RTC: Transients for awareness and JIT meta cache flushing by peterwilsoncc · Pull Request #11348 · WordPress/wordpress-develop

peterwilsoncc · 2026-03-25T00:02:50Z

This does a few things:

Moves awareness to the transients API: For sites with a persistent cache awareness will be stored in the object cache, the options table for site that don't
As the transients have a timeout set, they will not be autoloaded or invalidate the alloptions cache
Reduces awareness granularity to 10 seconds (rounded up), this changes the awareness timeout to be anywhere between 30 and 39 seconds which is close enough. It will also reduce database writes/cache invalidation
Introduces a filter wp_sync_awareness_timestamp_granularity to allow granularity to be modified by third party devs
Reverts r62099 to use the Meta APIs for syncing updates: I think this will be helpful as rooms meta data will no longer be updated following an awareness update. This will allow the cache to be used for rooms that have not had an update since the last polling update (eg, browser open while making lunch)
Introduces just in time cache invalidation for meta queries in WP_Query -- modifying the meta data will set a flag that the WP_Query post group will need to be invalidated next time a site makes a meta query
Introduces just in time cache invalidation for registered post meta. Modifying the meta data will set a flag to trigger cache invalidation next time a meta query is made in WP_Query.
Adds tests for above

Trac ticket: https://core.trac.wordpress.org/ticket/64696

Use of AI Tools

This Pull Request is for code review only. Please keep all other discussion in the Trac ticket. Do not merge this Pull Request. See GitHub Pull Requests for Code Review in the Core Handbook for more details.

github-actions · 2026-03-25T00:07:26Z

Test using WordPress Playground

The changes in this pull request can previewed and tested using a WordPress Playground instance.

WordPress Playground is an experimental project that creates a full WordPress instance entirely within the browser.

Some things to be aware of

All changes will be lost when closing a tab with a Playground instance.
All changes will be lost when refreshing the page.
A fresh instance is created each time the link below is clicked.
Every time this pull request is updated, a new ZIP file containing all changes is created. If changes are not reflected in the Playground instance,
it's possible that the most recent build failed, or has not completed. Check the list of workflow runs to be sure.

For more details about these limitations and more, check out the Limitations page in the WordPress Playground documentation.

Test this pull request with WordPress Playground.

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

github-actions · 2026-03-26T20:59:18Z

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

Core Committers: Use this line as a base for the props when committing in SVN:

Props peterwilsoncc, jorbin, czarate, desrosj, zieladam.

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

src/wp-includes/collaboration/class-wp-http-polling-sync-server.php

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

tests/phpunit/tests/collaboration/wpSyncPostMetaStorage.php

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

peterwilsoncc · 2026-03-27T01:16:17Z

I've pushed a few changes following reviews:

Following @aaronjorbin's review and suggestions in Slack

8020652 and some follow up commits: Introduces JIT meta invalidation to the register meta API so the JIT approach is only taken for specific meta data rather than all meta data. This is to help avoid unintentional consequences outside of RTC. As part of this change, I added _edit_lock for JIT invalidation. Discuss.
6948cde: Reuse a variable instead of recalling time()

Following @chriszarate's review:

25c571e: Restored a few changes in the original commit that were not related to switching to direct DB calls
f7e25c2: expanded docs on granularity filter

Additionally:

extended the transient timeout to a day following my own thought bubble
Link to the trac ticket in this bullet point as it didn't make it to trac the first time around https://core.trac.wordpress.org/ticket/64696

desrosj

Some small improvements for the inline docs, but this is looking promising so far.

src/wp-includes/meta.php

src/wp-includes/post.php

src/wp-includes/collaboration/class-wp-http-polling-sync-server.php

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

peterwilsoncc · 2026-03-27T03:37:39Z

src/wp-includes/post.php

+ * @param int    $object_id  Object ID for which the meta was updated.
+ * @param string $meta_key   Meta key that was updated.
+ */
+function wp_cache_set_needs_meta_query_flush( $meta_ids, $object_id, $meta_key ) {


A bit of help with naming things here would be great, same for the name of the cache key.

In an ideal world the cache key would be of positive intent, wp_query_meta_query_needs_flush or similar, but that would be problematic if an object cache's eviction policy drops the value.

