allow unicode modifiers after '

[simeonschaub]

As discussed in the triage call, this is technically breaking because we currently parse a'ᵀ as a' * ᵀ and a'ᵀb as a' * ᵀb, but it doesn't look like this is actually used in the wild. Still needs news.

fixes #34507

[JeffBezanson]

This should be handled in the lexer, by removing ' from the no-suffix list, then the parser test needs to be changed to use a SuffSet.

[simeonschaub]

Won't this cause any issues with character literals? Or is that handled differently?

[simeonschaub]

I think we could use read-operator directly for this though.

[simeonschaub]

This should be handled in the lexer, by removing ' from the no-suffix list, then the parser test needs to be changed to use a SuffSet.

When I remove ctrans-op form the no-suffix list, 'ᵀ' parses as (|'| |'ᵀ|), so there needs to be an additional argument to read-operator when parsing a call chain vs when starting to parse a new atom.

[JeffBezanson]

Oops, you're right.

[JeffBezanson]

One thing to think about is whether isoperator should return true for these.

[simeonschaub]

That's a good point! I think it should return true, because to the user, it acts just like ' does. I would say whether (operator? t) is true is mostly an implementation detail of the parser, but that for all practical purposes 'ᵀ is also an operator.

[simeonschaub]

Should I just change the Julia function isoperator, or should I add a check to the C function jl_is_operator to return true here?

[simeonschaub]

Other than my comment I think this should be good to go.

[simeonschaub]

@JeffBezanson Sorry for bothering you, but would you be willing to take another look and merge if ok?

[simeonschaub]

Bump. 🙂

[JeffBezanson]

Looks good, just needs a rebase.

[simeonschaub]

Ok?