Enums lose type information when serialised as part of a static map

[IceRidder]

Detailed description

When serializing with a TypeMap and one of the possible elements is an Enum it will lose its type

Ex:

#[StaticTypeMap(key: 'type', map: [
    'enum' => MyEnum::class,
    'object' =>  MyType::class,
])]

enum MyEnum: int implements MyInterface
{
    case A = 1;
    case B = 2;
}

class MyType implements MyInterface {
    public function __construct(
        public int $id = 1,
    )
    {
    }
}

class Element
{
    public function __construct(
        #[SequenceField(arrayType: MyInterface::class)]
        public array $elements = [],
    )
    {
    }
}

$element = new Element([MyEnum::A]);
$serde->serialize($element, 'json');

Will result in : {"elements":[1]} and thus de-serialization will fail

Context

I am trying to create a generic collection to serialize/de-serialize list of items

Possible implementation

Add map check in EnumExporter similar to ObjectExporter and if one is present serialize as {keyName: mixed, value: mixed}. I would not consider this a BC break as the feature was not working anyway.

Your environment

Include as many relevant details about the environment you experienced the bug in and how to reproduce it.

• Version used (e.g. PHP 5.6, HHVM 3): PHP 8.1
• Operating system and version (e.g. Ubuntu 16.04, Windows 7): Ubuntu 22.10

[agustingomes]

@Crell I started looking to this bug since I'm looking to adopt the library in a side project of mine.

This particular reproduction case seems a bit tricky. Unlike MyType class, which has the id property, MyEnum does not have any property, and as consequence, it would be a bit unclear what is the expected key when serializing.

From the possible implementation, I infer {"elements": ["type": "enum", "value": 1]} is the expected serialized outcome.

What do you think?

[Crell]

To be honest, I question the use case here. I don't know why you'd use an enum in a type map rather than... just an enum.

Were we to implement it, though, I believe @agustingomes is correct about what the serialized form would look like.

[agustingomes]

Ok. I'll take a deeper look to see if we can make this work.

[agustingomes]

@Crell I have a working solution for this case, but before I open a PR here, as part of this issue I needed to do a change upstream to make my initial solution work.

This is the crell/attributeutils PR: Crell/AttributeUtils#16

Based on the outcome of the PR above, we can decide if we want to continue fixing this use case.