PR: 2060.

I noticed that one more optimization wouldn't hurt the language loading code, replacing the expensive preg_replace() function which is called (2500 times) for all defined variables to look for translations.

PR 2059.

Program performance and performance measurement as we know it is a very tricky thing where intuition usually fools the programmer.

Knowing this, I decided to measure the code performance before and after the commit once again. And ... my goodness I didn't find any performance gain after replacing the seemingly heavy preg_replace() with a couple of simple, easy and fast strncmp() and substr(). Perhaps I measured without preloading or I was just so happy that the function disappeared from the hot functions and did not pay attention to the overall performance of the parent function. The PHP function call overhead is so high that the second call just ate up all the performance gain and the two “fast” functions work about the same as one “slow” function.

I've attached a graphical performance report for different versions (the unit on the charts is 10ns) of lang_load() function:

• preg_replace() - original code
• str_starts_with() + substr() - first version
• strncmp() + substr() - current version
• index + substr() - proposed in this pull request

The charts show that only the last variant has a performance gain. I did not evaluate memory, perhaps preg_replace() has more memory usage.

PR: 2061.

Wow you sure spent a lot of time profiling this, for a 0.7 ms gain ;-)
Thanks for your optimization efforts.

Basically, the idea is to get to know the code base and understand how things actually work. I also want to refresh my PHP knowledge and learn about debugging and profiling tools that I haven't used before (my main language is C++). Thanks for the opportunity and for being so patient.