From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Eisentraut <peter@eisentraut.org>
cc: pgsql-bugs <pgsql-bugs@postgresql.org>
Subject: Re: CREATE CAST allows creation of binary-coercible cast to range
 over domain
In-reply-to: <1537781.1724165175@sss.pgh.pa.us>
References: <076968e1-0852-40a9-bc0b-117cd3f0e43c@eisentraut.org>
 <1537781.1724165175@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Tue, 20 Aug 2024 10:46:15 -0400"
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="----- =_aaaaaaaaaa0"
Content-ID: <1611494.1724175016.0@sss.pgh.pa.us>
Date: Tue, 20 Aug 2024 13:30:39 -0400
Message-ID: <1611527.1724175039@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1611527.1724175039%40sss.pgh.pa.us>
Precedence: bulk

------- =_aaaaaaaaaa0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1611494.1724175016.1@sss.pgh.pa.us>

I wrote:
> I think it's an oversight that we allow binary-coercible casts
> involving range types at all.  There are no other nesting cases to
> worry about, because CreateCast already rejects binary-coercible casts
> applied to composites and arrays, explaining
>          * We know that composite, enum and array types are never binary-
>          * compatible with each other.  They all have OIDs embedded in them.
> But range types also embed their own OID, so I don't see why that
> concern doesn't apply to them.

In short, more or less as attached.  (I didn't bother with a
regression test, since none of the adjacent error checks are
covered either.)

I'm unsure whether to back-patch this.  If somebody were using
such a cast today, it might seem to mostly work, as long as the
two kinds of ranges had the same behavior.  So they might not
appreciate us breaking it in a minor release.  Maybe we should
put this into 17 but not further back.

			regards, tom lane


------- =_aaaaaaaaaa0
Content-Type: text/x-diff; name="reject-binary-coercions-on-ranges.patch";
	charset="us-ascii"
Content-ID: <1611494.1724175016.2@sss.pgh.pa.us>
Content-Description: reject-binary-coercions-on-ranges.patch
Content-Transfer-Encoding: quoted-printable

diff --git a/src/backend/commands/functioncmds.c b/src/backend/commands/fu=
nctioncmds.c
index 6593fd7d81..149642ab1b 100644
--- a/src/backend/commands/functioncmds.c
+++ b/src/backend/commands/functioncmds.c
@@ -1689,13 +1689,18 @@ CreateCast(CreateCastStmt *stmt)
 					 errmsg("source and target data types are not physically compatible"=
)));
 =

 		/*
-		 * We know that composite, enum and array types are never binary-
-		 * compatible with each other.  They all have OIDs embedded in them.
+		 * We know that composite, array, range and enum types are never
+		 * binary-compatible with each other.  They all have OIDs embedded in
+		 * them.
 		 *
 		 * Theoretically you could build a user-defined base type that is
-		 * binary-compatible with a composite, enum, or array type.  But we
-		 * disallow that too, as in practice such a cast is surely a mistake.
-		 * You can always work around that by writing a cast function.
+		 * binary-compatible with such a type.  But we disallow that too, as
+		 * in practice such a cast is surely a mistake.  You can always work
+		 * around that by writing a cast function.
+		 *
+		 * NOTE: if we didn't exclude every kind of container type for this
+		 * reason, we'd likely need to recursively apply all of these same
+		 * checks to the contained type(s).
 		 */
 		if (sourcetyptype =3D=3D TYPTYPE_COMPOSITE ||
 			targettyptype =3D=3D TYPTYPE_COMPOSITE)
@@ -1703,6 +1708,14 @@ CreateCast(CreateCastStmt *stmt)
 					(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
 					 errmsg("composite data types are not binary-compatible")));
 =

+		if (sourcetyptype =3D=3D TYPTYPE_RANGE ||
+			targettyptype =3D=3D TYPTYPE_RANGE ||
+			sourcetyptype =3D=3D TYPTYPE_MULTIRANGE ||
+			targettyptype =3D=3D TYPTYPE_MULTIRANGE)
+			ereport(ERROR,
+					(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
+					 errmsg("range data types are not binary-compatible")));
+
 		if (sourcetyptype =3D=3D TYPTYPE_ENUM ||
 			targettyptype =3D=3D TYPTYPE_ENUM)
 			ereport(ERROR,

------- =_aaaaaaaaaa0--