Re: Shared detoast Datum proposal

Hi,

Peter Smith <smithpb2250(at)gmail(dot)com> writes:

> 2024-01 Commitfest.
>
> Hi, This patch has a CF status of "Needs Review" [1], but it seems
> there were CFbot test failures last time it was run [2]. Please have a
> look and post an updated version if necessary.
>
> ======
> [1] https://commitfest.postgresql.org/46/4759/
> [2] https://cirrus-ci.com/github/postgresql-cfbot/postgresql/commitfest/46/4759

v5 attached, it should fix the above issue. This version also introduce
a data struct called bitset, which has a similar APIs like bitmapset but
have the ability to reset all bits without recycle its allocated memory,
this is important for this feature.

commit 44754fb03accb0dec9710a962a334ee73eba3c49 (HEAD -> shared_detoast_value_v2)
Author: yizhi.fzh <yizhi(dot)fzh(at)alibaba-inc(dot)com>
Date: Tue Jan 23 13:38:34 2024 +0800

shared detoast feature.

commit 14a6eafef9ff4926b8b877d694de476657deee8a
Author: yizhi.fzh <yizhi(dot)fzh(at)alibaba-inc(dot)com>
Date: Mon Jan 22 15:48:33 2024 +0800

Introduce a Bitset data struct.

While Bitmapset is designed for variable-length of bits, Bitset is
designed for fixed-length of bits, the fixed length must be specified at
the bitset_init stage and keep unchanged at the whole lifespan. Because
of this, some operations on Bitset is simpler than Bitmapset.

The bitset_clear unsets all the bits but kept the allocated memory, this
capacity is impossible for bit Bitmapset for some solid reasons and this
is the main reason to add this data struct.

Also for performance aspect, the functions for Bitset removed some
unlikely checks, instead with some Asserts.

[1] https://postgr.es/m/CAApHDvpdp9LyAoMXvS7iCX-t3VonQM3fTWCmhconEvORrQ%2BZYA%40mail.gmail.com
[2] https://postgr.es/m/875xzqxbv5.fsf%40163.com

I didn't write a good commit message for commit 2, the people who is
interested with this can see the first message in this thread for
explaination.

I think anyone whose customer uses lots of jsonb probably can get
benefits from this. the precondition is the toast value should be
accessed 1+ times, including the jsonb_out function. I think this would
be not rare to happen.

--
Best Regards
Andy Fan