On 03/01/2021 11:25, Martynas Jusevičius wrote:
I’ll see what RDF4J does.
I think ultimately the reasonable solution would be to have
<!ELEMENT graph (uri?, triple*)>
as the DTD.
Indeed - that's what's implemented!
Andy
On Sun, 3 Jan 2021 at 11.38, Andy Seaborne <a...@apache.org> wrote:
What do other systems do?
Everything else is secondary given that (W3C|HPLB)-trix is not a working
group, or community group, consensus.
(IIRC the Jena reader doesn't care about namespaces)
Andy
On 02/01/2021 17:48, Martynas Jusevičius wrote:
Hi,
some time ago we had an issue re. TriX:
https://issues.apache.org/jira/browse/JENA-1211
I was looking at the W3C DTD again, specifically for <graph> elements:
<!ELEMENT graph (uri, triple*)>
This made me wonder how the default graph should be represented if it
has no URI.
My understanding was that the default <graph> should have no <uri>,
but the W3C DTD contradicts this.
I was probably thinking about the HP's DTD version [1] where the
<graph> definition was different (though still looks wrong with an
unbounded number of URIs):
<!ELEMENT graph (uri*, triple*)>
A quick test with riot CLI converts N-Triples
<http://s> <http://b> <http://c> .
to
<trix xmlns="http://www.w3.org/2004/03/trix/trix-1/">
<graph>
<triple>
<uri>http://s</uri>
<uri>http://b</uri>
<uri>http://c</uri>
</triple>
</graph>
</trix>
Isn't this invalid in respect to the W3C DTD?
It's quite unfortunate that we have those 2 incompatible versions and
neither of them seems to get it right...
[1] https://www.hpl.hp.com/techreports/2004/HPL-2004-56.pdf
Martynas
atomgraph.com