and will be used instead of the CustomCore, so your code should probably inherit core.Core.

See Writing Custom Cores for details.

Atomic Children¶

module (unicode string; defaults to <Undefined>) – Path to the module containing the core definition.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element dataFormatter¶

A procedure application that renders data in a processed service.

These play the role of the renderer, which for datalink is usually trivial. They are supposed to take descriptor.data and return a pair of (mime-type, bytes), which is understood by most renderers.

When no dataFormatter is given for a core, it will return descriptor.data directly. This can work with the datalink renderer itself if descriptor.data will work as a nevow resource (i.e., has a renderHTTP method, as our usual products do). Consider, though, that renderHTTP runs in the main event loop and thus most not block for extended periods of time.

The following names are available to the code:

descriptor – whatever the DescriptorGenerator returned
args – all the arguments that came in from the web.

In addition to the usual names available to ProcApps, data formatters have:

Page – base class for resources with renderHTTP methods.
IRequest – the nevow interface to make Request objects with.
File(path, type) – if you just want to return a file on disk, pass its path and media type to File and return the result.
TemporaryFile(path, type) – as File, but the disk file is unlinked after use
soda – the protocols.soda module

May occur in Element datalinkCore.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element dataFunction¶

A procedure application that generates or modifies data in a processed data service.

All these operate on the data attribute of the product descriptor. The first data function plays a special role: It must set the data attribute (or raise some appropriate exception), or a server error will be returned to the client.

What is returned depends on the service, but typcially it’s going to be a table or products.*Product instance.

Data functions can shortcut if it’s evident that further data functions can only mess up (i.e., if the do something bad with the data attribute); you should not shortcut if you just think it makes no sense to further process your output.

To shortcut, raise either of FormatNow (falls though to the formatter, which is usually less useful) or DeliverNow (directly returns the data attribute; this can be used to return arbitrary chunks of data).

The following names are available to the code:

descriptor – whatever the DescriptorGenerator returned
args – all the arguments that came in from the web.

In addition to the usual names available to ProcApps, data functions have:

FormatNow – exception to raise to go directly to the formatter
DeliverNow – exception to raise to skip all further formatting and just deliver what’s currently in descriptor.data
File(path, type) – if you just want to return a file on disk, pass its path and media type to File and assign the result to descriptor.data.
TemporaryFile(path,type) – as File, but the disk file is unlinked after use
makeData – the rsc.makeData function
soda – the protocols.soda module
RemoteProduct(url) – 301-redirects to URL when rendered

May occur in Element datalinkCore.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element datalinkCore¶

A core for processing datalink and processed data requests.

The input table of this core is dynamically generated from its metaMakers; it makes no sense at all to try and override it.

See Datalink and SODA for more information.

In contrast to “normal” cores, one of these is made (and destroyed) for each datalink request coming in. This is because the interface of a datalink service depends on the request’s value(s) of ID.

The datalink core can produce both its own metadata and data generated. It is the renderer’s job to tell them apart.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

dataFormatter (contains Element dataFormatter) – Code that turns descriptor.data into a nevow resource or a mime, content pair. If not given, the renderer will be returned descriptor.data itself (which will probably not usually work).
dataFunctions (contains Element dataFunction and may be repeated zero or more times) – Code that generates of processes data for this core. The first of these plays a special role in that it must set descriptor.data, the others need not do anything at all.
descriptorGenerator (contains Element descriptorGenerator) – Code that takes a PUBDID and turns it into a product descriptor instance. If not given, //soda#fromStandardPubDID will be used.
inputKeys (contains Element inputKey and may be repeated zero or more times) – A parameter to one of the proc apps (data functions, formatters) active in this datalink core; no specific relation between input keys and procApps is supposed; all procApps are passed all arguments. Conventionally, you will write the input keys in front of the proc apps that interpret them.
inputTable (contains Element inputTable) – Description of the input data
metaMakers (contains Element metaMaker and may be repeated zero or more times) – Code that takes a data descriptor and either updates input key options or yields related data.
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element dbCore¶

A core performing database queries on one table or view.

DBCores ask the service for the desired output schema and adapt their output. The DBCore’s output table, on the other hand, lists all fields available from the queried table.

Atomic Children¶

distinct (boolean; defaults to ‘False’) – Add a ‘distinct’ modifier to the query?
groupBy (unicode string; defaults to None) – A group by clause. You shouldn’t generally need this, and if you use it, you must give an outputTable to your core.
limit (integer; defaults to None) – A pre-defined match limit (suppresses DB options widget).
namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
sortKey (unicode string; defaults to None) – A pre-defined sort order (suppresses DB options widget). The sort key accepts multiple columns, separated by commas.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element debugCore¶

a core that returns its arguments stringified in a table.

You need to provide an external input tables for these.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element descriptorGenerator¶

A procedure application for making product descriptors for PUBDIDs

Despite the name, a descriptor generator has to return (not yield) a descriptor instance. While this could be anything, it is recommended to derive custom classes from prodocols.datalink.ProductDescrpitor, which exposes essentially the columns from DaCHS’ product table as attributes. This is what you get when you don’t define a descriptor generator in your datalink core.

Before writing your own, see if one of the predefined descriptor generators work for you; see Descriptor Generators below.

The following names are available to the code:

pubDID – the pubDID to be resolved

args – all the arguments that came in from the web (these should not usually be necessary for making the descriptor and are completely unparsed at this point)

FITSProductDescriptor – the base class of FITS product descriptors

DLFITSProductDescriptor – the same, just for when the product table has a datalink.

ProductDescriptor – a base class for your own custom descriptors

DatalinkFault – use this when flagging failures

soda – contents of the soda module for convenience

If you made your pubDID using the getStandardPubDID rowmaker function, and you need no additional logic within the descriptor, the default (//soda#fromStandardPubDID) should do.

If you need to derive custom descriptor classes, you can see the base class under the name ProductDescriptor; there’s also FITSProductDescriptor and DatalinkFault in each proc’s namespace. If your Descriptor does not actually refer to something in the product table, it is likely that you want to set the descriptor’s suppressAutoLinks attribute to True. This will stop DaCHS from attempting to add automatic #this and #preview links.

May occur in Element datalinkCore.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element fancyQueryCore¶

A core executing a pre-specified query with fancy conditions.

Unless you select *, you must define the outputTable here; Weird things will happen if you don’t.

The queriedTable attribute is ignored.

Atomic Children¶

namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
query (unicode string; defaults to <Undefined>) – The query to execute. It must contain exactly one %s where the generated where clause is to be inserted. Do not write WHERE yourself. All other percents must be escaped by doubling them.
timeout (float; defaults to ‘5.0’) – Seconds until the query is aborted

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element fixedQueryCore¶

A core executing a predefined query.

This usually is not what you want, unless you want to expose the current results of a specific query, e.g., for log or event data.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
query (unicode string; defaults to <Undefined>) – The query to be executed. You must define the output fields in the core’s output table. The query will be macro-expanded in the resource descriptor.
timeout (float; defaults to ‘15.0’) – Seconds until the query is aborted
writable (boolean; defaults to ‘False’) – Use a writable DB connection?

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element inputTable¶

an input for a core.

These aren’t actually proper tables but actually just collection of the param-like inputKeys. They serve as input declarations for cores and services (where services derive their inputTDs from the cores’ ones by adapting them to the current renderer. Their main use is for the derivation of contextGrammars.

They can carry metadata, though, which is sometimes convenient when transporting information from the parameter parsers to the core.

For the typical dbCores (and friends), these are essentially never explicitly defined but rather derived from condDescs.

Do not read input values by using table.getParam. This will only give you one value when a parameter has been given multiple times. Instead, use the output of the contextGrammar (inputParams in condDescs). Only there you will have the correct multiplicities.

May occur in Element adqlCore, Element customCore, Element datalinkCore, Element dbCore, Element debugCore, Element fancyQueryCore, Element fixedQueryCore, Element nullCore, Element productCore, Element pythonCore, Element registryCore, Element scsCore, Element siapCutoutCore, Element ssapCore, Element tapCore, Element uploadCore.

Atomic Children¶

exclusive (boolean; defaults to ‘False’) – If true, context grammars built from this will raise an error if contexts passed in have keys not defined by this table

Structure Children¶

groups (contains Element group and may be repeated zero or more times) – Groups of inputKeys (this is used for form UI formatting).
inputKeys (contains Element inputKey and may be repeated zero or more times) – Input parameters for this table.

Element metaMaker¶

A procedure application that generates metadata for datalink services.

The code must be generators (i.e., use yield statements) producing either svcs.InputKeys or protocols.datalink.LinkDef instances.

metaMaker see the data descriptor of the input data under the name descriptor.

The data attribute of the descriptor is always None for metaMakers, so you cannot use anything given there.

Within MetaMakers’ code, you can access InputKey, Values, Option, and LinkDef without qualification, and there’s the MS function to build structures. Hence, a metaMaker returning an InputKey could look like this:

<metaMaker>
  <code>
    yield MS(InputKey, name="format", type="text",
      description="Output format desired",
        values=MS(Values,
        options=[MS(Option, content_=descriptor.mime),
              MS(Option, content_="text/plain")]))
  </code>
</metaMaker>

(of course, you should give more metadata – ucds, better description, etc) in production).

Note that InputKey-returning MetaMakers cannot rely on descriptor.pubDID to actually have a value; in particular SSA may construct cores to produce “direct” processing datalink descriptors. When you need a pubDID, just return if descriptor.pubDID is None.

Alternatively, yield link definitions, i.e., rows for the links response. You will usually define a semantics attribute for the meta maker then (this lets DaCHS use the right semantics in case something goes wrong), and you will usually create links from the descriptor so the pubDID will be right automatically:

<metaMaker semantics="#flat">
    <code>
       yield descriptor.makeLink(
         "http://example.org/flats/master-flat.fits",
         contentType="image/fits",
         description="The master flat for this epoch",
         contentLength=2048x2048*2)
    </code>
 </metaMaker>

It’s ok to yield None; this will suppress a Datalink and is convenient when some component further down figures out that a link doesn’t exist (e.g., because a file isn’t there). Note that in many cases, it’s more helpful to client components to handle such situations by yielding a DatalinkFault.NotFoundFault.

In addition to the usual names available to ProcApps, meta makers have:

MS – function to make DaCHS structures
InputKey – the class to make for input parameters
Values – the class to make for input parameters’ values attributes
Options – used by Values
LinkDef – a class to define further links within datalink services.
ProcLinkDef – a class to reference dlget services outside of the current service
DatalinkFault – a container of datalink error generators
soda – the soda module.

May occur in Element datalinkCore.

Atomic Children¶

code (unicode string; defaults to <Not given/empty>) – A python function body.
deprecated (unicode string; defaults to None) – A deprecation message. This will be shown if this procDef is being compiled.
doc (unicode string; defaults to ‘’) – Human-readable docs for this proc (may be interpreted as restructured text).
name (unicode string; defaults to <Not given/empty>) – A name of the proc. ProcApps compute their (python) names to be somewhat random strings. Set a name manually to receive more easily decipherable error messages. If you do that, you have to care about name clashes yourself, though.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
procDef (id reference; defaults to <Not given/empty>) – Reference to the procedure definition to apply
semantics (unicode string; defaults to ‘http://dc.g-vo.org/datalink#other’) – For link-generating metaMakers, the semantics makeLink or makeLinkFromFile will use unless given something else. This is also what will be used if the meta maker exceptions out. Note that the default is not good and is only set for backwards compatibility. You should override this.
type (One of: apply, dataFormatter, dataFunction, descriptorGenerator, iterator, metaMaker, mixinProc, pargetter, phraseMaker, regTest, rowfilter, sourceFields, t_t; defaults to None) – The type of the procedure definition. The procedure applications will in general require certain types of definitions.

Structure Children¶

bindings (contains Element bind and may be repeated zero or more times) – Values for parameters of the procedure definition
setups (contains Element setup and may be repeated zero or more times) – Setup of the namespace the function will run in

Element nullCore¶

A core always returning None.

This core will not work with the common renderers. It is really intended to go with coreless services (i.e. those in which the renderer computes everything itself and never calls service.runX). As an example, the external renderer could go with this.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element productCore¶

A core retrieving paths and/or data from the product table.

You will not usually mention this core in your RDs. It is mainly used internally to serve /getproduct queries.

It is instantiated from within //products.rd and relies on tables within that RD.

The input data consists of accref; you can use the string form of RAccrefs, and if you renderer wants, it can pass in ready-made RAccrefs. You can pass accrefs in through both an accref param and table rows.

The accref param is the normal way if you just want to retrieve a single image, the table case is for building tar files and such. There is one core instance in //products for each case.

The core returns a list of instances of a subclass of ProductBase above.

This core and its supporting machinery handles all the fancy product functionality (user authorization, cutouts, …).

Atomic Children¶

distinct (boolean; defaults to ‘False’) – Add a ‘distinct’ modifier to the query?
groupBy (unicode string; defaults to None) – A group by clause. You shouldn’t generally need this, and if you use it, you must give an outputTable to your core.
limit (integer; defaults to None) – A pre-defined match limit (suppresses DB options widget).
namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
sortKey (unicode string; defaults to None) – A pre-defined sort order (suppresses DB options widget). The sort key accepts multiple columns, separated by commas.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element pythonCore¶

A core doing computation using a piece of python.

See Python Cores instead of Custom Cores in the reference.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

coreProc (contains Element coreProc) – Code making the outputTable from the inputTable.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element registryCore¶

is a core processing OAI requests.

Its signature requires a single input key containing the complete args from the incoming request. This is necessary to satisfy the requirement of raising errors on duplicate arguments.

It returns an ElementTree.

This core is intended to work the the RegistryRenderer.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element scsCore¶

A core performing cone searches.

This will, if it finds input parameters it can make out a position from, add a _r column giving the distance between the match center and the columns that a cone search will match against.

If any of the conditions for adding _r aren’t met, this will silently degrade to a plain DBCore.

You will almost certainly want a:

<FEED source="//scs#coreDescs"/>

in the body of this (in addition to whatever other custom conditions you may have).

Atomic Children¶

distinct (boolean; defaults to ‘False’) – Add a ‘distinct’ modifier to the query?
groupBy (unicode string; defaults to None) – A group by clause. You shouldn’t generally need this, and if you use it, you must give an outputTable to your core.
limit (integer; defaults to None) – A pre-defined match limit (suppresses DB options widget).
namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
sortKey (unicode string; defaults to None) – A pre-defined sort order (suppresses DB options widget). The sort key accepts multiple columns, separated by commas.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element siapCutoutCore¶

A core doing SIAP plus cutouts.

It has, by default, an additional column specifying the desired size of the image to be retrieved. Based on this, the cutout core will tweak its output table such that references to cutout images will be retrieved.

The actual process of cutting out is performed by the product core and renderer.

Atomic Children¶

distinct (boolean; defaults to ‘False’) – Add a ‘distinct’ modifier to the query?
groupBy (unicode string; defaults to None) – A group by clause. You shouldn’t generally need this, and if you use it, you must give an outputTable to your core.
limit (integer; defaults to None) – A pre-defined match limit (suppresses DB options widget).
namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
sortKey (unicode string; defaults to None) – A pre-defined sort order (suppresses DB options widget). The sort key accepts multiple columns, separated by commas.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element ssapCore¶

A core doing SSAP queries.

This core knows about metadata queries, version negotiation, and dispatches on REQUEST. Thus, it may return formatted XML data under certain circumstances.

Interpreted Properties:

previews: If set to “auto”, the core will automatically add a preview column and fill it with the URL of the products-based preview. Other values are not defined.

Atomic Children¶

distinct (boolean; defaults to ‘False’) – Add a ‘distinct’ modifier to the query?
groupBy (unicode string; defaults to None) – A group by clause. You shouldn’t generally need this, and if you use it, you must give an outputTable to your core.
limit (integer; defaults to None) – A pre-defined match limit (suppresses DB options widget).
namePath (id reference; defaults to None) – Id of an element that will be used to located names in id references. Defaults to the queriedTable’s id.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.
queriedTable (id reference; defaults to <Undefined>) – A reference to the table this core queries.
sortKey (unicode string; defaults to None) – A pre-defined sort order (suppresses DB options widget). The sort key accepts multiple columns, separated by commas.

Structure Children¶

condDescs (contains Element condDesc and may be repeated zero or more times) – Descriptions of the SQL and input generating entities for this core; if not given, they will be generated from the table columns.
inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element tapCore¶

A core for the TAP renderer.

Atomic Children¶

original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Element uploadCore¶

A core handling uploads of files to the database.

It allows users to upload individual files into a special staging area (taken from the stagingDir property of the destination data descriptor) and causes these files to be parsed using destDD. Note that destDD must have updating="True" for this to work properly (it will otherwise drop the table on each update). If uploads are the only way updates into the table occur, source management is not necessary for these, though.

You can tell UploadCores to either insert or update the incoming data using the “mode” input key.

Atomic Children¶

destDD (id reference; defaults to <Undefined>) – Reference to the data we are uploading into. The destination must be an updating data descriptor.
original (id reference; defaults to None) – An id of an element to base the current one on. This provides a simple inheritance method. The general rules for advanced referencing in RDs apply.

Structure Children¶

inputTable (contains Element inputTable) – Description of the input data
outputTable (contains Element outputTable) – Table describing what fields are available from this core

Other Children¶

property (mapping of user-defined keywords in the name attribute to string values) – Properties (i.e., user-defined key-value pairs) for the element.

Predefined Macros ¶

Macro expansions in DaCHS start with a backslash, arguments are given in curly braces. What macros are available depends on the element doing the expansion; regrettably, not all strings are expanded, and at this point it’s not usually documented which are and which are not (though we hope DaCHS typically behaves “as expected”). If this bites you, complain to the authors and we promise we’ll give fixing this a higher priority.

Macro RSTcc0¶

\RSTcc0{self}{stuffDesignation}

expands to a declaration that stuffDesignation is available under CC-0.

This only works in reStructured text (though it’s still almost readable as source).

You’ll probably want to use the //procs#license-cc0 stream instead of this, as that also sets the rights URI.

Available in Element resource

Macro RSTccby¶

\RSTccby{self}{stuffDesignation}

expands to a declaration that stuffDesignation is available under CC-BY.

This only works in reStructured text (though it’s still almost readable as source).

You’ll probably want to use the //procs#license-cc-by stream instead of this, as that also sets the rights URI.

Available in Element resource

Macro RSTccbysa¶

\RSTccbysa{self}{stuffDesignation}

expands to a declaration that stuffDesignation is available under CC-BY-SA.

This only works in reStructured text (though it’s still almost readable as source).

You’ll probably want to use the //procs#license-cc-by-sa stream instead of this, as that also sets the rights URI.

Available in Element resource

Macro RSTservicelink¶

\RSTservicelink{self}{serviceId}{title=None}

a link to an internal service; id is <rdId>/<serviceId>/<renderer>, title, if given, is the anchor text.

The result is a link in the short form for restructured test.

Available in Element FEED, Element LFEED, Element LOOP, Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element inputTable, Element keyValueGrammar, Element mixinDef, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element outputTable, Element pdsGrammar, Element reGrammar, Element resRec, Element resource, Element rowmaker, Element rowsetGrammar, Element service, Element table, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro RSTtable¶

\RSTtable{self}{tableName}

adds an reStructured test link to a tableName pointing to its table info.

Macro colNames¶

\colNames{self}

returns an SQL-ready list of column names of this table.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element outputTable, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element table, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro curtable¶

\curtable{self}

returns the qualified name of the current table.

(this is identical to the macro qName, which you should prefer in new RDs.)

Available in Element outputTable, Element table

Macro decapitalize¶

\decapitalize{self}{aString}

returns aString with the first character lowercased.

Macro dlMetaURI¶

\dlMetaURI{self}{dlId}

returns a link to the datalink document for the current product.

This assumes you’re assigning standard pubDIDs (see also standardPubDID, which is used by this).

dlId is the XML id of the datalink service, which is supposed to be in the sameRD as the rowmaker.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro docField¶

\docField{self}{name}

returns an expression giving the value of the column name in the document row.

Available in Element rowmaker

Macro fullDLURL¶

\fullDLURL{self}{dlService}

returns a python expression giving a link to the full current data set retrieved through the datalink service.

You would write \fullDLURL{dlsvc} here, and the macro will expand into something like http://yourserver/currd/dlsvc/dlget?ID=ivo://whatever.

dlService is the id of the datalink service in the current RD.

This is intended for “virtual” data where the dataset is generated on the fly through datalink.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro fullPath¶

\fullPath{self}

returns an expression expanding to the full path of the current input file.

Available in Element rowmaker

Macro getConfig¶

\getConfig{self}{section}{name=None}

the current value of configuration item {section}{name}.

You can also only give one argument to access settings from the general section.

Macro getParam¶

\getParam{self}{parName}{default=''}

returns the string representation of the parameter parName.

This is the parameter as given in the table definition. Any changes to an instance are not reflected here.

If the parameter named does not exist, an empty string is returned. NULLs/Nones are rendered as NULL; this is mainly a convenience for obscore-like applications and should not be exploited otherwise, since it’s ugly and might change at some point.

If a default is given, it will be returned for both NULL and non-existing params.

Available in Element outputTable, Element table

Macro inputRelativePath¶

\inputRelativePath{self}{liberalChars='True'}

returns an expression giving the current source’s path relative to inputsDir

liberalChars can be a boolean literal (True, False, etc); if false, a value error is raised if characters that will result in trouble with the product mixin are within the result path.

In rowmakers fed by grammars with //products#define, better use @prodtblAccref.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro inputSize¶

\inputSize{self}

returns an expression giving the size of the current source.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro internallink¶

\internallink{self}{relPath}

an absolute URL from a path relative to the DC root.

Macro lastSourceElements¶

\lastSourceElements{self}{numElements}

returns an expression calling rmkfuncs.lastSourceElements on the current input path.

Available in Element rowmaker

Macro lower¶

\lower{self}{aString}

returns aString lowercased.

There’s no guarantees for characters outside ASCII.

Macro magicEmpty¶

\magicEmpty{self}{val}

returns __EMPTY__ if val is empty.

This is necessary when feeding possibly empty params from mixin parameters (don’t worry if you don’t understand this).

Macro metaSeq¶

\metaSeq{self}{metaKey}{default=''}{joiner='}{'}

returns all values of metaKey on the current macro expander joined by joiner.

This will be an empty string if there is no corresponding metadata (or default, if passed).

Macro metaString¶

\metaString{self}{metaKey}{default=None}

the value of metaKey on the macro expander.

This will raise an error when the meta Key is not available unless you give a default. It will also raise an error if metaKey is not atomic (i.e., single-valued). Use metaSeq for meta items that may have multiple values.

Because it’s sometimes useful, if the expander itself doesn’t have metadat, this goes up in the RD tree until it finds something that has metadata.

Macro nameForUCD¶

\nameForUCD{self}{ucd}

returns the (unique!) name of the field having ucd in this table.

If there is no or more than one field with the ucd in this table, we raise a ValueError.

Available in Element outputTable, Element table

Macro nameForUCDs¶

\nameForUCDs{self}{ucds}

returns the (unique!) name of the field having one of ucds in this table.

Ucds is a selection of ucds separated by vertical bars (|). The rules for when this raises errors are so crazy you don’t want to think about them. This really is only intended for cases where “old” and “new” standards are to be supported, like with pos.eq.*;meta.main and POS_EQ_*_MAIN.

If there is no or more than one field with the ucd in this table, we raise an exception.

Available in Element outputTable, Element table

Macro property¶

\property{self}{propName}

returns an expression giving the value of the property propName on the current DD.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro qName¶

\qName{self}

returns the qName of the table we are currently parsing into.

Available in Element outputTable, Element rowmaker, Element table

Macro quote¶

\quote{self}{arg}

returns the argument in quotes (with internal quotes backslash-escaped if necessary).

Macro rdIdDotted¶

\rdIdDotted{self}

the identifier for the current resource descriptor with slashes replaced with dots (so they work as the “host part” in URIs.

Macro reSub¶

\reSub{self}{pattern}{replacement}{string}

returns the string with the python RE pattern replaced with replacement.

This is directly handed through to python re.sub, so you can (but probably shouldn’t) play all the RE tricks you can in python (e.g., back references).

If you find yourself having to use reSub, you should regard that as an alarm sign that you’re probably doing it wrong.

Oh: closing curly braces can be included in the argument by backslash-escaping them.

Macro resdir¶

\resdir{self}

the input-relative resource directory of the current resource descriptor.

This never has a trailing slash.

Macro rootlessPath¶

\rootlessPath{self}

returns an expression giving the current source’s path with the resource descriptor’s root removed.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro rowsMade¶

\rowsMade{self}

returns an expression giving the number of records already returned by this row maker.

This number excludes failed and skipped rows.

Available in Element rowmaker

Macro rowsProcessed¶

\rowsProcessed{self}

returns an expression giving the number of records already delivered by the grammar.

Available in Element rowmaker

Macro sourceCDate¶

\sourceCDate{self}

returns an expression giving the timestamp for the create date of the current source.

Use dateTimeToJdn or dateTimeToMJD to turn this into JD or MJD (which is usually preferred in database tables). See also the sourceDate macro.

Available in Element rowmaker

Macro sourceDate¶

\sourceDate{self}

returns an expression giving the timestamp of the current source.

This is a timestamp of the modification date; use dateTimeToJdn or dateTimeToMJD to turn this into JD or MJD (which is usually preferred in database tables). See also the sourceCDate macro.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro splitPreviewPath¶

\splitPreviewPath{self}{ext}

returns an expression for the split standard path for a custom preview.

As standardPreviewPath, except that the directory hierarchy of the data files will be reproduced in previews. For ext, you should typically pass the extension appropriate for the preview (like {.png} or {.jpeg}).

See the introduction to custom previews for details.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro sql_standardPubDID¶

\sql_standardPubDID{self}{fromCol='accref'}

returns a SQL expression returning a DaCHS standard pubDID generated from the accref (or something overridden) column.

This is convenient in obscore or ssa views when the underlying table just has accrefs. If your code actually uses the pubDID to search in the table (and it probably shouldn’t), better use an actual column and index it.

Macro sqlquote¶

\sqlquote{self}{arg}

returns the argument as a quoted string, unless it is ‘NULL’ or None, in which case just NULL is returned.

Macro srcstem¶

\srcstem{self}

returns python code for the stem of the source file currently parsed in a rowmaker.

Example: if you’re currently parsing /tmp/foo.bar.gz, the stem is foo.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowmaker, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro standardPreviewPath¶

\standardPreviewPath{self}

returns an expression for the standard path for a custom preview.

This consists of resdir, the name of the previewDir property on the embedding DD, and the flat name of the accref (which this macro assumes to see in its namespace as accref; this is usually the case in //products#define, which is where this macro would typically be used).

As an alternative, there is the splitPreviewPath macro, which does not mogrify the file name. In particular, do not use standardPreviewPath when you have more than a few 1e4 files, as it will have all these files in a single, flat directory, and that can become a chore.

See the introduction to custom previews for details.

Available in Element binaryGrammar, Element cdfHeaderGrammar, Element columnGrammar, Element contextGrammar, Element csvGrammar, Element customGrammar, Element dictlistGrammar, Element embeddedGrammar, Element fitsProdGrammar, Element fitsTableGrammar, Element freeREGrammar, Element hdf5Grammar, Element keyValueGrammar, Element mySQLDumpGrammar, Element nullGrammar, Element odbcGrammar, Element pdsGrammar, Element reGrammar, Element rowsetGrammar, Element transparentGrammar, Element unionGrammar, Element voTableGrammar, Element xmlGrammar

Macro standardPubDID¶

\standardPubDID{self}

returns the “standard publisher DID” for the current product.

The publisher dataset identifier (PubDID) is important in protocols like SSAP and obscore. If you use this macro, the PubDID will be your authority, the path component ~, and the current value of @prodtblAccref. It thus will only work where products#define (or a replacement) is in action. If it isn’t, a normal function call getStandardPubDID(\inputRelativePath) would be an obvious alternative.

You can of course define your PubDIDs in a different way.

Available in Element rowmaker

Macro tablename¶

\tablename{self}

returns the unqualified name of the current table.

In most contexts, you will probably need to use the macro qName instead of this.

Available in Element outputTable, Element table

Macro tablesForTAP¶

\tablesForTAP{self}

returns a list of table names available for TAP querying.

This, really, is an implementation detail for the TAP service and might go away anytime.

Available in Element service

Macro upper¶

\upper{self}{aString}

returns aString uppercased.

There’s no guarantees for characters outside ASCII.

Mixins ¶

Mixins ensure a certain functionality on a table. Typically, this is used to provide certain guaranteed fields to particular cores. For many mixins, there are predefined procedures (both rowmaker applys and grammar rowfilters) that should be used in grammars and/or rowmakers feeding the tables mixing in a given mixin.

