On Wed, Jun 08, 2022 at 03:59:51PM -0700, David G. Johnston wrote:
> The described behavior didn't even make it into 14.0; and I have no clue
> where such a thing would be documented to guess if the lack was intentional
> or not. I can confirm that as of today the behavior in HEAD is to not
> decompress and recompress during rewrite. Frankly, this seems like the
> expected behavior absent any documentation to the contrary. Whether there
> is a use case for someone to code up a way to get a recompression to happen
> is another matter. Oddly, I kinda expected that "update cmdata set f1 =
> f1;" might do it but alas it does not (...set f1 = f1 || ''... works; I'm
> not quite sure how much worse that, immediately followed by vacuum full,
> would be compared to whatever it would take to do it all during the full
> vacuum...especially if one is changing every single row on the table).
See commit dbab0c0 about the reasons of this choice, with this thread:
https://www.postgresql.org/message-id/20210527003144.xxqppojoiwurc2iz@alap3.anarazel.de
The docs of ALTER TABLE .. SET COMPRESSION include a note on the
matter: https://www.postgresql.org/docs/14/sql-altertable.html
"This does not cause the table to be rewritten, so existing data may
still be compressed with other compression methods."
--
Michael