Allow type `T1?|T2?`

[Masquue]

T1?|T2? is now parsed as T1?, whose correct form should be (T1|T2)? or T1|T2|nil or field? T1|T2.

But many libraries (e.g. neovim plugins) are using this incorrect form.

This misuse is so prevalent that should we consider making this form legal?

[CppCXY]

If there are too many errors, they should be supported, then we would have to support too many cases. Why not let others fix the errors instead?

[CppCXY]

If we support this feature of yours, the type expression will become ambiguous. T1|T2?, will be parsed as (T1 | T2)? or T1|(T2?)

[Masquue]

If there are too many errors, they should be supported, then we would have to support too many cases.

I respect your argument and effort as a maintainer. But I think this misuse is not entirely groundless.

Currently T1|T2? and T1|nil|T2|nil(maybe unintentionally) are both accepted and mean the same. And according this comment,

in luals ? is also shorthand for marking the type as |nil

then T1?|T2? is a natural expansion of ? -> |nil to T1|nil|T2|nil.

T1?|T2 now is just T1?. I guess the parsing just stops after finding a ?.

T1|T2?, will be parsed as (T1 | T2)? or T1|(T2?)

As for ambiguity, syntactically ? should have the most precedence as a unary operator, thus the above should mean T1|(T2?). Semantically these two are identical, so as long as one type in a | list has a ? the final type has |nil.

I'm not requesting this feature, just offering some suggestions for your consideration. Again, I respect your decisions as maintainers.

[CppCXY]

My comments are not important and usually only represent my personal opinions. They will not influence the decisions of this project. I am not truly maintaining this project; I have my own projects to maintain. Based on the current state of maintenance, unless someone submits a PR to support it, nothing will change.