[12.x] perf: Optimize BladeCompiler

[rzv-me]

This PR reduces the execution time for getOpenAndClosingPhpTokens() by up to 5-8x and the compile() by up to 3x, based on my testing.

As the improvements are for the BladeCompiler, these benefit the most situations when the Blade Components are not cached (local and misconfigured production environments).

I used the Livewire starter kit on PHP 8.3 and my mac with M3 Max to test my changes.

page	v1	v2	v3	this version (v4)
complex page (92 calls)	43.06 ms	16.73 ms	13.98 ms	8.65 ms
billing page (51 calls)	18.74 ms	7.22 ms	6.22 ms	4.18 ms
dashboard (94 calls)	35.42 ms	13.69 ms	11.60 ms	7.57 ms
profile page (56 calls)	23.21 ms	8.97 ms	7.64 ms	5.05 ms

I went through a few iterations which I will add here:
v1: Original

        return (new Collection(token_get_all($contents)))
            ->pluck(0)
            ->filter(function ($token) {
                return in_array($token, [T_OPEN_TAG, T_OPEN_TAG_WITH_ECHO, T_CLOSE_TAG]);
            });

v2: Most obvious optimization, doing the pluck after the filter

        return (new Collection(token_get_all($contents)))
            ->filter(function ($token) {
                return in_array($token, [T_OPEN_TAG, T_OPEN_TAG_WITH_ECHO, T_CLOSE_TAG]);
            })
            ->pluck(0);

v3: Switching to foreach(), instead of using collection

        $tokens = [];
        foreach (token_get_all($contents) as $token) {
            if (in_array($token, [T_OPEN_TAG, T_OPEN_TAG_WITH_ECHO, T_CLOSE_TAG])) {
                $tokens[] = $token[0];
            }
        }
        return new Collection($tokens);

v4: Final optimization: Switch from in_array, to multiple if conditions

        $tokens = [];
        foreach (token_get_all($contents) as $token) {
            if ($token[0] === T_OPEN_TAG || $token[0] === T_OPEN_TAG_WITH_ECHO || $token[0] === T_CLOSE_TAG) {
                $tokens[] = $token[0];
            }
        }
        return new Collection($tokens);

[browner12]

is the bulk of the benefit from removing the in_array call that was running N times?

I'm curious if you would have got the same improvements sticking with the chained Collection calls, but removing the in_array.

[calebdw]

You might could get some additional performance gains by using a variable instead of repeatedly accessing the array:

        foreach (token_get_all($contents) as $token) {
            $token = $token[0];
            if ($token === T_OPEN_TAG || $token === T_OPEN_TAG_WITH_ECHO || $token === T_CLOSE_TAG) {
                $tokens[] = $token;
            }
        }

Alternatively, you might could use array_column to avoid the array access altogether:

        foreach (array_column(token_get_all($contents), 0) as $token) {
            if ($token === T_OPEN_TAG || $token === T_OPEN_TAG_WITH_ECHO || $token === T_CLOSE_TAG) {
                $tokens[] = $token;
            }
        }

[browner12]

been programming PHP for 20 years and TIL array_column exists 🤯

[rzv-me]

@browner12 the bulk of the imprvements come from 2 places:

• not doing the pluck before filtering, even better when you filter and pluck(0) at the same time
• not doing the in_array() when filtering, as the options are static and very limited (3)

@calebdw

• the first suggestion might be a bit better, but it also means there is an extra memory allocation
• as for the second suggestion, I suspect it might be the same as, doing the pluck(0) before the filter
  did you benchmark them?

[calebdw]

I didn't benchmark anything, just thinking out loud---performance is one of those things that usually defies intuition

[shaedrich]

        foreach (token_get_all($contents) as $token) {
            $token = $token[0];
            if ($token === T_OPEN_TAG || $token === T_OPEN_TAG_WITH_ECHO || $token === T_CLOSE_TAG) {
                $tokens[] = $token;
            }
        }

You can even avoid using indices altogether by using array destructuring:

        foreach (token_get_all($contents) as $token) {
-           $token = $token[0];
+           [$token] = $token;
            if ($token === T_OPEN_TAG || $token === T_OPEN_TAG_WITH_ECHO || $token === T_CLOSE_TAG) {
                $tokens[] = $token;
            }
        }

[donnysim]

You can also do

foreach (token_get_all($contents) as [$token]) {
```