I am working with Hibernate Search and finding it to be very cool, but I received an interesting request from my customer. For this particular search parameter, they want the search to be lenient between the number 1 (one) and the lowercase letter L as well as the letter O and the number 0 (zero).

I was wondering if Lucene offers any sort of query that would provide relevant functionality. If not, how would you go about implementing such a thing?

Thanks.

Hi,

the way to solve this is via a custom Analyzer. The idea would be to insert additional tokens when analyzing the text. For example, 'Oval' would be indexed as 'Oval', '0val', 'Ova1' and '0va1'. The idea is similar to the synonym engine described in Lucene in Action. Maybe this http://www.onjava.com/pub/a/onjava/2003/01/15/lucene.html helps as well (though the article is old).

--Hardy

I am new to Hibernate Search (and thus to Lucene), so I am unclear where to do what you describe. Is it at the Token level or at the Filter level or at the Analyzer level? (Let's assume that question makes sense.)

I am thinking a custom filter that gets called inside a custom analyzer that basically decorates StandardAnalyzer. Does that sound reasonable?

Any hints as to where the implementation should be are appreciated.

Incidentally, I can apply my custom analyzer at a field level that overrides the default analyzer specified (in my persistence.xml as this is a JPA app). Right?

Thanks.

Just had another thought: a RegexQuery.

To use your example, I could take the term "oval" and do something like this:

[o | 0]va[l | 1]

Ugly I know...basically "(lowercase O or the number zero) then lowercase V then lowercase A then (lowercase L or the number one)"

and pass that to a RegexQuery that will look at the index for "oval" generated by plain old StandardAnalyzer.

Would that work? If so, would there be a performance drawback compared to a more elegant approach?

Thanks.

Hi,

the Lucene FAQ has an example on how to write your own analyzer and yes you can annotate (specify) the analyzer on field level.

I recommend you are going down this path since it is the Lucene way :)

--Hardy