Horton

Re: Defects with invalid stats data for expressions in extended stats

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Chao Li <li.evan.chao@gmail.com>
Cc: Corey Huinker <corey.huinker@gmail.com>, Postgres hackers <pgsql-hackers@lists.postgresql.org>, Tomas Vondra <tomas@vondra.me>
Date: 2026-03-02T00:20:37Z
Lists: pgsql-hackers 
On Mon, Mar 02, 2026 at 07:39:45AM +0800, Chao Li wrote:
> But why do we still explicitly free the stats array? My concern is
> mostly about “partial free”: either free everything, or free nothing
> and let the memory context clean it up. “Partial free” tends to
> confuse code readers, and future readers may keep running into the
> same question. 

Not necessarity.  We have plenty of these patterns in the code with
such partial frees.  At the end, two things tend to take priority:
the readability of the code as well as its efficiency because repeated
pfree() calls can be more expensive than a memory context cleanup,
particularly in hot loops, though doing pfree() calls may be better in
some cases.

In this case, I see a better argument with the code clarity, as we use
the same array for the stats built with attributes and expressions.
We also cap the number of expressions to 8, which puts some control.
But at the end it does not really matter: in this context ANALYZE
makes sure that nothing crosses the processing of a single relation,
even with a transaction analyzes a lot of relations (same or
different).
--
Michael

Commits

  1. test_custom_types: Test module with fancy custom data types

  2. Fix set of issues with extended statistics on expressions