On 12/12/24 10:09, David Rowley wrote:
On Mon, 2 Dec 2024 at 17:18, Andrei Lepikhov wrote:
Patch 0002 looks helpful and performant. I propose to check 'relid > 0'
to avoid diving into 'foreach(lc, parse->rtable)' at all if nothing has
been found.
I did end up adding another fast path there, b
On Mon, 2 Dec 2024 at 17:18, Andrei Lepikhov wrote:
> Patch 0002 looks helpful and performant. I propose to check 'relid > 0'
> to avoid diving into 'foreach(lc, parse->rtable)' at all if nothing has
> been found.
I did end up adding another fast path there, but I felt like checking
relid > 0 was
On Mon, 2 Dec 2024 at 17:22, jian he wrote:
> regarding v10.
> you placed remove_useless_groupby_columns right after add_base_rels_to_query
> makes so much sense.
> so we can be safely use cached RelOptInfo->indexlist,
> RelOptInfo->notnullattnums
> overall it didn't find any issue.
Thanks for l
On Fri, Nov 29, 2024 at 5:20 PM David Rowley wrote:
>
> The reason I don't think is possible is that we have no infrastructure
> that allows us to tag functions or operators so that non-null input(s)
> mean non-null outputs. We only have strict, which means null input
> means null output. That's
On 29/11/2024 14:28, David Rowley wrote:
On Fri, 29 Nov 2024 at 19:36, Andrei Lepikhov wrote:
1. Thread reference in the patch comment doesn't work.
2. May it be possible to move remove_useless_groupby_columns in the
second commit? It would be easier to review the differences.
For #1, I rewro
On Fri, 29 Nov 2024 at 20:04, jian he wrote:
> I found that unique expression index can also be used for groupby
> column removal.
> so I implemented it, aslo added two tests on it.
> what do you think?
I think it's likely just not common enough to be worthwhile, plus,
unfortunately, I don't thi
On Fri, 29 Nov 2024 at 19:36, Andrei Lepikhov wrote:
> 1. Thread reference in the patch comment doesn't work.
> 2. May it be possible to move remove_useless_groupby_columns in the
> second commit? It would be easier to review the differences.
For #1, I rewrote the commit message.
I've split into
On Fri, Nov 29, 2024 at 2:36 PM Andrei Lepikhov wrote:
>
> > I forgot to do a local commit before sending v8. Fixed in the attached v9.
> 1. Thread reference in the patch comment doesn't work.
fixed.
I found that unique expression index can also be used for groupby
column removal.
so I implemente
On 11/29/24 09:39, David Rowley wrote:
On Fri, 29 Nov 2024 at 15:02, David Rowley wrote:
I've attached an updated patch that gets rid of the
get_unique_not_null_attnos() function completely and uses the
RelOptInfo.indexlist and RelOptInfo.notnullattnums.
I forgot to do a local commit before s
On Fri, 29 Nov 2024 at 15:02, David Rowley wrote:
> I've attached an updated patch that gets rid of the
> get_unique_not_null_attnos() function completely and uses the
> RelOptInfo.indexlist and RelOptInfo.notnullattnums.
I forgot to do a local commit before sending v8. Fixed in the attached v9.
On Fri, 29 Nov 2024 at 01:33, jian he wrote:
>
> On Thu, Nov 28, 2024 at 2:11 PM David Rowley wrote:
> > I think it'll make more sense to adjust some of the
> > existing tests to use a unique constraint instead of a PK and then
> > adjust a column's NOT NULL property to check that part of the cod
minor changes in get_unique_not_null_attnos:
* cosmetic changes
* if the returned list length is less than 2, no need sort.
From 58094504364d8a9eb51cbc77a75626aef39b17f0 Mon Sep 17 00:00:00 2001
From: jian he
Date: Fri, 29 Nov 2024 09:23:27 +0800
Subject: [PATCH v7 1/1] remove useless group by co
On Thu, Nov 28, 2024 at 2:11 PM David Rowley wrote:
>
> I think it'll make more sense to adjust some of the
> existing tests to use a unique constraint instead of a PK and then
> adjust a column's NOT NULL property to check that part of the code is
> working correctly.
>
looking around, i insert
On Thu, 28 Nov 2024 at 17:39, jian he wrote:
> v4 logic is to choose one with the least number of columns.
> if there is more than one with the least number of columns, simply
> choose the first one
> in the matched list.
The new code inside remove_useless_groupby_columns() is still more
complex
On Wed, Nov 27, 2024 at 5:30 PM David Rowley wrote:
>
> On Wed, 27 Nov 2024 at 19:51, jian he wrote:
> > v3 attached. in v3:
>
> 1. I find the following quite strange:
>
> postgres=# create table abcd (a int primary key, b int not null, c int
> not null, d int not null, unique(b));
> CREATE TABLE
On Wed, 27 Nov 2024 at 19:51, jian he wrote:
> v3 attached. in v3:
1. I find the following quite strange:
postgres=# create table abcd (a int primary key, b int not null, c int
not null, d int not null, unique(b));
CREATE TABLE
postgres=# explain select b,c from abcd group by b,c;
On Fri, Nov 22, 2024 at 9:24 PM jian he wrote:
>
> overall i come up with the attached patch.
in v2:
create table t1 (a int, b int not null, c int, unique(b, c));
explain(costs off) select count(*) from t1 group by b,c,a;
QUERY PLAN
--
HashAggregate
Group Key: b
-
On Thu, Sep 12, 2024 at 9:44 AM David Rowley wrote:
>
> On Sat, 30 Dec 2023 at 04:05, Zhang Mingli wrote:
> > So my patch make it easy: check unique index’s columns, it’s a valid
> > candidate if all of that have NOT NULL constraint.
> > And we choose a best one who has the least column numbers
Hi, all
I haven't paid attention to this topic in a long time, thanks all for the
advices, I will study them then update.
Thanks again.
Zhang Mingli
www.hashdata.xyz
On Sep 18, 2024 at 15:50 +0800, David Rowley , wrote:
> On Wed, 18 Sept 2024 at 19:28, Peter Eisentraut wrote:
> >
> > On 12.0
On Wed, 18 Sept 2024 at 19:28, Peter Eisentraut wrote:
>
> On 12.09.24 03:43, David Rowley wrote:
> > (Likely it could just look at pg_attribute.attnotnull instead)
>
> That won't work because you can't record dependencies on that. (This is
> one of the reasons for cataloging not-null constraints
On 12.09.24 03:43, David Rowley wrote:
On Sat, 30 Dec 2023 at 04:05, Zhang Mingli wrote:
So my patch make it easy: check unique index’s columns, it’s a valid candidate
if all of that have NOT NULL constraint.
And we choose a best one who has the least column numbers in
get_min_unique_not_null
On Sat, 30 Dec 2023 at 04:05, Zhang Mingli wrote:
> So my patch make it easy: check unique index’s columns, it’s a valid
> candidate if all of that have NOT NULL constraint.
> And we choose a best one who has the least column numbers in
> get_min_unique_not_null_attnos(), as the reason: less col
On Fri, Dec 29, 2023 at 11:05 PM Zhang Mingli wrote:
>
> Hi,
>
> This idea first came from remove_useless_groupby_columns does not need to
> record constraint dependencie[0] which points out that
> unique index whose columns all have NOT NULL constraints could also take the
> work with primary
Hi,
This idea first came from remove_useless_groupby_columns does not need to
record constraint dependencie[0] which points out that
unique index whose columns all have NOT NULL constraints could also take the
work with primary key when removing useless GROUP BY columns.
I study it and implemen
24 matches
Mail list logo
