| From: | James William Pye <flaw(at)rhid(dot)com> |
|---|---|
| To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
| Cc: | Hackers <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: SRFs ExecMakeTableFunctionResult |
| Date: | 2004-08-09 17:32:06 |
| Message-ID: | 20040809173206.GA18051@void.ph.cox.net |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
> Seems reasonable to me. A SRF function really ought to explicitly set
> isDone on every call anyway.
Aye, it seems reasonable, but a bit inconsistent with the effect of
ExecMakeFunctionResult, which does the same thing but bases the continuity of
the result gathering on the isDone pointer, which is set to
ExprMultipleResult if isDone is not ExprEndResult, thus making it continue
until rsinfo.isDone is explicitly set to ExprEndResult, unlike table
functions which will end on either SingleResult or EndResult.
(Around lines #941-984 in execQual.c)
Is this inconsistency desired?
My confusion came in when I implemented SRFs that worked with non-table
SRFs, and then table functions didn't work because I wasn't setting isDone
to MultipleResult every call.
--
Regards,
James William Pye
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Jan Wieck | 2004-08-09 17:44:59 | Re: VACUUM DELAY |
| Previous Message | Stefan Kaltenbrunner | 2004-08-09 17:23:41 | Tablespace issues (comment on ,moving indexes) |