| From: | Peter Geoghegan <pg(at)heroku(dot)com> |
|---|---|
| To: | Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> |
| Cc: | Pg Hackers <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: failures with tuplesort and ordered set aggregates (due to 5cefbf5a6c44) |
| Date: | 2015-02-20 21:01:17 |
| Message-ID: | CAM3SWZSm4GyfjBN1VmzzbtPXK7N53LVEKprJwkW8LBc=UkAMHg@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Fri, Feb 20, 2015 at 11:58 AM, Tomas Vondra
<tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
> This seems to happen because ordered_set_startup() calls
> tuplesort_begin_datum() when (use_tuples == true), which only sets
> 'onlyKey' and leaves (sortKeys == NULL). So 'mergeruns' fails because it
> does not expect that.
Oops. You're right. Attached patch fixes the issue, by making
tuplesort_begin_datum() consistent with other comparable routines for
other tuple cases. I think it makes sense to have the 'sortKeys' field
always set, and the 'onlyKey' field also set when that additional
optimization makes sense. That was what I understood the API to be, so
arguably this is a pre-existing issue with tuplesort_begin_datum().
Thanks for the report!
--
Peter Geoghegan
| Attachment | Content-Type | Size |
|---|---|---|
| fix_datumsort_fieled.patch | text/x-patch | 1.3 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Eric Grinstein | 2015-02-20 21:05:13 | Idea: GSoC - Query Rewrite with Materialized Views |
| Previous Message | Tomas Vondra | 2015-02-20 20:57:00 | Re: Re: Abbreviated keys for Datum tuplesort |