Optimize array_key_exists/in_array for empty array

[TysonAndre]

Make opcache replace the result with false if the array argument is
known to be empty.
This may be useful when a codebase has placeholders,
e.g. if (!in_array($method, self::ALLOWED_METHODS)) { return; }

In zend_inference.c: In php 8, array_key_exists will throw
a TypeError instead of returning null. I can create a separate PR for this if needed.

I didn't see any discussion of this optimization (for/against)
after a quick search on github, e.g. GH-3360

Potential future optimizations:

• convert in_array($needle, ['only one element'], true) to ===?
  (or == for strict=false)

• When the number of elements is less than 4, switch to looping instead of hash
  lookup. (exact threshold for better performance to be determined)

  Also support looping for in_array($value, [false, 'str', 2.5], true/false)

[TysonAndre]

001+ In function ExampleInArray::test (after calls):
002+ non-var op1 of 5 (ZEND_QM_ASSIGN) uses or defs an ssa var
003+ op1 use of 13 (CV $x) does not match op -3 of 5 (ZEND_QM_ASSIGN)
004+ non-var op1 of 11 (ZEND_QM_ASSIGN) uses or defs an ssa var
005+ op1 use of 13 (CV $x) does not match op -3 of 11 (ZEND_QM_ASSIGN)
006+ non-var op1 of 83 (ZEND_QM_ASSIGN) uses or defs an ssa var
007+ op1 use of 51 (TMP) does not match op -3 of 83 (ZEND_QM_ASSIGN)
008+ 
009+ In function ExampleInArray::test (after dce):
010+ non-var op1 of 5 (ZEND_QM_ASSIGN) uses or defs an ssa var
011+ op1 use of 13 (CV $x) does not match op -3 of 5 (ZEND_QM_ASSIGN)
001- Warning: Undefined variable: undef in %s on line 11

The tests are failing - I'm not familiar with the SSA implementation, and I don't know how to properly replace the function call with the result in all of the affected uses of the variable - the ZEND_QM_ASSIGN is a workaround to prototype this.

[nikic]

Once again, I feel like there should be a better place to do this ... possibly as a combination of TI, SCCP and DCE.

[nikic]

Part 1: We can DCE on array_key_exists whose result is not used: 9e403d6

[nikic]

Oops, I scrolled over the important part of the patch ... this does use SCCP. The main problem is the handling of errors -- SCCP has an assumption that any value we can compute, we can compute without triggering a diagnostic. To handle this cleanly, we should be separating out that assumption...

[TysonAndre]

Oops, I scrolled over the important part of the patch ... this does use SCCP. The main problem is the handling of errors -- SCCP has an assumption that any value we can compute, we can compute without triggering a diagnostic. To handle this cleanly, we should be separating out that assumption...

I'm assuming the only diagnostic to be concerned about in php 8.0 is the undefined variable warning - TypeError is now thrown by array_key_exists for invalid keys.

I agree that it'd be cleaner and help with eliminating dead code such as if (in_array($undef, [true])) { ... } - I'm not familiar enough with SCCP to know how to do that.

[cmb69]

Not sure how to proceed here, but at least the merge conflicts would need to be resolved.

[krakjoe]

I'm closing this on the basis that it's stale (and conflicting), and the conversation left questions open that were not addressed.

I can easily be wrong, if you feel very strongly this is worthwhile, please open a fresh PR against the master branch.