Riak Search "And" Queries inconsistency with Lucene

Tejus Parikh tejus at vijedi.net
Sat May 14 12:28:38 EDT 2011

I am considering using Riaksearch to replace a project that currently makes use of lucene. This project extensively uses 'AND' queries to return a very small subset of the total document store. Given that riaksearch provides a "solr" endpoint, I would have expected the queries to return the same results, but that does not appear to be the case.

I created a simple test containing three documents (described here as xml):
<field name="id">1</field>
<field name="topping_text">strawberry</field>
<field name="base_text">pastry</field>
<field name="id">2</field>
<field name="topping_text">blueberry</field>
<field name="base_text">waffle</field>
<field name="id">3</field>
<field name="topping_text">strawberry</field>
<field name="base_text">pancake</field>

I ran a few queries against this set of docs:
search-cmd search menu_index "topping_text:\"strawberry\""
** This returns 2 documents

search-cmd search menu_index "topping_text:\"strawberry\" and base_text:\"pastry\""
** This returns 2 documents, I think it should only return menu_index/1

search-cmd search menu_index "topping_text:\"strawberry\" and base_text:\"waffles\""
** This returns 3 documents, it shouldn't return any

For comparions sake, I ran the same queries against lucene and retrieved the results I expected.

I've uploaded my documents, the commands I used to create an index, and the sample lucene java code to: https://github.com/vijedi/Riak-Search--And--Query-Test 

Have I misinterpreted what is happening or is there a real difference here?


Tejus Parikh
tejusparikh.com | @vi_jedi

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20110514/046d744c/attachment.html>

More information about the riak-users mailing list