Hi hackers,
Please find attached a patch proposal to split index and table
statistics into different types of stats.
This idea has been proposed by Andres in a couple of threads, see [1]
and [2].
To sum up:
We currently track index and table types of statistics in the same
format (so that a number of the "columns" in index stats are currently
unused) and we rename column in views etc to make them somewhat sensible.
So that the immediate benefits to $SUBJECT are:
- have reasonable names for the fields
- shrink the current memory usage
The attached patch proposal:
- renames PGSTAT_KIND_RELATION to PGSTAT_KIND_TABLE
- creates a new PGSTAT_KIND_INDEX
- creates new macros: pgstat_count_index_fetch(),
pgstat_count_index_scan(), pgstat_count_index_tuples(),
pgstat_count_index_buffer_read() and pgstat_count_index_buffer_hit() to
increment the indexes related stats
- creates new SQL callable functions dedicated to the indexes that are
used in system_views.sql
It also adds basic tests in src/test/regress/sql/stats.sql for toast and
partitions (we may want to create a dedicated patch for those additional
tests though).
The fields renaming has not been done to ease the reading of this patch
(I think it would be better to create a dedicated patch for the renaming
once the split is done).
I'm adding a new CF entry for this patch.
Looking forward to your feedback,
Regards,
--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
[1]:
https://www.postgresql.org/message-id/flat/20221019181930.bx73kul4nbiftr65%40awork3.anarazel.de
[2]:
https://www.postgresql.org/message-id/20220818195124.c7ipzf6c5v7vxymc%40awork3.anarazel.de
