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

Andrei Lepikhov <lepihov@gmail.com>

From: Andrei Lepikhov <lepihov@gmail.com>

To: Alexander Korotkov <aekorotkov@gmail.com>

Cc: Alena Rybakina <a.rybakina@postgrespro.ru>, Ranier Vilela <ranier.vf@gmail.com>, Alexander Lakhin <exclusion@gmail.com>, jian he <jian.universality@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Nikolay Shaplov <dhyan@nataraj.su>, pgsql-hackers@lists.postgresql.org, pgsql-hackers@postgresql.org, Peter Geoghegan <pg@bowt.ie>, Marcos Pegoraro <marcos@f10.com.br>, teodor@sigaev.ru, Peter Eisentraut <peter@eisentraut.org>

Date: 2025-01-28T04:36:47Z

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 →

Make group_similar_or_args() reorder clause list as little as possible
- 775a06d44c04 18.0 landed
Allow usage of match_orclause_to_indexcol() for joins
- 627d63419e22 18.0 landed
Skip not SOAP-supported indexes while transforming an OR clause into SAOP
- 5bba0546eecb 18.0 landed
Remove the wrong assertion from match_orclause_to_indexcol()
- d4d11940df94 18.0 landed
Teach bitmap path generation about transforming OR-clauses to SAOP's
- ae4569161a27 18.0 landed
Transform OR-clauses to SAOP's during index matching
- d4378c0005e6 18.0 landed
Fix the value of or_to_any_transform_limit in postgresql.conf.sample
- 2af75e117478 17.0 landed
Transform OR clauses to ANY expression
- 72bd38cc99a1 17.0 landed
MergeAttributes code deduplication
- 64444ce071f6 17.0 cited
SEARCH and CYCLE clauses
- 3696a600e229 14.0 cited
Improve estimation of OR clauses using extended statistics.
- 25a9e54d2db3 14.0 cited
Teach btree to handle ScalarArrayOpExpr quals natively.
- 9e8da0f75731 9.2.0 cited
Revise collation derivation method and expression-tree representation.
- b310b6e31ce5 9.1.0 cited
Instead of trying to force WHERE clauses into CNF or DNF normal form,
- 9888192fb773 8.0.0 cited

On 1/27/25 16:50, Alexander Korotkov wrote:
> I expressed my point on this in [1].  We generally greedy about index 
> quals and there is no logic which prevent us from using a clause and 
> index qual because of its cost.  And there are many cases when this 
> causes regressions before d4378c0005.  One of examples from [1].
Ok,
Generally, I don't concern myself with the evaluation of individual 
subplans. As you mentioned, it should be a rare occurrence when this 
becomes important. My main concern is the shift in frequency of 
evaluations during execution for various reasons.
For example:

qsort(matches, n, sizeof(OrArgIndexMatch), or_arg_index_match_cmp);

To fit an index, the order of elements in the target array of the 
`ScalarArrayOpExpr` may change compared to the initial list of OR 
expressions. If there are indexes that cover the same set of columns but 
in reverse order, this could potentially alter the position of a 
Subplan. However, I believe this is a rare case; it is supported by the 
initial OR path and should be acceptable.

So, I do not have any further objections at this time.

-- 
regards, Andrei Lepikhov