Search across multiple collections

[bbarani <bb...@gmail.com>]

I am not sure the best way to search across multiple collection using SOLR
4.3. 

Suppose, each collection have their own config files and I perform various
operations on collections individually but when I search I want the search
to happen across all collections. Can someone let me know how to perform
search on multiple collections? Do I need to use sharding again?



--
View this message in context: http://lucene.472066.n3.nabble.com/Search-across-multiple-collections-tp4068469.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Search across multiple collections

[Erick Erickson <er...@gmail.com>]

You pretty much need to issue separate
queries against each collection and creatively
combine them. All of Solr's distributed search
stuff pre-supposes two things
1> the schemas are very similar
2> the types of docs in each collection are also
     very similar.

<2> is a bit subtle. If you store different kinds of
docs in different cores, then that statistics for
term frequency etc. will be different. There's some
work being done (I think) to support "distributed
tf/idf". But anyway, in this case the scores of the
docs from one collection will tend to dominate the
result set.

Or if you're talking about joining, see Anria's comments.

Best
Erick

On Wed, Jun 5, 2013 at 7:34 PM,  <ab...@innoventsolutions.com> wrote:
> hi
> I've successfully searched over several separate collections (cores with
> unique schemas) using this kind of syntax.  This demonstrates a 2 core
> search
>
> http://localhost:8983/solr/collection1/select?
> q=my phrase to search on&
> start=0&
> rows=25&
> fl=*,score&
> fq={!join+fromIndex=collection2+from=sku+to=sku}id:1571&
>
>
> I've split up the parameters so you see easily
> fq={!join+fromIndex=collection2+from=sku+to=sku}id:1571&
>
> --> collection1/select  = use the select requestHandler out of collection1
> as a base
> --> collection2 is the 2nd core : equivalent of a table join in SQL
> --> sku is the field shared in both collection1, and collection2
> --> id is the field I want to find the id=1571 in.
>
> Hope this helps
> Anria
>
>
>
>
> On 2013-06-05 16:17, bbarani wrote:
>>
>> I am not sure the best way to search across multiple collection using SOLR
>> 4.3.
>>
>> Suppose, each collection have their own config files and I perform various
>> operations on collections individually but when I search I want the search
>> to happen across all collections. Can someone let me know how to perform
>> search on multiple collections? Do I need to use sharding again?
>>
>>
>>
>> --
>> View this message in context:
>>
>> http://lucene.472066.n3.nabble.com/Search-across-multiple-collections-tp4068469.html
>> Sent from the Solr - User mailing list archive at Nabble.com.

Re: Search across multiple collections

[ab...@innoventsolutions.com]

hi
I've successfully searched over several separate collections (cores 
with unique schemas) using this kind of syntax.  This demonstrates a 2 
core search

http://localhost:8983/solr/collection1/select?
q=my phrase to search on&
start=0&
rows=25&
fl=*,score&
fq={!join+fromIndex=collection2+from=sku+to=sku}id:1571&


I've split up the parameters so you see easily
fq={!join+fromIndex=collection2+from=sku+to=sku}id:1571&

--> collection1/select  = use the select requestHandler out of 
collection1 as a base
--> collection2 is the 2nd core : equivalent of a table join in SQL
--> sku is the field shared in both collection1, and collection2
--> id is the field I want to find the id=1571 in.

Hope this helps
Anria



On 2013-06-05 16:17, bbarani wrote:
> I am not sure the best way to search across multiple collection using 
> SOLR
> 4.3.
> 
> Suppose, each collection have their own config files and I perform 
> various
> operations on collections individually but when I search I want the 
> search
> to happen across all collections. Can someone let me know how to 
> perform
> search on multiple collections? Do I need to use sharding again?
> 
> 
> 
> --
> View this message in context:
> http://lucene.472066.n3.nabble.com/Search-across-multiple-collections-tp4068469.html
> Sent from the Solr - User mailing list archive at Nabble.com.