Hi Erik:
It's not in the SolrJ library, but rather my use of it:
In my application code:
protected static final String SOLR_ALL_DOCS_QUERY = "*:*";
/*
* If no search terms provided, then return all neighbors.
* Results are to be returned in neighbor symbol alphabetical order.
*/
if (searchTerms == null) {
searchTerms = SOLR_ALL_DOCS_QUERY;
nodeQuery.addSortField("n_name", SolrQuery.ORDER.asc);
}
So, if no user search terms are provided, I search all documents (there are
other fqs in effect) and return them in name order.
That worked just fine. Then I read more about [e]dismax, and went and
configured:
<str name="q.alt">*:*</str>
Then I would get zero results. It's not a SolrJ issue though, as this request
in my browser also resulted in zero results:
http://localhost:8091/solr/ing-content/select/?qt=partner-tmo&fq=type%3Anode&fq=n_neighborof_id%3AING\:afa&q=*:*&rows=5&facet=true&facet.mincount=1&facet.field=n_neighborof_processExact&facet.field=n_neighborof_edge_type
That was due to the q=*:*. Once I set, say, q=cancer, I got results. So I
guess this is a [e]dismax thing? (partner-tmo is the name of my request
handler).
I solved my problem by net setting *:* in my application, and left q.alt=*:* in
place.
Hope this helps. Again, this is stock Solr 3.4.0, running the Apache war under
Tomcat 6.
Jeff
On Nov 22, 2011, at 8:05 AM, Erik Hatcher wrote:
>
> On Nov 22, 2011, at 09:55 , Jeff Schmidt wrote:
>> When using [e]dismax, does configuring q.alt=*:* and not specifying q affect
>> the performance/caching in any way?
>
> No different than using q=*:* with the "lucene" query parser.
> MatchAllDocsQuery is possibly the fastest query out there! (it simply
> matches documents in index order, all scores are 1.0)
>
>> As a side note, a while back I configured q.alt=*:*, and the application
>> (via SolrJ) still set q=*:* if no user input was provided (faceting). With
>> both of them set that way, I got zero results. (Solr 3.4.0) Interesting.
>
> Ouch. Really? I don't see in the code (looking at my trunk checkout) where
> there's any *:* used in the SolrJ library. Can you provide some details on
> how you used SolrJ? It'd be good to track this down as that seems like a bug
> to me.
>
> Erik
>
>
>>
>> Thanks,
>>
>> Jeff
>>
>> On Nov 22, 2011, at 7:06 AM, Erik Hatcher wrote:
>>
>>> If all you're doing is filtering (browsing by facets perhaps), it's
>>> perfectly fine to have q=*:*. MatchAllDocsQuery is fast (and would be
>>> cached anyway), so use *:* as appropriate without worries.
>>>
>>> Erik
>>>
>>>
>>>
>>> On Nov 22, 2011, at 07:18 , pravesh wrote:
>>>
>>>> Usually,
>>>>
>>>> Use the 'q' parameter to search for the free text values entered by the
>>>> users (where you might want to parse the query and/or apply
>>>> boosting/phrase-sloppy, minimum match,tie etc )
>>>>
>>>> Use the 'fq' to limit the searches to certain criterias like location,
>>>> date-ranges etc.
>>>>
>>>> Also, avoid using the q=*:* as it implicitly translates to
>>>> matchalldocsquery
>>>>
>>>> Regds
>>>> Pravesh
>>>>
>>>> --
>>>> View this message in context:
>>>> http://lucene.472066.n3.nabble.com/how-to-make-effective-search-with-fq-and-q-params-tp3527217p3527535.html
>>>> Sent from the Solr - User mailing list archive at Nabble.com.
>>>
>>
>>
>>
>> --
>> Jeff Schmidt
>> 535 Consulting
>> j...@535consulting.com
>> http://www.535consulting.com
>> (650) 423-1068
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
--
Jeff Schmidt
535 Consulting
j...@535consulting.com
http://www.535consulting.com
(650) 423-1068