Note that when a piece of metadata in a mixin gets in your way, you can selectively override attributes of columns and params by copying and changing them. For instance, if your mixin example#m" gives you a column ``flux, and you need to change its unit, you would say:

<table mixin="example#m">
  <column original="flux" unit="mJy"/>
</table>

The //epntap2#localfile-2_0 Mixin¶

Use this mixin if your epntap table is filled with local products (i.e., sources matches files on your hard disk that DaCHS should hand out itself). This will arrange for your products to be entered into the products table, and it will automatically compute file size, etc.

This wants a //products#define rowfilter in your grammar and a //epntap2#populate-localfile-2_0 apply in your rowmaker.

The //epntap2#table-2_0 Mixin¶

This mixin defines a table suitable for publication via the EPN-TAP protocol.

According to the standard definition, tables mixing this in should be called epn_core. The mixin already arranges for the table to be accessible by ADQL and be on disk.

This also causes the product table to be populated. This means that grammars feeding such tables need a //products#define row filter. At the very least, you need to say:

<rowfilter procDef="//products#define">
  <bind name="table">"\schema.epn_core"</bind>
</rowfilter>

If you absolutely cannot use //products#define, you will have to manually provide the prodtblFsize (file size in bytes), prodtblAccref (product URL), and prodtblPreview (thumbnail image or None) keys in what’s coming from your grammar.

Use the //epntap2#populate-2_0 apply in rowmakers feeding tables mixing this in.

The mixin has a parameter optional_columns, which accepts space-separated columns from the following extensions:

The APIS extension¶

north_pole_position:
detector_name:	Detector name
	North pole position angle with respect to celestial north pole
obs_mode:	Observing mode
opt_elem:	Optical element name
orientation:	Position angle of an image y-axis, in direct sense from north
platesc:	pixel angular size or platescale (on sky only)
target_primary_hemisphere:
	Primary observed hemisphere
target_secondary_hemisphere:
	Secondary observed hemisphere

The common extension¶

alt_target_name:
access_estsize:	Estimated file size in kbyte.
access_format:	File format type (RFC 6838 Media Type a.k.a. MIME type)
access_md5:	MD5 Hash for the file
access_url:	URL of the data file, case sensitive. If present, then access_format and access_estsize are mandatory.
acquisition_id:	ID of the data file/acquisition in the original archive
albedo:	Target albedo
	Provides alternative target name if more common (e.g. comets); multiple identifiers can be separated by hashes
altitude_fromshape_max:
	Max altitude of observed area above shape model / DTM
altitude_fromshape_min:
	Min altitude of observed area above shape model / DTM
bib_reference:	Bibcode or DOI preferred if available, or other bibliographic identifier or URL
campaign:	Name of the observational campaign
datalink_url:	Link to a datalink document for this dataset
dec:	Declination
earth_distance_max:
	Max Earth-target distance
earth_distance_min:
	Min Earth-target distance
external_link:	Web page providing more details on this granule
feature_name:	Secondary name (can be standard name of region of interest)
file_name:	Name of the data file only, case sensitive
filter:	Identifies filter in use, typically for images
flux:	Target flux
instrument_type:
	type of instrument
internal_reference:
	Related granule_uid(s) in the current service; hash-separated
local_time_max:	Max local time at observed region
local_time_min:	Min local time at observed region
magnitude:	Absolute magnitude. For small bodies, from HG magnitude system
messenger:	Vector of measured signal, including electromagnetic band, from http://www.ivoa.net/rdf/messenger
processing_level_desc:
	Describes specificities of the processing level
proposal_id:	Proposal identifier
proposal_pi:	Proposal principal investigator
proposal_target_name:
	target name as in proposal title
proposal_title:	Proposal title
publisher:	A short string identifying the entity running the data service used
ra:	Right ascension
radial_distance_max:
	Max distance from observed area to body center
radial_distance_min:
	Min distance from observed area to body center
coverage:	(ST)MOC footprint, valid for celestial, spherical, or body-fixed frames + time coverage
solar_longitude_max:
	Max Solar longitude Ls (location on orbit / season)
solar_longitude_min:
	Min Solar longitude Ls (location on orbit / season)
spatial_coordinate_description:
	ID of specific coordinate system and version or properties.
spatial_origin:	Defines the frame origin
species:	Identifies a chemical species, case sensitive
subobserver_latitude_max:
	Maximum sub-observer point latitude (sub-Earth for ground based observations)
subobserver_latitude_min:
	Minimum sub-observer point latitude (sub-Earth for ground based observations)
subobserver_longitude_max:
	Maximum sub-observer point longitude (sub-Earth for ground based observations)
subobserver_longitude_min:
	Minimum sub-observer point longitude (sub-Earth for ground based observations)
subsolar_latitude_max:
	Maximum sub-solar point latitude
subsolar_latitude_min:
	Minimum sub-solar point latitude
subsolar_longitude_max:
	Maximum sub-solar point longitude
subsolar_longitude_min:
	Minimum sub-solar point longitude
sun_distance_max:
	Max Sun-target distance
sun_distance_min:
	Min Sun-target distance
target_apparent_radius:
	Apparent radius of the target
target_description:
	Original target keywords
target_distance_max:
	Max observer-target distance
target_distance_min:
	Min observer-target distance
target_region:	Type of region or feature of interest
target_time_max:
	Max observing time in target frame
target_time_min:
	Min observing time in target frame
thumbnail_url:	URL of a thumbnail image with predefined size (png ~200 pix, for use in a client only)
time_refposition:
	Defines where the time is measured (e.g., ground vs. spacecraft). Defaults to the observer’s frame
time_scale:	Defaults to UTC in data services; takes values from http://www.ivoa.net/rdf/time_scale otherwise

The contributive works extension¶

observer_country:
observer_code:	Image observer’s username in service
	Image observer’s country of residence
observer_id:	Image observer numeric identifier in service
observer_institute:
	Observer institute
observer_lat:	Observer’s approximate latitude
observer_location:
	Broad, free-text location and geographic position of the observer or telescope. Can be used when the exact location cannot be released
observer_lon:	Observer’s approximate longitude
observer_name:	Observer name
original_publisher:
	Refers to the source of the data, e. g., in compilations of experimental data
producer_institute:
	Data producer institute, e. g., in compilations of experimental data
producer_name:	Data producer name, especially in compilations of experimental data

The events extension¶

event_cite:	Following VOEvent, this is one of followup, supersedes, retraction
event_status:	One of prediction, observation, or utility
event_type:	Type of event from a controlled vocabulary (meteor_shower, fireball, lunar_flash, comet_tail_crossing…)

The experimental spectroscopy extension¶

data_calibration_desc:
azimuth_max:	Max azimuth angle for illumination
azimuth_min:	Min azimuth angle for illumination
	Provides information on post-processing. Can be a hash list
geometry_type:	Type of observation, from a controlled vocabulary (cf. EPN-TAP specification). Can be a hash list
grain_size_max:	Max sample particle size
grain_size_min:	Min sample particle size
measurement_atmosphere:
	Describes experimental conditions. `vacuum` for measurements under vacuum
pressure:	Ambient pressure
sample_classification:
	Information related to class, sub-class, species… as hash list
sample_desc:	Describes the sample, its origin, and possible preparation. Can be a hash list
sample_id:	Additional ID of the sample, e.g., a specific fraction of a meteorite (in addition to target_name). Intended to refer to a pre-existing catalogue of a collection, will therefore contain a name/id mainly for local use
species_inchikey:
	Machine-readable description of the species involved. Can be a hash list
setup_desc:	Describes the experimental setup. Can be a hash list
spectrum_type:	Type of spectral observation, from a controlled vocabulary (under construction). Can be a hash list
temperature:	Ambient temperature

The maps extension¶

map_height:	Map size in px
map_projection:	The map projection, preferably as a FITS name or code, or parameters as a free string
map_scale:	Format TBD
map_width:	Map size in px
pixelscale_max:	Max pixel size at a surface
pixelscale_min:	Min pixel size at a surface

The orbital elements extension¶

semi_major_axis:
arg_perihel:	Argument of Perihelion, J2000.0
eccentricity:	Orbit eccentricity
epoch:	Epoch of interest
inclination:	Orbit inclination
long_asc:	Longitude of ascending node, J2000.0
mean_anomaly:	Mean anomaly at the epoch

The particle spectroscopy extension¶

particle_spectral_range_max:
	Upper bound of the mass/energy of the particles
particle_spectral_range_min:
	Lower bound of the mass/energy of the particles
particle_spectral_resolution_max:
	Worst resolution of the spectral range
particle_spectral_resolution_min:
	Best resolution of the spectral range
particle_spectral_sampling_step_max:
	Maximal separation of different values in the spectral range
particle_spectral_sampling_step_min:
	Minimal separation of different values in the spectral range
particle_spectral_type:
	The type of axis in use; this is one of `energy` (which is then in eV), `mass` (amu) or `mass/charge` (in amu/e). If you use any of this, please contact the DaCHS authors; this needs more work

The solar system objects extension¶

dynamical_class:
diameter:	Target diameter, or equivalent diameter for binary objects
	Class of small body, from a controlled vocabulary (see the EPN-TAP specification)
dynamical_type:	Subdivision of the dynamical class, from a controlled vocabulary
equatorial_radius:
	Equatorial radius of a solar system object.
mass:	Mass of object
mean_radius:	Mean radius of a solar system object.
polar_radius:	Polar radius of a solar system object.
sidereal_rotation_period:
	Object rotation rate
taxonomy_code:	Code for target taxonomy

This mixin has the following parameters:

Parameter optional_columns: Space-separated list of names of optional columns to include. For the column names available include see sect. 3 of the EPN-TAP specification or the list of the various sections above.
Parameter spatial_frame_type: Flavour of the coordinate system. Since this determines the units of the coordinates columns, this must be set globally for the entire dataset. Values defined by EPN-TAP and understood by this mixin include celestial, body, cartesian, cylindrical, spherical, none.

The //linetap#table-0 Mixin¶

This mixin makes a table suitable for publication as a LineTAP table.

It provides all standard columns, makes sure it is on disk and available through TAP, adds the most common indexes, and gives it the utype required by the standard.

It is recommended to fill this table with a rowmaker using the //linetap#populate-0 apply.

The //obscore#publish Mixin¶

Publish this table to ObsTAP.

This means mapping or giving quite a bit of data from the present table to ObsCore rows. Internally, this information is converted to an SQL select statement used within a create view statement. In consequence, you must give SQL expressions in the parameter values; just naked column names from your input table are ok, of course. Most parameters are set to NULL or appropriate defaults for tables mixing in //products#table.

Since the mixin generates script elements, it cannot be used in untrusted RDs. The fact that you can enter raw SQL also means you will get ugly error messages if you give invalid parameters.

Some items are filled from product interface fields automatically. You must change these if you obscore-publish tables not mixin in products.

Note: you must say dachs imp //obscore before anything obscore-related will work.

This mixin has the following parameters:

Parameter access_estsize: defaults to accsize/1024; The estimated size of the product in kilobytes. Only touch when you do not mix in products#table.
Parameter access_format: defaults to mime; The MIME type of the product file. Only touch if you do not mix in products.
Parameter access_url: defaults to accref; URL at which the product can be obtained. Leave as is for tables mixing in products.
Parameter calib_level: defaults to 2; Calibration level of data, a number between 0 and 3; for details, see http://dc.g-vo.org/tableinfo/ivoa.obscore#note-calib
Parameter createDIDIndex: defaults to False; Index whatever expression you give for the dataset identifier? You probably want this on tables with more than a few hundred datasets unless the DID is in a column that you already index anyway.
Parameter dataproduct_subtype: defaults to NULL; File subtype. Details pending
Parameter dataproduct_type: Data product type; one of image, cube, spectrum, sed, timeseries, visibility, event, or NULL if None of the above
Parameter em_max: defaults to NULL; Upper bound of wavelengths represented in the data set, in meters.
Parameter em_min: defaults to NULL; Lower bound of wavelengths represented in the data set, in meters.
Parameter em_res_power: defaults to NULL; Spectral resolution as lambda/delta lambda
Parameter em_ucd: defaults to NULL; UCD of the spectral axis as defined by the spectrum DM, plus a few values defined in obscore 1.1 for Doppler axes
Parameter em_xel: defaults to NULL; Number of samples along the spectral axis
Parameter facility_name: defaults to NULL; The institute or observatory at which the data was produced
Parameter instrument_name: defaults to NULL; The instrument that produced the data
Parameter o_ucd: defaults to NULL; UCD of the observable quantity, e.g., em.opt for wide-band optical frames.
Parameter obs_collection: defaults to 'unnamed'; A human-readable name for this collection. This should be short, so don’t just use the resource title
Parameter obs_creator_did: defaults to NULL; Global identifier of the data set assigned by the creator. Leave NULL unless the creator actually assigned an IVO id themselves.
Parameter obs_id: defaults to accref; Identifier of the data set. Only change this when you do not mix in products.
Parameter obs_publisher_did: defaults to __COMPUTE__; Global identifier of the data set. Leave __COMPUTE__ for tables mixing in products.
Parameter obs_title: defaults to NULL; A human-readable title of the data set.
Parameter pol_states: defaults to NULL; List of polarization states present in the data; if you give something, use the convention of choosing the appropriate from {I Q U V RR LL RL LR XX YY XY YX POLI POLA} and write them in alphabetical order with / separators, e.g. /I/Q/XX/.
Parameter pol_xel: defaults to NULL; Number of polarisation states in this product
Parameter preview: defaults to accref || '?preview=True'; The preview URL will point to the products service; if you don’t have any previews, you will have to set this to NULL manually.
Parameter s_dec: defaults to NULL; Center Dec
Parameter s_fov: defaults to NULL; Approximate diameter of region covered
Parameter s_pixel_scale: defaults to NULL; Size of a spatial pixel (in arcsec)
Parameter s_ra: defaults to NULL; Center RA
Parameter s_region: defaults to NULL; A polygon giving the spatial coverage of the data set; this must always be in ICRS. This is cast to an pgsphere spoly, which currently means that you have to provide an spoly (reference), too.
Parameter s_resolution: defaults to NULL; The (best) angular resolution within the data set, in arcsecs
Parameter s_xel1: defaults to NULL; Number of pixels along the first spatial axis
Parameter s_xel2: defaults to NULL; Number of pixels along the second spatial axis
Parameter source_table: defaults to '\qName'; The table this is mixed into (you probably don’t want to change this).
Parameter t_exptime: defaults to NULL; Total time of event counting. This simply is t_max-t_min for simple exposures.
Parameter t_max: defaults to NULL; MJD for the upper bound of times covered in the data set. See tMin
Parameter t_min: defaults to NULL; MJD for the lower bound of times covered in the data set (e.g. start of exposure). Use ts_to_mjd(ts) to get this from a postgres timestamp.
Parameter t_resolution: defaults to NULL; Temporal resolution (in seconds)
Parameter t_xel: defaults to NULL; Number of samples along the time axis
Parameter target_class: defaults to NULL; Class of target object(s). You should take whatever you put here from http://simbad.u-strasbg.fr/guide/chF.htx
Parameter target_name: defaults to NULL; Name of the target object.

The //obscore#publishObscoreLike Mixin¶

Publish a table that already has large parts of the obscore schema.

This has parameters called like the corresponding obscore columns that also default to taking their data columns named like them. Use this when you already have, by and large, and obscore structure in your source table.

Note that this will not do the right thing with product#table instances by default. For these, you will have to manually map access_url, access_format, and access_estsize.

This mixin has the following parameters:

Parameter access_estsize: defaults to access_estsize; Estimated size of data product
Parameter access_format: defaults to access_format; MIME type of the resource at access_url
Parameter access_url: defaults to access_url; The URL at which to obtain the data set.
Parameter calib_level: defaults to calib_level; Amount of data processing that has been applied to the data
Parameter createDIDIndex: defaults to False; Index whatever expression you give for the dataset identifier? You probably want this on tables with more than a few hundred datasets unless the DID is in a column that you already index anyway.
Parameter dataproduct_subtype: defaults to dataproduct_subtype; Data product specific type
Parameter dataproduct_type: defaults to dataproduct_type; High level scientific classification of the data product, taken from an enumeration
Parameter em_max: defaults to em_max; Maximal wavelength represented within the data set
Parameter em_min: defaults to em_min; Minimal wavelength represented within the data set
Parameter em_res_power: defaults to em_res_power; Spectral resolving power lambda/delta lamda
Parameter em_ucd: defaults to em_ucd; Nature of the product’s spectral axis (typically, em.freq, em.wl, or em.energy)
Parameter em_xel: defaults to em_xel; Number of elements (typically pixels) along the spectral axis.
Parameter facility_name: defaults to facility_name; Name of the facility at which data was taken
Parameter instrument_name: defaults to instrument_name; Name of the instrument that produced the data
Parameter o_ucd: defaults to o_ucd; UCD for the product’s observable
Parameter obs_collection: defaults to obs_collection; Name of a data collection (e.g., project name) this data belongs to
Parameter obs_creator_did: defaults to obs_creator_did; Dataset identifier assigned by the creator.
Parameter obs_id: defaults to obs_id; Unique identifier for an observation
Parameter obs_publisher_did: defaults to obs_publisher_did; Dataset identifier assigned by the publisher.
Parameter obs_title: defaults to obs_title; Free-from title of the data set
Parameter pol_states: defaults to pol_states; List of polarization states in the data set
Parameter pol_xel: defaults to pol_xel; Number of elements (typically pixels) along the polarization axis.
Parameter preview: defaults to preview; URL of a preview (low-resolution, quick-to-retrieve representation) of the data.
Parameter s_dec: defaults to s_dec; Dec of (center of) observation, ICRS
Parameter s_fov: defaults to s_fov; Approximate spatial extent for the region covered by the observation
Parameter s_pixel_scale: defaults to s_pixel_scale; Sampling period in world coordinate units along the spatial axis
Parameter s_ra: defaults to s_ra; RA of (center of) observation, ICRS
Parameter s_region: defaults to s_region; Region covered by the observation, as a polygon
Parameter s_resolution: defaults to s_resolution; Best spatial resolution within the data set
Parameter s_xel1: defaults to s_xel1; Number of elements (typically pixels) along the first spatial axis.
Parameter s_xel2: defaults to s_xel2; Number of elements (typically pixels) along the second spatial axis.
Parameter source_table: defaults to '\qName'; The table this is mixed into (you probably don’t want to change this).
Parameter t_exptime: defaults to t_exptime; Total exposure time
Parameter t_max: defaults to t_max; Upper bound of times represented in the data set
Parameter t_min: defaults to t_min; Lower bound of times represented in the data set
Parameter t_resolution: defaults to t_resolution; Minimal significant time interval along the time axis
Parameter t_xel: defaults to t_xel; Number of elements (typically pixels) along the time axis.
Parameter target_class: defaults to target_class; Class of the target object (star, QSO, …)
Parameter target_name: defaults to target_name; Object a targeted observation targeted

The //obscore#publishSIAP Mixin¶

Publish a PGS SIAP table to ObsTAP.

This works like //obscore#publish except some defaults apply that copy fields that work analogously in SIAP and in ObsTAP.

For special situations, you can, of course, override any of the parameters, but most of them should already be all right. To find out what the parameters described as “preset for SIAP” mean, refer to //obscore#publish.

Note: you must say dachs imp //obscore before anything obscore-related will work.

This mixin has the following parameters:

Parameter access_estsize: defaults to accsize/1024; The estimated size of the product in kilobytes. Only touch when you do not mix in products#table.
Parameter access_format: defaults to mime; The MIME type of the product file. Only touch if you do not mix in products.
Parameter access_url: defaults to accref; URL at which the product can be obtained. Leave as is for tables mixing in products.
Parameter calib_level: defaults to 2; Calibration level of data, a number between 0 and 3; for details, see http://dc.g-vo.org/tableinfo/ivoa.obscore#note-calib
Parameter createDIDIndex: defaults to False; Index whatever expression you give for the dataset identifier? You probably want this on tables with more than a few hundred datasets unless the DID is in a column that you already index anyway.
Parameter dataproduct_subtype: defaults to NULL; File subtype. Details pending
Parameter dataproduct_type: defaults to 'image'; preset for SIAP
Parameter em_max: defaults to bandpassHi; preset for SIAP
Parameter em_min: defaults to bandpassLo; preset for SIAP
Parameter em_res_power: defaults to NULL; Spectral resolution as lambda/delta lambda
Parameter em_ucd: defaults to NULL; UCD of the spectral axis as defined by the spectrum DM, plus a few values defined in obscore 1.1 for Doppler axes
Parameter em_xel: defaults to NULL; Number of samples along the spectral axis
Parameter facility_name: defaults to \sqlquote{\metaSeq{facility}{NULL}}; Default is to take from resource meta
Parameter instrument_name: defaults to instId; The instrument that produced the data
Parameter o_ucd: defaults to 'em.opt'; preset for SIAP; fix if you either know more about the band or if your images are not in the optical.
Parameter obs_collection: defaults to 'unnamed'; A human-readable name for this collection. This should be short, so don’t just use the resource title
Parameter obs_creator_did: defaults to NULL; Global identifier of the data set assigned by the creator. Leave NULL unless the creator actually assigned an IVO id themselves.
Parameter obs_id: defaults to accref; Identifier of the data set. Only change this when you do not mix in products.
Parameter obs_publisher_did: defaults to __COMPUTE__; Global identifier of the data set. Leave __COMPUTE__ for tables mixing in products.
Parameter obs_title: defaults to imageTitle; preset for SIAP
Parameter pol_states: defaults to NULL; List of polarization states present in the data; if you give something, use the convention of choosing the appropriate from {I Q U V RR LL RL LR XX YY XY YX POLI POLA} and write them in alphabetical order with / separators, e.g. /I/Q/XX/.
Parameter pol_xel: defaults to NULL; Number of polarisation states in this product
Parameter preview: defaults to accref || '?preview=True'; The preview URL will point to the products service; if you don’t have any previews, you will have to set this to NULL manually.
Parameter s_dec: defaults to centerDelta; preset for SIAP
Parameter s_fov: defaults to pixelScale[1]*pixelSize[1]; preset for SIAP; we use the extent along the X axis as a very rough estimate for the size. If you can do better, by all means do.
Parameter s_pixel_scale: defaults to pixelScale[1]*3600; preset for SIAP
Parameter s_ra: defaults to centerAlpha; preset for SIAP
Parameter s_region: defaults to coverage; preset for SIAP
Parameter s_resolution: defaults to pixelScale[1]*3600; preset for SIAP; this is just the pixel scale in one dimension. If that’s seriously wrong or you have uncalibrated images in your collection, you may need to be more careful here.
Parameter s_xel1: defaults to pixelSize[1]; preset for SIAP
Parameter s_xel2: defaults to pixelSize[2]; preset for SIAP
Parameter source_table: defaults to '\qName'; The table this is mixed into (you probably don’t want to change this).
Parameter t_exptime: defaults to NULL; Total time of event counting. This simply is t_max-t_min for simple exposures.
Parameter t_max: defaults to dateObs; preset for SIAP; if you want, change this to end of observation as available.
Parameter t_min: defaults to dateObs; preset for SIAP; if you want, change this to start of observation as available.
Parameter t_resolution: defaults to NULL; Temporal resolution (in seconds)
Parameter t_xel: defaults to NULL; Number of samples along the time axis
Parameter target_class: defaults to NULL; Class of target object(s). You should take whatever you put here from http://simbad.u-strasbg.fr/guide/chF.htx
Parameter target_name: defaults to NULL; Name of the target object.

The //obscore#publishSSAPMIXC Mixin¶

Publish a table mixing in //ssap#view (or the deprecated //ssap#mixc) to ObsTAP.

This works like the //obscore#publish mixin except some defaults apply that copy fields that work analogously in SSAP and in ObsTAP.

The columns already set in SSAP are marked as UNDOCUMENTED in the parameter list below. For special situations, you can, of course, override any of the parameters. To find out what they actually mean, mean, refer to the //obscore#publish mixin.

Note that this mixin does not set coverage (obscore: s_region). This is because although we could make a circle from ssa_location and ssa_aperture, circles are not allowed in DaCHS’ s_region (which has a fixed type of spoly). The recommended solution to still have s_region is to add (and index) a custom field; the //ssap#simpleCoverage will do this.

Note: you must say dachs imp //obscore before anything obscore-related will work.

This mixin has the following parameters:

Parameter access_estsize: defaults to accsize/1024; The estimated size of the product in kilobytes. Only touch when you do not mix in products#table.
Parameter access_format: defaults to mime; The MIME type of the product file. Only touch if you do not mix in products.
Parameter access_url: defaults to accref; URL at which the product can be obtained. Leave as is for tables mixing in products.
Parameter calib_level: defaults to 2; Calibration level of data, a number between 0 and 3; for details, see http://dc.g-vo.org/tableinfo/ivoa.obscore#note-calib
Parameter createDIDIndex: defaults to False; Index whatever expression you give for the dataset identifier? You probably want this on tables with more than a few hundred datasets unless the DID is in a column that you already index anyway.
Parameter dataproduct_subtype: defaults to NULL; File subtype. Details pending
Parameter dataproduct_type: defaults to ssa_dstype; Default should work for you
Parameter em_max: defaults to ssa_specend; Default should work for you
Parameter em_min: defaults to ssa_specstart; Default should work for you
Parameter em_res_power: defaults to ssa_specstart/ssa_specres; Default should work for you
Parameter em_ucd: defaults to \sqlquote{\getParam{ssa_spectralucd}}; Default should work for you
Parameter em_xel: defaults to ssa_length; Default should work for you
Parameter facility_name: defaults to \sqlquote{\metaSeq{facility}{NULL}}; Default should work for you
Parameter instrument_name: defaults to ssa_instrument; Default should work for you
Parameter o_ucd: defaults to \sqlquote{\getParam{ssa_fluxucd}}; Default should work for you
Parameter obs_collection: defaults to ssa_collection; Default should work for you
Parameter obs_creator_did: defaults to ssa_creatorDID; Default should work for you
Parameter obs_id: defaults to accref; Identifier of the data set. Only change this when you do not mix in products.
Parameter obs_publisher_did: defaults to ssa_pubdid; Default should work for you.
Parameter obs_title: defaults to ssa_dstitle; Default should work for you
Parameter pol_states: defaults to NULL; List of polarization states present in the data; if you give something, use the convention of choosing the appropriate from {I Q U V RR LL RL LR XX YY XY YX POLI POLA} and write them in alphabetical order with / separators, e.g. /I/Q/XX/.
Parameter pol_xel: defaults to NULL; Number of polarisation states in this product
Parameter preview: defaults to accref || '?preview=True'; The preview URL will point to the products service; if you don’t have any previews, you will have to set this to NULL manually.
Parameter s_dec: defaults to degrees(lat(ssa_location)); Default should work for you; for large data collections, consider having a separate Dec column with a q3c index.
Parameter s_fov: defaults to ssa_aperture; Default should work for you
Parameter s_pixel_scale: defaults to NULL; Size of a spatial pixel (in arcsec)
Parameter s_ra: defaults to degrees(long(ssa_location)); Default should work for you; for large data collections, consider having a separate RA column with a q3c index.
Parameter s_region: defaults to NULL; Use ssa_region when the table also mixes in //ssap#simpleCoverage
Parameter s_resolution: defaults to \getParam{ssa_spaceRes}{NULL}/3600.; Default should work for you
Parameter s_xel1: defaults to 1; You shouldn’t use SSA for cubes.
Parameter s_xel2: defaults to 1; You shouldn’t use SSA for cubes.
Parameter source_table: defaults to '\qName'; The table this is mixed into (you probably don’t want to change this).
Parameter t_exp_time: defaults to ssa_timeExt; Default should work for you
Parameter t_exptime: defaults to NULL; Total time of event counting. This simply is t_max-t_min for simple exposures.
Parameter t_max: defaults to ssa_dateObs+ssa_timeExt/43200.; Default should work for you
Parameter t_min: defaults to ssa_dateObs-ssa_timeExt/43200.; Default should work for you
Parameter t_resolution: defaults to NULL; Temporal resolution (in seconds)
Parameter t_xel: defaults to NULL; Number of samples along the time axis
Parameter target_class: defaults to ssa_targclass; Default should work for you
Parameter target_name: defaults to ssa_targname; Default should work for you

The //products#table Mixin¶

A mixin for tables containing “products”.

A “product” here is some kind of binary, typically a FITS file. The table receives the columns accref, accsize, owner, and embargo (which is defined in //products#prodcolUsertable).

By default, the accref is the path to the file relative to the inputs directory; this is also what /getproduct expects for local products. You can of course enter URLs to other places.

For local files, you are strongly encouraged to keep the accref URL- and shell-clean, the most important reason being your users’ sanity. Another is that obscore in the current implementation does no URL escaping for local files. So, just don’t use characters like like +, the ampersand, apostrophes and so on; the default accref parser will reject those anyway. Actually, try making do with alphanumerics, the underscore, the dash, and the dot, ok?

owner and embargo let you introduce access control. Embargo is a date at which the product will become publicly available. As long as this date is in the future, only authenticated users belonging to the group owner are allowed to access the product.

