They should be same as long as the same group heads are selected with both
queries. The CollapsingQParserPugin simply collapses the result set and
then forwards to lower collectors, so the DocSet created should always be
for the collapsed set.





Joel Bernstein
Search Engineer at Heliosearch


On Sun, Jul 13, 2014 at 11:17 PM, Umesh Prasad <umesh.i...@gmail.com> wrote:

> Hi Joel,
>      Actually I also have seen this. The counts given by groups.truncate
> and collapsingQParserPlugin differ.. We have a golden query framework for
> our product APIs and there we have seen differences in facet count given.
> One request uses groups.truncate and another collapsingQParser plugin and
> we have seen counts differ (By a small margin)
>     I haven't been able to isolate the issue to a unit test level, so I
> haven't raised a bug.
>
>
>
>
> On 12 July 2014 08:57, Joel Bernstein <joels...@gmail.com> wrote:
>
> > The CollapsingQParserPlugin currently supports facet counts that match
> > "group.truncate". This works great for some use cases.
> >
> > There are use cases though where "group.facets" counts are preferred. No
> > timetable yet on adding this feature for the CollapsingQParserPlugin.
> >
> > Joel Bernstein
> > Search Engineer at Heliosearch
> >
> >
> > On Thu, Jul 10, 2014 at 7:20 PM, shamik <sham...@gmail.com> wrote:
> >
> > > Are there any plans to release this feature anytime soon ? I think this
> > is
> > > pretty important as a lot of search use case are dependent on the facet
> > > count being returned by the search result. This issue renders renders
> the
> > > CollapsingQParserPlugin pretty much unusable. I'm now reverting back to
> > the
> > > old group query (painfully slow) since I can't use the facet count
> > anymore.
> > >
> > >
> > >
> > > --
> > > View this message in context:
> > >
> >
> http://lucene.472066.n3.nabble.com/RE-SOLR-6143-Bad-facet-counts-from-CollapsingQParserPlugin-tp4140455p4146645.html
> > > Sent from the Solr - User mailing list archive at Nabble.com.
> > >
> >
>
>
>
> --
> ---
> Thanks & Regards
> Umesh Prasad
>

Reply via email to