W3C home > Mailing lists > Public > public-cwm-bugs@w3.org > June 2005

Re: weirdness with literal or formula predicates

From: EspeonEefi <eefi@MIT.EDU>
Date: Thu, 16 Jun 2005 17:48:23 -0400
To: public-cwm-bugs@w3.org, syosi@MIT.EDU
Message-Id: <1118958503.6503.97.camel@localhost.localdomain>

On Fri, 2005-06-10 at 10:26 -0400, EspeonEefi wrote:
> Re: http://lists.w3.org/Archives/Public/public-cwm-bugs/2004Nov/0014
> 
> This seems to be fixed in latest CVS.
> 
> echo ':a "hello" :c .' | cwm
> exits with a traceback and "ValueError: Cannot have a literal as a
> predicate. This makes no sense, "hello""
> 
> echo ':a "hello" :c .' | cwm --n3 --rdf
> exits with a traceback and "ValueError: Cannot have a literal as a
> predicate. This makes no sense, "hello""
> 
> echo '{:a :b :c} {:e :f :g} :c . ' | cwm
> exits with a traceback and "ValueError: Cannot have a formula as a
> predicate. This makes no sense"
> 
> -- eefi
> 
> 

Well, after I poked some more at this bug, it seems that it's not
actually fixed after all. After consultation with Yosi, it seems the
code that checks for these cases happens during output (serialization)
rather than during parsing. Therefore, log:sematicsOrError will be
connected to invalid semantics, whereas we really want it to be
connected to an error statement.

I've checked in a test set in test/syntax/bad-preds*.n3 that should
probably be the benchmark for this bug. I've also added a (commented
out) test definition in regression.n3.

-- eefi
Received on Thursday, 16 June 2005 22:06:09 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:10:59 GMT