In addition, the mixin arranges for the products to be added to the system table products, which is important when delivering the files.

Tables mixing this in should be fed from grammars using the //products#define row filter.

The //scs#pgs-pos-index Mixin¶

A mixin adding a pgsphere index to the main spherical position for tables with separate RA and Dec columns.

You have to designate exactly one column with the ucds pos.eq.ra;meta.main pos.eq.dec;meta.main, respectively. These columns receive the positional index.

This should be used instead of q3cindex on new tables; it’s a bit slower than q3c, but it’s less funky, too.

If you’d like an index on other sorts of long/lat pairs, see the //scs#spoint-index-def STREAM.

The //scs#q3cindex Mixin¶

A mixin adding an index to the main equatorial positions.

In new RDs, use pgs-pos-index instead; we’d like to stop q3c support at some point.

This is what you usually want if your input data already has “sane” (i.e., ICRS or at least J2000) positions or you convert the positions manually.

You have to designate exactly one column with the ucds pos.eq.ra;meta.main pos.eq.dec;meta.main, respectively. These columns receive the positional index.

This will fail without the q3c extension to postgres.

The //siap2#pgs Mixin¶

A mixin pulling in all columns necessary to support SIAP2.

This is pulls in all obscore columns, including any you define locally in %#obscore-extracolumns. In DaCHS, we additionally have the columns coming from the products table. This latter fact means that the grammar filling tables mixing this in will need a //products#define rowfilter.

To feed these tables, use the //siap2#computePGS and the //siap2#setMeta applies in the rowmaker.

Added in 2.7.3.

The //slap#basic Mixin¶

This mixin is for tables serving SLAP services, i.e., tables with spectral lines. It does not contain all “optional” columns, hence the name basic. We’d do “advanced”, too, if there’s demand.

Use the //slap#fillBasic procDef to populate such tables.

The //ssap#hcd Mixin¶

Deprecated. use the //ssap#view mixin instead.

This mixin is for “homogeneous” data collections, where homogeneous means that all values in hcd_outpars are constant for all datasets in the collection. This is usually the case if they all come from one instrument.

Rowmakers for tables using this mixin should use the //ssap#setMeta proc application.

Do not forget to call the //products#define row filter in grammars feeding tables mixing this in. At the very least, you need to say:

<rowfilter procDef="//products#define">
  <bind name="table">"mySchema.myTableName"</bind>
</rowfilter>

This mixin has the following parameters:

Parameter collection: defaults to __NULL__; ivo id of the originating collection; ssa:DataID.Collection
Parameter creationType: defaults to __NULL__; Process used to produce the data (zero or more of archival, cutout, filtered, mosaic, projection, spectralExtraction, catalogExtraction); ssa:DataID.CreationType
Parameter creator: defaults to __NULL__; Creator designation; ssa:DataID.Creator
Parameter datasource: defaults to __NULL__; Generation type (typically, one survey, pointed, theory, custom, artificial); ssa:DataID.DataSource
Parameter fluxCalibration: Type of flux calibration (one of ABSOLUTE, RELATIVE, NORMALIZED, or UNCALIBRATED); ssa:Char.FluxAxis.Calibration
Parameter fluxSI: defaults to __NULL__; SI conversion factor for fluxes in the spectrum instance (not the SSA metadata) in Osuna-Salgado convention; ssa:Dataset.FluxSI (you probably want to leave this empty)
Parameter fluxUCD: defaults to phot.flux.density;em.wl; ucd of the flux column, like phot.count, phot.flux.density, etc. Default is for flux over wavelength; ssa:Char.FluxAxis.Ucd
Parameter fluxUnit: Flux unit used by the spectra and in SSA char metadata. This must be a VOUnit string (use a single blank if your spectrum is not calibrated).
Parameter instrument: defaults to __NULL__; Instrument or code used to produce these datasets; ssa:DataID.Instrument
Parameter publisher: defaults to \metaString{publisherID}{\metaString{publisher}}; Publisher IVO (by default taken from the DC config); ssa:Curation.Publisher
Parameter reference: defaults to __NULL__; URL or bibcode of a publication describing this data; ssa:Curation.Reference
Parameter refposition: defaults to UNKNOWN; Reference position for times given in this table. Choose from http://www.ivoa.net/rdf/refposition, leave as UNKNOWN if in doubt.
Parameter spectralCalibration: defaults to __NULL__; Type of wavelength Calibration (one of ABSOLUTE, RELATIVE, NORMALIZED, or UNCALIBRATED); ssa:Char.SpectralAxis.Calibration
Parameter spectralResolution: defaults to NaN; Resolution on the spectral axis; you must give this as FWHM wavelength in meters here. Approximate as necessary; ssa:Char.SpectralAxis.Resolution
Parameter spectralSI: defaults to __NULL__; SI conversion factor of frequency or wavelength in the spectrum instance (not the SSA metadata, they are all in meters); ssa:Dataset.SpectralSI (you probably want to leave this empty)
Parameter spectralUCD: defaults to em.wl; ucd of the spectral column, like em.freq or em.energy; default is wavelength; ssa:Char.SpectralAxis.Ucd
Parameter spectralUnit: Spectral unit used by the spectra (SSA char metadata always is wavelength in meters). This must be a VOUnit string (use a single blank if your spectrum is not calibrated).
Parameter statFluxError: defaults to __NULL__; Statistical error in flux; ssa:Char.FluxAxis.Accuracy.StatError
Parameter statSpaceError: defaults to __NULL__; Statistical error in position in degrees; ssa:Char.SpatialAxis.Accuracy.StatError
Parameter statSpectError: defaults to __NULL__; Statistical error in wavelength (units of specralSI); ssa:Char.SpectralAxis.Accuracy.StatError
Parameter sysFluxError: defaults to __NULL__; Systematic error in flux; ssa:Char.FluxAxis.Accuracy.SysError
Parameter sysSpectError: defaults to __NULL__; Systematic error in wavelength (in m); ssa:Char.SpectralAxis.Accuracy.SysError
Parameter timeSI: defaults to __NULL__; SI conversion factor for times in Osuna-Salgado convention; ssa:DataSet.TimeSI (you probably want to leave this empty)
Parameter timescale: defaults to TT; Timescale for times given in this table. Choose from http://www.ivoa.net/rdf/timescale. If in doubt, leaving this at TT is probably good enough for discovery.

The //ssap#mixc Mixin¶

Deprecated. use the //ssap#view mixin instead.

This mixin provides the columns and params for a common SSA service.

Rowmakers for tables using this mixin should use the //ssap#setMeta and the //ssap#setMixcMeta proc applications.

There are some limitations to the variability; in particular, all spectra must have the same types of axes (i.e., frequency, wavelength, or energy) with identical units. If you don’t have that, either leave the respective metadata empty or homogenize it before ingestion.

Do not forget to call the //products#define row filter in grammars feeding tables mixing this in. At the very least, you need to say:

<rowfilter procDef="//products#define">
  <bind name="table">"schema.table"</bind>
</rowfilter>

This mixin has the following parameters:

Parameter fluxSI: defaults to __NULL__; SI conversion factor for fluxes in the spectrum instance (not the SSA metadata) in Osuna-Salgado convention; ssa:Dataset.FluxSI (you probably want to leave this empty)
Parameter fluxUCD: defaults to phot.flux.density;em.wl; ucd of the flux column, like phot.count, phot.flux.density, etc. Default is for flux over wavelength; ssa:Char.FluxAxis.Ucd
Parameter fluxUnit: Flux unit used by the spectra and in SSA char metadata. This must be a VOUnit string (use a single blank if your spectrum is not calibrated).
Parameter refposition: defaults to UNKNOWN; Reference position for times given in this table. Choose from http://www.ivoa.net/rdf/refposition, leave as UNKNOWN if in doubt.
Parameter spectralSI: defaults to __NULL__; SI conversion factor of frequency or wavelength in the spectrum instance (not the SSA metadata, they are all in meters); ssa:Dataset.SpectralSI (you probably want to leave this empty)
Parameter spectralUCD: defaults to em.wl; ucd of the spectral column, like em.freq or em.energy; default is wavelength; ssa:Char.SpectralAxis.Ucd
Parameter spectralUnit: Spectral unit used by the spectra (SSA char metadata always is wavelength in meters). This must be a VOUnit string (use a single blank if your spectrum is not calibrated).
Parameter timeSI: defaults to __NULL__; SI conversion factor for times in Osuna-Salgado convention; ssa:DataSet.TimeSI (you probably want to leave this empty)
Parameter timescale: defaults to TT; Timescale for times given in this table. Choose from http://www.ivoa.net/rdf/timescale. If in doubt, leaving this at TT is probably good enough for discovery.

The //ssap#plainlocation Mixin¶

A mixin that adds ssa_location column to a table.

You probably want this in the source tables for //ssap#view tables. This will also index the column. At least if you later want to publish the data through obscore, you will also want the //ssap#simpleCoverage mixin if you mix this in.

Use the //ssap#fill-plainlocation apply to feed these.

The //ssap#sdm-instance Mixin¶

This mixin is intended for tables that get serialized into documents conforming to the Spectral Data Model 1, specifically to VOTables

The input to such tables comes from ssa tables (hcd, in this case). Their columns (and params) are transformed into params here.

The mixin adds two columns (you could add more if, e.g., you had errors depending on the spectral or flux value), spectral (wavelength or the like) and flux. Their metadata is taken from the ssa fields where available (ssa_fluxucd as flux UCD, ssa_fluxunit etc).

This mixin in action could look like this:

<table id="instance" onDisk="False">
  <mixin ssaTable="spectra"
    fluxUnit="Jy"
    >//ssap#sdm-instance</mixin>
</table>

The mixin thus defines a gazillion of params. This will almost always be filled using //ssap#feedSSAToSDM as explaned in SDM compliant tables

This mixin has the following parameters:

Parameter fluxDescription: defaults to The dependent variable of this spectrum (see the ucd for its physical meaning); Description for the flux column
Parameter fluxUCDOverride: Force UCD of the spectral column (don’t use this; fix your SSA table instead)
Parameter fluxUnitOverride: Force unit of the spectral column (don’t use this; fix your SSA table instead)
Parameter spectralDescription: defaults to The independent variable of this spectrum (see its ucd to figure out whether it's a wavelength, frequency, or energy); Description for the spectral column
Parameter spectralUCDOverride: Force UCD of the spectral column (don’t use this)
Parameter spectralUnitOverride: Force unit of the spectral column (don’t use this; fix your SSA table instead)
Parameter ssaTable: The SSAP (HCD) instance table to take the params from

The //ssap#simpleCoverage Mixin¶

A mixin furnishes a table with an ssa_region column giving a polygonal coverage. For SSA itself, that’s unnecessary, but it’s highly recommended if you have data with positional and aperture data and will publish it via obscore, too (which in turn is highly recommended).

The column will be filled with a hexagon approximating the aperture. This is done by //ssap#fill-plainlocation (or, historically, by //ssap#setMeta), so usually you’re all set with this mixin. We also create an index for the ssa_region field.

To make it visible in obscore, you must bind the ssa_region parameter of the //ssap#view mixin to ssa_region (so the column is in the SSAP table, and the coverage mixin par of the //obscore#publishSSAPMIXC mixin to ssa_region (so the value ends up in obscore’s s_region).

The //ssap#view Mixin¶

This mixin produces an SSA-ready relation as a view.

The idea is that you import your spectra into a table suitable for your particular data collection (but mixing in //products#define). You then fill the columns for an SSA response giving in each mixin parameter here either with a column reference (as a simple column name) or with a SQL literals (put strings into single quotes – sourcetable is the exception here). Save typing by having the final column names in the source table and using the copiedcolumns mixin par.

If you have positions for your spectra, you probably want to also mix in the //ssap#plainlocation mixin in the original table in order to have indexed positions in a way suitable for SSA queries.

In general, you will have to generate indices on the source table; postgres doesn’t support indices on views. If you can’t use the plainlocation mixin, please not that the the SSA engine expects spoints as the location (and these would be indexed like <index columns=”loc_col” method=”GIST”/>).

The mixin will automatically create an index over whatever you give for ssa_pubDID (if you give something).

This mixin has the following parameters:

Parameter accref: defaults to accref; Access key for the data
Parameter accsize: defaults to accsize; Size of the data in bytes
Parameter copiedcolumns: rowmaker/idmaps-like list of strings with column names or shell patterns to copy from sourcetable. These columns will automatically be taken over in the view. SSA columns taken over in this way cannot be overridden in mixin parameters; the mandatory mixin parameters ssa_spectralunit and ssa_fluxcalib have to be given (with ignored values) even if they are in copiedcolumns.
Parameter customcode: A SQL fragment with extra column definitions to go into the view. This must start with a comma (because it’s going to be appended to the select clause).
Parameter embargo: defaults to embargo; Date the data will become/became public
Parameter materialize: defaults to False; Set to True to make this a materialized view. For smallish tables, that is a reasonable thing to do that generally simplifies the handling a bit. On the other hand, the view will no longer follow the underlying tables.
Parameter mime: defaults to mime; MIME type of the file served
Parameter owner: defaults to owner; Owner of the data
Parameter refposition: defaults to UNKNOWN; Reference position for times given in this table. Choose from http://www.ivoa.net/rdf/refposition, leave as UNKNOWN if in doubt.
Parameter sourcetable: Reference to the table to build the view upon (i.e., the value of its id attribute). This mixin can only build onto a single table. To make an SSA table based on a join of multiple underlying table, define an intermediate table with a viewStatement of its own and use that as sourcetable.
Parameter ssa_aperture: defaults to NULL; Angular diameter of aperture
Parameter ssa_bandpass: defaults to NULL; Bandpass (i.e., rough spectral location) of this dataset; this should be the most appropriate term from the vocabulary http://www.ivoa.net/rdf/messenger.
Parameter ssa_binSize: defaults to NULL; Size of the typical spectral bin in meters of wavelength.
Parameter ssa_cdate: defaults to NULL; Dataset creation date as a datetime
Parameter ssa_collection: defaults to NULL; A short handle naming the collection this spectrum belongs to.
Parameter ssa_creationtype: defaults to \sqlquote{\metaString{ssap.creationType}{unknown}}; Process used to produce the data (archival, cutout, filtered, mosaic, projection, spectralExtraction, or catalogExtraction)
Parameter ssa_creator: defaults to NULL; Creator of the datasets included here.
Parameter ssa_creatorDID: defaults to NULL; Dataset identifier assigned by the creator
Parameter ssa_csysName: defaults to 'ICRS'; System RA and Dec are given in
Parameter ssa_cversion: defaults to NULL; Creator assigned version for this dataset (will be incremented when this particular item is changed).
Parameter ssa_datasource: defaults to \sqlquote{\metaString{ssap.dataSource}{unknown}}; Method of generation for the data (one of survey, pointed, theory, custom, artificial).
Parameter ssa_dateObs: defaults to NULL; Midpoint of exposure (MJD)
Parameter ssa_dstitle: defaults to NULL; A compact and descriptive designation of the dataset.
Parameter ssa_dstype: defaults to 'spectrum'; Type of data (spectrum, time series, etc)
Parameter ssa_fluxSI: defaults to NULL; It’s safe to ignore this.
Parameter ssa_fluxStatError: defaults to NULL; Statistical error in flux
Parameter ssa_fluxSysError: defaults to NULL; Systematic error in flux
Parameter ssa_fluxcalib: Type of flux calibration (ABSOLUTE, CALIBRATED, RELATIVE, NORMALIZED, or UNCALIBRATED).
Parameter ssa_fluxucd: defaults to 'phot.flux.density;em.wl'; UCD of the flux column
Parameter ssa_fluxunit: defaults to NULL; Unit of the flux column in the dataset. This is also the unit for the ssa_flux*Error columns (and thus better had not change between rows).
Parameter ssa_instrument: defaults to \sqlquote{\metaString{instrument}{unknown}}; Instrument or code used to produce these datasets
Parameter ssa_length: defaults to NULL; Number of points in the spectrum
Parameter ssa_location: defaults to NULL; You probably don’t want to manually handle this. Use the //ssap#plainlocation mixin on your metadata table instead and write ssa_location here.
Parameter ssa_model: defaults to 'Spectrum-1.0'; Data model name and version
Parameter ssa_pdate: defaults to \sqlquote{\today}; Date of (last) publication as a datetime
Parameter ssa_pubDID: defaults to \sql_standardPubDID; Dataset identifier assigned by the publisher
Parameter ssa_publisher: defaults to \sqlquote{\metaString{publisher}{unknown}}; Publisher of the datasets included here.
Parameter ssa_redshift: defaults to NULL; Redshift of target object
Parameter ssa_reference: defaults to \sqlquote{\metaString{source}{unknown}}; URL or bibcode of a publication describing this data.
Parameter ssa_region: defaults to not given; A region covered by the observation for this spectrum; this will be ssa_region when you mix //ssap#simpleCoverage into the original table. You’ll probably want to leave this at the default otherwise.
Parameter ssa_snr: defaults to NULL; Signal-to-noise ratio estimated for this dataset
Parameter ssa_spaceCalib: defaults to NULL; Type of calibration in spatial coordinates
Parameter ssa_spaceError: defaults to NULL; Statistical error in position
Parameter ssa_spaceRes: defaults to NULL; Spatial resolution of data
Parameter ssa_speccalib: defaults to NULL; Type of wavelength calibration
Parameter ssa_specend: defaults to NULL; Upper value of spectral coordinate
Parameter ssa_specext: defaults to (ssa_specend-ssa_specstart); Width of the spectrum
Parameter ssa_specmid: defaults to (ssa_specend+ssa_specstart)/2; Midpoint of region covered in this dataset
Parameter ssa_specres: defaults to NULL; Resolution (in meters of wavelength) on the spectral axis
Parameter ssa_specstart: defaults to NULL; Lower value of spectral coordinate
Parameter ssa_spectStatError: defaults to NULL; Statistical error in wavelength
Parameter ssa_spectSysError: defaults to NULL; Systematic error in wavelength
Parameter ssa_spectralSI: defaults to NULL; It’s safe to ignore this.
Parameter ssa_spectralucd: defaults to 'em.wl;obs.atmos'; UCD of the spectral column in the spectra served; when you have wavelengths, use em.wl for vacuum wavelengths, em.wl;obs.atmos for air wavelengths.
Parameter ssa_spectralunit: Unit of the flux column in the dataset (e.g., ‘Angstrom’). This is also the unit for the ssa_spectral*Error columns (and thus better had not change between rows).
Parameter ssa_targclass: defaults to NULL; Object class (star, QSO,…; use Simbad object classification http://simbad.u-strasbg.fr/simbad/sim-display?data=otypes if at all possible)
Parameter ssa_targetpos: defaults to NULL; Position of the intended target of the observation. You don’t usually need to give this. if you do, you probably want to write pgsphere.SPoint.fromDegrees(targ_ra, targ_dec)
Parameter ssa_targname: defaults to NULL; Common name of object observed.
Parameter ssa_timeExt: defaults to NULL; Exposure duration
Parameter ssa_timeSI: defaults to NULL; It’s safe to ignore this.
Parameter timescale: defaults to TT; Timescale for times given in this table. Choose from http://www.ivoa.net/rdf/timescale. If in doubt, leaving this at TT is probably good enough for discovery.
Parameter whereclause: A SQL fragment that fits after the source table (typically, a WHERE clause). Use this if you only want a subset of the original table to be visible through SSAP.

The //timeseries#phot-0 Mixin¶

Warning: The specs here are strongly in flux. The interface here is quite likely to change. If you use this mixin, please tell gavo@ari.uni-heidelberg.de so we can inform you of incompatible changes, and make sure you have robust regression tests in place.

A mixin for a simple photometric time series. Mix this in to get columns obs_time and phot_val, properly declared for the 2020 time series note.

Parameters marked with “SIL literal” can take a C-style token, a quoted string, or a reference to a column or param, where the name is prefixed by an @.

This mixin has the following parameters:

Parameter dependent_axes: defaults to [@phot]; Axes with values varying with time (only change if you define ones in addition to phot provided by the mixin; SIL literal)
Parameter effectiveWavelength: Central wavelength (or similar measure) for the passband used for the photometry, in meters (SIL literal)
Parameter filterIdentifier: Identifier for the passband the photometry is for (as defined by the Note). This is a SIL literal, and since it will probably contain characters like ‘/’, will likely have to quote it as in filterIdentifier='"Gaia/G"'.
Parameter independent_axes: defaults to [@obs_time]; Axes that describe the time (only change if you define ones in addition to the obs_time provided by the mixin; SIL literal)
Parameter latitude: defaults to __NULL__; Latitude/Dec of the position (SIL literal; you have to define params/columns yourself if you want to reference something here).
Parameter longitude: defaults to __NULL__; Longitude/RA of the position (SIL literal; you have to define params/columns yourself if you want to reference something here).
Parameter magnitudeSystem: defaults to __NULL__; Magnitude system used: Vega, AB,… (SIL literal)
Parameter phot_description: defaults to Flux; Description of the photometric coordinate.
Parameter phot_ucd: UCD of the photometric coordinate; something like phot.mag;em.opt.V or phot.flux;em.x-ray.hard
Parameter phot_unit: Unit of the photometric coordinate; something like mag, Jy, 1e-20 W/(Hz.m**2).
Parameter pos_epoch: defaults to __NULL__; Epoch you give the position for (SIL literal).
Parameter refframe: defaults to __NULL__; Frame of coordinates you may be giving (SIL literal).
Parameter refposition: Reference position for the time coordinate. Choose a value from http://www.ivoa.net/rdf/refposition (SIL literal).
Parameter time0: JD of the epoch (“day 0”) of the time coordinate; this is 0 for JD itself, 2400000.5 for MJD (SIL literal).
Parameter time_description: defaults to Time; Description of the time coordinate. We assume it’s in days; if you have something else, override the obs_time coordinate as described in mixins.
Parameter timescale: Time scale of the time coordinate. Choose a value from http://www.ivoa.net/rdf/timescale (SIL literal)
Parameter zeroPointFlux: defaults to __NULL__; Flux at the given zero point, in Jy; SIL literal).

Triggers ¶

In DaCHS, triggers are conditions on rows – either the raw rows emitted by grammars if they are used within grammars, or the rows about to be shipped to a table if they are used within tables. Triggers may be used recursively, i.e., triggers may contain more triggers. Child triggers are normally or-ed together.

Currently, there is one useful top-level trigger, the element ignoreOn. If an ignoreOn is triggered, the respective row is silently dropped (actually, you ignoreOn has a bail attribute that allows you to raise an error if the trigger is pulled; this is mainly for debugging).

The following triggers are defined:

Element and¶

A trigger that is true when all its children are true.

Atomic Children¶

name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Structure Children¶

triggers (contains any of keyPresent,keyMissing,keyNull,keyIs,not,and and may be repeated zero or more times) – One or more conditions joined by an implicit logical or. See Triggers for information on what can stand here.

Element keyIs¶

A trigger firing when the value of key in row is equal to the value given.

Missing keys are always accepted. You can define an SQL type; value will then be interpreted as a literal for this type, and this literal’s value will be compared against the key’s value. This is only needed for grammars like fitsProductGrammar that actually yield typed values.

Atomic Children¶

key (unicode string; defaults to <Undefined>) – Key to check
name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.
type (unicode string; defaults to ‘text’) – An SQL type the python equivalent of which the value should be converted to before checking.
value (unicode string; defaults to <Undefined>) – The string value to fire on.

Element keyMissing¶

A trigger firing if a certain key is missing in the dict.

This is equivalent to:

<not><keyPresent key="xy"/></not>

Atomic Children¶

key (unicode string; defaults to <Undefined>) – Key to check
name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Element keyNull¶

A trigger firing if a certain key is missing or NULL/None

Atomic Children¶

key (unicode string; defaults to <Undefined>) – Key to check
name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Element keyPresent¶

A trigger firing if a certain key is present in the dict.

Atomic Children¶

key (unicode string; defaults to <Undefined>) – Key to check
name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Element not¶

A trigger that is false when its children, or-ed together, are true and vice versa.

Atomic Children¶

name (unicode string; defaults to ‘unnamed’) – A name that should help the user figure out what trigger caused some condition to fire.

Structure Children¶

triggers (contains any of keyPresent,keyMissing,keyNull,keyIs,not,and and may be repeated zero or more times) – One or more conditions joined by an implicit logical or. See Triggers for information on what can stand here.

Renderers Available ¶

The following renderers are available for allowing and URL creation. The parameter style is relevant when adapting condDescs` or table based cores to renderers:

With clear, parameters are just handed through
With form, suitable parameters are turned into vizier-like expressions
With pql, suitable parameters are turned into their PQL counterparts, letting you specify ranges and such.

Unchecked renderers can be applied to any service and need not be explicitly allowed by the service.

The admin Renderer¶

This renderer’s parameter style is “clear”.

A renderer allowing to block and/or reload services.

This renderer could really be attached to any service since it does not call it, but it usually lives on //services/overview. It will always require authentication.

It takes the id of the RD to administer from the path segments following the renderer name.

By virtue of builtin vanity, you can reach the admin renderer at /seffe, and thus you can access /seffe/foo/q to administer the foo/q RD.

The api Renderer¶

This renderer’s parameter style is “dali”.

A renderer that works like a VO standard renderer but that doesn’t actually follow a given protocol.

Use this for improvised APIs. The default output format is a VOTable, and the errors come in VOSI VOTables. The renderer does, however, evaluate basic DALI parameters. You can declare that by including <FEED source=”//pql#DALIPars”/> in your service.

These will return basic service metadata if passed MAXREC=0.

The async Renderer¶

This renderer’s parameter style is “pql”.

A renderer speaking UWS.

This is for asynchronous execution of larger jobs. This is what is executed by the async renderer. It requests the worker system required from the service, which in turn obtains it from the core; these must hence cooperate with this to allow async operation.

See Custom UWSes for how to use this with your own cores.

The availability Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for a VOSI availability endpoint.

An endpoint with this renderer is automatically registered for every service. The answers can be configured using the admin renderer.

The capabilities Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for a VOSI capability endpoint.

An endpoint with this renderer is automatically registered for every service. The responses contain information on what renderers (“interfaces”) are available for a service and what properties they have.

This also doubles as a canary for authentication, which is why there are the somewhat complicated things in render; cf. https://wiki.ivoa.net/twiki/bin/view/IVOA/SSO_next

The coverage Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer returning various forms of a service’s spatial coverage.

This will return a 404 if the service doesn’t have a coverage.spatial meta (and will bomb out if that isn’t a SMoc).

Based on the accept header, it will return a PNG if the client indicates it’s interested in that or if it accepts text/html, in which case we assume it’s a browser; otherwise, it will produce a MOC in FITS format.

The custom Renderer¶

This renderer’s parameter style is “clear”.

A renderer defined in a python module.

To define a custom renderer write a python module and define a class MainPage inheriting from gavo.web.ServiceBasedPage.

This class basically is a gavo.formal.nevowc TemplatedPage, i.e., you can define loader, getChild, render, and so on.

To use it, you have to define a service with the resdir-relative path to the module in the customPage attribute and probably a nullCore. You also have to allow the custom renderer (but you may have other renderers, e.g., static).

If the custom page is for display in web browsers, define a class method isBrowseable(cls, service) returning true. This is for the generation of links like “use this service from your browser” only; it does not change the service’s behaviour with your renderer.

In general, avoid custom renderers. If you can’t, see the upstream twisted documentation on twisted.web.resource for how to write them.

The dali Renderer¶

This renderer’s parameter style is “clear”.

A meta-renderer for DALI-like multi-renderer services (sync, async, …)

This, for now, can only be used for creating registry records.

The dlasync Renderer¶

This renderer’s parameter style is “pql”.

A renderer for asynchronous datalink.

The dlget Renderer¶

This renderer’s parameter style is “clear”.

A renderer for data processing by datalink cores.

This must go together with a datalink core, nothing else will do.

This renderer will actually produce the processed data. It must be complemented by the dlmeta renderer which allows retrieving metadata.

The dlmeta Renderer¶

This renderer’s parameter style is “dali”.

A renderer for data processing by datalink cores.

This must go together with a datalink core, nothing else will do.

This renderer will return the links and services applicable to one or more pubDIDs.

See Datalink and SODA for more information.

The edition Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer representing a (tutorial-like) text document.

Not sure yet what I’ll do when people actually retrieve these.

This must have a meta accessURL with the document URI. It may have a sourceURL meta giving the VCS URI.

The examples Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for examples for service usage.

This renderer formats _example meta items in its service. Its output is XHTML compliant to VOSI examples; clients can parse it to, for instance, fill forms for service operation or display examples to users.

