From: Tom Lane <tgl@sss.pgh.pa.us>
> I think either the bit about rule_action is unnecessary, or most of
> the code immediately above this is wrong, because it's only updating
> flags in sub_action. Why do you think it's necessary to change
> rule_action in addition to sub_action?
Finally, I think I've understood what you meant. Yes, the current code seems to be wrong. rule_action is different
fromsub_action only when the rule action (the query specified in CREATE RULE) is INSERT SELECT. In that case,
rule_actionpoints to the entire INSERT SELECT, while sub_action points to the SELECT part. So, we should add the CTE
listand set hasModifyingCTE/hasRecursive flags in rule_action.
> Hm. So after looking at this more, the problem is that the rewrite
> is producing something equivalent to
>
> INSERT INTO bug6051_2
> (WITH t1 AS (DELETE FROM bug6051 RETURNING *) SELECT * FROM t1);
Yes. In this case, the WITH clause must be put before INSERT.
The attached patch is based on your version. It includes cosmetic changes to use = instead of |= for boolean variable
assignments. make check passed. Also, Greg-san's original failed test case succeeded. I confirmed that the
hasModifyingCTEof the top-level rewritten query is set to true now by looking at the output of debug_print_rewritten
anddebug_print_plan.
Regards
Takayuki Tsunakawa
