I’m using an dict_xsyn extension in postgres which returns multiple lexemes.
CREATE EXTENSION dict_xsyn; CREATE TEXT SEARCH DICTIONARY names_xsyn (TEMPLATE = 'xsyn_template'); CREATE SEARCH CONFIGURATION english_names ALTER MAPPING FOR asciiword, asciihword, hword_asciipart WITH names_xsyn, simple; SELECT to_tsquery('english_names', 'ed') -- returns: 'ed' & 'edward' & 'edwin' & 'edmond'
I would prefer if to_tsquery returned ORed lexemes so that I could match to other variants. e.g.
-- returns: 'ed' | 'edward' | 'edwin' | 'edmond'
I’ve tried ts_rewrite but that doesn’t seem to support swapping logical operators.
I’m doing this so that I can match partially matching sets.
I want ed to match edward and edwin, but I don’t want edward to match edwin. Here are my lexemes:
ed -> ed, edward, edwin
edward -> ed, edward
edwin -> ed, edwin
Using those lexmes to match ed to edward will only work with an or.
“ed, edward” @@ “ed | edward | edwin” will work
“ed, edward” @@ “ed & edward & edwin” won’t work
I don’t know how to do that exclusively with
dict_xsyn, other than by manipulating the output as text and the converting back, but I don’t think you want to do that anyway. I think it would be a performance nightmare.
You probably want to reverse this so that all nicknames get canonicalized down to a single variant, rather than exploded out to all variants.
alter TEXT SEARCH DICTIONARY names_xsyn (KEEPORIG=true, MATCHORIG=true, KEEPSYNONYMS=false, MATCHSYNONYMS=true);
But I have no idea why that is not the default, so perhaps I just fundamentally misunderstand the point of that extension.
Both the tsquery and the tsvector would be canonicalized the same way, so a query with ‘bobbie’ would match a document with ‘robert’ and vice versa.
But perhaps I just don’t understand your end goal. You can do what you are directly asking for with
ts_rewrite, but it is not done by building on top of
dict_xsyn. Is the example involving
supernovae & crab not just what you want?