public inbox for [email protected]  

help / color / mirror / Atom feed

From: David G. Johnston <[email protected]>
To: [email protected]
To: Pg Docs <[email protected]>
Subject: Re: 'CREATE OR REPLACE FUNCTION' behavior whenever a transaction is running
Date: Wed, 20 Apr 2022 13:52:38 -0700
Message-ID: <CAKFQuwaOE2jieS1CJ-mWRAopuEpu+f36te0YRh8+tVivqQw1ng@mail.gmail.com> (raw)
In-Reply-To: <[email protected]>
References: <[email protected]>

On Wed, Apr 20, 2022 at 12:50 PM PG Doc comments form <
[email protected]> wrote:

> The following documentation comment has been logged on the website:
>
> Page: https://www.postgresql.org/docs/14/sql-createfunction.html
> Description:
>
> The page does not mention what is the default behavior whenever a function
> is replaced while the same function is being used in another
> query/transaction.
>
> Does the query fail? Does the 'CREATE OR REPLACE' operation fail? Does the
> query finish before the function is replaced?
>
> As a side note, my interest in this came from our need to use 'C-Language
> Functions', aka binary functions, so it would be neat to also add what is
> the behavior on those as well, if there's any difference.
>

IIUC pg_proc is administered using MVCC behavior just like any other
table.  What you experience will be subject to your isolation mode but in
no case will a single command's execution see different versions nor will
such execution prevent the "replace"ment of the function with a newer
version.

Absent guidance to the contrary I wouldn't expect C language functions to
behave any differently than any others.  But that just pertains to the
"REPLACE" aspect.  You need to read the notes about C language functions to
get the rest of the story (and you likely wouldn't need to "replace" a
C-langauge function entry because of this different loading/compiling
mechanism). Namely:

https://www.postgresql.org/docs/current/xfunc-c.html#XFUNC-C-DYNLOAD

In short, C-language functions, referenced simply by name and module, and
pre-compiled, do not have a mechanism to invalidate the cache like pl/pgsql
functions do.  Other languages may involve yet other dynamics.

David J.

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: [email protected]
  Cc: [email protected], [email protected], [email protected]
  Subject: Re: 'CREATE OR REPLACE FUNCTION' behavior whenever a transaction is running
  In-Reply-To: <CAKFQuwaOE2jieS1CJ-mWRAopuEpu+f36te0YRh8+tVivqQw1ng@mail.gmail.com>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox