Implement Reflect for TypeId

[PROMETHIA-27]

Objective

Allows reflection and serialization of TypeIds. This is helpful for serializing and deserializing certain types that use TypeIds and are otherwise reflectable/serializable.

Solution

• Implement TypeIds as reflect values with Hash and PartialEq
• Hardcode TypeId serialization as an alternative to Value serialization, since TypeIds need access to the TypeRegistry to be serialized/deserialized and otherwise cannot access it. TypeIds are de/serialized as their corresponding type's name.

Issues:

• Hardcoding a new serialization method for one type is gross, especially when it's entirely possible other types might end up also needing a TypeRegistry for serialization.
• Serialization format can be simplified to TypeId("{type name here}") once 4561 hits

Alternative solutions:

• Don't implement reflect for TypeIds
  • Would entirely prevent behavior which is probably logically sound and useful
• Implement reflect without serialization
  • It should be sound to allow TypeIds to be serialized to a type name and back, so this doesn't seem like a good approach
• Allow any serialized types to access the TypeRegistry during de/serialization
  • Add it to thread local storage?
  • Pass it as a parameter?

[MrGVSV]

Looks good to me!

[MrGVSV]

Allow any serialized types to access the TypeRegistry during de/serialization

• Add it to thread local storage?
• Pass it as a parameter?

Reflecting on this more, I think you're right. It's probably best to allow the serialization method to access the registry (even if #4782 lands) as this gives the best control and flexibility for types going forward.

We could probably add that functionality to #4782 itself or create a new PR for it (I don't think it should be included as part of this one).

[PROMETHIA-27]

Agreed, that's a big change beyond the scope of this PR. I think it probably belongs in a new PR. I'm also not certain on how it should be done.

[MrGVSV]

I'm also not certain on how it should be done.

I’m in favor of the parameter method. The TLS solution will probably be a lot more controversial and a bigger change.

So just adding it as a parameter is probably simplest and most likely to be approved for now

[PROMETHIA-27]

I'll investigate that and open a PR if I can get it working

[alice-i-cecile]

@PROMETHIA-27, once this is rebased and the serialization fixed (remember to fix the PR description) I'm happy to merge this.

Other code LGTM.

[PROMETHIA-27]

Making a new PR for this soon:tm: