Regex for finding all namespaces in data - regex

Question

I need a regular expression (dubbed SOME_EXPRESSION below) that allows finding all namespaces for resources used as subject in a SPARQL 1.1 endpoint. The query should look like the following. How can I do this?
SELECT DISTINCT ?ns
WHERE
{
?s ?p ?o.
BIND(REPLACE(str(?s), SOME_EXPRESSION, "")) AS ?ns)
Filter(isURI(?s))
}

Answer 1

Since the harder part of this is processing the IRI strings, I'll show how you can do this for properties (which must be IRIs, so we don't need to check for isIRI). Adapting this to work with the IRIs of subjects won't be hard. However, there is one thing that needs some consideration: URIs for linked data typically (there's no hard requirement, but conventions do emerge) use prefixes that end in / or in #. Whether one is better than the other is the subject of plenty of debate and discussion (e.e., see section 4 of Cool URIs, or HashVsSlash). In general, you're going to want to replace everything after the final slash or hash with the final slash or hash. Since you can use groups in SPARQL's regex and replace, you can handle both cases with one replace:
select distinct ?ns where {
[] ?p [] .
bind( replace( str(?p), "(#|/)[^#/]*$", "$1" ) as ?ns )
}
This matches the regular expression (#|/)[^#/]*$ against the string form of the IRI, remembering # or / in the variable $1, and then grabs the rest of the characters (which must not contain # or /) up until the end of the string, and replaces the whole thing with $1, which is either # or /. For some data that I pulled from Linked Open British National Bibliography data, I get results like these:
$ sparql --query query.rq --data sample.nt
-----------------------------------------------------
| ns |
=====================================================
| "http://www.w3.org/2000/01/rdf-schema#" |
| "http://www.w3.org/1999/02/22-rdf-syntax-ns#" |
| "http://www.w3.org/2004/02/skos/core#" |
| "http://purl.org/ontology/bibo/" |
| "http://purl.org/dc/terms/" |
| "http://iflastandards.info/ns/isbd/elements/" |
| "http://www.bl.uk/schemas/bibliographic/blterms#" |
| "http://www.w3.org/2002/07/owl#" |
| "http://purl.org/NET/c4dm/event.owl#" |
-----------------------------------------------------
This seems like a reasonable set of namespace prefixes. In fact, when I look at the header of the RDF document, original namespaces included:
xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:skos="http://www.w3.org/2004/02/skos/core#"
xmlns:bibo="http://purl.org/ontology/bibo/"
xmlns:dct="http://purl.org/dc/terms/"
xmlns:isbd="http://iflastandards.info/ns/isbd/elements/"
xmlns:blt="http://www.bl.uk/schemas/bibliographic/blterms#"
xmlns:owl="http://www.w3.org/2002/07/owl#"
xmlns:event="http://purl.org/NET/c4dm/event.owl#"
As applied to your code, we end up with the following query. It's almost exactly what you wanted, since since there's just one regular expression that handles both cases (so just one thing to fill in for SOME_EXPRESSION. However, instead of replacing with "", you do have to replace with "$1". I hope that's not a terrible inconvenience, though.
SELECT DISTINCT ?ns
WHERE
{
?s ?p ?o.
BIND(REPLACE(str(?s), "(#|/)[^#/]*$", "$1") AS ?ns)
Filter(isURI(?s))
}
It's important to note, of course, that this is only a heuristic. A given IRI can be abbreviated using lots of different prefixes. This technique should give some relatively good results, though, because there are conventions that people tend to follow pretty well.