A little bit QoL cleaning up...

[Richardn2002]

...that I believe could benefit main before I finish ecs stuff on my own branch.

Put data def and parsing code to proper places:

• try_from() and associated error types for string->enum mappings should go under format/
• node data should all be defined under node/

[Speykious]

Not sold on most of these changes.

The introduction of the enums module is confusing because it doesn't mean anything in particular in regards to the implementation, it just stores various TryFrom implementations that happen to be currently used by the payload module. If you put it there because you want to signify that it's for deserializing, we might as well replace these impls with deserialize_[EnumType] functions within payload and remove enums, because a TryFrom implementation conveys more meaning than its current use case.

I disagree with moving TransformOffset and InterpMode to node::data. At this stage Matrix would get in there too. It would just put the structs further away from their implementation blocks. I think it's completely fine that someone coming to the codebase excepts TransformOffset to be in the transform module.

[Richardn2002]

Rationale for enum.rs: There are tons of reserved strings that carry a meaning according to the standard. These namings for enum variants make up a portion of such reserved strings. Namings for json fields make up the rest of such reserved strings.
And these reserved strings should all go under format/ apparently as they are related to the format. Any mistake (for example that angle length bug just resolved recently) or future extension of formats introducing new variants, should be able to be resolved in a way that is localized under format/.
Not to say all the UnknownXXXError error types are only used in payload.rs.

I am fine with file renames or putting the try_from to other places, but I believe these try_from s should be localized.

[Richardn2002]

On where to put TransformOffset and InterpolateMode, my rationale is, put all data that makes up nodes under node/, preparing for defining components under node/. And math merely provides helper functions on how to manipulate data, instead of being the definers of data.

Like, PhysicsModel and PhysicsState are under node, while actual physics calculations are outside, under physics.

[Speykious]

Looking great now :D
Love it when less code does the same thing.