In the recent discussion of bug #3883 we decided that for safety's sake, TRUNCATE, CLUSTER, and REINDEX ought to error out if there are any active scans on the table (or index in the case of REINDEX). This is essentially the same as the test currently applied by ALTER TABLE, which uses this code:
static void CheckTableNotInUse(Relation rel) { int expected_refcnt; expected_refcnt = rel->rd_isnailed ? 2 : 1; if (rel->rd_refcnt != expected_refcnt) ereport(ERROR, (errcode(ERRCODE_OBJECT_IN_USE), errmsg("relation \"%s\" is being used by active queries in this session", RelationGetRelationName(rel)))); if (AfterTriggerPendingOnRel(RelationGetRelid(rel))) ereport(ERROR, (errcode(ERRCODE_OBJECT_IN_USE), errmsg("cannot alter table \"%s\" because it has pending trigger events", RelationGetRelationName(rel)))); } I would like to export this routine and have it be used by all four commands, instead of duplicating this logic everywhere. However, that brings up the question of whether the error messages are generic enough for all four commands; and if not, how we want them to read. I'm tempted to rephrase both messages along the line of cannot %s \"%s\" because ... where the first %s is replaced by a SQL command name, viz ALTER TABLE, CLUSTER, etc. I'm not sure how nice this is for translation though. Also, with 8.3 release being so close, it's likely that any change would not get reflected into translations before release. I don't think that's a showstopper because these messages should hardly ever be seen by normal users anyway; but maybe it's a consideration. Comments, better ideas? regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 6: explain analyze is your friend