The examples make use of RDFa to convey semantic markup. To see what kind of semantics is contained, try http://www.w3.org/2012/pyRdfa/Overview.html and feed it the example URL of your service.

The default content of _example is ReStructuredText, and, really, not much else makes sense. An example for such a meta item can be viewed by executing gavo admin dumpDF //userconfig, in the tapexamples STREAM.

To support annotation of things within the example text, DaCHS defines several RST extensions, both interpreted text roles (used like :role-name:`content with blanks`) and custom directives (used to mark up blocks introduced by a single line like .. directive-name :: (the blanks before and after the directive name are significant).

Here’s the custom interpreted text roles:

dl-id: An publisher DID a service returns data for (used in datalink examples)
taptable: A (fully qualified) table name a TAP example query is (particularly) relevant for; in HTML, this is also a link to the table description.
genparam: A “generic parameter” as defined by DALI. The values of these have the form param(value), e.g., :genparam:`POS(32,4)`. Right now, not parentheses are allowed in the value. Complain if this bites you.

These are the custom directives:

tapquery: The query discussed in a TAP example.

Examples for how to write TAP examples are in the userconfig.rd distributed with DaCHS. Examples for Datalink examples can be found in the GAVO RDs feros/q and califa/q3.A

In addition, you can define moreExamples meta items. These point to further DALI-compliant examples document(s) and will typically be presented in a hierarchical fashion by clients. The content is either a URI to the examples document (when it starts with https?://) or a DaCHS-internal reference to a service with the examples. They should have a (short) title meta child to give clients a hint as to what the continuation is about, somewhat like this:

<meta>
  # more examples provided by the ex service in the RD rr/q
  moreExamples: rr/q#ex
  moreExamples.title: RegTAP
  # more examples provided by an external document
  moreExamples: http://ivoa.net/doc/obscore/tap-examples.xhtml
  moreExamples.title: ObsCore
</meta>

The external Renderer¶

This renderer’s parameter style is “clear”.

A renderer redirecting to an external resource.

These try to access an external publication on the parent service and ask it for an accessURL. If it doesn’t define one, this will lead to a redirect loop.

In the DC, external renderers are mainly used for registration of third-party browser-based services.

The fixed Renderer¶

This renderer’s parameter style is “clear”.

A renderer that renders a single template.

Use something like <template key="fixed">res/ft.html</template> in the enclosing service to tell the fixed renderer where to get this template from.

In the template, you can fetch parameters from the URL using something like <n:invisible n:data="parameter FOO" n:render="string"/>; you can also define new render and data functions on the service using customRF and customDF.

This is, in particular, used for the data center’s root page.

The fixed renderer is intended for non- or slowly changing content. It is annotated as cacheable, which means that DaCHS will in general only render it once and then cache it. If the render functions change independently of the RD, use the volatile renderer.

During development, users must add ?nocache=True to a fixed page URI to force DaCHS to reload the template.

Built-in services for such browser apps should go through the //run RD.

The form Renderer¶

This renderer’s parameter style is “form”.

The “normal” renderer within DaCHS for web-facing services.

It will display a form and allow outputs in various formats.

It also does error reporting as long as that is possible within the form.

The get Renderer¶

This renderer’s parameter style is “clear”.

The renderer used for delivering products.

This will only work with a ProductCore since the resulting data set has to contain products.Resources. Thus, you probably will not use this in user RDs.

The hips Renderer¶

This renderer’s parameter style is “clear”.

A static renderer with a few amenities for HiPS trees.

To make this work, set the service’s staticData property

The howtocite Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer that lets you format citation instructions.

The info Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer showing all kinds of metadata on a service.

This renderer produces the default referenceURL page. To change its appearance, override the serviceinfo.html template.

The mupload Renderer¶

This renderer’s parameter style is “form”.

A renderer allowing for updates to individual records using file uploads.

The difference to Uploader is that no form-redisplay will be done. All errors are reported through HTTP response codes and text strings. It is likely that this renderer will change and/or go away.

The pubreg.xml Renderer¶

This renderer’s parameter style is “clear”.

A renderer that works with registry.oaiinter to provide an OAI-PMH interface.

The core is expected to return a stanxml tree.

The qp Renderer¶

This renderer’s parameter style is “clear”.

The Query Path renderer extracts a query argument from the query path.

Basically, whatever segments are left after the path to the renderer are taken and fed into the service. The service must cooperate by setting a queryField property which is the key the parameter is assigned to.

QPRenderers cannot do forms, of course, but they can nicely share a service with the form renderer.

To adjust the results’ appreance, you can override resultline (for when there’s just one result row) and resulttable (for when there is more than one result row) templates.

The rdinfo Renderer¶

This renderer’s parameter style is “clear”.

A renderer for displaying various properties about a resource descriptor.

This renderer could really be attached to any service since it does not call it, but it usually lives on //services/overview.

By virtue of builtin vanity, you can reach the rdinfo renderer at /browse, and thus you can access /browse/foo/q to view the RD infos. This is the form used by table registrations.

In addition to all services, this renderer also links tableinfos for all non-temporary, on-disk tables defined in the RD. When you actually want to hide some internal on-disk tables, you can set a property internal on the table (the value is ignored).

The scs.xml Renderer¶

This renderer’s parameter style is “dali”.

A renderer for the Simple Cone Search protocol.

These do their error signaling in the value attribute of an INFO child of RESOURCE.

You must set the following metadata items on services using this renderer if you want to register them:

testQuery.ra, testQuery.dec – A position for which an object is present within 0.001 degrees.

The siap.xml Renderer¶

This renderer’s parameter style is “pql”.

A renderer for a the Simple Image Access Protocol.

These have errors in the content of an info element, and they support metadata queries.

For registration, services using this renderer must set the following metadata items:

sia.type – one of Cutout, Mosaic, Atlas, Pointed, see SIAP spec

You should set the following metadata items:

testQuery.pos.ra, testQuery.pos.dec – RA and Dec for a query that yields at least one image

testQuery.size.ra, testQuery.size.dec – RoI extent for a query that yields at least one image.

You can set the following metadata items (there are defaults on them that basically communicate there are no reasonable limits on them):

sia.maxQueryRegionSize.(long|lat)

sia.maxImageExtent.(long|lat)

sia.maxFileSize

sia.maxRecord (default dalHardLimit global meta)

The siap2.xml Renderer¶

This renderer’s parameter style is “dali”.

A renderer for SIAPv2.

In general, if you want a SIAP2 service, you’ll need something like the obscore view in the underlying table.

The slap.xml Renderer¶

This renderer’s parameter style is “pql”.

A renderer for the simple line access protocol SLAP.

For registration, you must set the following metadata on services using the slap.xml renderer:

There’s two mandatory metadata items for these:

slap.dataSource – one of observational/astrophysical, observational/laboratory, or theoretical
slap.testQuery – parameters that lead to a non-empty response. The way things are written in DaCHS, MAXREC=1 should in general work.

The soap Renderer¶

This renderer’s parameter style is “pql”.

A renderer for the simple line access protocol SLAP.

For registration, you must set the following metadata on services using the slap.xml renderer:

There’s two mandatory metadata items for these:

slap.dataSource – one of observational/astrophysical, observational/laboratory, or theoretical
slap.testQuery – parameters that lead to a non-empty response. The way things are written in DaCHS, MAXREC=1 should in general work.

The ssap.xml Renderer¶

This renderer’s parameter style is “pql”.

A renderer for the simple spectral access protocol.

For registration, you must set the following metadata for the ssap.xml renderer:

ssap.dataSource – survey, pointed, custom, theory, artificial

ssap.testQuery – a query string that returns some data; REQUEST=queryData is added automatically that describe the type of data served through the service. Will usually by spectrum, but timeseries is a realistic option.

Other SSA metadata includes:

ssap.creationType – archival, cutout, filtered, mosaic, projection, spectralExtraction, catalogExtraction (defaults to archival)

ssap.complianceLevel – set to “query” when you don’t deliver SDM compliant spectra; otherwise don’t say anything, DaCHS will fill in the right value.

It is recommended to set this metadata globally on the RD, as the SSA mixin can use that metadata to fill tables with sensible values without operator intervention.

Properties supported by this renderer:

datalink – if present, this must be the id of a datalink service that can work with the pubDIDs in this table (don’t use this any more, datalink is handled through table-level metadata now)

defaultRequest – by default, requests without a REQUEST parameter will be rejected. If you set defaultRequest to querydata, such requests will be processed as if REQUEST were given (which is of course sane but is a violation of the standard).

The static Renderer¶

This renderer’s parameter style is “clear”.

A renderer that just hands through files.

The standard operation here is to set a staticData property pointing to a resdir-relative directory used to serve files for. Indices for directories are created.

You can define a root resource by giving an indexFile property on the service. Note in particular that you can use an index file with an extension of shtml. This lets you use nevow templates, but since metadata will be taken from the global context, that’s probably not terribly useful. You are probably looking for the fixed renderer if you find yourself needing this.

The sync Renderer¶

This renderer’s parameter style is “dali”.

a DALI sync renderer.

In principle, this is just a shallow parser of the input parameter and renders tables as VOTables.

In practice, there are a few legacy hacks making this a bit more complicated after all.

The tableMetadata Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for a VOSI table metadata endpoint.

An endpoint with this renderer is automatically registered for every service. The responses contain information on the tables exposed by a given service.

The tableinfo Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for displaying table information.

Since tables don’t necessarily have associated services, this renderer cannot use a service to sit on. Instead, the table is being passed in as as an argument. There’s a built-in vanity tableinfo that sits on //dc_tables#show using this renderer (it could really sit anywhere else).

The tablenote Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for displaying table notes.

It takes a schema-qualified table name and a note tag in the segments.

This does not use the underlying service, so it could and will run on any service. However, you really should run it on __system__/dc_tables/show, and there’s a built-in vanity name tablenote for this.

The tables Renderer¶

This renderer’s parameter style is “clear”. This is an unchecked renderer.

A renderer for a VOSI table metadata endpoint.

An endpoint with this renderer is automatically registered for every service. The responses contain information on the tables exposed by a given service.

The upload Renderer¶

This renderer’s parameter style is “form”.

A renderer allowing for updates to individual records using file upload.

This renderer exposes a form with a file widget. It is likely that the interface will change.

The uws.xml Renderer¶

This renderer’s parameter style is “pql”.

A renderer speaking UWS.

See Custom UWSes for how to use this with your own cores.

The volatile Renderer¶

This renderer’s parameter style is “clear”.

A renderer rendering a single template with fast-changing results.

This is like the fixed renderer, except that the results are not cached.

Predefined Procedures ¶

Procedures available for rowmaker/parmaker apply¶

//epntap2#populate-2_0¶

Sets metadata for an epntap data set, including its products definition.

The values are left in vars, so you need to do manual copying, e.g., using idmaps=”*”.

In some descriptions below, you will see __replace_framed__. This means that the actual descriptions, units, and UCDs will depend on the value of spatial_frame_type in the //epntap2#table-2_0 mixin. After you have made a first (possibly severely incomplete) import of your table, you can see the actual metadata by opening http://localhost:8080/tableinfo/yourschema.epn_core.

Setup parameters for the procedure are:

Late parameter c1_resol_max: defaults to None; Resolution in the first coordinate, upper limit
Late parameter c1_resol_min: defaults to None; Resolution in the first coordinate, lower limit.
Late parameter c1max: defaults to None; __replace_framed__, upper limit
Late parameter c1min: defaults to None; __replace_framed__, lower limit.
Late parameter c2_resol_max: defaults to None; Resolution in the second coordinate, upper limit
Late parameter c2_resol_min: defaults to None; Resolution in the second coordinate, lower limit.
Late parameter c2max: defaults to None; __replace_framed__, upper limit
Late parameter c2min: defaults to None; __replace_framed__, lower limit.
Late parameter c3_resol_max: defaults to None; Resolution in the third coordinate, upper limit
Late parameter c3_resol_min: defaults to None; Resolution in the third coordinate, lower limit.
Late parameter c3max: defaults to None; __replace_framed__, upper limit
Late parameter c3min: defaults to None; __replace_framed__, lower limit.
Late parameter creation_date: defaults to None; Date of first entry of this granule
Late parameter dataproduct_type: defaults to None; The high-level organization of the data product, from a controlled vocabulary (e.g., ‘im’ for image, sp for spectrum). Multiple terms may be used, separated by # characters.
Late parameter emergence_max: defaults to None; Emergence angle during data acquisition, upper limit
Late parameter emergence_min: defaults to None; Emergence angle during data acquisition, lower limit.
Late parameter granule_gid: Common to granules of same type (e.g. same map projection, or geometry data products). Can be alphanumeric.
Late parameter granule_uid: Internal table row index, which must be unique within the table. Can be alphanumeric.
Late parameter incidence_max: defaults to None; Incidence angle (solar zenithal angle) during data acquisition, upper limit
Late parameter incidence_min: defaults to None; Incidence angle (solar zenithal angle) during data acquisition, lower limit.
Late parameter index_: defaults to \rowsMade; A numeric reference for the item. By default, this is just the row number. As this will (usually) change when new data is added, you should override it with some unique integer number specific to the data product when there is such a thing.
Late parameter instrument_host_name: Name of the observatory or spacecraft that the observation originated from; for ground-based data, use IAU observatory codes, http://www.minorplanetcenter.net/iau/lists/ObsCodesF.html, for space-borne instruments use http://nssdc.gsfc.nasa.gov/nmc/
Late parameter instrument_name: defaults to None; Service providers are invited to include multiple values for instrument_name, e.g., complete name + usual acronym. This will allow queries on either ‘VISIBLE AND INFRARED THERMAL IMAGING SPECTROMETER’ or VIRTIS to produce the same reply.
Late parameter measurement_type: defaults to None; UCD(s) defining the data, with multiple entries separated by hash (#) characters.
Late parameter modification_date: defaults to None; Date of last modification (used to handle mirroring)
Late parameter obs_id: Associates granules derived from the same data (e.g. various representations/processing levels). Can be alphanumeric, may be the ID of original observation.
Late parameter phase_max: defaults to None; Phase angle during data acquisition, upper limit
Late parameter phase_min: defaults to None; Phase angle during data acquisition, lower limit.
Late parameter processing_level: defaults to None; CODMAC calibration level; see the et_cal note http://dc.g-vo.org/tableinfo/titan.epn_core#note-et_cal for what values are defined here
Late parameter release_date: defaults to None; Start of public access period
Late parameter s_region: defaults to None; A spatial footprint of a dataset located on a spherical coordinate system. Currently, this is fixed to be a spherical polygon (fill it with something like pgsphere.SPoly.fromDALI([@long1, @lat1, @long2, @lat2,…], all coordinates in degrees).
Late parameter service_title: defaults to None; Title of resource (an acronym really, will be used to handle multiservice results)
Late parameter spectral_range_max: defaults to None; Spectral range (frequency), upper limit
Late parameter spectral_range_min: defaults to None; Spectral range (frequency), lower limit.
Late parameter spectral_resolution_max: defaults to None; Spectral resolution, upper limit
Late parameter spectral_resolution_min: defaults to None; Spectral resolution, lower limit.
Late parameter spectral_sampling_step_max: defaults to None; Spectral sampling step, upper limit
Late parameter spectral_sampling_step_min: defaults to None; Spectral sampling step, lower limit.
Late parameter target_class: defaults to "UNKNOWN"; The type of the target; choose from asteroid, dwarf_planet, planet, satellite, comet, exoplanet, interplanetary_medium, ring, sample, sky, spacecraft, spacejunk, star
Late parameter target_name: Name of the target object, preferably according to the official IAU nomenclature. As appropriate, take these from the exoplanet encyclopedia http://exoplanet.eu, the meteor catalog at http://www.lpi.usra.edu/meteor/, the catalog of stardust samples at http://curator.jsc.nasa.gov/stardust/catalog/
Late parameter target_region: defaults to None; This is a complement to the target name to identify a substructure of the target that was being observed (e.g., Atmosphere, Surface). Preferably take terms from the UAT http://www.ivoa.net/rdf/uat.
Late parameter time_exp_max: defaults to None; Integration time of the measurement, upper limit
Late parameter time_exp_min: defaults to None; Integration time of the measurement, lower limit.
Late parameter time_max: defaults to None; Acquisition stop time (in JD), as UTC at time_refposition
Late parameter time_min: defaults to None; Acquisition start time (in JD), as UTC at time_refposition
Late parameter time_refposition: defaults to "TOPOCENTER"; Indicates where the time is measured, can be a planet or a spacecraft.
Late parameter time_sampling_step_max: defaults to None; Sampling time for measurements of dynamical phenomena, upper limit
Late parameter time_sampling_step_min: defaults to None; Sampling time for measurements of dynamical phenomena, lower limit.
Late parameter time_scale: defaults to "UTC"; Time scale used for the various times, as given by IVOA’s STC data model. Choose from TT, TDB, TOG, TOB, TAI, UTC, GPS, UNKNOWN

//epntap2#populate-localfile-2_0¶

Use this apply when you use the //epntap2#localfile-2_0 mixin. This will only (properly) work when you use a //products#define rowfilter; if you have that, this will work without further configuration.

Setup parameters for the procedure are:

Late parameter creation_date: defaults to \sourceCDate; A timestamp giving the dataset’s creation time as a datetime object

//linetap#populate-0¶

Fills the columns of a LineTAP table, typically created using the //linetap#table-0 mixin. The values are left in vars, so you need to copy them into the finished record, probably through idmaps="*".

Setup parameters for the procedure are:

Late parameter einstein_a: defaults to None; Einstein A coefficient of the radiative transition.
Late parameter element: defaults to None; Element name for atomic transitions, NULL otherwise.
Late parameter inchi: International Chemical Identifier InChI.
Late parameter inchikey: The InChi key (hash) generated from inchi.
Late parameter ion_charge: Total charge (ionisation level) of the emitting particle.
Late parameter line_reference: defaults to None; Reference to the source of the line data; this could be a bibcode, a DOI, or a plain URI.
Late parameter lower_energy: defaults to None; Energy of the lower state (in J)
Late parameter mass_number: defaults to None; Number of nucleons in the atom or molecule
Late parameter method: defaults to None; Method the wavelength was obtained with (XSAMS controlled vocabulary)
Late parameter title: Human-readable line designation.
Late parameter upper_energy: defaults to None; Energy of the upper state (in J)
Late parameter vacuum_wavelength: Vacuum wavelength of the transition (in Angstrom)
Late parameter vacuum_wavelength_error: defaults to None; Total error in vacuum_wavelength (in Angstrom)
Late parameter xsams_uri: defaults to None; A URI for a full XSAMS description of this line.

//procs#dictMap¶

Maps input values through a dictionary.

The dictionary is given in its python form here. This apply only operates on the rawdict, i.e., the value in vars is changed, while nothing is changed in the rowdict.

Setup parameters for the procedure are:

Parameter default: defaults to KeyError; Default value for missing keys. Leave this at KeyError to raise an error. Set it to base.NotGiven to make DaCHS retain unmapped values when they don’t appear in the mapping.
Parameter key: Name of the input key to map
Parameter mapping: Python dictionary literal giving the mapping

//procs#fullQuery¶

runs a free query against the data base and enters the first result record into vars.

locals() will be passed as data, so you can define more bindings and refer to their keys in the query.

Setup parameters for the procedure are:

Parameter errCol: defaults to '<unknown>'; a column name to use when raising a ValidationError on failure.
Parameter query: an SQL query

//procs#mapValue¶

is an apply proc that translates values via a utils.NameMap

Destination may of course be the source field (though that messes up idempotency of macro expansion, which shouldn’t usually hurt).

The format of the mapping file is:

<target key><tab><source keys>

where source keys is a whitespace-seperated list of values that should be mapped to target key (sorry the sequence’s a bit unusual).

A source key must be encoded quoted-printable. This usually doesn’t matter except when it contains whitespace (a blank becomes =20) or equal signs (which become =3D).

Here’s an example application for a filter that’s supposed to translate some botched object names:

<apply name="cleanObject" procDef="//procs#mapValue">
  <bind name="destination">"cleanedObject"</bind>
  <bind name="failuresMapThrough">True</bind>
  <bind name="value">@preObject</bind>
  <bind name="sourceName">"flashheros/res/namefixes.txt"</bind>
</apply>

The input could look like this, with a Tab char written as ” <TAB> ” for clarity:

alp Cyg <TAB> aCyg alphaCyg
Nova Cygni 1992 <TAB> Nova=20Cygni=20'92 Nova=20Cygni

Setup parameters for the procedure are:

Parameter destination: name of the field the mapped value should be written into
Parameter failuresAreNone: defaults to False; Rather than raise an error, yield NULL for values not in the mapping
Parameter failuresMapThrough: defaults to False; Rather than raise an error, yield the input value if it is not in the mapping (this is for ‘fix some’-like functions and only works when failureAreNone is False)
Parameter logFailures: defaults to False; Log non-resolved names?
Parameter sourceName: An inputsDir-relative path to the NameMap source file.
Late parameter value: The value to be mapped.

//procs#resolveObject¶

Resolve identifiers to simbad positions.

It caches query results (positive as well as negative ones) in cacheDir. To avoid flooding simbad with repetetive requests, it raises an error if this directory is not writable.

It leaves J2000.0 positions as floats in the simbadAlpha and simbadDelta variables.

Setup parameters for the procedure are:

Late parameter identifier: The identifier to be resolved.
Parameter ignoreUnknowns: defaults to True; Return Nones for unknown objects? (if false, ValidationErrors will be raised)
Parameter logUnknowns: defaults to False; Write unresolved object names to the info log

//procs#simpleSelect¶

Fill variables from a simple database query.

The idea is to obtain a set of values from the data base into some columns within vars (i.e., available for mapping) based on comparing a single input value against a database column. The query should always return exactly one row. If more rows are returned, the first one will be used (which makes the whole thing a bit of a gamble), if none are returned, a ValidationError is raised.

Setup parameters for the procedure are:

Parameter assignments

mapping from database column names to vars column names, in the format {<db colname>:<vars name>}”

Parameter column

the column to compare the input value against

Parameter errCol

defaults to '<unknown>';: UNDOCUMENTED

Parameter table

name of the database table to query

Late parameter val

UNDOCUMENTED

//siap2#computePGS¶

Computes the spatial coverage of an image based on WCS keys it looks for in the rawdict.

The minimum is CRVAL1, CRVAL2, CRPIX1, CRPIX2, CRVAL1, CRVAL2, CUNIT1, CUNIT2, NAXISn, and CDi_j or CDELTn. Interpretation of more WCS may or may not happen. You can override the indexes of the spatial axes using naxis, which will of course change the parameter names, too.

Records without or with insufficient wcs keys are furnished with all-NULL spatial columns if the missingIsError setup parameter is False, else they bomb out with a DataError (the default).

This writes directly into rowdict; do not use idmaps="*" on rowmakers with computePGS.

Added in 2.7.3.

Setup parameters for the procedure are:

Parameter missingIsError: defaults to True; Throw an exception when no WCS information can be located.
Parameter naxis: defaults to "1,2"; Comma-separated list of integer axis indices (1=first) to be considered for WCS

//siap2#getBandFromFilter¶

sets the bandpassId, bandpassUnit, bandpassRefval, bandpassHi, and bandpassLo from a set of standard band Ids.

The bandpass ids known are contained in a file supplied by DaCHS that you should consult for supported values by running dachs admin dumpDF data/filters.txt.

If you pass in an unknown filter name, no keys will be generated, but no diagnostics will be emitted either. Make sure to dachs info on the imported table if you expect no NULLs in the bandpass columns.

All values filled in here are in meters.

If this is used, it must run after //siap#setMeta since setMeta clobbers our result fields.

Added in 2.7.3.

Setup parameters for the procedure are:

Parameter sourceCol: defaults to None; Name of the column containing the filter name; leave at default None to take the band from result[‘bandpassId’], where such information would be left by siap#setMeta.

//siap2#setMeta¶

Fills non-spatial information in an obscore record for an image.

If you define the bandpasses yourself, do not change bandpassUnit and give all values in metres.

For optical images, we recommend to fill out bandpassId and then let the //siap2#getBandFromFilter apply compute the actual limits.

Do not use idmaps="*" when using this procDef; it writes directly into result, and you would be clobbering what it does.

The proc parameters use the obscore names whereever possible, but accept most of the names of the version 1 //siap#setMeta and the //obscore#publishSIAP mixins. easy migration.

Added in 2.7.3.

Setup parameters for the procedure are:

Late parameter bandpassId: defaults to None; a rough indicator of the bandpass, like Johnson bands. Only give if you want //siap2#getBandFromFilter to infer em_min and em_max
Late parameter calib_level: defaults to 2; Calibration level of the image described here (see obscore docs for details). Images having WCS will generally have a 2 or 3 (if heavily resampled or stacked) here.
Late parameter dataproduct_subtype: defaults to None; Closer specification of the sort of data this row is for. It is almost always a good idea to leave this at None
Parameter dataproduct_type: defaults to "image"; Type of data in this table. Must be a constant string taken from https://www.ivoa.net/rdf/product-type/
Late parameter dateObs: defaults to None; The midpoint of the observation; this can either be a datetime instance, or a float>1e6 (a julian date) or something else (which is then interpreted as an MJD). This is ignored if you give t_min and t_max
Late parameter em_max: defaults to None; Upper value of the wavelength covered, in meters (you usually want to use //siap2#getBandFromFilter to fill this from bandpassId).
Late parameter em_min: defaults to None; Lower value of the wavelength covered, in meters (you usually want to use //siap2#getBandFromFilter to fill this from bandpassId).
Late parameter em_res_power: defaults to None; Spectral resolving power λ/Δλ. Unless you have a spectral cube, leave this at None.
Parameter em_ucd: defaults to None; Nature of the spectral axis in this data collection (when you have a spectral cube, probably one of em.freq, em.wl, or em.energy)
Late parameter em_xel: defaults to 1; Number of spectral pixels in this image.
Late parameter facility_name: defaults to base.getMetaText(targetTable, "facility", default=None); Name of the facility at which data was taken.
Late parameter instrument_name: defaults to base.getMetaText(targetTable, "instrument", default=None); Name of the instrument that produced that data.
Parameter o_ucd: defaults to "phot.count"; UCD for the observable in this data collection
Late parameter obs_collection: A shorthand for the data collection this image belongs to. Make it really terse. Freetext otherwise
Late parameter obs_creator_did: defaults to None; Identifier assigned by the creator. This is rarely used.
Late parameter obs_id: defaults to None; An identifier for an observation if that has yielded multiple rows. Keep this None otherwise
Late parameter obs_publisher_did: defaults to \standardPubDID; Dataset identifier assigned by you. If you want to re-publish this dataset in obscore, make this non-NULL.
Late parameter obs_title: defaults to None; This should, in as few characters as possible, convey some idea what the image will show (e.g., instrument, object, bandpass)
Late parameter pol_states: defaults to None; Polarisation states in this image. See Obscore if you have polarimetric observations.
Late parameter pol_xel: defaults to None; Number of polarisation pixels in this image.
Late parameter t_exptime: defaults to None; Total time in seconds the detector was collecting photons.
Late parameter t_max: defaults to None; MJD the last photons on this image were received. If you leave this as None, DaCHS will try to fill it up from dateObs and t_exptime.
Late parameter t_min: defaults to None; MJD the first photons on this image were received. If you leave this as None, DaCHS will try to fill it up from dateObs and t_exptime.
Late parameter t_resolution: defaults to None; Minimal significant resolution on a time axis. Unless you have a space-time cube, leave at None.
Late parameter t_xel: defaults to 1; Number of time pixels in this image (make this None for stacked exposures.
Late parameter target_class: defaults to None; Class of the target of a targeted observation. Choose from http://www.ivoa.net/rdf/object-type (or have the vocabulary extended if your object type is missing)
Late parameter target_name: defaults to None; Target of a targeted observation. Choose something Simbad can resolve if you can.

//slap#fillBasic¶

This apply is intended for rowmakers filling tables mixing in //slap#basic. It populates vars for all the columns in there; you’ll normally want idmaps=”*” with this apply.

For most of its parameters, it will take them for same-named vars, so you can slowly build up its arguments through var elements.

Setup parameters for the procedure are:

Late parameter chemical_element: defaults to @chemical_element; Element that makes the transition. It’s probably ok to dump molecule names in here, too.
Late parameter final_level_energy: defaults to @final_level_energy; Energy of the final state
Late parameter final_name: defaults to @final_name; Designation of the final state
Late parameter id_status: defaults to "identified"; Identification status; this would be identified or unidentified plus possibly uncorrected (but read the SLAP spec for that).
Late parameter initial_level_energy: defaults to @initial_level_energy; Energy of the initial state
Late parameter initial_name: defaults to @initial_name; Designation of the initial state
Late parameter linename: defaults to @linename; A brief designation for the line, like ‘H alpha’ or ‘N III 992.973 A’.
Late parameter pub: defaults to @pub; Publication this came from (use a bibcode).
Late parameter wavelength: defaults to @wavelength; Wavelength of the transition in meters; this will typically be an expression like int(@wavelength)*1e-10

//ssap#feedSSAToSDM¶

feedSSAToSDM takes the current rowIterator’s sourceToken and feeds it to the params of the current target. sourceTokens must be an SSA rowdict (as provided by the sdmCore). Further, it takes the params from the sourceTable argument and feeds them to the params, too.

