there is basic support of GROUPING SETS. This patch isn't ready for
production! It's only material for discus. Parser and executor are
done. Planner part should be completely rewritten. I don't understand
well to planner, and I haven't idea that parts should be shared with
groupby. Implementation is based on using hash tables that holds
result. It implements push mode via new node feeder.
Parser is ugly - there was two problems - 1. collision '(' ... ')'
between set spec and ROW constructor, 2. identificator cube shouldn't
be keyword (compatibility with cube contrib module).
I invite any help or comments. Without help I am not able to prepare
this patch for last commitfest. Help is necessary in two parts -
documentation, planner and testing. This version is based on CVS HEAD
before CTE patch!
* only hash aggregates are supported
* doe's somebody understand what's mean ALL in GROUP BY CLAUSE?
Description: application/x-gzip (24.8 KB)
pgsql-hackers by date
|Next:||From: Richard Huxton||Date: 2008-10-17 11:17:49|
|Subject: Re: Cross-column statistics revisited|
|Previous:||From: Vladimir Sitnikov||Date: 2008-10-17 09:30:50|
|Subject: Re: contrib/pg_stat_statements|