Fix deserialization of redacted aliases events

[DevinR528]

I split the first commit into 2, I got the hang of I think!

[jplatte]

Nice! :D

[jplatte]

I really don't want us hardcoding specific event types into the macro code. I guess we could always fall back to deserializing from an empty object instead of doing that only for m.room.aliases. That doesn't feel like a very clean solution though. Ideas? ^^°

[DevinR528]

Yeah I kinda figured you wouldn't be thrilled with that. Well after thinking about it all evening yesterday the only thing I came up with was adding another trait method to RedactedEventContent something like has_optional_content() -> bool that has a default impl of returning false?

Also documenting that the RedactedEventContent trait is for internal macro use and users who are extending matrix events otherwise it should be ignored. I almost marked it as doc(hidden) but there are cases where it's needed.

[jplatte]

Yeah I kinda figured you wouldn't be thrilled with that. Well after thinking about it all evening yesterday the only thing I came up with was adding another trait method to RedactedEventContent something like has_optional_content() -> bool that has a default impl of returning false?

Okay, maybe we should just go at this in fully generality. I think that would mean adding is_empty also, and maybe has_optional_content would instead be a third possible value returned from has_deserialize_fields? As in: make it return enum HasDeserializeFields { Yes, No, Optional }.

Also documenting that the RedactedEventContent trait is for internal macro use and users who are extending matrix events otherwise it should be ignored. I almost marked it as doc(hidden) but there are cases where it's needed.

Actually, since we only need custom_redacted and manual RedactedEventContent implementations for supporting multiple room versions, I'd prefer for it to be #[doc(hidden)] for its methods / associated functions to be #[doc(hidden)], then we can also iterate on that without breaking the public API. Or is there some scenario you can think of where a user would need to implement it manually for a custom event? I would expect all custom events to be redacted fully. (pre-spec would be a different beast, but for that we've now got a different mechanism luckily, thanks to you!)

[jplatte]

Just two doc tiny errors that I will apply, this is looking good.

[jplatte]

GitHub is having issues, I rebased/squashed and pushed manually.

[DevinR528]

Cool!