All this probably only makes sense in parmakers when making tables mixing in //ssap#sdm-instance in data children of sdmCores.

//ssap#fill-plainlocation¶

This mixin fills the columns added by the plainlocation mixin with values generated from ra, dec, and aperture.

Setup parameters for the procedure are:

Late parameter aperture: defaults to None; Size of the aperture in degrees; if you leave this at None, no ssa_region will be generated.
Late parameter dec: ICRS Dec of aperture center
Late parameter ra: ICRS RA of aperture center

//ssap#setMeta¶

Sets metadata for an SSA data collection, including its products definition.

Since this is only useful with the deprecated hcd and mixc mixins, this should no longer be used.

The values are left in vars, so you need to do manual copying, e.g., using idmaps=”*”, or, if you need to be more specific, idmaps=”ssa_*”.

Setup parameters for the procedure are:

Late parameter alpha: defaults to None; right ascension of target (ICRS degrees); ssa:Char.SpatialAxis.Coverage.Location.Value.C1
Late parameter aperture: defaults to None; angular diameter of aperture (expected in degrees); ssa:Char.SpatialAxis.Coverage.Bounds.Extent
Late parameter bandpass: defaults to None; bandpass (i.e., rough spectral location) of this dataset; ssa:DataID.Bandpass
Late parameter cdate: defaults to None; date the file was created (or processed; optional); this must be either a string in ISO format, or you need to parse to a timestamp yourself; ssa:DataID.Date
Late parameter creatorDID: defaults to None; id given by the creator (leave out if not applicable); ssa:DataID.CreatorDID
Late parameter cversion: defaults to None; creator assigned version for this file (should be incremented when it is changed); ssa:DataID.Version
Late parameter dateObs: defaults to None; observation midpoint (you can give a datetime, a string in iso format, a jd, or an mjd, the latter two being told apart by comparing against 1e6)
Late parameter delta: defaults to None; declination of target (ICRS degrees); ssa:Char.SpatialAxis.Coverage.Location.Value.C2
Late parameter dstitle: a title for the data set (e.g., instrument, filter, target in some short form; must be filled in); ssa:DataID.Title
Late parameter length: defaults to None; Number of samples in the spectrum; ssa:Dataset.Length
Late parameter pdate: defaults to datetime.datetime.utcnow(); date the file was last published (in general, the default is fine); ssa:Curation.Date
Late parameter pubDID: Id provided by the publisher (i.e., you); this is an opaque string and must be given; ssa:Curation.PublisherDID
Late parameter redshift: defaults to None; source redshift; ssa:Target.Redshift
Late parameter snr: defaults to None; signal-to-noise ratio estimated for this dataset; ssa:Derived.SNR
Late parameter specend: defaults to None; upper bound of wavelength interval (in meters); ssa:Char.SpectralAxis.Coverage.Bounds.Stop
Late parameter specext: defaults to None; (ignored; only present for compatibility, computed from specstart and specend)
Late parameter specmid: defaults to None; (ignored; only present for compatibility, computed from specstart and specend)
Late parameter specstart: defaults to None; lower bound of wavelength interval (in meters); ssa:Char.SpectralAxis.Coverage.Bounds.Start
Late parameter targclass: defaults to None; object class (star, QSO,…); ssa:Target.Class
Late parameter targname: defaults to None; common name of the object observed; ssa:Target.Name
Late parameter timeExt: defaults to None; exposure time (in seconds); ssa:Char.TimeAxis.Coverage.Bounds.Extent

//ssap#setMixcMeta¶

Sets metadata for an SSA data set from mixed sources. This will only work sensibly in cooperation with setMeta

Since //ssap#mixc is deprecated, there is no reason to use this in new RDs.

As with setMeta, the values are left in vars; if you did as recommended with setMeta, you’ll have this covered as well.

Setup parameters for the procedure are:

Late parameter binSize: defaults to None; Bin size on the spectral axis in m
Late parameter collection: defaults to None; IOVA id of the originating data collection (leave empty if you don’t know what this is about)
Late parameter creationType: defaults to None; Process used to produce the data (zero or more of archival, cutout, filtered, mosaic, projection, spectralExtraction, catalogExtraction, concatenated by commas); ssa:DataID.CreationType
Late parameter creator: defaults to "Take from RD"; Creator/Author
Late parameter datasource: defaults to None; Generation type (typically, one survey, pointed, theory, custom, artificial); ssa:DataID.DataSource
Late parameter dstype: defaults to "spectrum"; Type of data. The only defined value currently is Spectrum, but you may get away with TimeSeries; ssa:Dataset.Type
Late parameter fluxCalib: defaults to None; Type of flux calibration (one of ABSOLUTE, RELATIVE, NORMALIZED, or UNCALIBRATED); ssa:Char.FluxAxis.Calibration
Late parameter fluxStatError: defaults to None; Statistical error for flux in units of fluxUnit
Late parameter fluxSysError: defaults to None; Systematic error for flux in units of fluxUnit
Late parameter instrument: defaults to "Take from RD"; Instrument or code used to produce this dataset; ssa:DataID.Instrument
Late parameter publisher: defaults to "Take from RD"; Publisher IVO; ssa:Curation.Publisher
Late parameter reference: defaults to "Take from RD"; URL or bibcode of a publication describing this data.
Late parameter specCalib: defaults to None; Type of wavelength Calibration (one of ABSOLUTE, RELATIVE, NORMALIZED, or UNCALIBRATED); ssa:Char.SpectralAxis.Calibration
Late parameter specres: defaults to None; Resolution on the spectral axis; you must give this as FWHM wavelength in meters here. This will default to binSize if not given; ssa:Char.SpectralAxis.Resolution
Late parameter spectStatError: defaults to None; Statistical error for the spectral coordinate in m
Late parameter spectSysError: defaults to None; Systematic error for the spectral coordinate in m

Procedures available for grammar rowfilters¶

//procs#expandComma¶

A row generator that reads comma separated values from a field and returns one row with a new field for each of them.

Setup parameters for the procedure are:

Parameter destField: Name of the column the individual columns are written to
Parameter srcField: Name of the column containing the full string

//procs#expandDates¶

is a row generator to expand time ranges.

The finished dates are left in destination as datetime.datetime instances

Setup parameters for the procedure are:

Parameter dest: defaults to 'curTime'; name of the column the time should appear in
Parameter end: the end date(time)
Late parameter hrInterval: defaults to 24; difference between generated timestamps in hours
Parameter start: the start date(time), as either a datetime object or a column ref

//procs#expandIntegers¶

A row processor that produces copies of rows based on integer indices.

The idea is that sometimes rows have specifications like “Star 10 through Star 100”. These are a pain if untreated. A RowExpander could create 90 individual rows from this.

Setup parameters for the procedure are:

Parameter endName: column containing the end value
Parameter indName: name the counter should appear under
Parameter startName: column containing the start value

//products#define¶

Enters the values defined by the product interface into a grammar’s result.

See the documentation on the //products#table mixin. In short: you will always have to touch table (to the name of the table this row is managed in).

If you don’t serve FITS images, you will also have to set mime. Use a media type like “image/jpeg” or “text/csv” here as appropriate. If not set, this defaults to “image/fits” (which is, we claim, suitable for cubes and certain spectra, too); for FITS binary tables, use application/fits.

Everything else is optional: You may want to set preview and preview_mime if DaCHS can’t do previews of your stuff automatically. What’s left is for special situations.

This will create the keys prodblAccref, prodtblOwner, prodtblEmbargo, prodtblPath, prodtblFsize, prodtblTable, prodtblMime, prodtblPreview, prodtbleMime, and prodtblDatalink keys in rawdict – you can refer to them in the usual @foo way, which is sometimes useful even outside products processing proper (in particular for prodtblAccref).

Setup parameters for the procedure are:

Late parameter accref: defaults to \inputRelativePath{False}; an access reference (this usually is the input-relative path; only file names well-behaved in URLs are accepted here by default for easier operation with ObsTAP)
Late parameter datalink: defaults to None; Legacy. Do not use.
Late parameter embargo: defaults to None; for proprietary data, the date the file will become public
Late parameter fsize: defaults to \inputSize; the size of the input in bytes
Late parameter mime: defaults to 'image/fits'; MIME-type for the product
Late parameter owner: defaults to None; for proprietary data, the owner as a gavo creds-created user
Late parameter path: defaults to \inputRelativePath{True}; the inputs-relative path to the product file (change at your peril)
Late parameter preview: defaults to 'AUTO'; file path to a preview. Set to None if no preview is available. The default ‘AUTO’ will make DaCHS try to make a preview itself, which currently fails for anything but FITS and jpeg. You can also put in a URL if you have pre-computed. If using a local path, see Precomputed Previews before writing something yourself.
Late parameter preview_mime: defaults to None; MIME-type for the preview (if there is one).
Parameter table: the table this product is managed in. You must fill this in, and don’t forget the quotes.

Procedures available for datalink cores¶

//datalink#fromtable¶

A descriptor generator simply pulling a row from a database table. This row is made available as the .metadata attribute. You also must give a field that the generator will pull a URL from; the generator arranges things so that the default dlget will simply redirect there.

Note that the #this and #preview links DaCHS normally makes for descriptors from products are not added automatically here. Try to at least provide #this.

Setup parameters for the procedure are:

Late parameter didPrefix: defaults to base.NotGiven; A URI to complete an identifier with with if it looks like a non-URI. Leave empty to not do pubDID completion.
Late parameter idColumn: The name of the table column to match the ID against (as a python value).
Late parameter tableName: The name of the table to pull the metadata row from (as a python value).

//soda#fits_doWCSCutout¶

A fairly generic FITS cutout function.

It expects some special attributes in the descriptor to allow it to decode the arguments. These must be left behind by the metaMaker(s) creating the parameters.

This is axisNames, a dictionary mapping parameter names to the FITS axis numbers or the special names WCSLAT or WCSLONG. It also expects a skyWCS attribute, a wcs.WCS instance for spatial cutouts.

Finally, descriptor must have a list attribute slices, containing zero or more tuples of (fits axis, lowerPixel, upperPixel); this allows things like BAND to add their slices obtained from parameters in standard units.

The .data attribute must be a pyfits hduList, as generated by the fits_makeHDUList data function.

//soda#fits_formatHDUs¶

Formats pyfits HDUs into a FITS file.

This all works in memory, so for large FITS files you’d want something more streamlined.

//soda#fits_genDesc¶

A data function for SODA returning the a fits descriptor.

This has, in addition to the standard stuff, a hdr attribute containing the primary header as pyfits structure.

The functionality of this is in its setup, getFITSDescriptor. The intention is that customized DGs (e.g., fixing the header) can use this as an original.

Setup parameters for the procedure are:

Parameter accrefPrefix: defaults to None; A prefix for the accrefs the parent SODA service works on. Calls on all other accrefs will be rejected with a 403 forbidden. You should always include a restriction like this when you make assumptions about the FITSes (e.g., what axes are available).
Parameter contentQualifier: defaults to "#image"; An identifier for the sort of data #this is. Choose from http://www.ivoa.net/rdf/product-type in case the default does not quite work for you (version 2.8+).
Parameter descClass: defaults to FITSProductDescriptor; The descriptor class to use. The default is fine for vanilla FITS files, but when you deliver datalinks through the product table, you’ll have to use DLFITSDescriptor here. Also, you can define a descriptor yourself in the setup (inherit from FITSDescriptor).
Parameter qnd: defaults to True; Pass 0 or False here to not use DaCHS fast header reader here. This is necessary to properly handle compressed FITS images – but it entails the risk that astropy magic will mogrify the header, and it may be dramatically slower in some circumstances.

//soda#fits_makeBANDMeta¶

Yields standard BAND params.

DaCHS should in general be smart enough to convert between common spectral units (like MHz, or keV, or whatever) and the meter that SODA BAND expects. If your files do to give VOUnits-parseable units on the spectral axis’ CUNIT header, use the wavelengthOverride param.

This adds specToMeter, meterToSpec, and spectralAxis attributes to the descriptor for later use by fits_makeBANDSlice.

Setup parameters for the procedure are:

Parameter fitsAxis: defaults to 3; FITS axis index (1-based) of the spectral axis
Parameter wavelengthOverride: defaults to None; Override for the FITS unit given for the wavelength (for when it is botched or missing; leave at None for taking it from the header); this is a python literal.

//soda#fits_makeBANDSlice¶

Computes a cutout for the parameters added by makeBANDMeta.

This must sit in front of doWCSCutout.

This also reuses internal state added by makeBANDMeta, so this really only makes sense together with it.

//soda#fits_makeHDUList¶

An initial data function to construct a pyfits hduList and make that into a descriptor’s data attribute.

This wants a descriptor as returned by fits_genDesc.

There’s a hack here: this sets a dataIsPristine boolean on descriptor that’s made false when one of the fits manipulators change something. If that’s true by the time the formatter sees it, it will just push out the entire file. So, if you use this and insert your own data functions, make sure you set dataIsPristine accordingly.

Setup parameters for the procedure are:

Parameter crop: defaults to True; Cut away everything but the primary HDU? (This is unconditionally suppresed for compressed FITSes and when operations are on a non-primary extension).

//soda#fits_makeWCSParams¶

A metaMaker that generates parameters allowing cutouts along the various WCS axes in physical coordinates.

This uses astropy.wcs for the spatial coordinates and tries to figure out what these are with some heuristics. For the remaining coordinates, it can set up separate, manual transformations. However, since in general, they will correspond to standard SODA parameters (like BAND; there are other streams of that in this RD), you have to explictly request that by mentioning the fits axis index in axisMetaOverrides, minimally like this:

<bind key="axisMetaOverrides">{3: {}}</bind>

This will generate a parameter from the FITS metadata of axis 3. You can override param attributes in the dictionary that is the value of the axis key.

The metaMaker leaves an axisNames mapping in the descriptor. This is important for the fits_doWCSCutout, and replacement metaMakers must do the same.

The meta maker also creates a skyWCS attribute in the descriptor if successful, containing the spatial transformation only. All other transformations, if present, are in miscWCS, by a dict mapping axis labels to the fitstools.WCS1Trans instances.

Setup parameters for the procedure are:

Parameter axisMetaOverrides: defaults to {}; A python dictionary mapping fits axis indices (1-based) to dictionaries of inputKey constructor arguments; for spatial axes, use the axis name instead of the axis index.
Parameter stcs: defaults to None; A QSTC expression describing the STC structure of the parameters. This is currently ignored and will almost certainly look totally different when STC2 finally comes around. Meanwhile, don’t bother.

//soda#fromStandardPubDID¶

A descriptor generator for SODA that builds a ProductDescriptor for PubDIDs that have been built by getStandardsPubDID (i.e., the path part of the IVOID is a tilde, with the products table accref as the query part).

Setup parameters for the procedure are:

Parameter accrefPrefix: defaults to None; A prefix for the accrefs the parent SODA service works on. Calls on all other accrefs will be rejected with a 403 forbidden. You should always include a restriction like this when you make assumptions about the FITSes (e.g., what axes are available).
Parameter contentQualifier: defaults to None; An identifier for the sort of data #this is. Choose from http://www.ivoa.net/rdf/product-type if you want to give clients a hint what other client might be suitable to work with that link (version 2.8+).

//soda#generateProduct¶

A data function for SODA that returns a product instance. You can restrict the mime type of the product requested so the following filters have a good idea what to expect.

Setup parameters for the procedure are:

Parameter requireMimes: defaults to frozenset(); A set or sequence of mime type strings; when given, the data generator will bail out with ValidationError if the product mime is not among the mimes given.

//soda#sdm_genData¶

A data function for SODA returning a spectral data model compliant table that later data functions can then work on. As usual for generators, it uses the implicit PUBDID argument.

Setup parameters for the procedure are:

Parameter builder: Full reference (like path/rdname#id) to a data element building the SDM instance table as its primary table.

//soda#sdm_genDesc¶

A data function for SODA returning the product row corresponding to a PubDID within an SSA table.

The descriptors generated have an ssaRow attribute containing the original row in the SSA table.

Setup parameters for the procedure are:

Parameter contentQualifier: defaults to "#spectrum"; The sort of data that #this is; you might want to override this if you are abusing this for time series or so (version 2.8+).
Late parameter descriptorClass: defaults to ssap.SSADescriptor; The SSA descriptor class to use. You’ll need to override this if the dc.products path doesn’t actually lead to the file (see custom generators). This class must have an fromSSAResult constructor.
Parameter ssaTD: Full reference (like path/rdname#id) to the SSA table the spectrum’s PubDID can be found in.
Parameter useAccref: defaults to True; By default, this genDesc will parse the pubDID passed in and match with the accref, which in DaCHS typically is indexed. This don’t work if you have exotic pubDID schemes; in that case, set this to False

//soda#trivialFormatter¶

The tivial formatter for SODA processed data – it just returns descriptor.data, which will only work it it works as a nevow resource.

If you do not give any dataFormatter yourself in a SODA core, this is what will be used.

Predefined Streams ¶

Streams are recorded RD elements that can be replayed into resource descriptors using the FEED active tag. They do, however, support macro expansion; if macros are expanded, you need to given them values in the FEED element (as attributes). What attributes are required should be mentioned in the following descriptions for those predefined streams within DaCHS that are intended for developer consumption.

Datalink-related Streams¶

//soda#sdm_plainfluxcalib¶

A stream inserting a data function and its metadata generator to do select flux calibrations in SDM data. This expects sdm_generate (or at least parameters.data as an SDM data instance) as the generating function within the SODA core.

Clients can select “RELATIVE” as FLUXCALIB, which does a normalization to max(flux)=1 here. Everything else is rejected right now.

This probably is more an example of how to write such a thing then genuinely useful.

//soda#sdm_cutout¶

A stream inserting a data function and its metaMaker to do cutouts in SDM data. This expects sdm_generate (or at least parameters.data as an SDM data instance) as the generating function within the SODA core.

The cutout limits are always given in meters, regardless of the spectrum’s actual units (as in SSAP’s BAND parameter).

//soda#sdm_format¶

A formatter for SDM data, together with its input key for FORMAT.

//soda#fits_genKindPar¶

This stream should be included in FITS-handling SODA services; it adds parameter and code to just retrieve the FITS header to the core.

For this to work as expected, it must be immediately before the formatter.

//soda#fits_genPixelPar¶

This stream should be included in FITS-handling SODA services; it add parameters and code to perform cut-outs along pixel coordinates.

//soda#fits_standardDLFuncs¶

Pulls in all “standard” SODA functions for FITSes, including cutouts and header retrieval.

You can give an stcs attribute (for fits_makeWCSParams); for this doesn’t make sense because STCS cannot express the SODA parameter structure.

For spectral cubes, you can give a spectralAxis attribute here containing the fits axis index (1..n) of the spectral axis. If you don’t, no BAND cutout will be generated. If you do, you may want to fix wavelengthOverride (default is to take what the FITS says).

For velocity cubes, there is, analogously, velocityAxis and and velocityOverride. This will generate a (non-SODA) VELOCITY parameter (using m/s).

For axisMetaOverrides, see //soda#fits_makeWCSParams (except it’s an attribute here rather than a proc param).

To work, this needs a descriptor generator; you probably want //soda#fits_genDesc here.

Defaults for macros used in this stream:

axisMetaOverrides: ‘{}’
spectralAxis: ‘0’
stcs: ‘’
velocityAxis: ‘0’
velocityOverride: ‘None’
wavelengthOverride: ‘None’

//soda#fits_standardBANDCutout¶

Adds metadata and data function for one axis containing wavelengths.

(this could be extended to cover frequency and energy axes, I guess)

To use this, give the fits axis containing the spectral coordinate in the spectralAxis attribute; if needed, you can override the unit in wavelengthUnit (if the unit in the header is somehow bad or missing; don’t use quotes here).

This must be included physically before fits_doWCSCutout. Otherwise, no cutout will be performed.

Defaults for macros used in this stream:

spectralAxis: ‘0’
wavelengthOverride: ‘’

Streams for building conditions on dbCores¶

//procs#rangeCond¶

A condDesc that expresses a range and has an InputKey each for min and max.

Specify the following macros when replaying:

name – the column name in the core’s queried table
groupdesc – a terse phrase describing the range. This will be used in the description of both the input keys and the group
grouplabel – a label (include the unit, it is not taken from InputKey) written in front of the form group

groupdesc has to work after “Range of”, “Lower bound of”, and “Upper bound of”. Do not include a concluding period.

//procs#negatableBoolean¶

A condDesc over a boolean column.

By default, DaCHS does not distinguish between False and NULL when auto-generating input keys from boolean columns. That is, you can check a checkbox and will get everything where the column is true. If you do not check it, however, you will get everything, i.e., rows in which the column is any of NULL, True, or False. That’s pretty much the semantics of HTML checkboxes.

However, when you want people to be able to explicitly say “this should be off“, you need to be a bit more cunning. That is what this stream does; when you have a boolean column my_bool, you can have:

<FEED source="//procs#negatableBoolean" column="my_bool"/>

next to your other condDescs in the dbCore. This will then produce a three-way selection between Yes, No, and ANY.

//ssap#hcd_condDescs¶

This stream defines the condDescs for an SSA service based on one of the mixins defined here.

//scs#coreDescs¶

This stream inserts three condDescs for SCS services on tables with pos.eq.(ra|dec).main columns; one producing the standard SCS RA, DEC, and SR parameters, another creating input fields for human consumption, and finally MAXREC.

//pql#DALIPars¶

This stream includes the standard DALI service parameters (RESPONSEFORMAT, MAXREC, VERB). For services available through IVOA-standard protocols (renderers scs.xml, siap.xml, ssap.xml, and also api), this is included automatically, so you will not usually have to manually FEED this.

//pql#DALIUpload¶

This stream includes a DALI UPLOAD parameter. This is purely declarative for now, as the interpretation is done anyway for VO renderers and not at all otherwise.

Other Streams¶

//procs#declare-indexes-from¶

Tells DaCHS to copy the index declarations from tables a view derives from (meaning; feeding this probably does not make any sense outside of a view declaration).

Since DaCHS does not have a good way to guess which tables you derive from (in the end), you have to give blank-separated RD#id references to them in the sourceTables attribute.

The proc will declare all indexes over columns that appear to be in the View (by name). That’s a rough heuristics which works in most cases. If it doesn’t work for you, you can always declare the indexes manually.

Note that this will only see columns that are declared lexically before the FEED; you will thus typically want to include this at the end of the view definition.

Example:

<FEED source="//procs#declare-indexes-from"
  sourceTables="res1/q#table1 res14#othertable"/>

//procs#license-cc0¶

Include this stream with a @what (a short phrase saying what is licensed) to make your resource licensed under Creative Commons-0 (a.k.a. public domain). This will generate the rights and rightsURI metadata items. It needs to live in the toplevel /resource element.

Example:

<FEED source="//procs#license-cc0" what="the HSOY catalogue"/>

//procs#license-cc-by¶

Include this stream with a @what (a short phrase saying what is licensed) to make your resource licensed under Creative Commons Attribution (CC-BY). This will generate the rights and rightsURI metadata items. It needs to live in the toplevel /resource element.

Example:

<FEED source="//procs#license-cc-by" what="the HSOY catalogue"/>

//procs#license-cc-by-sa¶

Include this stream with a @what (a short phrase saying what is licensed) to make your resource licensed under Creative Commons Attribution Share Alike (CC-BY-SA). This will generate the rights and rightsURI metadata items. It needs to live in the toplevel /resource element.

Example:

<FEED source="//procs#license-cc-by-sa" what="the HSOY catalogue"/>

//obscore#obscore-columns¶

The columns of a (standard) obscore table. This can be used to define a “native” obscore table (as opposed to the more usual mixins below that expose standard products via obscore.

Even if you are sure you want to do this, better ask again…

//ssap#atomicCoords¶

A stream for form-based service’s VOTables to include simple RA and Dec rather than normal ssa_location.

SSA services get that from the core and don’t need this.

//ssap#obscore-time-index¶

Include this stream in an ssa-like table feeding obscore. It will add an index useful for querying against obscore t_min and t_max.

This uses the dateObs and timeExt attributes which are preset for what the columns are called in SSA tables. dateObs must be a column reference because we declare the index to be on it. timeExt can be an expression, too, and no index will be declared on it.

Defaults for macros used in this stream:

dateObs: ‘ssa_dateObs’
timeExt: ‘ssa_timeExt’

//echelle#ssacols¶

Additional columns for SSA metadata tables describing Echelle spectra.

//scs#splitPosIndex¶

Adds an index over the separate long and lat columns. This is currently implemented using q3c; DaCHS will move to pgsphere entirely at some point. RDs using this will not need an update then. This will by default cluster according to the index. Then the rare cases when you don’t want this, add a cluster=”False” in the FEED.

Since long and lat may be expressions, this will not automatically declare index/columns. You should therefore pass in a comma-separated list of column names in the columns attribute in order to inform clients of the index.

Examples:

<FEED source="//scs#splitPosIndex" long="ra" lat="dec"
  columns="ra,dec"/>

<FEED source="//scs#splitPosIndex"
  long="long(ssa_location)" lat="lat(ssa_location)"
  columns="ssa_location"/>

Defaults for macros used in this stream:

cluster: ‘True’
columns: ‘’

//scs#spoint-index-def¶

Definition of an index over spoint(ra, dec); give this parameters long and lat naming the corresponding columns (in degrees).

Or use the spoint-index mixin.

This will also cluster the table according to this index, which is almost always what you want.

This needs to be replayed within a table (or something that defines a tablename macro).

Configuration Reference ¶

DaCHS’ basic configuration is done through one or more INI-style files, which are parsed using Python’s configparser module; in case of doubt on features and syntax, refer to the Python reference documentation of the Python version you are using.

Some of the more common items are discussed in the tutorial (tutorial.html#configuration-settings).

DaCHS first looks for the configuration in /etc/gavo.rc, and on production sites, it is recommended to only use this file to avoid surprises. For special situations and development systems, the following extra features are available:

DaCHS will also pick up configuration from .gavorc in the calling user’s home directory. Configuration there is merged with what is in the main configuration. You can use this to, for instance, run a second, development-type server on a production instance (obviously, you will have to run it on a different port), or for configuring extra debugging features, or whatever.
You can override the location of /etc/gavo.rc using the GAVOSETTINGS environment variable. This is intended mainly for when everything DaCHS-related should be in a single, non-OS directory, such as when the DaCHS binaries sit in a container that should be readily exchangable. Another usage is when, on a development system, you want to be able to run DaCHS using the resources of the production system (but that is, obviously, a bit dangerous if the DaCHS versions on development and production are significantly different).
You can even override the location of the per-user configuration (i.e., ~/.gavorc) using GAVOCUSTOM. This is mainly for quickly switching between configurations on development systems and thus probably irrelevant for operators.

The configuration items available are, by section:

Section [general]¶

Paths and other general settings.

cacheDir: path relative to rootDir; defaults to ‘cache’ – Path to the DC’s persistent scratch space
configDir: path relative to rootDir; defaults to ‘etc’ – Path to the DC’s non-ini configuration (e.g., DB profiles)
defaultProfileName: string; defaults to ‘’ – Deprecated and ignored.
future: set of strings; defaults to ‘’ – A set of strings naming experimental features to enable in the server. You only want to touch this if you are working on proving proposed improvements.
gavoGroup: string; defaults to ‘gavo’ – Name of the unix group that administers the DC
group: string; defaults to ‘gavo’ – Name of the group that may write into the log directory
inputsDir: path relative to rootDir; defaults to ‘inputs’ – Path to the DC’s data holdings
logDir: path relative to rootDir; defaults to ‘logs’ – Path to the DC’s logs (should be local)
logLevel: value from the list debug, error, info, warning; defaults to ‘info’ – How much should be logged?
maintainerAddress: string; defaults to ‘’ – An e-mail address to send reports and warnings to; this could be the same as contact.email; in practice, it is shown in more technical circumstances, so it’s adviable to have a narrower distribution here.
platform: string; defaults to ‘’ – Platform string (can be empty if inputsDir is only accessed by identical machines)
rdblacklist: set of strings; defaults to ‘’ – File names of blacklisted RDs. These are matched against the ends of file names, so be careful you do not over-block; in general, you should use something like ‘/resname/q.rd’ or so. You can also blacklist RD ids (which are compared literally). Separate multiple entries by commas.
rootDir: string; defaults to ‘/var/gavo’ – Path to the root of the DC file (all other paths may be relative to this
sendmail: string; defaults to ‘sendmail -t’ – Command that reads a mail from stdin, taking therecipient address from the mail header, and transfers the mail (this is for sending mails to the administrator). This command is processed by a shell (generally running as the server user), so you can do tricks if necessary.
stateDir: path relative to rootDir; defaults to ‘state’ – Path to the DC’s state information (last imported,…)
tempDir: path relative to rootDir; defaults to ‘tmp’ – Path to the DC’s scratch space (should be local)
uwsWD: path relative to rootDir; defaults to ‘state/uwsjobs’ – Directory to keep uws jobs in. This may need lots of space if your users do large queries
webDir: path relative to rootDir; defaults to ‘web’ – Path to the DC’s web related data (docs, css, js, templates…)
xsdclasspath: shell-type path; defaults to ‘None’ – Classpath necessary to validate XSD using an xsdval java class. You want GAVO’s VO schemata collection for this. Deprecated, we’re now using libxml2 for validation.

