Horton

Re: Table AM Interface Enhancements

Alexander Korotkov <aekorotkov@gmail.com>

View thread View on postgr.es
From: Alexander Korotkov <aekorotkov@gmail.com>
To: Mark Dilger <mark.dilger@enterprisedb.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2023-11-25T17:47:57Z
Lists: pgsql-hackers 
On Fri, Nov 24, 2023 at 5:18 PM Mark Dilger
<mark.dilger@enterprisedb.com> wrote:
> > On Nov 23, 2023, at 4:42 AM, Alexander Korotkov <aekorotkov@gmail.com> wrote:
>
>
> > 0006-Generalize-table-AM-API-for-INSERT-.-ON-CONFLICT-v1.patch
> >
> > Provides a new table AM API method to encapsulate the whole INSERT ...
> > ON CONFLICT ... algorithm rather than just implementation of
> > speculative tokens.
>
> I *think* I understand that you are taking the part of INSERT..ON CONFLICT that lives outside the table AM and pulling it inside so that table AM authors are free to come up with whatever implementation is more suited for them.  The most straightforward way of doing so results in an EState parameter in the interface definition.  That seems not so good, as the EState is a fairly complicated structure, and future changes in the executor might want to rearrange what EState tracks, which would change which values tuple_insert_with_arbiter() can depend on.

I think this is the correct understanding.

> Should the patch at least document which parts of the EState are expected to be in which states, and which parts should be viewed as undefined?  If the implementors of table AMs rely on any/all aspects of EState, doesn't that prevent future changes to how that structure is used?

New tuple tuple_insert_with_arbiter() table AM API method needs EState
argument to call executor functions: ExecCheckIndexConstraints(),
ExecUpdateLockMode(), and ExecInsertIndexTuples().  I think we
probably need to invent some opaque way to call this executor function
without revealing EState to table AM.  Do you think this could work?

> > 0008-Let-table-AM-insertion-methods-control-index-inse-v1.patch
> >
> > Allows table AM to skip index insertions in the executor and handle
> > those insertions itself.
>
> The new parameter could use more documentation.
>
> > 0009-Custom-reloptions-for-table-AM-v1.patch
> >
> > Enables table AMs to define and override reloptions for tables and indexes.
>
> This could use some regression tests to exercise the custom reloptions.

Thank you for these notes.  I'll take this into account for the next
patchset version.

------
Regards,
Alexander Korotkov

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