There is no need for a new component, just implement a new
SolrSpellChecker. The component itself is not much more than a thin
wrapper around an actual implementation.
I'm not sure about the scanning though. The Main SC most people use is
the Lucene one, and that searches an index that it makes out of the
source index. Have a look at the Lucene implementation for more
understanding.
-Grant
On Feb 21, 2009, at 20:38, Lance Norskog <goks...@gmail.com> wrote:
The SpellCheckComponent
<https://issues.apache.org/jira/browse/SOLR-698>scans the term
database for likely words. This does not scale up to giant
indexes. It also does not allow creating separate spelling DBs for
subsets
of the index. I want to create an alternate spell checker
implementation
that does a query and builds the index from terms found.
How would you put this together? It would be easy to do a simple
Lucene
search, since this would be configured by the solr developer, but
this seems
too limiting. Is making a new SearchComponent the right technique?
http://wiki.apache.org/solr/SpellCheckComponent
http://wiki.apache.org/solr/SearchComponent
--
Lance Norskog
goks...@gmail.com
650-922-8831 (US)
