Re: full-text search : How to improve ? (solved)

Hi Norbert:

I did a little more poking around in the whoosh indexer after the
native and xapian indexers operated as you wanted.

Looking in indexer-whoosh, I saw:
 
        q = query.And([query.FuzzyTerm("content", word.lower())
                        for word in wordlist])

doh. The FuzzyTerm changes the search term to do a fuzzy match.  I
either forgot or never realized that was the case. Also I was the one
who added Whoosh. It's not native to the alpine image.

I'm tempted to just document this in the installation document and
recommend using xapian or native explicitly if you don't want fuzzy
matching. I don't want to disable it since none of the other engines
have fuzzy matching. I also don't want to add a new flag (and
non-fuzzy matching code) to enable/disable fuzzy matching.

If in the future we enable database full text search (e.g. FTS5 in
sqlite, trigram search in postgresql or similar in mysql), some of
those have fuzzy search available. Fuzzy search on/off can be added at
that time.

But I could be convinced to support a fuzzy on/off setting now.

Also I just pushed a new dockerfile to the repo that includes
xapian. I used the method for building xapian for the travisci service
and deleted the build only python packages which are 50MB.

--
				-- rouilj
John Rouillard
===========================================================================
My employers don't acknowledge my existence much less my opinions.