Horton

Re: POC, WIP: OR-clause support for indexes

Alena Rybakina <a.rybakina@postgrespro.ru>

View thread
From: Alena Rybakina <a.rybakina@postgrespro.ru>
To: Alexander Korotkov <aekorotkov@gmail.com>
Cc: pgsql-hackers@postgresql.org, Peter Geoghegan <pg@bowt.ie>, Robert Haas <robertmhaas@gmail.com>, "Finnerty, Jim" <jfinnert@amazon.com>, Marcos Pegoraro <marcos@f10.com.br>, Andrey Lepikhov <a.lepikhov@postgrespro.ru>, teodor@sigaev.ru, Ranier Vilela <ranier.vf@gmail.com>, Tomas Vondra <tomas.vondra@enterprisedb.com>, Peter Eisentraut <peter@eisentraut.org>
Date: 2023-11-28T03:52:20Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Make group_similar_or_args() reorder clause list as little as possible

  2. Allow usage of match_orclause_to_indexcol() for joins

  3. Skip not SOAP-supported indexes while transforming an OR clause into SAOP

  4. Remove the wrong assertion from match_orclause_to_indexcol()

  5. Teach bitmap path generation about transforming OR-clauses to SAOP's

  6. Transform OR-clauses to SAOP's during index matching

  7. Fix the value of or_to_any_transform_limit in postgresql.conf.sample

  8. Transform OR clauses to ANY expression

  9. MergeAttributes code deduplication

  10. SEARCH and CYCLE clauses

  11. Improve estimation of OR clauses using extended statistics.

  12. Teach btree to handle ScalarArrayOpExpr quals natively.

  13. Revise collation derivation method and expression-tree representation.

  14. Instead of trying to force WHERE clauses into CNF or DNF normal form,

Attachments

On 25.11.2023 19:13, Alexander Korotkov wrote:
> On Sat, Nov 25, 2023 at 1:10 PM Alena Rybakina
> <a.rybakina@postgrespro.ru> wrote:
>> On 25.11.2023 04:13, Alexander Korotkov wrote:
>>
>> It seems to me there is a confusion.  I didn't mean we need to move
>> conversion of OR-expressions to ANY into choose_bitmap_and() function
>> or anything like this.  My idea was to avoid degradation of plans,
>> which I've seen in [1].  Current code for generation of bitmap paths
>> considers the possibility to split OR-expressions into distinct bitmap
>> index scans.  But it doesn't consider this possibility for
>> ANY-expressions.  So, my idea was to enhance our bitmap scan
>> generation to consider split values of ANY-expressions into distinct
>> bitmap index scans.  So, in the example [1] and similar queries
>> conversion of OR-expressions to ANY wouldn't affect the generation of
>> bitmap paths.
>>
>> Thanks for the explanation, yes, I did not understand the idea correctly at first. I will try to implement something similar.
> Alena, great, thank you.  I'm looking forward to the updated patch.
>
I wrote the patch (any_to_or.diff.txt), although it is still under 
development (not all regression tests have been successful so far), it 
is already clear that for a query where a bad plan was chosen before, it 
is now choosing a more optimal query plan.

postgres=# set enable_or_transformation =on;
SET
postgres=# explain select * from test where (x = 1 or x = 2) and y = 100;
                                                   QUERY PLAN
--------------------------------------------------------------------------------------------------------------
  Bitmap Heap Scan on test  (cost=8.60..12.62 rows=1 width=12)
    Recheck Cond: (((y = '100'::double precision) AND (x = 1)) OR ((y = 
'100'::double precision) AND (x = 2)))
    ->  BitmapOr  (cost=8.60..8.60 rows=1 width=0)
          ->  Bitmap Index Scan on test_x_1_y  (cost=0.00..4.30 rows=1 
width=0)
                Index Cond: (y = '100'::double precision)
          ->  Bitmap Index Scan on test_x_2_y  (cost=0.00..4.30 rows=1 
width=0)
                Index Cond: (y = '100'::double precision)
(7 rows)

While I'm thinking how to combine it now.

BTW, while I was figuring out how create_index_paths works and creating 
bitmapscan indexes, I think I found a bug with unallocated memory (fix 
patch is bugfix.diff.txt). Without a fix here, it falls into the crust 
at the stage of assigning a value to any of the variables, specifically, 
skip_lower_stop and skip_nonnative_saop. I discovered it when I forced 
to form a bitmapindex plan for ANY (any_to_or.diff.txt). I'm causing a 
problem with my OR->ANY conversion patch.

-- 
Regards,
Alena Rybakina
Postgres Professional