From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
cc: Sami Imseih <samimseih@gmail.com>, Christoph Berg <myon@debian.org>,
        Lukas Fittl <lukas@fittl.com>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
        ma lz <ma100@hotmail.com>
Subject: Re: query_id: jumble names of temp tables for better pg_stat_statement UX
In-reply-to: <Z-M32Cm7FM-1gpxB@paquier.xyz>
References: <461405.1742691859@sss.pgh.pa.us> <Z9-sUvq8pE5uinEu@paquier.xyz> <CAA5RZ0t4XQ_SANiJ5VfuoKnAuxMaY2ggQWoGKkHo+U2H+Sh-Sw@mail.gmail.com> <1189112.1742869660@sss.pgh.pa.us> <CAA5RZ0uNofEXfEfNw3uRN3D3oXkFPQ_s+huLLHMKR_+MCk8RPQ@mail.gmail.com> <Z-IoMKL8I-r-xsuW@paquier.xyz> <CAP53PkzAPrHNHquT76mVjfStrz0B2hKhF+CbAZ_w9QJUJ_1oJQ@mail.gmail.com> <Z-JU8EmGdlnNEGot@paquier.xyz> <Z-LOknecFpjVjhkP@msg.df7cb.de> <CAA5RZ0sy8gZzK0qNScY-C_jOpqQb6Y3EQaXHu0Ti9pon-4jJBg@mail.gmail.com> <Z-M32Cm7FM-1gpxB@paquier.xyz>
Comments: In-reply-to Michael Paquier <michael@paquier.xyz>
	message dated "Wed, 26 Mar 2025 08:10:16 +0900"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1324035.1742945060.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Tue, 25 Mar 2025 19:24:20 -0400
Message-ID: <1324036.1742945060@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/1324036.1742945060%40sss.pgh.pa.us>
Precedence: bulk

Michael Paquier <michael@paquier.xyz> writes:
> [ v6-0001-Add-custom-query-jumble-function-for-RangeTblEntr.patch ]

Couple of minor review comments ...

* In 5ac462e2b, this bit:

        # node type
-       if (($t =3D~ /^(\w+)\*$/ or $t =3D~ /^struct\s+(\w+)\*$/)
+       if ($query_jumble_custom)
+       {
+           # Custom function that applies to one field of a node.
+           print $jff "\tJUMBLE_CUSTOM($n, $f);\n";
+       }
+       elsif (($t =3D~ /^(\w+)\*$/ or $t =3D~ /^struct\s+(\w+)\*$/)

fails to honor $query_jumble_ignore as the other if-branches do.
Perhaps it's unlikely that a node would have both query_jumble_custom
and query_jumble_ignore annotations, but still, the script would emit
completely incorrect code if it did.  Also, the "# node type" comment
should probably be moved down to within the first "elsif" block.

* In the v6 patch, this doesn't read very smoothly:

+     * Expanded reference names.  This uses a custom query jumble functio=
n so
+     * as the table name is included in the computation, not its list of
+     * columns.

Perhaps better

+     * Expanded reference names.  This uses a custom query jumble functio=
n so
+     * that the table name is included in the computation, but not its li=
st of
+     * columns.

* Also, here:

+   considered the same. However, if the alias for a table is different
+   for semantically similar queries, these queries will be considered
+   distinct.

I'd change "semantically similar queries" to "otherwise-similar
queries"; I think writing "semantically" will just confuse people.

Otherwise LGTM.

			regards, tom lane