Optimize prettifyTestMethodName

[kamil-tekiela]

This should optimise the performance of this method by 10x. Regex is much more suitable for a task like this than a loop which reconstructs the string character by character. ucfirst is twice as fast as strtoupper with character substitution. Since this is hot code, the optimisation has a significant impact on the runtime.

One behavioural change is that the second regex also skips the digits at the start of the string. Since the return value is trimmed, it makes sense not to add a space at the beginning of the string.

[staabm]

Please share how you benchmark it

[kamil-tekiela]

Please share how you benchmark it

Using a very simple handwritten benchmark. https://3v4l.org/iH2iu/rfc#vgit.master

I have also verified that there is an impact by executing the change with phpMyAdmin's test suite.

I am sure you can verify the changes much more scientifically.

[staabm]

I don't need a scientific one.

How much does it speedup end-to-end on phpMyAdmin?

Thank you

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.01%. Comparing base (1aba907) to head (271c88a).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #6474      +/-   ##
============================================
- Coverage     96.01%   96.01%   -0.01%     
+ Complexity     7352     7343       -9     
============================================
  Files           798      798              
  Lines         22627    22619       -8     
============================================
- Hits          21726    21718       -8     
  Misses          901      901              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kamil-tekiela]

I don't need a scientific one.

How much does it speedup end-to-end on phpMyAdmin?

Thank you

As measured by xDebug, it's about 2s. Without it, it's difficult to measure because of the unpredictability of the tests due to I/O and HTTP requests. I estimate it's in the range of 100-200 ms for a suite that takes around 35s to execute.

[staabm]

I see, thanks for the indication.

the initial PR description sounded more dramatic, so its nice to have a real number on a real project.

as the implementation also will be simplified by the change, I would welcome the change.

thank you.