Re: Index-only scan for btree_gist turns bpchar to char

On Fri, 07 Jan 2022 at 03:21, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I looked at this and it does seem like it might work, as per attached
> patch.  The one thing that is troubling me is that the opclass is set
> up to apply gbt_text_same, which is formally the Wrong Thing for bpchar,
> because the equality semantics shouldn't be quite the same.  But we
> could not fix that without a module version bump, which is annoying.
> I think that it might not be necessary to change it, because
>
> (1) There's no such thing as unique GIST indexes, so it should not
> matter if the "same" function is a bit stricter than the datatype's
> nominal notion of equality.  It's certainly okay for that to vary
> from the semantics applied by the consistent function --- GIST has
> no idea that the consistent function is allegedly testing equality.
>
> (2) If all the input values for a column have been coerced to the same
> typmod, then it doesn't matter because two values that are equal after
> space-stripping would be equal without space-stripping, too.
>
> However, (2) doesn't hold for an existing index that the user has failed
> to REINDEX, because then the index would contain some space-stripped
> values that same() will not say are equal to incoming new values.
> Again, I think this doesn't matter much, but maybe I'm missing
> something.  I've not really dug into what GIST uses the same()
> function for.
>
> In any case, if we do need same() to implement the identical
> behavior to bpchareq(), then the other solution isn't sufficient
> either.
>
> So in short, it seems like we ought to do some compatibility testing
> and see if this code misbehaves at all with an index created by the
> old code.  I don't particularly want to do that ... any volunteers?
>

Thanks for your patch, it looks good to me.  I'm not sure how to test this.

-- 
Regrads,
Japin Li.
ChengDu WenWu Information Technology Co.,Ltd.