Mutating functions as keyword arguments (yet more parser space sensitivity)

[timholy]

I doubt there's anything to be done here, but just FYI here's a fun way to play head games with the parser:

julia> mysum(A; transform! = identity!) = sum(transform!(A))
mysum (generic function with 1 method)

julia> mysum(A; transform!=identity!) = sum(transform!(A))
ERROR: syntax: invalid keyword argument syntax "transform!=identity!" (expected assignment)

The issue is that != is being parsed as not-equals.

[JeffBezanson]

Yes, a!=b is ambiguous, and we have to resolve it in favor of !=, since comparing values is much more common than identifier names with !.

[johnmyleswhite]

I know I've said this before and people might hate hearing me rant again, but we probably should just enforce spaces between operators and operands. We have enough places where the language allows an ambiguity between the ends of identifiers and the start of operators that people are almost certainly going to keep tripping on these issues indefinitely far into the future.

[timholy]

Given that the error message is already pretty good, I'll close this.

The space around operators might be a good way to go, but I do fear how much it might cause some of my more complicated calculations to grpw.

[tkelman]

There's also the issue that whitespace is an operator in the case of horizontal concatenation, which maybe just maybe could get figured out and changed as part of the great breaking array reshuffle of 0.4. If only keyboards had a few more delimiters available.