Caching false can be problematic with some implementations so whatever the cache key is named, it needs to make sense that true is used to indicate that the cache does not need a JIT flush. Sigh.

aaronjorbin

Thank you @peterwilsoncc ! Overall, I think this is moving the code in a very positive direction.

aaronjorbin · 2026-03-27T02:30:19Z

src/wp-includes/collaboration.php

+		)
+	);
+
+	// @todo This should probably go elsewhere.


wp_create_initial_post_meta makes sense to me and has a wp_is_collaboration_enabled() gate if we want WP_Sync_Post_Meta_Storage::SYNC_UPDATE_META_KEY to be conditionally registered.

tests/phpunit/tests/query/metaQuery.php

aaronjorbin · 2026-03-27T03:41:30Z

src/wp-includes/meta.php

+ *                                              array with 'schema' or 'prepare_callback' keys instead of a boolean.
+ *     @type bool       $revisions_enabled      Whether to enable revisions support for this meta_key. Can only be used when the
+ *                                              object type is 'post'.
+ *     @type bool       $jit_cache_invalidation Whether to enable just-in-time cache invalidation for this meta key. When enabled, the cache


@kittenkamala: This addition to post_meta should likely get a mention in the misc changes devnote.

aaronjorbin · 2026-03-27T04:47:21Z

tests/phpunit/tests/rest-api/rest-sync-server.php

+		);
+
+		$original_wpdb = $wpdb;
+		$proxy_wpdb    = new class( $original_wpdb, $storage_post_id, $injected_update ) {


Not blocking and I know this isn't as new as the green makes it look, but I don't love this living inside the test function. It's also copied in multiple places.

aaronjorbin · 2026-03-27T05:22:24Z

src/wp-includes/collaboration.php

+	// @todo This should probably go elsewhere.
+	register_post_meta(
+		'',
+		'_edit_lock',


As part of this change, I added _edit_lock for JIT invalidation. Discuss.

I think this makes sense. "Is this only used for writing purposes and not for display purposes" feels like it could be a good rule of thumb.

aaronjorbin · 2026-03-27T05:43:31Z

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

If this is storing things in both post-meta and transients, should it (and the class) have it's name changed? It also means that if a new DB table is added later, this class can be continue to be used as the interface for extenders.

It could also be made it clear in a devnote (or inline comment?) that interacting with sync related data should be done through this class, not through any hooks called as a result of how it implements the storage. i.e. don't count on update_postmeta firing forever.

This feels inline with the comment on the class being "Core class that provides an interface for storing and retrieving sync updates and awareness data during a collaborative session."

adamziel · 2026-03-27T13:30:23Z

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

+		 * transient longer than the awareness can avoid adding new entries in the options
+		 * table unnecessarily.
+		 */
+		set_transient( self::AWARENESS_TRANSIENT_PREFIX . ":{$room_hash}", $awareness, DAY_IN_SECONDS );


set_transient has a race condition where we might end up calling add_option multiple times. There's probably a million different possible outcomes depending on whether we have object cache, is it persistent, etc. and it's important we don't end up inserting states A and B and then end up reading A but updating B, or reading A on some requests but B on other requests.

Is the race description you are talking about https://core.trac.wordpress.org/ticket/63450 or something different?

adamziel · 2026-03-27T13:34:59Z

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

-			if ( null !== $decoded ) {
-				$updates[] = $decoded;
-			}
+			$updates[] = maybe_unserialize( $row->meta_value );


Is there any way we can avoid relying on serialize()? I really liked how we dealt with JSON before, it removes an entire class of serialize()-related security vulnerabilities.

adamziel · 2026-03-27T13:35:54Z

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php

-		}
-
-		$awareness = json_decode( $meta_value, true );
+		$room_hash = md5( $room ); // Not used for cryptographic purposes.


any way we can avoid all the hashing? It could add up to substantial load when there's enough users of this feature

any way we can avoid all the hashing? It could add up to substantial load when there's enough users of this feature

@adamziel There isn't as we can't use the room ids in slugs due to WP sanitization and transients have a maximum lenght we need to avoid. We could aim for a less intensive hashing mechanism if you have a recommendation.

Coming from a place of curiosity: Is md5 really that expensive? It seems to be a bit slower than serializing a simple array:

> php -v PHP 8.5.4 (cli) (built: Mar 10 2026 23:15:23) (NTS) Copyright (c) The PHP Group Built by Homebrew Zend Engine v4.5.4, Copyright (c) Zend Technologies with Zend OPcache v8.5.4, Copyright (c), by Zend Technologies > php -r '$start = microtime(true); for ($i = 0; $i < 1000000; $i++) { md5("test" . $i); } $end = microtime(true); echo "Time: " . ($end - $start) . " seconds\n";' Time: 0.12329506874084 seconds > php -r '$start = microtime(true); for ($i = 0; $i < 1000000; $i++) { serialize([["foo" => "test" . $i]]); } $end = microtime(true); echo "Time: " . ($end - $start) . " seconds\n";' Time: 0.099403858184814 seconds

Also, we only call md5 once per room per request.

adamziel · 2026-03-27T13:37:18Z

src/wp-includes/class-wp-query.php

 			$join   .= $clauses['join'];
 			$where  .= $clauses['where'];
+
+			if ( ! wp_cache_get( 'wp_query_meta_query_updated', 'post_meta' ) ) {


Can we add a thorough writeup in an inline docblock that goes through the entire data flow and the reasoning behind this? Every time I see it it takes a lot of effort to unpack

…st_meta functions." This reverts commit e77b83c.

Co-Authored-By: aaronjorbin <jorbin@git.wordpress.org>

Co-Authored-By: chriszarate <czarate@git.wordpress.org>

…ear.

