Enums: new function enum_exists

[weirdan]

psalm.dev/r/9b5935609d

Guide to editing callmaps: Editing callmaps.

[weirdan]

psalm.dev/r/9b5935609d

[psalm-github-bot]

I found these snippets:

https://psalm.dev/r/9b5935609d

<?php

enum Status: int
{
    case FOO;
    case BAR = 42;
}

enum_exists(Status::FOO); //returns true
enum_exists(Status::BAZ); //returns false

Psalm output (using commit 916b098):

ERROR: ParseError - 3:6 - Syntax error, unexpected T_STRING on line 3

ERROR: ParseError - 5:5 - Syntax error, unexpected T_CASE on line 5

[brayniverse]

I should be able to do this one as well if you want @weirdan?

[weirdan]

Sure. Although this one will need some more work, as it's similar to class_exist() for which Psalm has special treatment here and there:

$ rg "'class_exists'" src
src/Psalm/Internal/Codebase/Functions.php
452:            'class_exists', // impure by virtue of triggering autoloader

src/Psalm/Internal/Analyzer/Statements/Expression/AssertionFinder.php
1772:            && strtolower($stmt->name->parts[0]) === 'class_exists'

src/Psalm/Internal/Analyzer/Statements/Expression/Call/NamedFunctionCallHandler.php
81:        if ($function_id === 'class_exists') {

src/Psalm/Internal/PhpVisitor/Reflector/ExpressionResolver.php
327:        if ($function->name->parts === ['class_exists']

src/Psalm/Internal/Analyzer/Statements/Expression/Call/ArgumentsAnalyzer.php
169:            if ($method_id === 'class_exists'

[orklah]

If you're stuck, just ping us, we'll help :)

[brayniverse]

Interesting, I'll give it a crack tonight and let you know if I get stuck.

[brayniverse]

I had a first attempt last night but got a little lost. I'll try again tonigh and tell you where I'm getting lost if I don't make any progress.

[brayniverse]

Hey, sorry I've gone quiet. I haven't had enough time to get into this. Please let anyone else pick this up if they want to do it. Whenever I do have time I'll pick something else up.

[DannyvdSluijs]

I took a stab at this one and was able create some additional test cases in tests/FunctionCallTest.php based on the class_exists ones that are already there based on the red-green principle. The test fail, showing the lack of support for enum_exists. So my journey continued.

I started to fill in the gaps you pointed out.

• Related to the AssertionFinder I added the hasEnumExistsCheck function similar to the hasClassExistsCheck and hasTraitExistsCheck.
• But when needing to implement the function around https://github.com/DannyvdSluijs/psalm/blob/b664850cdc15cefd8f0022fbf97e59529175934b/src/Psalm/Internal/Analyzer/Statements/Expression/AssertionFinder.php#L765 I got suck in deciding what to do. Basically enums are a valid class, calling class_exists on a Enum class returns true but the other way around isn't true. So you typically would like to annotate it as enum-string
• When I arrived at src/Psalm/Internal/Analyzer/Statements/Expression/Call/ArgumentsAnalyzer.php it became hazy and I was uncertain at this point if I was following down the correct path. Some inputs on what is needed would be welcomed. Should I expand the context to know if it is inside an enum_exists call or should we alias it to the class_exists behaviour that already is there?

Also from the documentation I noticed the Callmap and Callmap_81_delta should be adjusted which was a walk in the park, assuming it is nothing more that making sure the signature is available in the callmap. As that is a straight up copy from class_exists it was easy.

You can see my fork at https://github.com/DannyvdSluijs/psalm/tree/6431-new-function--enum_exists if you're interested and it would help. The test are still failing.

[orklah]

I got suck in deciding what to do. Basically enums are a valid class, calling class_exists on a Enum class returns true but the other way around isn't true. So you typically would like to annotate it as enum-string

Your call. You could probably go the way of interface-string by creating an enum-string which would be an alias of class-string. You could also go the way of trait-string who is a fully fledged type TTraitString. I would probably tend towards the second one myself, this will probably more flexible to handle future specific behaviors for enums

it became hazy and I was uncertain at this point if I was following down the correct path

Please try without adding anything here. I'm not actually sure why we have specific handling here, we'll see if we can make the tests pass without it

Seems like a good start :)