Index Only Scan support for cube
Hi, hackers!
Here's a small patch that implements fetch function necessary for
Index Only Scans that use cube data type.
I reuse function g_cube_decompress() instead of creating new function
g_cube_fetch().
Essentially, they both have to detoast data.
How do you think, is it better to create a shallow copy of
g_cube_decompress instead?
Any other suggestions on the functionality?
This
CREATE TABLE SOMECUBES AS SELECT CUBE(X,X+1) C FROM GENERATE_SERIES(1,100) X;
CREATE INDEX SOMECUBES_IDX ON SOMECUBES USING GIST(C);
SET ENABLE_SEQSCAN = FALSE;
EXPLAIN (COSTS OFF ) SELECT C FROM SOMECUBES WHERE C<@CUBE(30,40);
now produces
Index Only Scan using somecubes_idx on somecubes
Index Cond: (c <@ '(30),(40)'::cube)
instead of
Index Scan using somecubes_idx on somecubes
Index Cond: (c <@ '(30),(40)'::cube)
Best regards, Andrey Borodin, Octonica.
Attachments:
cubefetch.patchapplication/octet-stream; name=cubefetch.patchDownload+35-2
Hi hackers!
23 мая 2017 г., в 14:53, Andrew Borodin <borodin@octonica.com> написал(а):
Here's a small patch that implements fetch function necessary for
Index Only Scans that use cube data type.
Tom Lane have just commited d3a4f89 (Allow no-op GiST support functions to be omitted) Thanks, Tom! : )
"Index Only Scan support for cube" patch now is obsolete. I'm working on another similar patch for contribs to support GiST IOS and remove no-op support functions.
Best regards, Andrey Borodin.
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
On Wed, Sep 20, 2017 at 8:26 AM, Andrey Borodin <x4mmm@yandex-team.ru>
wrote:
Hi hackers!
23 мая 2017 г., в 14:53, Andrew Borodin <borodin@octonica.com>
написал(а):
Here's a small patch that implements fetch function necessary for
Index Only Scans that use cube data type.Tom Lane have just commited d3a4f89 (Allow no-op GiST support functions to
be omitted) Thanks, Tom! : )
"Index Only Scan support for cube" patch now is obsolete. I'm working on
another similar patch for contribs to support GiST IOS and remove no-op
support functions.
Good.
BTW, some strangeness of g_cube_decompress() catch my eye. It compares
results of two evaluations of same expression DatumGetNDBOXP(entry->key).
NDBOX *key = DatumGetNDBOXP(entry->key);
if (key != DatumGetNDBOXP(entry->key))
In fact it's correct, because it compares results of two detoasting. If
datum isn't toasted then results would be the same. And if data is toasted
then results would be two different allocation of detoasted datum.
However, we do extra detoasting here.
For example, see gbt_var_decompress(). There is no extra detoasting here.
GBT_VARKEY *key = (GBT_VARKEY *) PG_DETOAST_DATUM(entry->key);
if (key != (GBT_VARKEY *) DatumGetPointer(entry->key))
------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company