Change the position of notoptions lookup in get_option

[spacedmonkey]

Added in [4855] and #3726 the notoptions cache was added to get_option. This cache was designed to save database queries, do option name that are none not to exist in the database / cache.

However, the ordering this functions present a performance problem. In get_option the look for if a option does not exist, is done before if it does. It is extremely uncommon for a get_option, for a option name that does not exist. Most times it is called, it for a option that does exist. This means the call to notoptions cache is called pointless.

Moving this check, before the database lookup would improve performance.

Trac ticket: https://core.trac.wordpress.org/ticket/58277

[spacedmonkey]

500 runs on TT1

	Trunk - TT1	PR - TT1	Trunk - TT1 - Memcached	PR - TT1 - Memcached
Response Time (median)	73.16	73.1	73.62	72.49
wp-load-alloptions-query (median)	0.7	0.69	0.68	0.68
wp-before-template (median)	28.21	28.28	28.29	28.04
wp-before-template-db-queries (median)	2.76	2.78	2.76	2.75
wp-template (median)	37.73	37.64	38.11	37.62
wp-total (median)	65.91	66.1	66.63	65.87
wp-template-db-queries (median)	6.29	6.34	6.39	6.37

500 runs on TT3

	Trunk - TT3	PR - TT3	Trunk - TT3 - Memcached	PR - TT3 - Memcached
Response Time (median)	109.18	108.92	108.14	107.78
wp-before-template (median)	58.54	58.76	55.08	54.61
wp-template (median)	44.77	44.6	47.25	47.47
wp-total (median)	103.52	103.45	102.76	102.47

[spacedmonkey]

In a block theme, this saves 100s of calls to cache.

PR - 1302 vs Trunk - 1949.

[spacedmonkey]

Trunk - TT1

[spacedmonkey]

PR - TT1

[azaozz]

Replying to spacedmonkey:

In a block theme, this saves 100s of calls to cache.

Yep, this change makes sense imho. I'm a bit unsure on how to best test the result. Seems it would have to compare getting existing and non-existing options (that would trigger DB connections). Perhaps something like get an existing option 1000 times, and 5 different non-existing options 3 times each, etc.

In any case reducing the number of cache lookups is a good thing, even if the speed improvements are minimal.

[spacedmonkey]

Added unit test.

@costdev @felixarntz Feedback actioned. Ready for review.

[spacedmonkey]

In 56595:

Options, Meta APIs: Optimize get_option by relocating notoptions cache lookup.

In the get_option function, a cache lookup for the notoptions key is performed, which stores an array of keys for options known not to exist. This optimization prevents repeated database queries when certain options are requested. However, the cache lookup for notoptions was conducted before checking if the requested option exists in the cache. Given that it's more likely that the option does exist, this commit reorders the checks to first verify the option's existence in the cache before confirming its absence. This adjustment reduces redundant queries and also eliminates an unnecessary cache lookup, improving overall performance.

Props spacedmonkey, costdev, flixos90, azaozz.
Fixes #58277.

Committed https://core.trac.wordpress.org/changeset/56595

[peterwilsoncc]

In 59631:

Options/Meta APIs: Optimize cache hits for non-existent options.

Optimize the order of checking the various options caches in get_option() to prevent hitting external caches each time it is called for a known non-existent option.

The caches are checked in the following order when getting an option:

1. Check the alloptions cache first to prioritize existing loaded options.
2. Check the notoptions cache before a cache lookup or DB hit.
3. Check the options cache prior to a DB hit.

Follow up to [56595].

Props adamsilverstein, flixos90, ivankristianto, joemcgill, rmccue, siliconforks, spacedmonkey.
Fixes #62692.
See #58277.