Simplify sanitize_key() and use it in more places

[ryan]

sanitize_key() does some unnecessary work that can be eliminated.

sanitize_user() is still being called in places where sanitize_key() should be used and can burn up to a percent of page load time. It is called 26 times for a default install. Using a simplified sanitize_key() instead of sanitize_user() will save a little time.

[ryan]

[15635]

[ryan]

sanitize_key() could be simplified a little more, but I want to audit how plugins are using it and what kind of strings are being used from post type and status keys.

[sboisvert]

If I may just add to this, I encountered some very unexpected behavior when I started adding actions to sanitize_user since I wasn't aware that it was being used to sanitize the initial post types and many other things and when I added checks to make sure my username's followed a certain convention, all hell broke loose.
I therefore strongly support this. Thank you Ryan.

[jane]

@ryan: what are your thoughts following the plugin audit? Should decide and patch as needed asap if wanting to do it for 3.1 by freeze.

[WraithKenny]

Does this mean that $post_type in register_post_type() only supports lowercase alpha-numeric names?

[WraithKenny]

To clarify: Plugins that register "preFix_Post_Type" will become "prefixposttype" and break.

[WraithKenny]

Seems setting the correct map_meta_cap value corrected my issues (See #14122), and from what I can tell, the old way disallowed Upper also no problem. (I should of looked at the patch more closely, sorry :)

[WraithKenny]

Use sanitize_key in register_taxonomy for compatibilty with register_post_type

[WraithKenny]

Use sanitize_key in register_taxonomy and account for arrays

[WraithKenny]

Use sanitize_key for wpmu_create_blog in ms-functions

[WraithKenny]

I did a search on sanitize_user, and everything left (17 results) looks like it actually does deal with users.

[ryan]

[16822]

[ryan]

(In [16848]) Go back to sanitize_user() so that dots are allowed. Props ocean90. see #14910

[nacin]

Suggesting we revert the rest of [16822] in #15982.