Hi Ken,
Not collapsing on null field values is not possible in the patch.
However you can if you want to fix this in the patch it is a really
small change. Assuming that you're using the default collapsing
algorithm you can add the following piece of code in the
NonAdjacentDocumentCollapser.java file:
if (currentValue == null) {
continue;
}
Place it in the doCollapsing method after the following statement:
String currentValue = values.lookup[values.order[currentId]];
This makes sure that documents that have no value in the collapse
field are not collapsed.
Field collapsing has a big impact on your search times and also on
memory usage to a lesser extend. It can increase search times up to 10
times, but this depends per situation. As indexes get bigger this
becomes a bigger problem. Also using field collapsing in a distributed
environment can cause problems. This is due that collapse information
is not shared between shards, resulting in incorrect collapse results.
They only work around for this problem I know is, that you 'll have to
make sure that the groups are distributes evenly between shards and
that a group's documents are not spread across shards.
Other then that there are no further major issues with this patch.
Many people are using this patch in their Solr setups, but it is a
patch so you 'll have to keep that in mind. There are efforts to put
grouping functionality into Solr (without patching) in SOLR-236's
child issues, so keep an eye on these issues.
Cheers,
Martijn
On 3 July 2010 19:20, osocurious2 <ken.fos...@realestate.com> wrote:
>
> <boost>
>
> I wanted to extend my question some. My original question about collapsing
> null fields is still open, but in trying to research elsewhere I see a lot
> of angst about the Field Collapse functionality in general. Can anyone
> summarize what the current state of affairs is with it? I'm on Solr 1.4,
> just the latest release build, not any current builds. Field Collapse seems
> to be in my build because I could do single field collapse just fine (hence
> my null field question). However there seems to be talk of problems with
> Field Collapse that aren't fixed yet. What kinds of issues are people
> having? Should I avoid Field Collapse in a production app for now? (tricky
> because I'm merging my schema with a third party tool schema and they are
> using Field Collapse).
>
> Any insight would be helpful, thanks
> Ken
> --
> View this message in context:
> http://lucene.472066.n3.nabble.com/Field-Collapse-question-tp939118p940923.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
