Horton

Re: ISN extension - wrong volatility level for isn_weak() function

Viktor Holmberg <v@viktorh.net>

View thread
From: Viktor Holmberg <v@viktorh.net>
To: Daniel Gustafsson <daniel@yesql.se>, Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-bugs@lists.postgresql.org
Date: 2025-03-14T14:28:01Z
Lists: pgsql-bugs

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. contrib/isn: Make weak mode a GUC setting, and fix related functions.

  2. Update contrib/seg for new scalarlesel/scalargesel selectivity functions.

I haven’t checked the source code, but yes the isn_weak feature has some footgun potential. As it doesn’t respect transactions, but rather sets a flag on session level, it’s easy for the “isn weakness” to leak out into a connection pool.
Still the feature is very useful as I work with many suppliers who give us product lists full of invalid EANs, which still need to be ingested.

/Viktor
On 14 Mar 2025 at 14:16 +0000, Tom Lane <tgl@sss.pgh.pa.us>, wrote:
> Daniel Gustafsson <daniel@yesql.se> writes:
> > On 14 Mar 2025, at 12:49, Viktor Holmberg <v@viktorh.net> wrote:
> > > The isn_weak function in the isn extension reports the wrong value if you look at it inside a transaction:
>
> > I wonder if this should really be marked VOLATILE instead as it has a side
> > effect.
>
> Indeed. This whole area seems really poorly considered. The comment
> in isn--1.1.sql says
>
> -- isn_weak(boolean) - Sets the weak input mode.
> -- This function is intended for testing use only!
>
> despite which it's documented at length in isn.sgml. On the other
> hand, so far as I can find it's tested nowhere, which means none
> of the "weak mode" code is getting exercised.
>
> regards, tom lane