On Mon, Oct 26, 2020 at 5:03 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Julien Rouhaud <rjuju123@gmail.com> writes:
> > A french user recently complained that with an index created using
> > gin_trgm_ops (or gist_trgm_ops), you can use the index with a clause
> > like
> > col LIKE 'something'
> > but not
> > col = 'something'
>
> Huh, I'd supposed we did that already.
>
> > even though both clauses are technically identical. That's clearly
> > not a high priority thing to support, but looking at the code it seems
> > to me that this could be achieved quite simply: just adding a new
> > operator = in the opclass, with an operator strategy number that falls
> > back doing exactly what LikeStrategyNumber is doing and that's it.
> > There shouldn't be any wrong results, even using wildcards as the
> > recheck will remove any incorrect one.
>
> I think you may be overoptimistic about being able to use the identical
> code path without regard for LIKE wildcards; but certainly it should be
> possible to do this with not a lot of new code. +1.
Well, that's what I was thinking too, but I tried all the possible
wildcard combinations I could think of and I couldn't find any case
yielding wrong results. As far as I can see the index scans return at
least all the required rows, and all extraneous rows are correctly
removed either by heap recheck or index recheck.
I'm attaching a patch POC pach with regression tests covering those
combinations. I also found a typo in the 1.4--1.5 pg_trgm upgrade
script, so I'm also attaching a patch for that.