These are the sources for the IVOA vocabularies as specified in Vocabularies in the VO 2. This repository is maintained by the chairs of the Semantics Working Group; other persons usually have no reason to interact with it.

Vocabulary users would use the built vocabularies from http://www.ivoa.net/rdf. To propose changes, please follow the VEP process.

The rest of the README essentially is a crib sheet for the maintainers, and it is to be read together with Appendix A of the Vocabularies recommendation.

Repository Organisation

In this repository, each vocabulary is in a subdirectory of its own; the directory structure copies what is eventually on ivoa.net; vocabulary names with several segments ("datalink/core", "voresource/relationship") are supported but discouraged.

In each subdirectory, a file terms.csv contains the actual terms (there are exceptions, in particular for SKOS input).

Per-vocabulary metadata is given in a single file in the root of the repository called vocabs.conf. This, in particular, controls the versioning of the vocabularies. See below for a discussion of its format.

Operation

To build a single vocabulary, run:

python3 convert.py <vocabulary name>

where <vocabulary name> is one of the keys in vocabs.conf.

This will leave a deployable hierarchy for the vocabulary in the build subdirectory. For instance, after calling:

python3 convert.py refposition

there will be a subdirectory build/refposition, containing an htaccess configured for ivoa.net, and the actual vocabulary artefacts in a child named for the vocabulary version.

See below for actual deployment.

Defining Vocabularies

By default, our vocabularies are kept in CSV files. The CSV separator is ";". Since our descriptions will have lots of commas in them, this helps to keep the necessity for quoting low.

The columns or the CSVs are:

subject; level; label; description; more_relations

The fields mean:

term:	This is the actual, machine-readable vocabulary term. Please only use letters, digits, underscores, and dashes here. It is not intended for human consumption as such, although in the VO we prefer terms that make sense in English.
level:	This is used for simple input of wider/narrower relationships. It must be 1 for “root” terms. Terms with level of 2 that follow a root term becomes its children (“narrower”); you can nest, i.e., have terms of level 3 below terms of level 2. Note that this means the order of rows must be preserved in our CSV files: Do not sort them and then save them.
label:	This is a short, human-readable label for the term. In the VO, this is generally derived fairly directly from subject, ususally by inserting blanks at the right places and fixing capitalisation.
description:	This is a longer explanation of what the term means. We do not support any markup here, not even paragraphs, so there is probably a limit how much can be communicated here.
more_relations:	This column enables the declaration of non-hierarchical relationships and contains whitespace-separated declarations. Each declaration has the form property[(term)]. See below for the common properties supported here. Plain terms are references within the vocabulary, but CURIEs with known prefixes or full URIs are admitted, too.

Non-ASCII characters are allowed in label and description; files must be in UTF-8.

It is recommended to have, in addition to the terms.csv, a README in each vocabulary directory; this can contain internal notes, todo items, backwards compatibility considerations, and the like.

Defining Vocabulary Metadata

See Vocabularies in the VO 2, Appendix A.2.

In addition to what's there, in 2024-03, we're prototyping a topconcepts key, which is a space-separated list of vocabulary terms that should become objects in (vocab, skos:hasTopConcept, .) triples.

Deployment

To have a local installation, decide on a location from where you'll serve your vocabularies; since we currently generate apache .htaccess files only, this should be served by an apache HTTP server. Then, call make local with ROOT_URI set to that root URL, e.g.,:

python3 convert.py --root-uri http://localhost/rdf ALL

In order for the .htaccess configuration to work, the toplevel directory has to be configured to interpret FileInfo content in the apache configuration. This would look like this:

<Directory /var/www/docs/rdf>
      AllowOverride FileInfo
      Options Indexes FollowSymLinks
</Directory>

Also, mod_rewrite must be enabled.

If you are the administrator of the IVOA semantics repository, it is recommended to work like this:

1. Edit vocabs.conf, update the timestamp of the respective vocabulary to today (if applicable)
2. Do the edit to the vocabulary source; don't forget the ivoasem:preliminary property on newly added terms.
3. Run python3 convert.py <vocname>, paying attention to any diagnostics.
4. Inspect the files newly left in build as to whether they reflect the intended changes.
5. Commit the result. In the commit message, specify what caused the change ("mail by person@example.org"; "TCG telecon of 2020-02-11")
6. Run sshfs semantics@ivoa.info:rdf ivoa-repo (you'll have to mkdir ivoa-repo if it doesn't exist yet, of course)
7. Run python3 convert.py --dest-dir ivoa-repo <vocname> to update the vocabulary itself.
8. Update the vocabulary index: python3 make-rdf-index.py ivoa-repo
9. Unmount the vocabulary repo: fusermount -u ivoa-repo.
10. Update https://wiki.ivoa.net/twiki/bin/view/IVOA/VEPs as appropriate.
11. Run the validator from vocino2/validator on the new vocabulary: ../vocinvo/validator/vocvalidator.py http://www.ivoa.net/rdf/...

(Markus has a script rebuild.sh that does steps (6)-(9))