Section [adql]¶

(ignored, only left for backward compatibility)

webDefaultLimit: integer; defaults to ‘2000’ – (ignored, only present for backwards compatibility; use [async]defaultMAXREC instead.

Section [async]¶

Settings concerning TAP, UWS, and friends

csvDialect: string; defaults to ‘excel’ – CSV dialect as defined by the python csv module used when writing CSV files.
defaultExecTime: integer; defaults to ‘3600’ – Default timeout for UWS jobs, in seconds
defaultExecTimeSync: integer; defaults to ‘60’ – Timeout for synchronous TAP/UWS jobs, in seconds.
defaultLifetime: integer; defaults to ‘172800’ – Default time to destruction for UWS jobs, in seconds
defaultMAXREC: integer; defaults to ‘20000’ – Default match limit for ADQL queries via the UWS/TAP
hardMAXREC: integer; defaults to ‘20000000’ – Hard match limit (i.e., users cannot raise MAXREC or TOP beyond that) for ADQL queries via the UWS/TAP
maxSlowPollWait: integer; defaults to ‘300’ – Maximal time a UWS 1.1-WAIT request will delay the response. This should be smaller than what you have as timeout on outgoing connections.
maxTAPRunning: integer; defaults to ‘2’ – Maximum number of TAP jobs running at a time
maxUserUWSRunningDefault: integer; defaults to ‘2’ – Maximum number of user UWS jobs running at a time

Section [db]¶

Settings concerning database access.

adqlProfiles: set of strings; defaults to ‘untrustedquery’ – Name(s) of profiles that get access to tables opened for ADQL
defaultLimit: integer; defaults to ‘100’ – Default match limit for DB queries
dumpSystemTables: boolean; defaults to ‘False’ – Dump the tables from //users and //system to stateDir/system_tables once a day?
indexWorkMem: integer; defaults to ‘2000’ – Megabytes of memory to give to postgres while making indices. Set to roughly half your RAM when you have big tables.
interface: string; defaults to ‘psycopg2’ – Don’t change
maintainers: set of strings; defaults to ‘admin’ – Name(s) of profiles that should have full access to gavo imp-created tables by default
managedExtensions: list of strings; defaults to ‘pg_sphere’ – Name(s) of postgres extensions gavo upgrade -e should watch
msgEncoding: string; defaults to ‘utf-8’ – Encoding of the messages coming from the database
poolSize: integer; defaults to ‘2’ – Number of connections in DaCHS’ postgres connection pools.
profilePath: shell-type path; defaults to ‘~/.gavo:$configDir’ – Path for locating DB profiles
queryProfiles: set of strings; defaults to ‘trustedquery’ – Name(s) of profiles that should be able to read gavo imp-created tables by default

Section [ivoa]¶

The interface to the Greater VO.

authority: string; defaults to ‘x-unregistred’ – The authority id for this DC; this has no leading ivo://
dalDefaultLimit: integer; defaults to ‘10000’ – Default match limit on SCS/SSAP/SIAP queries
dalHardLimit: integer; defaults to ‘1000000’ – Hard match limit on SCS/SSAP/SIAP queries (be careful: due to the way these protocols work, the results cannot be streamed, and the results have to be kept in memory; 1e7 rows requiring 1k of memory each add up to 10 Gigs…)
oaipmhPageSize: integer; defaults to ‘500’ – Default number of records per page in the OAI-PMH interface
registerAlternative: boolean; defaults to ‘False’ – Give access URLs for the alternative protocol (https when serverURL is http and vice versa) as a mirrorURL? If you’re listening to HTTPS, this is probably a good idea.
sdmVersion: value from the list 1, 2; defaults to ‘1’ – Obsolete (SDM version 2 is shelved). Don’t use.
validOAISets: list of strings; defaults to ‘ivo_managed, vosi, local’ – Comma-separated list of OAI sets DaCHS should accept for publication. This must always include ivo_managed and vosi, and you will want local if you run a web interface.
VOSITableDetail: value from the list max, min; defaults to ‘max’ – Default level of detail to return on the VOSI endpoint (change to min when you have more than 100 or tables).
votDefaultEncoding: value from the list binary, td; defaults to ‘binary’ – Default ‘encoding’ for VOTables in many places (like the DAL responses; this can be user-overridden using the _TDENC local HTTP parameter.

Section [ui]¶

Settings concerning the local user interface

Section [web]¶

Settings related to serving content to the web.

adaptProtocol: boolean; defaults to ‘True’ – Adapt internal absolute links to http/https by request method used. You must switch this off if running behind a reverse proxy.
adminpasswd: string; defaults to ‘’ – Password for online administration, leave empty to disable
alternateHostnames: list of strings; defaults to ‘’ – A comma-separated list of hostnames this server is also known under. Only set this if you’re running https. With this, you can handle a situation where your data center can be reached as both example.org and www.example.org.
bindAddress: string; defaults to ‘127.0.0.1’ – Interface to bind to
corsOriginPat: string; defaults to ‘’ – A regular expression for URLs from which to authorise cross-origin requests. This is matched, i.e., the RE must account for the whole URL including the schema. Example: https?://example.com/apps/.*.
enableTests: boolean; defaults to ‘False’ – Enable test pages (don’t if you don’t know why)
favicon: path relative to webDir; defaults to ‘None’ – Webdir-relative path to a favicon; this overrides the default of a scaled version of the logo.
graphicMimes: list of strings; defaults to ‘image/fits,image/jpeg,application/x-votable+xml;content=datalink’ – Media types considered as graphics (for SIAP, mostly)
ignore-uir-header: boolean; defaults to ‘False’ – Do not honour clients’ upgrade-insecure-requests headers if https is enabled. Outside of development or testing environments, there is generally little reason to use this.
jsSource: boolean; defaults to ‘False’ – If True, Javascript will not be minified on delivery (this is for debugging)
logFormat: value from the list combined, default; defaults to ‘default’ – Log format to use. Default doesn’t log IPs, user agents, or referrers and thus should be ok in terms of not processing personal data, which in turn means you probably don’t have to declare anything in EU jurisdictions.
maxPreviewWidth: integer; defaults to ‘300’ – Ignored, only present for backward compatibility
maxSyncUploadSize: integer; defaults to ‘500000’ – Maximal size of file uploads for synchronous TAP in bytes.
maxUploadSize: integer; defaults to ‘20000000’ – Maximal size of (mainly TAP) file uploads in async requests in bytes; sync requests use maxSyncUploadSize.
nevowRoot: path fragment; defaults to ‘/’ – Path fragment to the server’s root for operation off the server’s root; this must end with a slash (setting this will currently break essentially the entire web interface. If you must use it, contact the authors and we will fix things.)
operatorCSS: string; defaults to ‘’ – URL of an operator-specific CSS. This is included as the last item and can therefore override rules in the distributed CSS.
preloadPublishedRDs: boolean; defaults to ‘False’ – Preload all RDs of services you’ve published. This is mainly helpful when code in such RDs might, for instance, lock and such failures don’t suddenly occur in operation.
preloadRDs: list of strings; defaults to ‘’ – RD ids to preload at the server start. Load time of RDs listed here goes against startup time, so only do this for RDs that have execute children that should run regularly. For everything else consider preloadPublishedRDs.
previewCache: path relative to webDir; defaults to ‘previewcache’ – Webdir-relative directory to store cached previews in
realm: string; defaults to ‘X-Unconfigured’ – Authentication realm to be used (currently, only one, server-wide, is supported)
root: string; defaults to ‘__system__/services/root/fixed’ – The path of the the root page of the data center. The default renders the root.html template. On single-service installations, you could direct this to that single service, or you could use the ADQL query form.
serverFDLimit: integer; defaults to ‘4000’ – A hard limit of the number of file handles DaCHS should try to set. This will only take effect if DaCHS is started as root. Otherwise, DaCHS will just adjust its soft limit to the external limit irrespective of this setting.
serverPort: integer; defaults to ‘8080’ – Port to bind the http port of the server to; https, if present at all, is always on 443.
serverURL: string; defaults to ‘http://localhost:8080’ – URL fragment used to qualify relative URLs where necessary. Note that this must contain the port the server is accessible under from the outside if that is not 80; nonstandard ports are not supported for https. If you offer both http and https, use your preferred protocol here. If you change this on a running server, you must run dachs pub -a to update internal and external links.
sitename: string; defaults to ‘Unnamed data center’ – A short name for your site
sqlTimeout: integer; defaults to ‘15’ – Default timeout for non-TAP database queries (which can in general be overridden by users setting _TIMEOUT).
templateDir: path relative to webDir; defaults to ‘templates’ – webDir-relative location of global nevow templates
user: string; defaults to ‘gavo’ – Run server as this user.

Code in DaCHS ¶

This section contains a few general points for python code embedded in DaCHS. Most of the material applies to procedure definitions (Element apply, Element dataFormatter, Element dataFunction, Element descriptorGenerator, Element iterator, Element metaMaker, Element processEarly, Element processLate, Element pargetter, Element phraseMaker, Element regTest, Element rowfilter, Element sourceFields) as well as to other pieces of code, such as in Element customGrammar, Element customCore, or Custom Pages.

More information on what names DaCHS would like you to see are available in Functions Available For Row Makers and The DaCHS API.

Importing Modules¶

To keep the various resources as separate from each other as possible, DaCHS does not manipulate Python’s import path. However, one frequently wants to have library-like modules providing common functionality or configuration in a resdir (the conventional place for these would be in res/).

To import these, use api.loadPythonModule(path). Path, here, is the full path to the file containing the python code, but without the .py. When you have the RD, the conventional pattern is:

mymod, _ = api.loadPythonModule(rd.getAbsPath("res/mymod"))

instead of import mymod.

As you can see loadPythonModule returns a tuple; you’re very typically only interested in the first element.

Note in particular that for modules loaded in this way, the usual rule that you can just import modules next to you does not apply. To import a modules “next to” you without having to go through the RD, use the special form:

siblingmod, _ = api.loadPythonModule("siblingmod", relativeTo=__file__)

instead of import siblingmod. This will take the directory part for what’s in relativeTo (here, the module’s own path) and make a full path out of the first argument to pull the modules from there.

Database Queries¶

You often need to query the database from DaCHS-related code. To get connections, use DaCHS’ connection pool, and to make sure you return the connections to the pool when done, only use them through context managers. Depending on what you need to do, there are four pools you might be interested in:

getTableConn – a sane default, lets you look at all tables but not write (anything). This is what DaCHS uses to run queries derived from normal DAL requests.
getUntrustedConn – a connection that has the privileges of a remote user coming in through TAP.
getWritableAdminConn – a connection that lets you read and write about everywhere. When using this, you must explicitly commit the connection, or your changes will be lost.
getAdminConn – like getWritableAdminConn, but will automatically be committed as it is returned to the pool (unless an exception happened, in which case the connection will be closed without committing it).

With DaCHS’ connections, you usually will not obtain cursors but directly use one of the

query(q, args={}, timeout=None) or
queryToDict(q, args={}, timeout=None, caseFixer=None) methods.

The query q has psycopg2-style placeholders (... WHERE mag<%(maglim)s) which are filled, again psycopg2-like from the args dictionary. timeout is given in seconds.

Both return an iterator; for query, this yields a tuple per row, for queryToDict, that is dictionaries, the keys of which are the lowercased column names (or something hard to predict for expressions in the select clause not sporting an AS). In case you need mixed-case keys (we recommend you avoid that), you can pass in a caseFixer dictionary that maps the lowercased names to their mixed-case versions.

Note that the queries passed will not be executed unless you start consuming the iterator returned. This means that you can only use them when you actually get back a result. Instead of the query methods, use execute for statements that do not return anything.

In sum, the typical database query in the vicinity of DaCHS would look like this:

count_sum = 0
with base.getTableConn() as conn:
  for row in conn.queryToDicts("SELECT * FROM sch.my_table"):
     count_sum += row["count"]
     print(row)

or, for queries not returning anything:

with base.getWritableAdminConn() as conn:
  conn.execute("DROP TABLE obsolete_table")

In case you wondered: Yes, in the past we have experimented with abstracting away the SQL. And we found it doesn’t make the code any more robust, just a lot harder to figure out.

Data Descriptors ¶

Most basic information on data descriptors is contained in tutorial.html. The material here just covers some advanced topics.

Updating Data Descriptors¶

By default, dachs imp will try to drop all tables made by the data descriptors selected. For “growing” data, that is suboptimal, since typicaly just a few new datasets need to be added to the table, and re-ingesting everything else is just a waste of time and CPU.

To accomodate such situations, DaCHS allows to add an updating="True" attribute to a data element; updating DDs will create tables that do not exist but will not drop existing ones.

Using fromdb on ignoreSources¶

Updating DDs will still run like normal DDs and thus import everything matching the DD’s sources. Thus, after the second import you would have duplicate records for sources that existed during the first import.

To avoid that, you (usually) need to ignore existing sources (see Element ignoreSources). In the typical case, where a dataset’s accref is just the inputs-relative path to the dataset’s source, that is easily accomplished through the fromdb attribute of ignoreSources; its value is a database query that returns the inputs-relative paths of sources to ignore.

Hence, unless you are playing games with the accrefs (in which case you are probably smart enough to figure out how to adapt the pattern), the following speficiation will exactly import all FITS files within the data subdirectory of the resdir that haven’t been ingested into the mydata table during the last run, either because they’ve not been there or because there were skipped during an import -c:

<data id="import" updating="true">
  <sources pattern="data/*.fits">
    <ignoreSources fromdb="select accref from \schema.mydata"/>
  </sources>

  <fitsProdGrammar>
    <rowfilter procDef="//products#define">
      <bind key="table">"\schema.mydata"</bind>
    </rowfilter>
  </fitsProdGrammar>

  <make table="mydata">
    <!-- your rowmaker here -->
  </make>
</data>

Note that fromdb can be combined with fromfiles and pattern; whatever is specified in the latter two will always be ignored.

To completely re-import such a table – for instance after a table schema change or because the whole data collection has been re-processed –, just run dachs drop on the DD and run import as usual.

It is probably a good idea to occasionally run dachs imp -I on tables updated in this way to optimise the indices (a REINDEX <tablename> in a database shell will do, too).

Using fromdbUpdating on ignoreSources¶

Sometimes reprocessing happens quite frequently to a small subset of the datasets in a resource. In that case, it would again be a waste to tear down the entire thing just to update a handful of records.

For such situations, there is the fromdbUpdating attribute of ignoreSources. As with fromdb, this contains a database query, but in addition to the accref, this query has to return a timestamp. A source is then only ignored if this timestamp is not newer than the disk file’s one. If that timestamp is the mtime of the file in the original import, the net effect is that files that have been modified since that import will be re-ingested.

There is a catch, though: You need to make sure that the record ingested previously is removed from the table. Typically, you can do that by defining accref as a primary key (if that’s not possible because you are generating multiple records with the same accref, there is nothing wrong with using a compound primary key). This will, on an attempted overwrite, cause an IntegrityError, and you can configure DaCHS to turn this into an overwrite using the table’s forceUnique and dupePolicy attributes.

The following snippet illustrates the technique:

<table id="withdate" mixin="//products#table" onDisk="True"
    primary="accref"
    forceUnique="True"
    dupePolicy="overwrite">
  <column name="mtime" type="timestamp"
    ucd="time;meta.file"
    tablehead="Timestamp"
    description="Modification date of the source file."/>
  <!-- your other columns -->
</table>

<data id="import" updating="True">
  <sources pattern="data/*.fits">
    <ignoreSources
      fromdbUpdating="select accref, mtime from \schema.withdate"/>
  </sources>
  <fitsProdGrammar>
    <rowfilter procDef="//products#define">
      <bind key="table">"\schema.withdate"</bind>
    </rowfilter>
  </fitsProdGrammar>
  <make table="withdate">
    <rowmaker>
      <map key="mtime">datetime.datetime.utcfromtimestamp(
        os.path.getmtime(\fullPath))</map>
      <!-- other rowmaker rules -->
    </rowmaker>
  </make>
</data>

Again, this can be combined with the other attributes of ignoreSources; in effect, whatever is ignored from them is treated as if their modification dates were in the future.

Harvesting from Remote Databases¶

When using Element odbcGrammar, identifying what is already ingested clearly cannot use sources. Instead, you will have to pick added or modified records in some other way. Realistically, you should keep some monotonously increasing value on both sides. Ideally, it would be a transaction id, because for these, it’s clear whether or not something has actually been transferred. In a pinch, a unix timestamp will do, too.

Particular care should be taken when harvesting from databases to avoid duplicate rows when a re-harvest fetches the “same” record twice for whatever reason. You should probably designate a primary key and specify a dupePolicy like this:

<table id="mirror" onDisk="True"
  primary="obid"
  dupePolicy="overwrite">

In case some other table holds foreign keys into your table, it is wise to think hard whether the dupePolicy should really be dropOld (cf. Element table).

Your data element will use an odbc grammar with a computed query (available on DaCHS newer than 2.5). The example in Element makeyQuery shows the basics. The important part is to consider the case when the local table does not exist yet (as it will on the original import). Dealing with this as shown in the example lets you use the same data element for imports and updates.

Metadata ¶

Various elements support the setting of metadata through meta elements. Metadata is used for conveying RMI-style metadata used in the VO registry. See [RMI] for an overview of those. We use the keys given in RMI, but there are some extensions discussed in RMI-style Metadata.

The other big use of meta information is for feeding templates. Those “local” keys should all start with an underscore. You are basically free to use those as you like and fetch them from your custom templates. The predefined templates already have some meta items built in, discussed in Template Metadata.

So, metadata is a key-value mapping. Keys may be compound like in RMI, i.e., they may consist of period-separated atoms, like publisher.address.email. There may be multiple items for each meta key.

Defining Metadata¶

In RDs, there are two ways to define metadata: Meta elements and meta streams; the latter are also used in defaultmeta.txt.

Meta Elements¶

These look like normal XML elements and have a mandatory name attribute, a meta key relative to the element’s root . The text content is taken as the meta value; child meta elements are legal.

An optional attribute for all meta elements is format (see Meta Formats).

Typed meta elements can have further attributes; these usually can also be given as meta children with the same name.

Usually, metadata is additive; add a key twice and you will have a sequence of two meta values. To remove previous content, prefix the meta name with a bang (!). Here is an example:

<resource>
  <!-- a simple piece of metadata -->
  <meta name="title">A Meta example</meta>

  <!-- repeat a meta thing for a sequence (caution: not everything
    is repeatable in all output formats -->
  <meta name="subject">Examples</meta>
  <meta name="subject">DaCHS</meta>

  <!-- Hierarchical meta can be set nested -->
  <meta name="creator">
    <meta name="name">Nations, U.N.</meta>
    <meta name="logo">http://un.org/logo.png</meta>
  </meta>
  <meta name="creator">
    <meta name="name">Neumann, A.E.</meta>
  </meta>

  <!-- @format lets you specify extra markup; make sure you
  have consistent initial indentation. -->
  <meta name=description" format="rst">
    This resource is used in the `DaCHS reference docs`_

    .. _DaCHS reference Docs: http://docs.g-vo.org/DaCHS
  </meta>

  <!-- you can contract "deeper" trees in paths -->
  <meta name="contact.email">gavo@ari.uni-heidelberg.de</meta>

  <!-- typed meta elements can have additional attributes -->
  <meta name="uses" ivoId="ivo://org.gavo.dc/DaCHS"
    >DaCHS server sortware</meta>

  <!-- To overwrite a key set before, prefix the name with a bang. -->
  <meta name="!title">An improved Meta example</meta>
</resource>

The resulting meta structure is like this:

+-- title
|     +---- "An improved Meta example
|
+-- subject
|     +---- "Examples
|     +---- "DaCHS
|
+-- creator
|     +----- name
|     |       +---- "Nations, U.N.
|     +----- logo
|     |       +---- "http://un.org/logo.png
+-- creator
|     +----- name
|             +---- "Neumann, A.E.
|
+-- description
|     +----- [formatted text, "This resource..."]
|
+-- contact
|      +----- email
|               +----- "gavo@ari.uni-heidelberg.de
|
+-- uses
       +----- "DaCHS server software
       +----- ivoId
                +----- "ivo://org.gavo.dc/DaCHS

Stream Metadata¶

In several places, most notably in the defaultmeta.txt file and in meta elements without a name attribute, you can give metadata as a “meta stream”. This is just a sequence of lines containing pairs of <meta key> and <meta value>.

In addition, there are comments, empty lines, continuations, forced overwriting, and format selection.

Continuation lines work by ending a line with a backslash. The following line separator and all blanks and tabs following it are then ignored. Thus, the following two meta keys end up having identical values:

meta1: A contin\
  uation line needs \
    a blank if you wan\
t one.
meta2: A continuation line needs a blank if you want one

Note that whitespace behind a backslash prevents it from being a continuation character. That is, admittedly, a bit of a trap.

Other than their use as continuation characters, backslashes have no special meaning within meta streams as such. Within meta elements, however, macros are expanded after continuation line processing if the meta parent knows how to expand macros. This lets you write things like:

<meta>
  creationDate: \metaString{authority.creationDate}
  managingOrg:ivo://\getConfig{ivoa}{authority}
</meta>

Comments and empty lines are easy: Empty lines are allowed, and a comment is a line with a hash (#) as its first non-whitespace character. Both constructs are ignored, and you can even continue comments (though you should not).

When you repeat a key, metadata is being added. Hence,:

subject: active-galaxies
subject: meteorites

will lead to two keywords in the subject meta. Sometimes you instead want to overwrite what is already there, in particular for meta that needs to be unique:

!title: A revised title

will make sure that there is just a single title meta, and it is “A revised title”.

Finally, stream meta has format=plain by default. To select raw or reStructuredText format, prefix the value with raw: or rst:, respectively:

_sidebarlocal: raw:<div class="sidebarnote">\
  <a href="/adql">Try ADQL</a> to query our data.</div>
description: rst:This is *nice*, **important** data.

Meta inheritance¶

When you query an element for metadata, it first sees if it has this metadata. If that is not the case, it will ask its meta parent. This usually is the embedding element. It wil again delegate the request to its parent, if it exists. If there is no parent, configured defaults are examined. These are taken from rootDir/etc/defaultmeta, where they are given as colon-separated key-value pairs, e.g.,

publisher: The GAVO DC team
publisherID: ivo://org.gavo.dc
contact.name: GAVO Data Center Team
contact.address: Moenchhofstrasse 12-14, D-69120 Heidelberg
contact.email: gavo@ari.uni-heidelberg.de
contact.telephone: ++49 6221 54 1837
creator.name: GAVO Data Center
creator.logo: http://vo.ari.uni-heidelberg.de/docs/GavoTiny.png

The effect is that you can give global titles, descriptions, etc. in the RD but override them in services, tables, etc. The configured defaults let you specify meta items that are probably constant for everything in your data center, though of course you can override these in your RD elements, too.

In HTML templates, missing meta usually is not an error. The corresponding elements are just left empty. In registry documents, missing meta may be an error.

Meta formats¶

Metadata must work in registry records as well as in HTML pages and possibly in other places. Thus, it should ideally be given in formats that can be sensibly transformed into the various formats.

DaCHS knows four input formats:

literal: The textual content of the element will not be touched. In HTML, it will end up in a div block of class literalmeta.
plain: The textual content of the element will be whitespace-normalized, i.e., whitespace will be stripped from the start and the end, runs of blanks and tabs are replaced by a single blank, and empty lines translate into paragraphs. In HTML, these blocks com in plainmeta div elements.
rst: The textual content of the element is interpreted as ReStructuredText. When requested as plain text, the ReStructuredText itself is returned, in HTML, the standard docutils rendering is returned.
raw: The textual content of the element is not touched. It will be embedded into HTML directly. You can use this, probably together with CDATA sections, to embed HTML – the other formats should not contain anything special to HTML (i.e., they should be PCDATA in XML lingo). While the software does not enforce this, raw content should not be used with RMI-type metadata. Only use it for items that will not be rendered outside of HTML templates.

Macros in Meta Elements¶

Macros will be expanded in meta items using the embedding element as macro processors (i.e., you can use the macros defined by this element).

Typed Meta Elements¶

While generally the DC software does not care what you put into meta items and views them all as strings, certain keys are treated specially. The following meta keys trigger some special behaviour:

_dataUpdated

A meta value representing a timestamp.

Accessing it, you will get a formatted ISO/DALI string. You can construct them with both strings (that we’ll try to parse and bomb if that’s not possible) and datetime.datetime objects.

_example

A MetaValue to keep VOSI examples in.

All of these must have a title, which is also used to generate references.

These also are in reStructuredText by default, and changing that probably makes no sense at all, as these will always need interpreted text roles for proper markup.

Thus, the usual pattern here is:

<meta name="_example" title="An example for _example">
  See docs_

  .. _docs: http://docs.g-vo.org
</meta>

_metadataUpdated

A meta value representing a timestamp.

Accessing it, you will get a formatted ISO/DALI string. You can construct them with both strings (that we’ll try to parse and bomb if that’s not possible) and datetime.datetime objects.

_news

A meta value representing a “news” items.

The content is the body of the news. In addition, they have date, author, and role children. In plain text, you would write:

_news: Frobnicated the quux.
_news.author: MD
_news.date: 2009-03-06
_news.role: updated

In XML, you would usually write:

<meta name="_news" author="MD" date="2009-03-06">
  Frobnicated the quux.
</meta>

_news items become serialised into Registry records despite their leading underscores. role then becomes the date’s role.

_related

A meta value containing a link and optionally a title

In plain text, this would look like this:

_related:http://foo.bar
_related.title: The foo page

In XML, you can write:

<meta name="_related" title="The foo page"
  ivoId="ivo://bar.org/foo">http://foo.bar</meta>

or, if you prefer:

<meta name="_related">http://foo.bar
   <meta name="title">The foo page</meta></meta>

These values are used for _related (meaning “visible” links to other services).

For links within you data center, use the internallink macro, the argument of which the the “path” to a resource, i.e. RD path/service/renderer; we recommend to use the info renderer in such links as a rule. This would look like this:

<meta name="_related" title="Aspec SSAP"
  >\internallink{aspec/q/ssa/info}</meta>

cites

A meta value containing an ivo-id and a name of a related resource.

The sort of relationsip is encoded in the meta name, where the terms are defined in the vocabuary http://www.g-vo.org/rdf/voresource/relationship_type (where there are minor lexical deviations from identifiers there to DaCHS’ meta names).

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isServedBy and isServiceFor are somewhat special cases, as the service attribute of data publications automatically takes care of them; so, you shouldn’t usually need to bother with these two manually.

continues

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

creationDate

A meta value representing a timestamp.

Accessing it, you will get a formatted ISO/DALI string. You can construct them with both strings (that we’ll try to parse and bomb if that’s not possible) and datetime.datetime objects.

creator.logo

A MetaValue corresponding to a small image.

These are rendered as little images in HTML. In XML meta, you can say:

<meta name="_somelogo" type="logo">http://foo.bar/quux.png</meta>

derivedFrom

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

doi

A MetaValue for a DOI.

This lets people construct DOI meta with or without a doi: prefix. It also creates landing page links in HTML.

hasPart

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

info

A meta value for info items in VOTables.

In addition to the content (which should be rendered as the info element’s text content), it contains an infoName and an infoValue.

They are only used internally in VOTable generation and might go away without notice.

isContinuedBy

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isDerivedFrom

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isIdenticalTo

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isNewVersionOf

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isPartOf

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isPreviousVersionOf

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isServedBy

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isServiceFor

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isSourceOf

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isSupplementTo

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

isSupplementedBy

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

logo

A MetaValue corresponding to a small image.

These are rendered as little images in HTML. In XML meta, you can say:

<meta name="_somelogo" type="logo">http://foo.bar/quux.png</meta>

mirrorOf

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

note

A meta value representing a “note” item.

This is like a footnote, typically on tables, and is rendered in table infos.

The content is the note body. In addition, you want a tag child that gives whatever the note is references as. We recommend numbers.

Contrary to other meta items, note content defaults to rstx format.

Typically, this works with a column’s note attribute.

