Bug/performance: Repetitive calls to file_get_contents

[aristath]

While running tests with Xdebug & web grind, I identified repetitive calls to file_get_contents() from the get_default_block_editor_settings() function.
That function repeatedly gets the contents of the /css/dist/block-editor/default-editor-styles.css without any implementation to avoid those.

This is an easy bugfix, and involves using a static variable in the function. Details and screenshots in the PR.

Fixes repeat calls to file_get_contents

Testing on WP 6.1 (trunk), on the front-page of a site using Xdebug & webgrind: (values in milliseconds)

Before the patch:

After the patch:

Invocation count goes down from 181 to 93. Total self-cost goes down from 160ms to ~2ms

Trac ticket: https://core.trac.wordpress.org/ticket/56637

[aristath]

Edit: There was an error in the 2nd screenshot above.
Total self-cost goes down from 160ms to 93ms, not 1.88 (I forgot to switch from % to ms values in the after image)
Screenshot & text updated in the PR, but these changes don't get reflected in the comment above from the Github-sync bot.

@aristath Left one comment. Does that make sense to you?

[mukesh27]

Is it specific to Editor or General component ticket?

[aristath]

Replying to mukesh27:

Is it specific to Editor or General component ticket?

General. The get_default_block_editor_settings() function runs on the frontend as well (it's used in the JSON resolver), so the impact is global and not just the editor

[mukesh27]

Ok thanks.

​PR #3318 looks good to me.
Thanks @aristath

[SergeyBiryukov]

In 54291:

Block Editor: Remove repetitive calls to file_get_contents() in block editor settings.

The get_default_block_editor_settings() function used to repeatedly get the default-editor-styles.css file contents without any implementation to avoid this.

This commit utilizes a static variable to remove repetitive calls made during the same request. In tests ran on the front page of a site using Xdebug & Webgrind, the total file_get_contents() invocation count goes down from 181 to 93, and total self cost (the time that the function is responsible for) goes down from 160 ms to 93 ms.

Follow-up to [52042].

Props aristath, mukesh27.
Fixes #56637.

Thanks for the PR! Merged in r54291.