From: Tom Lane <tgl@sss.pgh.pa.us>
To: Marcelo Fernandes <marcefern7@gmail.com>
cc: pgsql-general@lists.postgresql.org
Subject: Re: Increasing a NUMERIC column precision doesn't cause a table rewrite. Why?
In-reply-to: <CAM2F1VNniBN2RK9+DD8Pwz4tphE-1w3hZgtwZF-6c08--6JLiw@mail.gmail.com>
References: <CAM2F1VNniBN2RK9+DD8Pwz4tphE-1w3hZgtwZF-6c08--6JLiw@mail.gmail.com>
Comments: In-reply-to Marcelo Fernandes <marcefern7@gmail.com>
	message dated "Mon, 03 Nov 2025 21:24:00 +1300"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <329226.1762187472.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 03 Nov 2025 11:31:12 -0500
Message-ID: <329227.1762187472@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/329227.1762187472%40sss.pgh.pa.us>
Precedence: bulk

Marcelo Fernandes <marcefern7@gmail.com> writes:
> After doing some playing around with different operations against a nume=
ric
> column, namely: increasing/decreasing the precision or increasing/decrea=
sing
> the scale, I noticed that the table is rewritten in all cases except whe=
n
> increasing the precision number.

Yup.

> For example, is there a catalog table I can have a look to see how these=
 values
> are defined, along with some form of documentation to understand how the=
y are
> interpreted by Postgres?

You'd have to look into the source code:

https://github.com/postgres/postgres/blob/c4067383cb2c155c4cfea2351036709e=
2ebb3535/src/backend/utils/adt/numeric.c

The first few hundred lines of numeric.c are largely comments that
you'd find relevant.  Also see numeric_support():

https://github.com/postgres/postgres/blob/c4067383cb2c155c4cfea2351036709e=
2ebb3535/src/backend/utils/adt/numeric.c#L1179

ALTER TABLE skips table rewrite if the datatype conversion expression
is simplified to a no-op by a support function such as
numeric_support.

			regards, tom lane