Implement bind_namespaces strategy for Prefix.cc

[cthoyt]

Summary of changes

This PR implements a web-dependent loader for prefix-namespace definitions from Prefix.cc by parsing its context document (https://prefix.cc/context.jsonld)

Demo

The Gene Ontology is a biomedical ontology describing biological processes, cellular locations, and cellular components.
It is typically abbreviated with the prefix "go" and uses PURLs issued by the Open Biological and Biomedical Ontologies (OBO) Foundry. Prefix.cc has an entry for GO that uses its preferred OBO PURL.

import rdflib

graph = rdflib.Graph(bind_namespaces="cc")

prefix_map = {prefix: str(ns) for prefix, ns in graph.namespaces()}
assert "go" in prefix_map
assert prefix_map["go"] == "http://purl.obolibrary.org/obo/GO_"
assert graph.qname("http://purl.obolibrary.org/obo/GO_0032571") == "go:0032571"

Checklist

• Checked that there aren't other open pull requests for
  the same change.
• Added tests for any changes that have a runtime impact.
• Checked that all tests and type checking passes.
• For changes that have a potential impact on users of this project:
  • Updated relevant documentation to avoid inaccuracies.
  • Considered adding additional documentation.
  • Considered adding an example in ./examples for new features.
  • Considered updating our changelog (CHANGELOG.md).
• Considered granting push permissions to the PR branch,
  so maintainers can fix minor issues and keep your PR up to date.

[aucampia]

One concern here is that there are some valid "@context" values [ref] that will not be handled correctly here, it may be that they don't appear in prefix.cc, but it may be good to confirm that somehow.

[cthoyt]

yes, it's the case that there are no special values in this context dictionary based on the way that the context is constructed (https://github.com/cygri/prefix.cc/blob/cbc85c00e59e00cf4fee697374109fdd9027231a/templates/format/jsonld.php) and the strict requirements on prefixes (though right now I am having a hard time finding where it's documented that these have to be lowercase strings of alphanumeric characters length <= 10)

[aucampia]

You will need to add cc to

rdflib/rdflib/_type_checking.py

Line 29 in 4aaaa6f

_NamespaceSetString = PyLiteral["core", "rdflib", "none"]

And also update

rdflib/rdflib/namespace/__init__.py

Lines 373 to 375 in 4aaaa6f

	* cc:
	* using prefix bindings from prefix.cc which is a online prefixes database
	* not implemented yet - this is aspirational

[aucampia]

Will try to get this into the next release, I will make any changes needed directly to your branch.

[cthoyt]

@aucampia thank you for the feedback! Feel free to make any edits you think are appropriate or request I make some updates.

[aucampia]

@cthoyt I will likely make another patch release soon, so will hold back a bit on merging this.

[aucampia]

This looks good to me, though I may add a warning that indicates that the "cc" selector is offered as a best-effort feature and may stop working if prefix.cc stops working or if they change their interface.

[niklasl]

Why use the TLD name for prefix.cc? cc is also the conventional prefix for creativecommons, so this might be confusing.

(I would also prefer having this as a utility (e.g. graph.bind_namespaces(util.get_prefix_cc()) rather than a flag to Graph. That would be more explicit and lets the user control network access and caching.)

[cthoyt]

@niklasl I am happy to do whatever the RDFLib team decides on, but this interface and nomenclature was already predefined, so I just filled it in with an implementation as suggested.

I agree that since prefix.cc relies on a network connection that this is a bit of a tricky situation.

[niklasl]

Actually, you can achieve the same effect using RDFLib as is, using just:

from rdflib import Graph

graph = Graph()
graph.parse('https://prefix.cc/context.jsonld')

for pfx, ns in graph.namespaces():
    print(pfx, ns)

This yields one very interesting difference though: the go prefix won't work, as it ends in a _, which is not treated as a namespace prefix in JSON-LD 1.1, since it does not en with a URI gen-delim (it has to be explicitly declared using "@prefix": true in the context to be treated as a prefix anyway).

[aucampia]

Since we had this in the docstring, I'm going to merge it (possibly with some minor changes), though possibly it would have been better in rdflib._contrib, in future I will channel functionality like this there instead.

[aucampia]

Since we had this in the docstring, I'm going to merge it (possibly with some minor changes), though possibly it would have been better in rdflib._contrib, in future I will channel functionality like this there instead.

Actually, I think it would be better to just move this into rdflib._contrib, some ways in which we can implement this is:

• Change type of bind_namespaces to also accept a mapping of prefix to string (similar to rdflib.namespace._NAMESPACE_PREFIXES_RDFLIB), add a function to rdflib._contrib that returns such a mapping for prefix.cc, this would then be used as follow:

  from rdflib._contrib import get_prefixcc_mapping
  from rdflib import Graph
  
  g = Graph(bind_namespaces=rdflib._contrib.get_prefixcc_mapping())

• Change type of bind_namespaces to also accept a function that returns a mapping of prefix to string

  from rdflib._contrib.namespace_providers import prefix_cc
  from rdflib import Graph
  
  g = Graph(bind_namespaces=prefix_cc)

  rdflib._contrib.namespace_providers.prefix_cc would then return the mapping when called.

Maybe something better can also be done, but I would rather have it in rdflib._contrib than explain in the doc string that we cannot cover this with normal guarantees for our public interface stability.

[aucampia]

PRs to V6 is closed until further notice. See this for more details:

• RDFLib V6 and V7 #2395

[aucampia]

PRs to V6 is closed until further notice. See this for more details:

• RDFLib V6 and V7 #2395

We will be open for PRs again once this is resolved:

• docs: add guidelines for breaking changes #2402