Re: Next Steps with Hash Indexes - Mailing list pgsql-hackers

From Peter Geoghegan
Subject Re: Next Steps with Hash Indexes
Date
Msg-id CAH2-WznD+m-=E4AyzD6R7iA3xd3KiDsbDk9CWu1Le2xgkA2R7w@mail.gmail.com
Whole thread Raw
In response to Next Steps with Hash Indexes  (Simon Riggs <simon.riggs@enterprisedb.com>)
List pgsql-hackers
On Thu, Jul 15, 2021 at 9:41 AM Simon Riggs
<simon.riggs@enterprisedb.com> wrote:
> It would be very desirable to allow Hash Indexes to become Primary Key
> Indexes, which requires both
>   amroutine->amcanunique = true;
>   amroutine->amcanmulticol = true;

Why do you say that? I don't think it's self-evident that it's desirable.

In general I don't think that hash indexes are all that compelling
compared to B-Trees. In practice the flexibility of B-Trees tends to
win out, even if B-Trees are slightly slower than hash indexes with
certain kinds of benchmarks that are heavy on point lookups and have
no locality.

I have no reason to object to any of this, and I don't object. I'm just asking.

-- 
Peter Geoghegan



pgsql-hackers by date:

Previous
From: Michael Paquier
Date:
Subject: Re: Quirk of pg_temp schemas ...
Next
From: Michael Paquier
Date:
Subject: Re: [PATCH] test/ssl: rework the sslfiles Makefile target