Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-30 Thread Bruce Momjian
On Sun, May 21, 2017 at 11:01:57PM -0400, Chapman Flack wrote: > ? A transformOpclassLike function could verify that foo and the opcintype > of int4_ops have the same typlen and typbyval, and that the operators and > support procs are backed by C functions, and return a list of > CREATE OPERATOR

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-22 Thread Chapman Flack
On 05/22/17 18:39, Alvaro Herrera wrote: > Chapman Flack wrote: >> CREATE INDEX ON foo USING btree ( bar, baz ALSO quux ); > > INCLUDING: > https://www.postgresql.org/message-id/56168952.4010...@postgrespro.ru I'd buy that. -Chap -- Sent via pgsql-hackers mailing list

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-22 Thread Alvaro Herrera
Chapman Flack wrote: > That was what gave me the idea in the first place, which then > I realized could be more generally useful. If I could say > something like > > CREATE INDEX ON foo USING btree ( bar, baz ALSO quux ); > > so that only bar and baz are compared in insertion and search, > but

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-22 Thread Chapman Flack
On 05/22/2017 05:16 AM, Amit Kapila wrote: > Agreed, but even if we have any such syntax, making it work for hash > indexes is tricky, because we currently store the hash code in the > index, not the original hash index key. That was what gave me the idea in the first place, which then I realized

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-22 Thread Amit Kapila
On Mon, May 22, 2017 at 8:31 AM, Chapman Flack wrote: > On 05/19/17 11:41, Tom Lane wrote: > >> No, nobody's done anything about allowing hash indexes to support >> uniqueness AFAIK. I don't have a clear picture of how much work >> it would be, but it would likely be more

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-21 Thread Chapman Flack
On 05/19/17 11:41, Tom Lane wrote: > No, nobody's done anything about allowing hash indexes to support > uniqueness AFAIK. I don't have a clear picture of how much work > it would be, but it would likely be more than trivial effort; I see what you mean. Because of the way hash values are

Re: [HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-19 Thread Tom Lane
Chapman Flack writes: > Was my guess about the reason right? Does this PG10 announcement > also mean it will be possible to use UNIQUE constraints with some > pure-identifier, no-natural-ordering type that supports only hashing? No, nobody's done anything about allowing

[HACKERS] PG10 Crash-safe and replicable Hash Indexes and UNIQUE

2017-05-19 Thread Chapman Flack
Hi, The item on hash indexes reminded me of an old comment from years ago that I put in the code of the first custom PG datatype I ever built at $work: COMMENT ON OPERATOR CLASS puid_ops USING btree IS 'As puids are only identifiers, there is no obvious reason to define ordering operators or