[Mulgara-dev] Lost phase token

[Paul Gearon]

Just another quick note....

On Fri, Sep 18, 2009 at 9:01 AM, James Leigh <james-nospam at leighnet.ca> wrote:
> I also get a lot of "Ignoring solutions" warnings, but I am not sure
> what they mean as the query results seem to be correct.
>
> 17-Sep-2009 6:27:26 PM org.apache.log4j.Category warn
> WARNING: Ignoring solutions for [gn3350 gn33 $obj $ctx]
> 17-Sep-2009 6:27:26 PM org.apache.log4j.Category info
> INFO: Creating Answer: 4166458, xa=0
>
> 17-Sep-2009 6:27:26 PM org.apache.log4j.Category warn
> WARNING: Ignoring solutions for non-local graph [gn3350 gn33 $obj $ctx]

This is because you have a variable graph that hasn't been bound when
the query engine gets to it. Theoretically this means that is has to
apply the remainder of the pattern to every graph in existence. That's
OK for the graphs being stored locally, but it's potentially infinite
for graphs on the web (particularly of the http://... type).

So for resolvers like the HTTP resolver, we can't do anything. If one
of these resolvers is asked for an unbound graph pattern then it will
return nothing, rather than the "theoretically correct" answer of
*everything*. When this happens, a warning is logged to say what has
happened.

Looking at it, it shouldn't be a WARNING, since it's (mostly) OK. It's
probably better as a debug message.

Regards,
Paul