MIME-Version: 1.0
In-Reply-To: 
 <CAMsGm5c7F-mRshFO=TOvnX+9WgmTPnzRMTnfqjOMOCfWj4GWEQ@mail.gmail.com>
References: 
 <CAMsGm5c7F-mRshFO=TOvnX+9WgmTPnzRMTnfqjOMOCfWj4GWEQ@mail.gmail.com>
From: "David G. Johnston" <david.g.johnston@gmail.com>
Date: Fri, 10 Apr 2026 20:53:39 -0700
Message-ID: 
 <CAKFQuwYX7a5A=Kg3xtsGq51Jqe5VWLNH9Dos-LxvVwFgQJ5Hwg@mail.gmail.com>
Subject: Re: Possible mismatch between behaviour and documentation in CREATE
 FUNCTION
To: Isaac Morland <isaac.morland@gmail.com>
Cc: PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
Content-Type: multipart/alternative; boundary="000000000000f0ddb9064f2731c5"
Archived-At: 
 <https://www.postgresql.org/message-id/CAKFQuwYX7a5A%3DKg3xtsGq51Jqe5VWLNH9Dos-LxvVwFgQJ5Hwg%40mail.gmail.com>
Precedence: bulk

--000000000000f0ddb9064f2731c5
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Friday, April 10, 2026, Isaac Morland <isaac.morland@gmail.com> wrote:

> The documentation says that only BEGIN ATOMIC parses the function body at
> definition time. Whereas AS parses the function body at execution time:
>
> https://www.postgresql.org/docs/current/sql-createfunction.html
>
> (and look at the explanation of sql_body)
>

Possibly some documentation tweaks are in order.  I haven=E2=80=99t looked =
at it in
depth.  The optional checks are able to be disabled:

https://www.postgresql.org/docs/current/runtime-config-client.html#RUNTIME-=
CONFIG-CLIENT-STATEMENT

While the result of the required parsing of a function body is implying
that such parsing is saved to the catalogs so that creation-time object
references are retained.  While the others are re-parsed during execution
and object references retained only for the lifetime of the query.

David J.

--000000000000f0ddb9064f2731c5
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Friday, April 10, 2026, Isaac Morland &lt;<a href=3D"mailto:isaac.morlan=
d@gmail.com">isaac.morland@gmail.com</a>&gt; wrote:<br><blockquote class=3D=
"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding=
-left:1ex"><div dir=3D"ltr">The documentation says that only BEGIN ATOMIC p=
arses the function body at definition time. Whereas AS parses the function =
body at execution time:<div><br></div><div><a href=3D"https://www.postgresq=
l.org/docs/current/sql-createfunction.html" target=3D"_blank">https://www.p=
ostgresql.org/<wbr>docs/current/sql-<wbr>createfunction.html</a><br></div><=
div><br></div><div>(and look at the explanation of sql_body)</div></div></b=
lockquote><div><br></div><div>Possibly some documentation tweaks are in ord=
er.=C2=A0 I haven=E2=80=99t looked at it in depth.=C2=A0 The optional check=
s are able to be disabled:</div><div><br></div><div><a href=3D"https://www.=
postgresql.org/docs/current/runtime-config-client.html#RUNTIME-CONFIG-CLIEN=
T-STATEMENT">https://www.postgresql.org/docs/current/runtime-config-client.=
html#RUNTIME-CONFIG-CLIENT-STATEMENT</a><br></div><div><br></div><div>While=
 the result of the required parsing of a function body is implying that suc=
h parsing is saved to the catalogs so that creation-time object references =
are retained.=C2=A0 While the others are re-parsed during execution and obj=
ect references retained only for the lifetime of the query.</div><div><br><=
/div><div>David J.</div><div><br></div><div></div><blockquote class=3D"gmai=
l_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left=
:1ex">
</blockquote>

--000000000000f0ddb9064f2731c5--