array_key_last: improving performance

[h4kuna]

First PR is here, where was discussion.

Q: The issue is that this mutates the internal pointer.
A: created benchmark and implementation does not change internal pointer.

Q: about performance for small vs big arrays
A: look like implementation is every time faster

This implementation is approx twice faster than cycle foreach.

Benchmark is on my gist.

Results on my machine.

array_key_last_symfony
0.1387140750885

array_key_last
0.055586814880371

[BackEndTea]

end changes the internal pointer of the array. And the 'official' 7.3 array_key_last (and first), don't change the internal pointer.

The foreach approach was taken to make sure that the polyfill behaves the same as the official version.

I should learn how to read better 😅

[nicolas-grekas]

Actually it doesn't change the internal pointer, because the "end" operates on a copy of the array (argument is passed by value and not by reference to the function)

[stof]

but on PHP 5.x, the reference mismatch forces to create a copy of the array, and that's also not the behavior of the core method.

[stof]

your benchmark is checking time, but does not check memory

[h4kuna]

@stof you have right. Could you check if the benchmarks are relevant?

Array size is 2000, on site is execute limit 3s for php.

Memory usage

Here is benchmark.

php 5.6.29

Array size: 2000
array_key_last_symfony: time: 1.5190160274506 s, memory: 80 B
array_key_last: time: 0.70938301086426 s, memory: 112 B

php 7+ return this result

Array size: 2000
array_key_last_symfony: time: 0.29880118370056 s, memory: 0 B
array_key_last: time: 0.13791918754578 s, memory: 0 B

Peak of memory

array_key_last

• php 7.0.1: 140.03 kB
• php 5.6.29: 468.34 kB

array_key_last_symfony

• php 7.0.1: 140.02 kB
• php 5.6.29: 468.34 kB

[nicolas-grekas]

Thank you @h4kuna.