On Wed, Apr 11, 2012 at 4:37 PM, jmlucjav <jmluc...@gmail.com> wrote: > Just to be sure, reproduced this with example config from 3.5. >
Regardless of your tokenizer, be aware that with this version of solr its going to split up terms based on 'identifier rules' (including splitting on whitespace). This is because suggesters go thru the ordinary spellchecker framework. If you are trying to autosuggest actual phrases, have a look at http://wiki.apache.org/solr/Suggester#Tips_and_tricks which describes how to set this up along with example configurations. More information is available in https://issues.apache.org/jira/browse/SOLR-3143 Essentially this provides a QueryConverter thats hopefully more suitable for autosuggesters, it just passes the whole entire input to your query analyzer, and its your responsibility to do whatever you need there to extract the 'meat' of the query for autosuggest. The example configuration linked from the wiki page is just that and uses some regexps to try to imitate what google's does (discarding operators like +/- but still keeping the whole thing as a phrase). You will need Solr 3.6 for this..., but its on its way out. -- lucidimagination.com
