Cache incrementors for get_bookmarks()

[ryan]

Make use of a caching incrementor and store queries in individual cache buckets to avoid memory exhaustion.

Pattern this after #23167, which does the same thing for get_pages().

See #23173 for an explanation of the motivation behind this.

[wonderboymusic]

these are all marked 3.7-early

[nacin]

Needs a patch.

[jipmoors]

meta cache implementation (based upon get_posts cache)

[jipmoors]

I had a slight problem with figuring out how to create a unit-test that would consider the current 'defaults' from get_bookmarks, since that data is only available in the function itself.

This is the most important part in the generation of the cache key, so it will be very difficult to test without knowing this and not have it break when the defaults might be modified in the future.

Any advice/suggestions on that?

[SergeyBiryukov]

Replying to jipmoors:

This is the most important part in the generation of the cache key, so it will be very difficult to test without knowing this and not have it break when the defaults might be modified in the future.

Link manager is disabled for new installs (see #21307), so it's very unlikely for those defaults to change.

[jipmoors]

Replying to SergeyBiryukov:

Link manager is disabled for new installs (see #21307), so it's very unlikely for those defaults to change.

Alright, I'll write a test for it then and put some additional commentary in there. Thanks for the reply.

[jipmoors]

Unit tests

[jipmoors]

Please be critical about the unit tests, these are a couple of my first ones so I'm sure there is room for improvement.
On a side note (not an excuse): the bookmarks are kind of old and should not take up too much energy and time I think.

[jipmoors]

I'm getting some fails on the unit tests. Could somebody give me some pointers on those?

[desrosj]

I'm going to close this out as a wontfix.

The Link Manager was essentially deprecated in Core when it was disabled by default for new installs, and hidden entirely when no links were present on a site updating in WP 3.5 (see #21307).

I've also opened #56362 to transition the related code out of Core and into a canonical plugin. If it's determined in the future that this is beneficial once transitioned to plugin form, then an issue should be opened upstream where that code is managed.