[wp-trac] [WordPress Trac] #22286: Enhancing backslashit performance

[WordPress Trac]

#22286: Enhancing backslashit performance
-------------------------+------------------------------
 Reporter:  jbutkus      |       Owner:
     Type:  enhancement  |      Status:  new
 Priority:  normal       |   Milestone:  Awaiting Review
Component:  Formatting   |     Version:
 Severity:  normal       |  Resolution:
 Keywords:  has-patch    |
-------------------------+------------------------------

Comment (by jbutkus):

 [comment:2 @nacin] - thank you, for your comment. I have updated diff with
 array-style brackets for char-in-string accessor.

 As for first character comparison - it is based on ASCII table, and
 compiles to C code directly.
 We are saving method call (JMP, more likely - long one, in assembly),
 which would occur if ''is_numeric'', or ''strpbrk'' would be used.

 And yes - ''is_numeric'' is slower by order of magnitude. Mostly because
 {{{is_numeric( '1.8e26' )}}} is '''true''', but it has to evaluate rather
 large number, thus causing a call to mathematical co-processor, instead of
 relying on general CPU.

 As for ''strpbrk'' - it maps to C equivalent directly (with sanity checks,
 only).
 Although it is also slower, than direct char comparison using order mask.
 Mostly we save the iteration, which occurs during char-list traversal in
 ''strpbrk''.

 And what you observe, when {{{$string[0] <= '9'}}} is ommited is true not
 only in PHP, but in C as well:

 {{{
 #include <stdio.h>

 int main( int argc, const char* argv[] ) {
     unsigned char single_letter = 'a';

     if ( argc > 1 ) {
         single_letter = *(*(argv+1));
     }

     if ( single_letter >= '0' /*&& single_letter <= '9'*/ ) { // notice,
 that second check is commented out
         printf( "It is a digit: %c\n", single_letter );
     } else {
         printf( "It is NOT a digit: %c\n", single_letter );
     }

     return 0;
 }
 }}}

 If we compile the program, and run it as following:
 ./tester a
 It would say '''It is a digit: a'''.

 Although with {{./tester $}} it would say '''It is NOT a digit: $'''.

 And if we un-comment the following part: {{{&& single_letter <= '9'}}} it
 would respond as expected: '''It is NOT a digit: a''' vs. '''It is a
 digit: 5'''.

-- 
Ticket URL: <http://core.trac.wordpress.org/ticket/22286#comment:3>
WordPress Trac <http://core.trac.wordpress.org/>
WordPress blogging software