Hi, Thomas Munro and Laurenz Albe.
Since I didn't subscribe to the psql-hackers mailing list before this bug
was raised, please forgive me for not being able to reply to this email by
placing the email message below.
https://www.postgresql.org/message-id/flat/[email protected]
I forbid to create indexes on whole-row expression in the following patch.
I'd like to hear your opinions.
--
Best Wishes,
ywgrit
diff --git a/src/backend/commands/indexcmds.c b/src/backend/commands/indexcmds.c
index cd23ab3b25..e4451b1d36 100644
--- a/src/backend/commands/indexcmds.c
+++ b/src/backend/commands/indexcmds.c
@@ -572,9 +572,18 @@ DefineIndex(Oid relationId,
Oid root_save_userid;
int root_save_sec_context;
int root_save_nestlevel;
+ ListCell *lc;
root_save_nestlevel = NewGUCNestLevel();
+ foreach (lc, stmt->indexParams)
+ {
+ IndexElem *ielem = castNode(IndexElem, lfirst(lc));
+ if (IsA(ielem->expr, Var) && castNode(Var, ielem->expr)->varattno == 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
+ errmsg("cannot create index on whole-row expression of table '%s'", ielem->indexcolname)));
+ }
/*
* Some callers need us to run with an empty default_tablespace; this is a
* necessary hack to be able to reproduce catalog state accurately when
