Added regression test #3062
Added regression test #3062
Conversation
| @@ -201,7 +200,8 @@ public static function union(Type ...$types): Type | |||
| if ($types[$i] instanceof StringType && !$types[$i] instanceof ClassStringType) { | |||
| $hasGenericScalarTypes[ConstantStringType::class] = true; | |||
| } | |||
| if ($types[$i] instanceof EnumCaseObjectType) { | |||
| $enumCases = $types[$i]->getEnumCases(); | |||
|
This pull request has been marked as ready for review. |
|
after this PR we now have both cases fast: |
|
will be AFK for a few hours |
src/Type/ObjectType.php
Outdated
| @@ -94,6 +94,9 @@ class ObjectType implements TypeWithClassName, SubtractableType | |||
|
|
|||
| private ?string $cachedDescription = null; | |||
|
|
|||
| /** @var array<class-string, list<EnumCaseObjectType>> */ | |||
| private static array $enumCasesCache = []; | |||
There was a problem hiding this comment.
Needs to be reset in resetCaches()
src/Type/ObjectType.php
Outdated
| foreach ($this->subtractedType->getEnumCases() as $enumCase) { | ||
| $subtracted[$enumCase->getEnumCaseName()] = true; | ||
| $className = $classReflection->getName(); | ||
| if (array_key_exists($className, self::$enumCasesCache)) { |
There was a problem hiding this comment.
The cache key needs to be $this->describeCache().
staabm
force-pushed
the
fix2
branch
3 times, most recently
from
69a9068
to
168d145
Compare
|
this one is good to go. I will have a look at the other remaining open PRs after this one is merged, whether there is still meaningful impact |
cs Update ObjectType.php
| } | ||
|
|
||
| $errors = $this->runAnalyse(__DIR__ . '/data/bug-10979.php'); | ||
| $this->assertCount(80, $errors); |
There was a problem hiding this comment.
Can you please fix the example so that it's valid and does not report any errors? And make sure it still reproduces the performance problem.
I'm worrying that the number 80 is going to fluctuate and is going to fail quite often as we work on PHPStan.
There was a problem hiding this comment.
we now have a zero errors repro
|
Thank you! |
closes phpstan/phpstan#10979
reverting the fix for #2985 does not regress the test of said PR.I need to dive deeper why the test no longer reproduces the perf problem we had at the time #2985 was created
I had a typo locally .. investigating..