In XML, you would usually write:

<meta name="note" tag="1">
  Better ignore this.
</meta>

referenceURL

A meta value containing a link and optionally a title

In plain text, this would look like this:

_related:http://foo.bar
_related.title: The foo page

In XML, you can write:

<meta name="_related" title="The foo page"
  ivoId="ivo://bar.org/foo">http://foo.bar</meta>

or, if you prefer:

<meta name="_related">http://foo.bar
   <meta name="title">The foo page</meta></meta>

These values are used for _related (meaning “visible” links to other services).

<meta name="_related" title="Aspec SSAP"
  >\internallink{aspec/q/ssa/info}</meta>

relatedTo

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

servedBy

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

serviceFor

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

source

A MetaValue that may contain bibcodes, which are rendered as links into ADS.

subject

A MetaValue translating UAT terms to their labels in HTML.

This follows the assumption that HTML is what humans look at, anything else is either computers or nerds.

uses

A meta value containing an ivo-id and a name of a related resource.

All relationship metas should look like this (using isSupplementTo as an example; while an ivoId is not mandatory, it rarely makes sense to declare a relationship without it):

isSupplementTo: GAVO TAP service
isSupplementTo.ivoId: ivo://org.gavo.dc

votlink

A MetaValue serialized into VOTable links (or, ideally, analogous constructs).

This exposes the various attributes of VOTable LINKs as href linkname, contentType, and role. You cannot set ID here; if this ever needs referencing, we’ll need to think about it again. The href attribute is simply the content of our meta (since there’s no link without href), and there’s never any content in VOTable LINKs).

You could thus say:

votlink: http://docs.g-vo.org/DaCHS
votlink.role: doc
votlink.contentType: text/html
votlink.linkname: GAVO DaCHS documentation

Additionally, there is creator, which is really special (at least for now). When you set creator to a string, the string will be split at semicolons, and for each substring a creator item with the respective name is generated. This may sound complicated but really does about what your would expect when you write:

<meta name="creator">Last, J.; First, B.; Middle, I.</meta>

Metadata in Standard Renderers¶

Certain meta keys have a data center-internal interpretation, used in renderers or writers of certain formats. These keys should always start with an underscore. Among those are:

_noresultwarning:
_intro:	used by the standard HTML template for explanatory text above the search form.
_bottominfo:	used by the standard HTML template for explanatory text below the search form.
_related:	used in the standard HTML template for links to related services. As listed above, this is a link, i.e., you can give a title attribute.
_longdoc:	used by the service info renderer for an explanatory piece of text of arbitrary length. This will usually be in ReStructuredText, and we recommend having the whole meta body in a CDATA section.
_news:	news on the service. See above at Typed Meta Elements.
_warning:	used by both the VOTable and the HTML table renderer. The content is rendered as some kind of warning. Unfortunately, there is no standard how to do this in VOTables. There is no telling if the info elements generated will show anywhere.
	displayed by the default response template instead of an empty table (use it for things like “No Foobar data for your query”)
_type:	on Data instances, used by the VOTable writer to set the `type` attribute on `RESOURCE` elements (to either “results” or “meta”). Probably only useful internally.
superseded:	in RDs or services, marks them as superseded, which generally makes them inaccessible. The body of this meta should provide pointers to where the new version(s) of the resources might be found (cf. tutorial.html#deleting-resources).
_plotOptions:	typically set on services, this lets you configure the initial appearance of the javascript-based quick plot. The value must be a javascript dictionary literal (like `{"xselIndex": 2}`) unless you’re trying CSS deviltry (which you could, using this meta; then again, if you can inject RDs, you probably don’t need CSS attacks). Keys evaluated include: xselIndex – 0-based index of the column plotted on the x-axis (default: 0) yselIndex – 0-based index of the column plotted on the y-axis (default: length of the column list; that’s “histogram on y) usingIndex – 0-based index of the plotting style selector. For now, that’s 0 for points and 1 for lines.

RMI-Style Metadata¶

For services (and other things) that are registered in the Registry, you must give certain metadata items (and you can give more), where we take their keys from [RMI]. We provide a explanatory leaflet for data providers. The most common keys – used by the registry interface and in part by HTML and VOTable renderers – include:

_metadataUpdated:
title:	this should in general be given separately on the resource, each table, and each service. In simple cases, though, you may get by by just having one global title on the resource and rely on metadata inheritance.
shortName:	a string that should indicate what the service is in 16 characters or less.
creationDate:	Use ISO format with time, UTC only, like this: 2007-10-04T12:00:00Z
_dataUpdated:	The timestamp of the last successful `dachs imp`, again in DALI/ISO format.
	Timestamp when the metadata was last updated. On RDs, that’s the timestamp of the RD source, on published things, it’s the timestamp of the last `dachs pub`.
subject:	A subject keyword. By VOResource 1.1, these should be taken from http://www/ivoa.net/rdf/uat
rights:	freetext copyright notice. See tutorial.html#Licensing for details.
rights.rightsURI:
	machine readable license URI. Take from https://spdx.org/licenses/ if possible.
source:	bibcodes will be expanded to ADS links here.
referenceURL:	again, a link, so you can give a title for presentation purposes. If you give no referenceURL, the service’s info page will be used.
creator.name:	this should be the name of the “author” of the data set. If you set this, you may want to override creator.logo as well. For persons, always use the form “Last, F.I.“; this saves all components the unsolvable problem to tell first from last names, and sorting these strings will naturally yield the sequence people expect. Also, if you have multiple creators, better just set `creator` as discussed in Typed Meta Elements.
type:	one of Other, Archive, Bibliography, Catalog, Journal, Library, Simulation, Survey, Transformation, Education, Outreach, EPOResource, Animation, Artwork, Background, BasicData, Historical, Photographic, Press, Organisation, Project, Registry – it’s optional and we doubt its usefulness. You may repeat the content type if you need to; see also [RMI], sect. 3.3.
contentLevel:	addresse(s) of the data: Research, Amateur, General
facility:	no IVOA ids are supported here yet, but probably this should change.
coverage:	see the special section
service-specific metadata (for SIA, SCS, etc.):
	see the documentation of the respective cores.
utype:	tables (and possibly other items) can have utypes to signify their role in specific data models. For tables, this utype gets exported to the tap_schema.
identifier:	this is the IVOID of the resource, usually generated by DaCHS. Do not override this unless you know what you are doing (which at least means you know how to make DaCHS declare an authority and claim it). If you do override the identifier of a service that’s already published, make sure you run `dachs admin makeDeletedRecord <previous identifier>` (before or after the `dachs pub` on the resource, or the registries will have two copies of your record, one of which will not be updated any more; and that would suck for Registry users.
published_identifer:
	Like identifier, except it will be None if the resource does not look like it is (destined to be) published.
mirrorURL:	add these on publication to declare mirrors for a service. Only do so if you actually manage the other service. If you list the service’s own accessURL here, it will be filtered from this registry record; this is so you can use the same RD on the primary site and the mirror.
_example:	A DALI example. See tutorials.html#writing-examples
moreExamples:	A URI for an additional DALI examples document. These get translated to DALI continuation-s.
tableset:	A DaCHS reference to a table to include in a registry tableset (new in 2.7.3). This currently will only be interpreted in document-typed resources.

While you can set any of these in etc/defaultmeta.txt, the following items are usually set there:

publisher
publisherID
contact.name
contact.address
contact.email
contact.telephone

Coverage Metadata¶

Coverage metadata lets clients get a quick idea of where in space, time, and electromagnetic spectrum the data within a resource is. Obviously, this information is particularly important for resource discovery in registries.

Not all resources have coverages on all axes; a service validator, say, probably has no physical coverage at all, and a theoretical spectral service may just have meaningful spectral coverage.

There are two meta keys pertinent to coverage metadata:

coverage.waveband – One of Radio, Millimeter, Infrared, Optical, UV, EUV, X-ray, Gamma-ray, and you can have multiple waveband specifications. As this information is quite regularly used in discovery you should make sure to define it if applicable.
coverage.regionOfRegard – in essence, the “pixel size” of the service in degrees. If, for example, your service gives data on a lattice of sampling points, the typical distance of such points should be given here. You will not usually specify this unless your „pixel size” is significantly larger than about an arcsec.

The legacy coverage.profile meta key should not be used any more.

To give proper, numeric STC coverage, use the Element coverage.

It has three children, one each for the spatial, spectral, and temporal axes. For spectral and temporal, just add as many intervals as necessary. Do not worry about gaps in the temporal coverage: it is not necessary that the coverage is “tight”; as long as there is a reasonable expectation that data could be there, it’s fine to declare coverage. Hence, for ground-based observations, there is no need to exclude intervals of daylight, bad weather, or even maintenance downtime.

Intervals are given as in VOTable tabledata, i.e., as two floating point numbers separated by whitespace. There are no (half-) open intervals – just use insanely small or large numbers if you really think you need them.

For spatial coverage, a single spatial element should be given. It has to contain a MOC in ASCII serialisation. Recent versions of Aladin can generate those, or you can write SQL queries to have them computed by sufficiently new versions of pgsphere. Most typically, you will use updater elements to fill spatial coverage (see below).

A complete coverage element would thus look like this:

<coverage>
  <spectral>3.8e-07 5.2e-07</spectral>
  <temporal>18867 27155</temporal>
  <spatial>
    4/2068
    5/8263,8268-8269,8271,8280,8323,8326,8329,9376,9378
    6/33045-33047,33049,33051,33069,33080-33081,33083,33104-33106,
      33112,33124-33126,33128-33130,33287,33289,33291,33297-33299,
      33313,33315,33323-33326,33328-33330,37416,37418,37536
  </spatial>
</coverage>

In general computing coverage is a tedious task. Hence, DaCHS has rules to compute it for many common cases (SSAP, SIAP, Obscore, catalogs with usable UCDs). Because coverage calculations can run for a long time, they are not performed online. Instead, DaCHS updates coverage elements when the operator runs dachs limits. In the simplest case, operators add:

<coverage>
  <updater sourceTable="data"/>
  <spectral/>
  <temporal/>
  <spatial/>
</coverage>

into an RD with a table named data. Currently, this must be lexically below the table element, but if this isn’t fixed to allow the location of the coverage element near the rest of the metadata near the top of the RD, complain fiercely.

Operators then run dachs limits q (assuming the RD is called q.rd), and DaCHS will fill out the three coverage elements (in case you want to fix them: the heuristics it uses to do that are in gavo.user.info).

In this construction, DaCHS will overwrite any previous content in the coverage child elements. If you want to fill out some coverage items manually and have DaCHS only compute, say, the spatial coverage, don’t give the sourceTable attribute (which essentially says: “grab as much coverage from the referenced table as you can”) but rather the specialised spaceTable. This is particularly useful if you want to annotate ”holes” in your temporal coverage. For instance, if your resource contains two fairly separate campaigns (which DaCHS does not currently realise automatically):

<coverage>
  <updater spaceTable="main"/>
  <spatial/>
  <temporal>45201 45409</temporal>
  <temporal>54888 55056</temporal>
</coverage>

Due to limitations of pgsphere, DaCHS does not currently take into account the size of the items in a database table. While that is probably all right for spectra and catalogs, for images this might lose significant coverage, as DaCHS only uses the centers of the images and just marks the containing healpix of the selected MOC order. The default MOC order is 6 (a resolution of about a degree). Until we properly deal with polygons, make sure to increase the MOC order to at least the order of magnitude of the images in an image service, like this:

<coverage>
 <updater sourceTable="main" mocOrder="4"/>
 <spatial/>
</coverage>

If you know your resource only contains relatively few but compact patches, you may also want to increase mocOrder (spatial resolution doubles when you increase mocOrder by one).

Display Hints ¶

Display hints use an open vocabulary. As you add value formatters, you can evaluate any display hint you like. Display hints understood by the built-in value formatters include:

displayUnit

use the value of this hint as the unit to display a value in.

spectralUnit

as displayUnit, expect non-linear transformations between length, frequency, and energy are also supported, assuming they refer to electromagnitic radiation (otherwise it makes absolutely no sense to convert Joule to meter, say).

nopreview

if this key is present with any value, no HTML code to generate previews when mousing over a link will be generated.

sepChar

a separation character for sexagesimal displays and the like.

sf

“Significant figures” – length of the mantissa for this column. Will probably be replaced by a column attribute analogous to what VOTable does.

type

a key that gives hints what to do with the column. Values currently understood include:

bar: display a numeric value as a bar of length value pixels.
bibcode: display the value as a link to an ADS bibcode query.
checkmark: in HTML tables, render this column as empty or checkmark depending on whether the value is false or true to python.
humanDate: display a timestamp value or a real number in either yr (julian year), d (JD, or MJD if DaCHS guesses it’s mjd; that’s unfortunately arcane still), or s (unix timestamp) as an ISO string.
humanDay: display a timestamp or date value as an ISO string without time.
humanTime: display values as h:m:s.
keephtml: lets you include raw HTML. In VOTables, tags are removed.
product: treats the value as a product key and expands it to a URL for the product (i.e., typically image). This is defined in protocols.products. This display hint is also used by, e.g., the tar format to identify which columns should contribute to the tar file.
dms: format a float as degree, minutes, seconds.
simbadlink: formats a column consisting of alpha and delta as a link to query simbad. You can add a coneMins displayHint to specify the search radius.
hms: force formatting of this column as a time (usually for RA).
url: makes value a link in HTML tables. The anchor text will be the last element of the path part of the URL, or, if given, the value of the anchorText property of the column (which is for cases when you want a constant text like “Details”). If you need more control over the anchor text, use an outputField with a formatter.
imageURL: makes value the src of an image. Add width to force a certain image size.

noxml

if ‘true’ (exactly like this), do not include this column in VOTables.

Note that not any combination of display hints is correctly interpreted. The interpretation is greedy, and only one formatter at a time attempts to interpret display hints.

Data Model Annotation ¶

In the VO, data models are used when simple, more or less linear annotation methods like UCDs do not provide sufficient expressive power. Or well, they should be used. As of early 2017, things are, admittedly, still a mess.

DaCHS lets you annotate your data in dm elements; the annotation will then be turned into standard VOTable annotation (when that’s defined). Sometimes, the structured references provided by the DM annotation are useful elsewhere, too – the first actual use of this framework was the geojson serialisation discussed below.

We first discuss SIL, then its use in actual data models. At least skim over the next section – it sucks to discover the SIL grammar by trial and error.

Old-style STC annotation is not discussed here. If you still want to do it (and for now, you have to if you want any STC annotation – sigh), check out the terse discussion in the tutorial

Annotation Using SIL¶

Data model annotation in DaCHS is done using SIL, the Simple Instance Language. It essentially resembles JSON, but all delimiters not really necessary for our use case have been dropped, and type annotation has been added.

The elements of SIL are:

The NULL literal, __NULL__. Attributes that are set to this literal are elided. This is mostly useful in the connection with data model annotation within mixins.
Atomic Values. For SIL, everything is a string (it’s a problem of DM validation to decide otherwise). When your string consists exclusively of alphanumerics and [._-], you can just write it in SIL. Otherwise, you must use double quotes. as in SQL, write two double quotes to include a literal double quote. So, valid literals in SIL are
- 2.3e-3
- red
- "white and blue"
- """Yes,"" the computer said."
- "could write (type:foo) {bar. baz} here" (elements of SIL are protected in quoted literals)
Invalid literals include:
- http://www.g-vo.org (: and / may not occur in literals)
- red, white and blue (no blanks and commas)
- 22" (no single quotes)
Plain Identifiers. These are C-like identifiers (a letter or an underscore optionally followed by letters, number or underscores).
Comments. SIL comments are classical C-style comments (/*...*/). They don’t nest yet, but they probably will at some point, so don’t write /* within a comment.
Object annotation. This is like a dictionary; only plain identifiers are allowed as keys. So, an object looks like this:
```
{ foo: bar
  longer: "This is a value with blanks in it"
}
```
Note again that no commas or quotes around the keys are necessary (or even allowed).
Sequences. This is like a list. Members can be atomic or objects, but they have to be homogeneous (SIL doesn’t enforce this by grammatical means, though. Here is an object with two sequences:
```
{ seq1: [3 4 5 "You thought these were numbers? They're strings!"]
  seq2: [
    { seq_index: 0 value: 3.3}
    { seq_index: 1 value: 0.3}
  ]
}
```
References. The point of SIL is to say things about column and param instances. Both of them (and other dm instances, tables, and in principle anything else in RDs) can be referenced from within SIL. A reference starts with an @ and is then a normal DaCHS cross identifier (columns and params within a table can be referenced by name only, columns take precedence on name clashes). If you use odd characters in your RD names or in-RD identifiers, think again: only [._/#-] are allowed in such references. Here is an object with some valid references:
```
{ long: @raj2000  /* a column in the enclosing table */
  lat: @dej2000
  system: @//systems#icrs /* could be a dm instance in a
    DaCHS-global RD;, this does *not* exist yet */
  source: @supercat/q#main /* perhaps a table in another RD */
}
```
Casting. You can (and sometimes have to) give explicit types in the SIL annotation. Types look like C-style casts. The root of a SIL annotation must always have a cast; that allows DaCHS to figure out what it is, which is essential for validation (and possibly inference of defaults and such). You can cast both single objects and sequences. Here’s an example that actually validates for DaCHS’ SIL (which the examples above wouldn’t because they’re missing the root annotation):
```
(testdm:testclass) {  /* cast on root: mandatory */
  attr1 {   /* no cast here; DaCHS can infer attr1's type if necessary */
    attr2: val
  }
  seq: (testdm:otherclass)[  /* Sequence cast: */
    {attr1: a}  /* all of these are now treated as testdm:otherclass */
    {attr1: b}
    {attr1: c}]}
```

Photometry annotation¶

To produce photcal groups as per the 2020 Timeseries Note and perhaps later specs, use an annotation like this:

<dm>
  (phot:PhotCal) {
    filterIdentifier: "Gaia/G"
    zeroPointFlux: \zeroPointFlux
    magnitudeSystem: Vega
    effectiveWavelength: \effectiveWavelength
    value: @phot
  }
</dm>

– where phot would be the column containing the photometry.

When – as is most likely when you want to have the – you are using the //timeseries#phot-0 mixin, you would only give such a group when you have multiple photometry systems in one light curve (which you should avoid). For the first photometry column, this declaration is done by the mixin.

GeoJSON annotation¶

To produce GeoJSON output (as supported by DaCHS’ TAP implementation), DaCHS needs to know what the “geometry“ in the sense of GeoJSON is. Furthermore, DaCHS keeps supporting declaring reference systems in the crs attribute, as the planetology community uses it.

The root class of the geojson DM is geojson:FeatureCollection. It has up to two attributes (crs and feature), closely following the GeoJSON structure itself. The geometry is defined in feature’s geometry attribute. All columns not used for geometry will end up in GeoJSON properties.

So, a complete GeoJSON annotation, in this case for an EPN-TAP table, could look like this:

<table>
   <dm>
    (geojson:FeatureCollection){
      crs: (geojson:CRS) {
          type: name
          properties: (geojson:CRSProperties) {
            name: "urn:x-invented:titan"}}}}
      feature: {
        geometry: {
          type: sepsimplex
          c1min: @c1min
          c2min: @c2min
          c1max: @c1max
          c2max: @c2max }}}
  </dm>

  <mixin
    spatial_frame_type="body"/>
</table>

Yes, the use type attributes is a bit of an abomination, but we wanted the structure to follow GeoJSON in spirit.

The crs attribute could also be of type link, in which case the properties would have attributes href and type; we’re not aware of any applications of this in planetology, though. crs is optional (but standards-compliant GeoJSON clients will interpret your coordinates as WGS84 on Earth if you leave it out).

For geometry, several values for type are defined by DaCHS, depending on how the GeoJSON geometry should be constructed from the table. Currently defined types include (complain if you need something else, it’s not hard to add):

sepcoo – this is for a spherical point with separate columns for the two axes. This needs latitude and longitude attributes, like this:

<dm>
  (geojson:FeatureCollection){
    feature: {
      geometry: {
        type: sepcoo
        latitude: @lat
        longitude: @long }}}
</dm>

seppoly – this constructs a spherical polygon out of column references. These have the form c_n_m, where m is 1 or 2, and n is counted from 1 up to the number of points. DaCHS will stop collecting points as soon as it doesn’t find an expected key. If you find yourself using this, check your data model. An example:

<dm>
  (geojson:FeatureCollection){
    feature: {
      geometry: {
        type: seppoly /* a triangle of some kind */
        c1_1: @rb0
        c1_2: @rb1
        c2_1: @lb0
        c2_2: @lb1
        c3_1: @t0
        c3_2: @t1 }}}
</dm>

sepsimplex – this constructs a spherical box-like thing from minimum and maximum values. It has c[12](min|max) keys as in EPN-TAP. As a matter of fact, a fairly typical annotation for EPN-TAP would be:

<dm>
  (geojson:FeatureCollection){
    feature: {
      geometry: {
        type: sepsimplex
        c1min: @c1min
        c2min: @c2min
        c1max: @c1max
        c2max: @c2max }}}
</dm>

geometry – this constructs a geometry from a pgsphere column. Since GeoJSON doesn’t have circles, only spoint and spoly columns can be used. They are referenced from the value key. For instance, obscore and friends could use:
```
<dm>
  (geojson:FeatureCollection) {
    feature:  {
        geometry:  {
          type: geometry
          value: @s_region }}}
  </dm>
```

DaCHS’ Service Interface ¶

Even though normal users should rarely be confronted with too many of the technical details of request processing in DaCHS, it helps to have a rough comprehension in order to understand several user-visible details.

In DaCHS’ architecture, a service is essentially a combination of a core and a renderer. The core is what actually does the query or the computation, the renderer adapts input and outputs to what a protocol or interface expects. While a service always has exactly one core (could be a nullCore, though), it can support more than one renderer, although the parameters in all renderers are, within reason, about the same, within reason.

However, parameters on a form interface will typcially be interpreted differently from a VO interface on the same core. For instance, ranges on the form interface are written as 1 .. 3 (VizieR compliance), on an SSA 1.x interface 1/3 (“PQL” prototype), and on a datalink dlget interface “1 2” (DALI 1.1 style). The extreme of what probably still makes sense is the core search core that replaces SCS’s RA, DEC, and SR with an entirely different set of parameters perhaps better suited for interactive, browser-based usage.

Cores communicate their input interface by defining an input table, which is essentially a sequence of input keys, which in turn essentially work like params: in particular, they have all the standard metadata like units, ucds, etc. Input tables, contrary to what their name might suggest, have no rows. They can hold metadata, though, which is sometimes convenient to pass data between parameter parsers and the core.

When a request comes in, the service first determines the renderer responsible. It then requests an inputTable for that renderer from the core. The core, in turn, will map each inputKey in its inputTable through a renderer adaptor as returned from svcs.inputdef.getRendererAdaptor; this inspects the renderer.parameterStyle, which must be taken from the svcs.inputdef._RENDERER_ADAPTORS’ keys (currently form, pql, dali). inputKeys have to have the adaptToRenderer property set to True to have them adapted. Most automatically generated inputKeys have that; where you manually define inputKeys, you would have to set the property manually if you want that behaviour (and know that you want it; outside of table-based cores, it is unikely that you do).

Core Args¶

The input table, together with the raw arguments coming from the client, is then used to build a svcs.CoreArgs instance, which in turn takes the set of input keys to build a context grammar. The core args have the underlying input table (with the input keys for the metadata) in the inputTD attribute, the parsed arguments in the dictionary args.

For each input key args maps its name to a value; context grammars are case-semisensitive, meaning that case in the HTTP parameter names is in general ignored, but if a parameter name matching case is found, it is preferred. Yes, ugly, but unfortunately the VO has started with case-insensitive parameter names. Sigh.

The values in args are a bit tricky:

each raw parameter given must parse with a single inputKey’s parse. For instance, if an inputKey is a real[2], it will be parsed as a float array.
if no raw parameter is given for an input key, its value will be None.
when an inputKey specifies multiplicity=”multiple”, the non-None value in the core args is a list. Each list item is something that came out of the inputKey’s parser (i.e., it could be another list for array-valued parameters).
when an inputKey specifies multiplicity=”single”, the value in the core args is a single value of whatever inputKey parses (or None for missing parameters). This is even true when a parameter has been given multiple times; while currently, the last parameter will win, we don’t guarantee that.
when an inputKey specifies multiplicity=”force-single”, DaCHS works as in the single case, except that multiple specification will lead to an error.
when an inputKey does not specify multiplicity, DaCHS will infer the desired multiplicity from various hints; essentially, enumerated parameters (values/options given in some way) have multiplicity multiple, everything else multiplicity single. It is wise not to rely on this behaviour.

These rules are independent of the type of core and hold for pythonCores or whatever just as for the normal, table-based cores. For these (and they are what users are mostly concerned with), special rules and shortcuts apply, though.

Table-based cores¶

Conddescs and input keys: Defining the input parameters¶

You will usually deal with cores querying database tables – dbCore, ssapCore, etc. For these, you will not normally define an inputTable, as it is being generated by the software from condDescs.

To create simple constraints, just buildFrom the columns queried:

<condDesc buildFrom="myColumn"/>

(the names are resolved in the core’s queried table). DaCHS will automatically adapt the concrete parameter style is adapted to the renderer – in the web interface, there are vizier-like expressions, in protocol interfaces, you get fields understanding expressions, either as in SSAP (for the pql parameter style) or as defined in DALI (the dali parameter style).

This will generate query fields that work against data as stored in the database, with some exceptions (columns containing MJDs will, for example, be turned into VizieR-like date expressions for web forms).

Since in HTML forms, astronomers often ask for odd units and then want to input them, too, DaCHS will also honor the displayUnit display hint for forms. for instance, if you wrote:

<table id="ex1">
  <column name="minDist"
    unit="deg"
    displayHint="displayUnit=arcsec"/>
  ...

<dbCore queriedTable="ex1">
  <condDesc buildFrom="minDist"/>
  ...

then the form renderer would declare the minDist column to take its values in arcsecs and do the necessary conversions, while minDist would properly work with degrees in SCS or TAP.

For object lists and similar, it is frequently desirable to give the possible values (unless there are too many of those; these will be translated to option lists in forms and to metadata items for protocol services and hence be user visible). In this case, you need to change the input key itself. You can do this by deriving the input key from the column and assign it to a condDesc, like this:

<condDesc>
   <inputKey original="source">
     <values fromdb="source from plc.data"/>
   </inputKey>
 </condDesc>

Use the showItems="n" attribute of input keys to determine how many items in the selector are shown at one time.

If you want your service to fail if a parameter is not given, declare the condDesc as required:

<condDesc buildFrom="myColumn" required="True"/>

(you can also declare individual an inputKey as required).

If, on the other hand, you want DaCHS to fill in a default if the user provides no value, give a default to the input key using the values child:

<condDesc>
  <inputKey original="radius">
    <values default="0.5"/>
  </inputKey>
<condDesc>

Sometimes a parameter shouldn’t be defaulted in a protocol request (perhaps to satisfy an external contract), while the web interface should pre-fill a sensible choice. In that case, use the defaultForForm property:

<condDesc>
  <inputKey original="radius">
    <property key="defaultForForm">0.5</property>
  </inputKey>
<condDesc>

DaCHS will also interpret min and max attributes on the input keys (and the columns they are generated from) to generate input hints; that’s a good way to fight the horror vacui users have when there’s an input box and they have no idea what to put there. The best way to deal with this, however, is to not change the input keys but the columns themselves, as in:

<table id="ex1">
  <column name="mjd" type="double precision"
    ...>
    <values min="" max=""/>
...
<dbCore queriesTable="ex1">
  <condDesc buildFrom="mjd"/>

You will typically leave min and max empty and run:

dachs limits q#ex

when the table contents change; this will make DaCHS update the values in the RD itself.

Phrasemakers: Making custom queries¶

CondDescs will generate SQL adapted to the type of their input keys, which; as you can imagine, for cases like the VizieR expressions, that’s not done in a couple of lines. However, there are times when you need custom behaviour. You can then give your conddescs a phraseMaker, a piece of python code generating a query and adding parameters:

<condDesc>
   <inputKey original="confirmed" multiplicity="single">
     <property name="adaptToRenderer">False</property>
   </inputKey>
   <phraseMaker>
     <code>
       if inPars.get(inputKeys[0].name, False):
         yield "confirmed IS NOT NULL"
     </code>
   </phraseMaker>
 </condDesc>

PhraseMakers work like other code embedded in RDs (and thus may have setup). inPars gives a dictionary of the input parameters as parsed by the inputDD according to multiplicity. inputKeys contains a sequence of the conddesc’s inputKeys. By using their names as above, your code will not break if the parameters are renamed.

It is usually a good idea to set the property adaptToRenderer to False in such cases – you generally don’t want DaCHS to use its standard rules for input key adaptation as discussion above because that will typically change what ends up in inPars and hence break your code for some renderers.

