If you really need wildcards on both ends, I don't think edgengram will
work, you need plain ngrams. The trick is that the query side needs
to make the 2-grams into phrases. BTW, I think you'd be fine with
bigrams.
Best
Erick
On Thu, Aug 8, 2013 at 1:47 PM, meena.sri...@mathworks.com <
meena
Thanks for your responses, helped me to understand the issue. Digged through
the documentation and now I am implementing EdgeNGramFilterFactory to see
how fastly can I improve wild card searches.
Compare timings in the following cases:
- Without the wildcard
- With suffix wild card only - test*
- With reverse wild card filter factory and two separate terms - *test OR
test*
On Thu, Aug 8, 2013 at 8:15 PM, meena.sri...@mathworks.com <
meena.sri...@mathworks.com> wrote:
> Index size is arou
Hello!
In general, wildcard queries can be expensive. If you need wildcard
queries, you can try the EdgeNGram -
http://wiki.apache.org/solr/AnalyzersTokenizersTokenFilters#solr.EdgeNGramFilterFactory
--
Regards,
Rafał Kuć
Sematext :: http://sematext.com/ :: Solr - Lucene - ElasticSearch
> In
Sometimes bolding comes through my e-mail as *, so is *test* with the
asterisk on each end really what you're doing? Assuming so, this
will inevitably be slow. It must iterate through all the terms in the field
to see if any of them match. This is generally a bad practice.
You can to go n-grams if
