On Tue, Oct 15, 2019 at 3:35 PM Alan Hodgson <ahodgson@lists.simkin.ca> wrote:
> My company has found the pg_trm extension to be more useful for partial text searches than the full text functions. I
don'tknow specifically how it might help with your hyphens but it would be worth testing. The docs actually suggest
usingthem in conjunction in some cases.
We actually do use pg_trgm already for the names/titles of things.
Indexing the content with a trigram index and then doing
LOWER(content) LIKE '%789-xyz%' would certainly work, but
1. we'd have to do a little bit of finagling if we wanted to match on
word boundaries (don't match '6789-xyza' in the above example)
2. trigram indexes are pretty huge for long documents, which is why we
currently only use them for names/titles
We may give up and just use pg_trgm for contents if nothing else works
out but it feels like the text search lexer is _so_ close to what we
want.