Re: Turtle "coverage" test suite issues with N-Triples

* Gregory Williams <greg@evilfunhouse.com> [2013-03-04 12:41-0500]
> Gregg alerted to me to the fact that I had overlooked the "coverage" part of the test suite. In trying to run those, I've run into another problem that I wish to comment on.
> 
> The two eval tests:
> 
> #literal_with_FORM_FEED
> #literal_with_BACKSPACE
> 
> are based on a comparison with N-Triples files that seem to rely on the new, unpublished N-Triples grammar (with the use of escaped forms \b and \f in strings). However, I don't see that this has anything to do with testing of Turtle, and wonder why those files don't use the serialization format from the existing N-Triples grammar (using the \u escaped form). Moreover, even when/if a new N-Triples document is published, I don't believe it's wise to base the new Turtle test suite on the also new N-Triples grammar. The N-Triples files in the Turtle test suite should conform to the existing grammar.

I like using ye olde n-triples for this (and in fact, think it should have a name, like "n-triples"). I originally committed
https://dvcs.w3.org/hg/rdf/file/a9221c4323e3/rdf-turtle/coverage/tests/literal_with_BACKSPACE.nt
which has
  <http://a.example/s> <http://a.example/p> "\u0008" .

I think Pierre-Antoine has a different notion of what's an appropriate escape form: https://dvcs.w3.org/hg/rdf/rev/3bb586f7bd9a . Pierre-Antoine, why did you prefer "\b" to "\u0008"?


> thanks,
> .greg
> 
> 

-- 
-ericP

Received on Monday, 4 March 2013 19:47:48 UTC