Horton

Re: [WIP]Vertical Clustered Index (columnar store extension) - take2

Timur Magomedov <t.magomedov@postgrespro.ru>

View thread
From: Timur Magomedov <t.magomedov@postgrespro.ru>
To: Peter Smith <smithpb2250@gmail.com>, Tomas Vondra <tomas@vondra.me>
Cc: "Aya Iwata (Fujitsu)" <iwata.aya@fujitsu.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2025-06-04T13:47:31Z
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. Mark ItemPointer arguments as const throughout

  2. Reorganize GUC structs

  3. Eliminate XLOG_HEAP2_VISIBLE from vacuum phase III

  4. Update various forward declarations to use typedef

  5. Pathify RHS unique-ification for semijoin planning

  6. Revert "Don't lock partitions pruned by initial pruning"

  7. For inplace update, send nontransactional invalidations.

  8. Remove nearly-unused SizeOfIptrData macro.

Attachments

Hello Peter,

I've tried running VCI, the idea looks great to me. Thank you and
Iwata-san for working on this feature. Looked at the source code of VCI
module, the patch is huge but here are some ideas I hope could be
useful for community. Made some patches on top of v4 that can be
squashed into future iteration of VCI module.

0001-Removed-unnecessary-preload-libraries-checks.patch
Please correct me if I'm wrong but there is no need for adding VCI to
"session_preload_libraries" in case it is in "shared_preload_libraries"
already.

0002-Meson-fixes.patch
Fixed some ghost files in Meson build, added VCI to
contrib/meson.build.

0003-Removed-unused-OS-specific-CAS-functions.patch
Having own VCI-specific CAS looks scary to me, fortunately looks like
it was unused and can be removed.

0004-Fixed-segfault-in-case-there-is-no-hottable.patch
The most exciting one, faced it while playing around with queries from
ClickBench benchmark. Segfault was caused by accessing
aggstate->hottable even in case aggstate->hottable is NULL, using
aggstate->hashtable instead fixes this.

Also there are some ideas of making the VCI module patch smaller and
simpler for review:

First of all, we have this hottable optimization which is enabled but
surrounded by "#ifdef VCI_ENABLE_HOT_HASH_TABLE ... #endif". I think
this could be a separate patch on top of VCI module patch, say "Hot
hashtable optimization".

Second, there is some OS-specific code that looks like used only in
case ENABLE_WOSROS_TRANS_DEFERRAL is defined. This feature is unclear
to me, probably it can be better reviewed if separated to another
"Deferral of WOS->ROS transforamtion" patch that would introduce new
GUC and some OS-specific code. So we can be sure other VCI code has no
OS-specific parts.

-- 
Regards,
Timur Magomedov