[wp-trac] [WordPress Trac] #22951: Performance enhancements for esc_url()

[WordPress Trac]

#22951: Performance enhancements for esc_url()
-------------------------------------------------+-------------------------
 Reporter:  markjaquith                          |       Owner:  schlessera
     Type:  defect (bug)                         |      Status:  assigned
 Priority:  normal                               |   Milestone:  Future
Component:  Formatting                           |  Release
 Severity:  normal                               |     Version:  2.8
 Keywords:  reporter-feedback has-patch 2nd-     |  Resolution:
  opinion                                        |     Focuses:
                                                 |  performance
-------------------------------------------------+-------------------------
Changes (by schlessera):

 * keywords:  needs-patch reporter-feedback needs-unit-tests => reporter-
     feedback has-patch 2nd-opinion


Comment:

 @peterwilsoncc Sorry for not getting back to you sooner.

 I had some time during a contributor day and made a first attempt at
 improving performance. Basically, I just add a regex in front of the bad
 protocol detection to return early for http(s) URLs. I'm sure there might
 be other performance enhancements at other (earlier) points in the code
 execution, but these will also have more potential impact on the security
 implications than what I've currently done.

 With the change in place, the existing unit tests in group `kses` still
 pass successfully.

 The regex I've taken is the @diegoperini one (with a slight modification
 to remove ftp) from the comparison here: https://mathiasbynens.be/demo
 /url-regex . It is pretty long and feature-complete, but due to the way
 the regex is evaluated, it is actually fast enough to provide a real
 improvement.

 I've done some (not 100% scientific) benchmarks to compare the current
 version of `esc_url` from trunk against my modified version. For all
 benchmarks, both versions run on the same machine from within the same
 process, and only the relative change is stored, to get rid of
 fluctuations due to allocations of server resources as good as possible.
 You can see a table with the benchmark results here:
 https://docs.google.com/spreadsheets/d/1XK5lgAIDMlj6lQSTrelXdD5eh6wAqJuLMS1605iM618/edit?usp=sharing

 The benchmark summary: In a normal use case, this change should speed up
 the `esc_url` function by 25-30%. In a best case scenario, the speed-up
 can be as high as 40%, while the absolute worst case scenario (basically,
 dealing with 100% bad protocol links) produces a slow-down of 2%, which I
 deem is acceptable given the very low probability of this ever being the
 case.

 The test script I've used to run the benchmarks can be found here:
 https://gist.github.com/schlessera/d4dd4db71d59dfee6ee6c5b9a0ca8a33

--
Ticket URL: <https://core.trac.wordpress.org/ticket/22951#comment:16>
WordPress Trac <https://core.trac.wordpress.org/>
WordPress publishing platform