Horton

Re: Table AM Interface Enhancements

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Alexander Korotkov <aekorotkov@gmail.com>
Cc: Melanie Plageman <melanieplageman@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Pavel Borisov <pashkin.elfe@gmail.com>, Japin Li <japinli@hotmail.com>, Mark Dilger <mark.dilger@enterprisedb.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Michael Paquier <michael@paquier.xyz>, Heikki Linnakangas <hlinnaka@iki.fi>, Thomas Munro <thomas.munro@gmail.com>, Nazir Bilal Yavuz <byavuz81@gmail.com>
Date: 2024-04-15T20:10:57Z
Lists: pgsql-hackers 
Hi,

On 2024-04-12 01:04:03 +0300, Alexander Korotkov wrote:
> 1) If we just apply my revert patch and leave c6fc50cb4028 and
> 041b96802ef in the tree, then we get our table AM API narrowed.  As
> you expressed the current API requires block numbers to be 1:1 with
> the actual physical on-disk location [2].  Not a secret I think the
> current API is quite restrictive.  And we're getting the ANALYZE
> interface narrower than it was since 737a292b5de.  Frankly speaking, I
> don't think this is acceptable.

As others already pointed out, c6fc50cb4028 was committed quite a while
ago. I'm fairly unhappy about c6fc50cb4028, fwiw, but didn't realize that
until it was too late.


> In token of all of the above, is the in-tree state that bad? (if we
> abstract the way 27bc1772fc and dd1f6b0c17 were committed).

To me the 27bc1772fc doesn't make much sense on its own. You added calls
directly to heapam internals to a file in src/backend/commands/, that just
doesn't make sense.

Leaving that aside, I think the interface isn't good on its own:
table_relation_analyze() doesn't actually do anything, it just sets callbacks,
that then later are called from analyze.c, which doesn't at all fit to the
name of the callback/function.  I realize that this is kinda cribbed from the
FDW code, but I don't think that is a particularly good excuse.

I don't think dd1f6b0c17 improves the situation, at all. It sets global
variables to redirect how an individual acquire_sample_rows invocation
works:
void
block_level_table_analyze(Relation relation,
						  AcquireSampleRowsFunc *func,
						  BlockNumber *totalpages,
						  BufferAccessStrategy bstrategy,
						  ScanAnalyzeNextBlockFunc scan_analyze_next_block_cb,
						  ScanAnalyzeNextTupleFunc scan_analyze_next_tuple_cb)
{
	*func = acquire_sample_rows;
	*totalpages = RelationGetNumberOfBlocks(relation);
	vac_strategy = bstrategy;
	scan_analyze_next_block = scan_analyze_next_block_cb;
	scan_analyze_next_tuple = scan_analyze_next_tuple_cb;
}

Notably it does so within the ->relation_analyze tableam callback, which does
*NOT* not actually do anything other than returning a callback.  So if
->relation_analyze() for another relation is called, the acquire_sample_rows()
for the earlier relation will do something different.  Note that this isn't a
theoretical risk, acquire_inherited_sample_rows() actually collects the
acquirefunc for all the inherited relations before calling acquirefunc.

This is honestly leaving me somewhat speechless.

Greetings,

Andres Freund

Commits

  1. Remove extra comment at TableAmRoutine.scan_analyze_next_block

  2. revert: Generalize relation analyze in table AM interface

  3. Revert: Allow table AM to store complex data structures in rd_amcache

  4. Revert: Allow table AM tuple_insert() method to return the different slot

  5. Revert: Allow locking updated tuples in tuple_update() and tuple_delete()

  6. Revert: Let table AM insertion methods control index insertion

  7. Revert: Custom reloptions for table AM

  8. Provide a way block-level table AMs could re-use acquire_sample_rows()

  9. Custom reloptions for table AM

  10. Use streaming I/O in ANALYZE.

  11. Revert "Custom reloptions for table AM"

  12. Let table AM insertion methods control index insertion

  13. Generalize relation analyze in table AM interface

  14. Improve error message for tts_(virtual|minimal)_is_current_xact_tuple

  15. Add comments on some MinimalTupleSlots methods usage

  16. Add TupleTableSlotOps.is_current_xact_tuple() method

  17. Allow table AM tuple_insert() method to return the different slot

  18. Allow table AM to store complex data structures in rd_amcache