Note again that parameters not given will have the value None throughout. The will be present in inPars, though, so do not try things like "myName" in inPars – that’s always true.

Phrase makers must yield zero or more SQL fragments; multiple SQL fragments are joined in conjunctions (i.e., end up in ANDed conditions in the WHERE clause). If you need to OR your fragments, you’ll have to do that yourself. Use the base.joinOperatorExpr(operator, operands) for robustness to construct ORs.

Since you are dealing with raw SQL here, never include material from inPars directly in the query strings you return – this would immediately let people do SQL injections at least when the input key’s type is text or similar. Instead, use the getSQLKey function as in this example:

<condDesc>
  <inputKey original="hdwl" multiplicity="single"/>
  <phraseMaker>
    <code>
      ik = inputKeys[0]
      destRE = "^%s\\.[0-9]*$"%inPars[ik.name]
      yield "%s ~ (%%(%s)s)"%(ik.name,
        base.getSQLKey("destRE", destRE, outPars))
    </code>
  </phraseMaker>
</condDesc>

getSQLKey takes a suggested name, a value and a dictionary, which within phrase makers always is outPars. It will enter value with the suggested name as key into outPars or change the suggested name if there is a name clash. The generated name will be returned, and that is what is entered in the SQL statement.

The outPars dictionary is shared between all conddescs entering into a query. Hence, if you do anything with it except passing it to base.getSQLKey, you’re voiding your entire warranty.

Here’s how to define a condDesc doing a full text search in a column:

<condDesc>
  <inputKey original="source" description="Words from the catalog
    description, e.g., author names or title words.">
    <property name="adaptToRenderer">False</property>
  </inputKey>
  <phraseMaker>
    <code>
      yield ("to_tsvector('english', source)"
        " @@ plainto_tsquery('english', %%(%s)s)")%(
      base.getSQLKey("source", inPars["source"], outPars))
    </code>
  </phraseMaker>
</condDesc>

Incidentally, this would go with an index definition like:

<index columns="source" method="gin"
  >to_tsvector('english', source)</index>

Grouping Input Keys¶

For special effects, you can group inputKeys. This will make them show up under a common label and in a single line in HTML forms. Other renderers currently don’t do anything with the groups.

Here’s an example for a simple range selector:

<condDesc>
  <inputKey name="el" type="text" tablehead="Element"/>

  <inputKey name="mfmin" tablehead="Min. Mass Fraction \item">
    <property name="cssClass">a_min</property>
  </inputKey>

  <inputKey name="mfmax" tablehead="Max. Mass Fraction \item">
    <property name="cssClass">a_max</property>
  </inputKey>

  <group name="mf">
    <description>Mass fraction of an element. You may leave out
      either upper or lower bound.</description>
    <property name="label">Mass Fraction between...</property>
    <property name="style">compact</property>
  </group>
</condDesc>

You will probably want to style the result of this effort using the service element’s customCSS property, maybe like this:

<service...>
  <property name="customCSS">
    input.a_min {width: 5em}
    input.a_max {width: 5em}
    input.formkey_min {width: 6em!important}
    input.formkey_max {width: 6em!important}
    span.a_min:before { content:" between "; }
    span.a_max:before { content:" and "; }
    tr.mflegend td {
      padding-top: 0.5ex;
      padding-bottom: 0.5ex;
      border-bottom: 1px solid black;
    }
  </property>
</service>

See also the entries on multi-line input, selecting input fields with a widget, and customizing generated SCS conditions in DaCHS’ howto document.

Output tables¶

When determining what columns to include in a response from a table-based core, DaCHS follows relatively complicated rules because displays in the browser and almost anywhere else are subject to somewhat different constraints. In the following, when wie talk about “VOTable”, we refer to all tabular formats produced by DaCHS (FITS binary, CSV, TSV…).

The column selection is influenced by:

Verbosity. This is controlled by the VERB parameter (1..3) or preferentially verbosity (1..30). Only columns with verbLevel not exceeding verbosity (or, if not given, VERB*10) are included in the result set. This, in particular, means that columns with verbLevel larger than 30 are never automatically included in output tables (but they can be manually selected for HTML using _ADDITEM).
Output Format. While VOTable takes the core’s output table and apply the verbosity filter, HTML uses the service’s output table as the basis from which to filter columns. On the other hand, in HTML output the core output table is used to create the list of potential additional columns.
votableRespectsOutputTable. This is a property on services that makes DaCHS use the service’s output table even when generating VOTable output if it is set to True. Write:
```
<property name="votableRespectsOutputTable">True</property>
```
in your service element to enable this behaviour.
_ADDITEM. This parameter (used by DaCHS’ web interface) lets users select columns not selected by the current settings or the service’s output table. _ADDITEM is ignored in VOTable unless in HTML mode (which is used in transferring web results via SAMP).
noxml. Columns can be furnished with a displayHint="noxml=true", and they will never be included in VOTable output; use this when you use complex formatters to produce HTML displays.
_SET. DaCHS supports “column sets”, for instance, to let users select certain kinds of coordinates. See apfs/res/apfs_new.rd` for an example. Essentially, when defining an output table, each output field gets a sets attribute (default: no set; use ALL to have the column included in all outputs). Then, add a _SET service parameter (use values do declare the available sets). Note that the _SET parameter changes VOTable column selection to votableRespectsOutputTable mode as discussed above. Services that use column sets should therefore set the property manually for consistency whether or not clients actually pass _SET.

Sorry for this mess; all this had, and by and large still has, good reasons.

The Vanity Map¶

DaCHS’ URL scheme leads to somewhat clunky URLs that, in particular, reflect the file system underneath. While this doesn’t matter to the VO registry, it is possibly unwelcome when publishing URLs outside of the VO. To overcome it, you can define “vanity names”, single path elements that are mapped to paths.

These mappings are read from the file $GAVO_ROOT/etc/vanitynames.txt. The file contains lines of the format:

<target> <key> [<option>]

Target is a path that must not include nevowRoot and must not start with a slash (unless you’re going for very special effects).

Key normally is a single path element (i.e., a string without a slash). If this path element is found in the first segment, it is replaced with the segments in target.

<option> can only be !redirect or empty right now.

If it is !redirect, <key> may be a path fragment (as opposed to a single path element); leading and trailing slashes are ignored. If the enire query path matches this key, a redirect to this key is generated. This is intended to let you shut down services and introduce replacements. If the incoming URL contains a query, it will be appended to the replacement URL. Thus, even stored queries or forms can potentially work across such a redirect.

You can also (ab)use the redirect option to give vanity names, but since the target will show up in the browser address line, normal maps are highly preferred. The only time normal maps don’t work for this is when the resource directory is identical to the vanity name (you’ll get an endless loop then), so you should avoid that situation.

Empty lines and #-on-a-line-comments are allowed in the input.

As an example, here’s the vanity map that DaCHS had builtin as of version 2.1:

__system__/products/p/get getproduct
__system__/products/p/dlasync datalinkuws
__system__/services/registry/pubreg.xml oai.xml
__system__/services/overview/external odoc
__system__/dc_tables/show/tablenote tablenote
__system__/dc_tables/show/tableinfo tableinfo
__system__/services/overview/admin seffe
__system__/services/overview/rdinfo browse
__system__/tap/run/tap tap
__system__/adql/query/form adql !redirect
__system__/run/genrd genrd

Note again that <key> must be a single path element only.

Writing Custom Cores ¶

While DaCHS provides cores for many common operations – in particular, database queries and wrapped external binaries –, there are of course services that need to do things not covered by what the shipped cores do. A common case is wrapping external binaries.

Many such cases still follow the basic premise of services: GET or POST parameters in, something table-like out. You should then use custom cores, which then still let you use normal DaCHS renderers (in particular form and api/sync). When that doesn’t cut it, you’ll need to use a custom renderer.

While a custom core is defined in a separate module – this also helps debugging since you can run it outside of DaCHS –, there’s also the python core that keeps the custom code inside of the RD. This is very similar; Python Cores instead of Custom Cores explains the differences.

The following exposition is derived from the times service in the GAVO data center, a service wrapping some FORTRAN code wrapping SOFA (yes, we’re aware that we would directly use SOFA through astropy; that’s not the point here). Check out the sources at http://svn.ari.uni-heidelberg.de/svn/gavo/hdinputs/apfs; the RD is times.rd.

Defining a Custom Core¶

In an RD, a custom core is very typically just written with a reference to a defining module:

<customCore module="res/timescore"/>

The path is relative to the resdir, and you don’t include the module’s extension (DaCHS uses normal python module resolution, except for temporarily extending the search path with the enclosing directory). You can, in principle, declare the core’s interface in that element, but that’s typically not a good idea (see below).

The above declaration means you will find the core itself in res/timescore.py.

Ideally, you’ll just use the DaCHS API in the core, since we try fairly hard to keep that api constant. The timescore doesn’t quite follow that rule because it wants to expand VizieR expressions, which normal services probably won’t do.

DaCHS expects the custom core under the name Core. Thus, the centerpiece of the module is:

from gavo import api

class Core(api.Core):

The core needs an InputTable and an OutputTable like all cores. You could define it in the resource descriptor like this:

<customCore id="createCore" module="bin/create">
  <inputTable>
    <inputKey .../>
  </inputTable>
  <outputTable>
    <column name="itemsAdded" type="integer" tablehead="Items added"/>
  </outputTable>
</customCore>

It’s preferable to define at least the input in the code, though, since it’s more likely to be kept in sync with the code in that case. Embedding the definitions is done using the class attribute inputTableXML:

class Core(core.Core):
  inputTableXML = """
    <inputTable>
      <inputKey name="ut1" type="vexpr-date" multiplicity="single"
        tablehead="UT1"
        description="Date and time (UT1)" ucd="time.epoch;meta.main"/>
      <inputKey name="interval" type="integer" multiplicity="single"
        tablehead="Interval"
        unit="s" ucd="time.interval"
        description="Interval between two sets of computed values"
      >3600</inputKey>
    </inputTable>
  """

There is also outputTableXML, which you should use if you were to compute stuff in some lines of Python, since then the fields are directly defined by the core itself.

However, the case of timescore is fairly typical: There is some, essentially external, resource that produces something that needs to be parsed. In that case, it’s a better idea to define the parsing logic in a normal RD data item. Its table then is the output table of the core. In the times example, the output of timescompute is described by the build_result data item in times.rd:

<table id="times">
  <column name="ut1" type="timestamp" tablehead="UT1"
    ucd="time.epoch;meta.main" verbLevel="1"
    description="Time and date (UT1)" displayHint="type=humanDate"/>
  <column name="gmst" type="time" tablehead="GMST"
    verbLevel="1" description="Greenwich mean sidereal time"
    xtype="adql:TIMESTAMP" displayHint="type=humanTime,sf=4"/>
  <column name="gast" type="time" tablehead="GAST"
    verbLevel="1" description="Greenwich apparent sidereal time"
    xtype="adql:TIMESTAMP" displayHint="type=humanTime,sf=4"/>
  <column name="era" type="double precision" tablehead="ERA"
    verbLevel="1" description="Earth rotation angle"
    displayHint="type=dms,sf=3" unit="deg"/>
</table>

<data id="build_result" auto="False">
  <reGrammar>
    <names>ut1,gmst,gast,era</names>
  </reGrammar>
  <make table="times">
    <rowmaker>
      <map dest="gmst">parseWithNull(@gmst, parseTime, "None")</map>
      ...
    </rowmaker>
  </make>
</data>

So, the core needs to say “my output table has the structure of #times”.

As usual with DaCHS structures, you should not override the constructor, as it is defined by a metaclass. Instead, Cores call, immediately after the XML parse (technically, as the first thing of their completeElement method), a method called initialize. This is where you should set the output table. For the times core, this looks like this:

def initialize(self):
  self.outputTable = api.OutputTableDef.fromTableDef(
    self.rd.getById("times"), None)

Of course, you are not limited to setting the output table there; as initialize is only called once while parsing, this is also a good place to perform expensive, one-time operations like reading and parsing larger external resources.

Giving the Core Functionality¶

To have the core do something, you have to override the run method, which has to have the following signature:

run(service, inputTable, queryMeta) -> stuff

The stuff returned will usually be a Table or Data instance (that need not match the outputTable definition – the latter is targeted at the registry and possibly applications like output field selection). The standard renderers also accept a pair of mime type and a string containing some data and will deliver this as-is. With custom renderers, you could return basically anything you want.

Services come up with some idea of the schema of the table they want to return and adapt tables coming out of the core to this. Sometimes, you want to suppress this behaviour, e.g., because the service’s ideas are off. In that case, set a noPostprocess attribute on the table to any value (the TAP core does this, for instance).

In service you get the service using the core; this may make a difference since different services can use the same core and could control details of its operations through properties, their output table, or anything else.

The inputTable argument is the CoreArgs instance discussed in Core Args. Essentially, you’ll usually use its args attribute, a dictionary mapping the keys defined by your input table to values or lists of them.

The queryMeta argument is discussed in Database Options.

In the times example, the parameter interpretation is done in an extra function (which helps testability when there’s a bit more complex things going on):

def computeDates(args):
  """yields datetimes at which to compute times from the ut1/interval
  inputs in coreArgs args.
  """
  interval = args["interval"] or 3600
  if args["ut1"] is None:
    yield datetime.datetime.utcnow()
    return

  try:
    expr = vizierexprs.parseDateExpr(args["ut1"])
    if expr.operator in set([',', '=']):
      for c in expr.children:
        yield c

    elif expr.operator=='..':
      for c in expandDates(expr.children[0],
          expr.children[1], interval):
        yield c

    elif expr.operator=="+/-":
      d0, wiggle = expr.children[0], datetime.timedelta(
        expr.expr.children[1])
      for c in expandDates(d0-wiggle, d0+wiggle):
        yield c

    else:
      raise api.ValidationError("This sort of date expression"
        " does not make sense for this service", colName="ut1")
  except base.ParseException, msg:
    raise api.ValidationError(
      "Invalid date expression (at %s)."%msg.loc,
      colName="ut1")

While the details of the parameter parsing and expansion don’t really matter, note now exceptions are mapped to a ValidationError and give a colName – this lets the form renderer display error messages next to the inputs that caused the failure.

The next thing timescore does is build some input, which in this case is fairly trivial:

input = "\n".join(utils.formatISODT(date) for date in dates)+"\n"

If your input is more complex or you need input files or similar, you want to be a bit more careful. In particular, do not change directory (or, equivalently, use the utils.sandbox context manager); this may confuse the server, and in particular will break the first time two requests are served simultaneously: The core runs within the main process, and that can only have one current directory.

Instead, in such situations, make a temporary directory and manually place your inputs in there. The spacecore (http://svn.ari.uni-heidelberg.de/svn/gavo/hdinputs/sp_ace/res/spacecore.py) shows how this could look like, including tearing the stuff down safely when done (the runSpace function).

For the timescore, that is not necessary; you just run the wrapped program using standard subprocess functionality:

computer = service.rd.getAbsPath("bin/timescompute")

pipe = subprocess.Popen([computer],
    stdin=subprocess.PIPE, stdout=subprocess.PIPE, close_fds=True,
    cwd=os.path.dirname(computer))
data, errmsg = pipe.communicate(input)
if pipe.returncode:
  raise api.ValidationError("The backend computing program failed"
    " (exit code %s).  Messages may be available as"
    " hints."%pipe.returncode,
    "ut1",
    hint=errmsg)

Note that with today’s computers, you shouldn’t need to worry about streaming input or output until they are in the dozens of megabytes (in which case you should probably think hard about a custom UWS and keep the files in the job’s working directories).

To turn the program’s output into a table, you use the data item defined in the RD:

return api.makeData(
  self.rd.getById("build_result"),
  forceSource=StringIO(data))

When the core defines the data itself, you would skip makeData. Just directly produce the rowdicts and make the output table directly from the rows:

rows = [{"foo": 3*i, "bar": 8*i} for i in range(30)]
return api.TableForDef(self.outputTable, rows=rows)

Database Options¶

The standard DB cores receive a “table widget” on form generation, including sort and limit options. To make the Form renderer output this for your core as well, define a method wantsTableWidget() and return True from it.

The queryMeta that your run method receives has a dbLimit key. It contains the user selection or, as a fallback, the global db/defaultLimit value. These values are integers.

So, if you order a table widget, you should do something like:

cursor.execute("SELECT .... LIMIT %(queryLimit)s",
  {"queryLimit": queryMeta["dbLimit"],...})

In general, you should warn people if the query limit was reached; a simple way to do that is:

if len(res)==queryLimit:
  res.addMeta("_warning", "The query limit was reached.  Increase it"
    " to retrieve more matches.  Note that unsorted truncated queries"
    " are not reproducible (i.e., might return a different result set"
    " at a later time).")

where res would be your result table. _warning metadata is displayed in both HTML and VOTable output, though of course VOTable tools will not usually display it.

Python Cores instead of Custom Cores¶

If you only have a couple of lines of python, you don’t have to have a separate module. Instead, use a python core. In it, you essentially have the run method as discussed in Giving the Core Functionality in a standard procApp. The advantage is that interface and implementation is nicely bundled together. The following example should illustrate the use of such python cores; note that rsc already is in the procApp’s namespace:

<pythonCore>
  <inputTable>
    <inputKey name="opre" description="Operand, real part"
      required="True"/>
    <inputKey name="opim" description="Operand, imaginary part"
      required="True"/>
    <inputKey name="powers" description="Powers to compute"
      type="integer" multiplicity="multiple"/>
  </inputTable>
  <outputTable>
    <outputField name="re" description="Result, real part"/>
    <outputField name="im" description="Result, imaginary part"/>
    <outputField name="log"
      description="real part of logarithm of result"/>
  </outputTable>

  <coreProc>
    <setup imports="cmath"/>
    <code>
      powers = inputTable.args["powers"]
      if not powers:
        powers = [1,2]
      op = complex(inputTable.args["opre"],
        inputTable.args["opim"])

      rows = []
      for p in powers:
        val = op**p
        rows.append({
          "re": val.real,
          "im": val.imag,
          "log": cmath.log(val).real})

      return api.TableForDef(self.outputTable, rows=rows)
    </code>
  </coreProc>
</pythonCore>

Regression Testing ¶

Introduction¶

Things break – perhaps because someone foolishly dropped a database table, because something happened in your upstream, because you changed something or even because we changed the API (if that’s not mentioned in Changes, we owe you a beverage of your choice). Given that, having regression tests that you can easily run will really help your peace of mind.

Therefore, DaCHS contains a framework for embedding regression tests in resource descriptors. Before we tell you how these work, some words of advice, as writing useful regression tests is an art as much as engineering.

Don’t overdo it. There’s little point in checking all kinds of functionality that only uses DaCHS code – we’re running our tests before committing into the repository, and of course before making a release. If the services just use condDescs with buildFrom and one of the standard renderers, there’s little point in testing beyond a request that tells you the database table is still there and contains something resembling the data that should be there.

Don’t be over-confident. Just because it seems trivial doesn’t mean it cannot fail. Whatever code there is in the service processing of your RD, be it phrase makers, output field formatters, custom render or data functions, not to mention custom renderers and cores, deserves regression testing.

Be specific. In choosing the queries you test against, try to find something that won’t change when data is added to your service, when you add input keys or when doing similar maintenance-like this. Change will happen, and it’s annoying to have to fix the regression test every time the output might legitimately change. This helps with the next point.

Be pedantic. Do not accept failing regression tests, even if you think you know why they’re failing. The real trick with useful testing is to keep “normal” output minimal. If you have to “manually” ignore diagnostics, you’re doing it wrong. Also, sometimes tests may fail “just once”. That’s usually a sign of a race condition, and you should really try to figure out what’s going on.

Make it fail first. It’s surprisingly easy to write no-op tests that run but won’t fail when the assertion you think you’re making is no longer true. So, when developing a test, assert something wrong first, make sure there’s some diagnostics, and only then assert what you really expect.

Be terse. While in unit tests it’s good to test for maximally specific properties so failing unit tests lead you on the right track as fast as possible, in regression tests there’s nothing wrong with plastering a number of assertions into one test. Regression tests actually make requests to a web server, and these are comparatively expensive. The important thing here is that regression testing is fast enough to let you run them every time you make a change.

Writing Regression Tests¶

DaCHS’ regression testing framework is organized a bit along the lines of python’s unittest and its predecessors, with some differences due to the different scope.

So, tests are grouped into suites, where each suite is contained in a regSuite element. These have a (currently unused) title and a boolean attribute sequential intended for when the tests contained must be executed in the sequence specified and not in parallel. It defaults to false, which means the requests are made in random order and in parallel, which speeds up the test runs and, in particular, will help uncover race conditions.

On the other hand, if you’re testing some sort of interaction across requests (e.g., make an upload, see if it’s there, remove it again), this wouldn’t work, and you must set sequential=”True”. Keep these sequential suites as short as possible. In tests within such suites (and only there), you can pass information from one test to the following one by adding attributes to self.followUp (which are available as attributes of self in the next test). If you need to manipulate the next URL, it’s at self.followUp.url.content_. For the common case of a redirect to the url in the location header (or a child thereof), there’s the pointNextToLocation(child="") method of regression tests. In the tests that are manipulated like this, the URL given in the RD should conventionally be overridden in the previous test. Of course, additional parameters, httpMethods, etc, are still applied in the manipulated url element.

Regression suites contain tests, represented in regTest elements. These are procDefs (just like, e.g., rowmakery apply), so you can have setup code, and you could have a library of parametrizable regTests procDefs that you’d then turn into regTests by setting their parameters. We’ve not found that terribly useful so far, though.

You must given them a title, which is used when reporting problems with them. Otherwise, the crucial children of these are url and, as always with procDefs, code.

Here are some hints on development:

Give the test you’re just developing an id; at the GAVO DC, we’re usually using cur; that way, we run variations of dachs test rdId#cur, and only the test in question is run.
After defining the url, just put an assert False into the test code. Then run dachs test -Devidence.xml rdId#cur or similar. Then investigate evidence.xml (possibly after piping through xmlstarlet fo) for stable and strong indicators that things are working.
If you get a BadCode for a test you’re just writing, the message may not always be terribly helpful. To see what’s actually bugging python, run dachs --debug test ... and check dcInfos.

RegTest URLs¶

The url element encapsulates all aspects of building the request. In the simplest case, you just can have a simple URL, in which case it works as an attribute, like this:

<regTest title="example" url="svc/form">
  ...

URLs without a scheme and a leading slash are interpreted relative to the RD’s root URL, so you’d usually just give the service id and the renderer to be applied. You can also specify root-relative and fully specified URLs as described in the documentation of the url element.

White space in URLs is removed, which lets you break long URLs as convenient.

You could have GET parameters in this URL, but that’s inconvient due to both XML and HTTP escaping. So, if you want to pass parameters, just give them as attributes to the element:

<regTest title="example">
  <url RA="10" DEC="-42.3" SR="1" parSet="form">svc/form</url>

The parSet=form here sets up things such that processing for the form renderer is performed – our form library nevow formal has some hidden parameters that you don’t want to repeat in every URL.

To easily translate URLs taken from a browser’s address bar or the form renderer’s result link, you can run dachs totesturl and paste the URLs there. Note that totesturl fails for values with embedded quotes, takes only the first value of repeated parameters and is a over-quick hack all around. Patches are gratefully accepted.

The url element hence accepts arbitrary attributes, which can be a trap if you think you’ve given values to url’s private attributes and mistyped their names. If uploads or authentication don’t seem to happen, check if your attribute ended up the in the URL (which is displayed with the failure message) and fix the attribute name; most private url attributes start with http. If you really need to pass a parameter named like one of url’s private attributes, pass it in the URL if you can. If you can’t because you’re posting, spank us. After that, we’ll work out something not too abominable .

If you have services requiring authentication, use url’s httpAuthKey attribute. We’ve introduced this to avoid having credentials in the RD, which, after all, should reside in a version control system which may be (and in the case of GAVO’s data center is) public. The attribute’s value is a key into the file ~/.gavo/test.creds, which contains, line by line, this key, a username and a password, e.g.:

svc1 testuser notASecret
svc2 regtest NotASecretEither

A test using this would look like this:

<regTest title="Authenticated user can see the light">
  <url httpAuthKey="svc1">svc1/qp/light.txt</url>
  <code>
    self.assertHTTPStatus(200)
  </code>
</regTest>

By default, a test will perform a GET request. To change this, set the httpMethod attribute. That’s particularly important with uploads (which must be POSTed).

For uploads, the url element offers two facilities. You can set a request payload from a file using the postPayload attribute (the path is interpreted relative to the resource directory), but it’s much more common to do a file upload like browsers do them. Use the httpUpload element for this, as in:

<url> <httpUpload name="UPLOAD"
  fileName="remote.txt">a,b,c</httpUpload> svc1/async </url>

(which will work as if the user had selected a file remote.txt containing “a,b,c” in a browser with a file element named UPLOAD), or as in:

<url>
  <httpUpload name="UPLOAD" fileName="remote.vot"
    source="res/sample.regtest"/>
  svc1/async
</url>

(which will upload the file referenced in source, giving the remote server the filename remote.vot). The fileName attribute is optional.

Finally, you can pass arbitrary HTTP headers using the httpHeader element. This has an attribute key; the header’s value is taken from the element content, like this:

<url postPayload="res/testData.regtest" httpMethod="POST">
  <httpHeader key="content-type">image/jpeg</httpHeader>
  >upload/custom</url>

RegTest Tests¶

Since regression tests are just procDefs, the actual assertions are contained in the code child of the regTest. The code in there sees the test itself in self, and it can access

self.data (the response content as a byte string),
self.headers (a sequence of header name, value pairs; note that you should match the names case-insensitively here),
self.status (the HTTP response code),
self.requestTime (the time the URI request actually took; this will be None for failed requests), and
the URL actually retrieved in self.url.httpURL.

Incidentally, that last name is right; the regression framework only supports http, and it’s not terribly likely that we’ll change that.

You should probably only access those attributes in a pinch and instead use the pre-defined assertions, which are methods on the test objects as in pyunit – conventional assertions are clearer to read and less likely to break if fixes to the regression test API become necessary. If you still want to have custom tests, raise AssertionErrors to indicate a failure.

Here’s a list of assertion methods defined right now:

assertHTTPStatus(self, expectedStatus)

checks whether the request came back with expectedStatus.

assertHasStrings(self, *strings)

checks that all its arguments are found within content.

If string arguments are passed, they are utf-8 encoded before comparison. If that’s not what you want, pass bytes yourself.

assertHeader(self, key, value)

checks that header key has value in the response headers.

keys are compared case-insensitively, values are compared literally.

assertLacksStrings(self, *strings)

checks that all its arguments are not found within content.

assertValidatesXSD(self)

checks whether the returned data are XSD valid.

This uses DaCHS built-in XSD validator with the built-in schema files; it hence will in general not retrieve schema files from external sources.

assertXpath(self, path, assertions)

checks an xpath assertion.

path is an xpath (as understood by lxml), with namespace prefixes statically mapped; there’s currently v2 (VOTable 1.2), v1 (VOTable 1.1), v (whatever VOTable version is the current DaCHS default), h (the namespace of the XHTML elements DaCHS generates), m (the provisional MIVOT namespace) and o (OAI-PMH 2.0).<

DaCHS Reference Documentation¶

Resource Descriptor Element Reference¶

Element apply¶

Atomic Children¶

Structure Children¶

Element bind¶

Atomic Children¶

Element column¶

Atomic Children¶

Structure Children¶

Other Children¶

Element columnRef¶

Atomic Children¶

Element condDesc¶

Atomic Children¶

Structure Children¶

Element coverage¶

Atomic Children¶

Structure Children¶

Element customDF¶

Atomic Children¶

Element customRF¶

Atomic Children¶

Element data¶

Atomic Children¶

Structure Children¶

Other Children¶

Element DEFAULTS¶

Element dm¶

Atomic Children¶

Element EDIT¶

Atomic Children¶

Structure Children¶

Element events¶

Atomic Children¶

Structure Children¶

Element execute¶

Atomic Children¶

Structure Children¶

Other Children¶

Element foreignKey¶

Atomic Children¶

Element group¶

Atomic Children¶

Structure Children¶

Other Children¶

Element httpUpload¶

Atomic Children¶

Element ignoreOn¶

Atomic Children¶

Structure Children¶

Element ignoreSources¶

Atomic Children¶

Element index¶

Atomic Children¶

Element inputKey¶

Atomic Children¶

Structure Children¶

Other Children¶

Element job¶

Atomic Children¶

Structure Children¶

Element lateEvents¶

Atomic Children¶

Structure Children¶

Element macDef¶

Atomic Children¶

Element make¶

Atomic Children¶

Structure Children¶

Element map¶

Atomic Children¶

Element mixinDef¶

Atomic Children¶

Structure Children¶

Element mixinPar¶

Atomic Children¶

Element option¶

Atomic Children¶

Element outputField¶

DaCHS Reference Documentation ¶

Resource Descriptor Element Reference ¶