Bug #74933 [ReO->Csd]: MBstring functions are much slower when called with encoding parameter

Edit report at https://bugs.php.net/bug.php?id=74933&edit=1 ID: 74933 Updated by: nikic@php.net Reported by: reinir dot puradinata at gmail dot com Summary: MBstring functions are much slower when called with encoding parameter -Status: Re-Opened +Status: Closed Type: Bug Package: mbstring related Operating System: Windows PHP Version: 7.1.7 Assigned To: nikic Block user comment: N Private report: N New Comment: This is fixed in master with the introduction of an encoding cache, which avoids the expensive lookup if it's the same as the encoding that was used last, which should be the norm. Yet to be seen if these changes will be in PHP 7.2 as well. Previous Comments: ------------------------------------------------------------------------ [2017-07-18 08:15:42] nikic@php.net There are better ways to look up an encoding than doing an O(n) search: https://github.com/php/php-src/blob/6053987bc27e8dede37f437193a5cad448f99bce/ext/mbstring/libmbfl/mbfl/mbfl_encoding.c#L224 This should be turned into a lookup table. Additionally caching the last used encoding probably makes sense. ------------------------------------------------------------------------ [2017-07-18 07:53:24] jhdxr@php.net Thank you for taking the time to write to us, but this is not a bug. Please double-check the documentation available at http://www.php.net/manual/ and the instructions on how to report a bug at http://bugs.php.net/how-to-report.php Whenever you passed in a string, php has to converted it into internal enum type, and it's why the second case is so slow in your test file. ------------------------------------------------------------------------ [2017-07-17 01:08:57] reinir dot puradinata at gmail dot com Description: ------------ Several MBstring functions are much slower when called with encoding parameter than without encoding parameter. Functions that exhibit this behavior are mb_strlen, mb_substr, mb_strpos, mb_strrpos. For more information see: https://stackoverflow.com/questions/45028018/a/45107408 Test script: --------------- mb_internal_encoding("UTF-8"); echo "without encoding parameter:\n"; $a = microtime(true); for($i=0; $i<100000; $i++){ $n = mb_strlen("あえいおう"); } $a = microtime(true)-$a; echo "- mb_strlen: ".number_format($a*1000)." ms\n"; echo "\nwith encoding parameter:\n"; $b = microtime(true); for($i=0; $i<100000; $i++){ $n = mb_strlen("あえいおう", "UTF-8"); } $b = microtime(true)-$b; echo "- mb_strlen: ".number_format($b*1000)." ms (".number_format(($b-$a)*100/$a)."% slower)\n"; Expected result: ---------------- Because the character encoding is UTF-8 in both cases, I expected to see similar performance. In other words, 0% slower or close to it. Actual result: -------------- When called with encoding parameter, performance drops greatly. Example output from the test script: without encoding parameter: - mb_strlen: 14 ms with encoding parameter: - mb_strlen: 585 ms (4,186% slower) ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=74933&edit=1