Co-authored-by: Jonathan Desrosiers <359867+desrosj@users.noreply.github.com>

peterwilsoncc changed the title ~~Transients for awareness~~ Transients for awareness and JIT meta cache flushing Mar 26, 2026

peterwilsoncc force-pushed the 64696-using-core-apis branch 2 times, most recently from 2977e90 to f3467c2 Compare March 26, 2026 02:31

peterwilsoncc changed the title ~~Transients for awareness and JIT meta cache flushing~~ RTC: Transients for awareness and JIT meta cache flushing Mar 26, 2026

peterwilsoncc commented Mar 26, 2026

View reviewed changes

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php Outdated Show resolved Hide resolved

peterwilsoncc force-pushed the 64696-using-core-apis branch from 0ddf83c to bfb758e Compare March 26, 2026 20:44

peterwilsoncc marked this pull request as ready for review March 26, 2026 20:59

aaronjorbin reviewed Mar 26, 2026

View reviewed changes

src/wp-includes/collaboration/class-wp-http-polling-sync-server.php Outdated Show resolved Hide resolved

chriszarate reviewed Mar 26, 2026

View reviewed changes

src/wp-includes/collaboration/class-wp-http-polling-sync-server.php Outdated Show resolved Hide resolved

chriszarate reviewed Mar 26, 2026

View reviewed changes

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php Show resolved Hide resolved

chriszarate reviewed Mar 26, 2026

View reviewed changes

tests/phpunit/tests/collaboration/wpSyncPostMetaStorage.php Show resolved Hide resolved

chriszarate reviewed Mar 26, 2026

View reviewed changes

src/wp-includes/collaboration/class-wp-sync-post-meta-storage.php Outdated Show resolved Hide resolved

peterwilsoncc force-pushed the 64696-using-core-apis branch 3 times, most recently from 660c289 to be9e5e4 Compare March 27, 2026 01:07

peterwilsoncc force-pushed the 64696-using-core-apis branch from be9e5e4 to 74dc594 Compare March 27, 2026 01:16

desrosj reviewed Mar 27, 2026

View reviewed changes

peterwilsoncc commented Mar 27, 2026

View reviewed changes

aaronjorbin reviewed Mar 27, 2026

View reviewed changes

adamziel reviewed Mar 27, 2026

View reviewed changes

peterwilsoncc added 4 commits March 28, 2026 08:31

Revert "Real-time collaboration: Use prepared queries instead of *_po…

8294ce7

…st_meta functions." This reverts commit e77b83c.

Use transient for awareness.

fe23aa5

Use JIT flushing of post meta cache.

0895f1d

Use serialization in core functions.

36a4157

peterwilsoncc and others added 20 commits March 28, 2026 08:31

Relocate deterministic ordering of transient.

bc66ed4

JIT meta data cache clearing tests.

def3081

Use data provider for various forms of meta queries.

997baf4

Allow granularity to be filtered.

161de7b

CS: Whitespace.

1960088

Introduce JIT meta data.

ffa2ae3

Update docblocks.

0702b65

Throw doing_it_wrong for cache invalidation for non-posts.

ecef2a4

Reuse current time variable for current time().

8e02ea2

Co-Authored-By: aaronjorbin <jorbin@git.wordpress.org>

Document why reduced granularity is a good thing.

d7811c3

Co-Authored-By: chriszarate <czarate@git.wordpress.org>

Restore things not related to direct DB calls

b3e2c87

Co-Authored-By: chriszarate <czarate@git.wordpress.org>

Let’s keep rooms for a day to account for shared rooms.

bbd0e5b

Register foo for JIT invalidation in tests.

18a90c3

Test JIT invalidation doesn’t affect unregistered meta.

4feef96

meta registered without JIT invalidation.

f03158f

Add jit_cache_invalidation to expected values.

65028a5

There’s always one.

225a8a3

This year has been a bit of a month; this month has been a bit of a y…

f25c715

…ear.

Apply suggestions from code review

69d041a

Co-authored-by: Jonathan Desrosiers <359867+desrosj@users.noreply.github.com>

Protect against DivisionByZeroError.

59995f5

peterwilsoncc force-pushed the 64696-using-core-apis branch from b57a482 to 59995f5 Compare March 27, 2026 21:32

peterwilsoncc added 2 commits March 28, 2026 08:57

Actually use data providers.

8834de0

Docblocks for data provider.

383b708

Conversation

peterwilsoncc commented Mar 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Use of AI Tools

Uh oh!

github-actions bot commented Mar 25, 2026

Test using WordPress Playground

Some things to be aware of

Uh oh!

Uh oh!

github-actions bot commented Mar 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

peterwilsoncc commented Mar 27, 2026

Uh oh!

desrosj left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aaronjorbin left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

peterwilsoncc commented Mar 25, 2026 •

edited

Loading

github-actions bot commented Mar 26, 2026 •

edited

Loading