cygri
commented
9 years ago A bit of progress towards fixing this.
- I've tightened up the regex for namespace URIs, and made it more readable (see source).
- I've added a little test script that can be run on the command line.
There's an ABNF for IRIs in RFC 3987. The problem is that the ucschar seen in the iunreserved production is not reflected in the $subdomain and $path parts of the prefix.cc regex. So what needs to be done:
- Modify the regex to allow the full
ucscharrange, assuming UTF-8 encoding. PHP'spreg_matchhas aumodifier that should be helpful here. - Conveniently, the database column for storing namespace URIs already is UTF-8 capable, so nothing to do there.
- Check that the
prefix.cc/foaf.gostyle and reverse lookups work correctly; some URI-to-IRI or IRI-to-URI conversion may be necessary otherwise. - Perhaps make a pass over source code and output strings to make sure we say “IRI” where we should.
I'm trying to add a prefix bgcat http://bg.dbpedia.org/resource/Категория:
prefix.cc says "URI must start with http:// or https://, end in / or : or #." But it does (ends in ":")! Maybe it's disagreeing with the Cyrilllic characters?
Richard: I did a few tests and you're right, it's the non-URI characters.