parsing ambiguity of .! ?

[Sacha0]

With

julia> versioninfo()
Julia Version 0.6.0-dev.2673
Commit e24faa5* (2017-02-08 21:01 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin15.6.0)
  CPU: Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, ivybridge)

julia> [[true] !([true])]
1×2 Array{Bool,2}:
 true  false

I expected [[true] .!([true])] to function identically, but instead this yields a binary ! broadcast

julia> [[true] .!([true])]
ERROR: MethodError: no method matching !(::Bool, ::Bool)
Closest candidates are:
  !(::Bool) at bool.jl:36
Stacktrace:
 [1] broadcast_t(::Function, ::Type{Any}, ::Tuple{Base.OneTo{Int64}}, ::CartesianRange{CartesianIndex{1}}, ::Array{Bool,1}, ::Array{Bool,1}) at ./broadcast.jl:255
 [2] broadcast_c at ./broadcast.jl:298 [inlined]
 [3] broadcast(::Function, ::Array{Bool,1}, ::Array{Bool,1}) at ./broadcast.jl:406

julia> expand(:([[true] .!([true])]))
:((Base.vect)((Base.broadcast)(!, (Base.vect)(true), (Base.vect)(true))))

likewise e.g.

julia> [true .!true]
ERROR: MethodError: no method matching !(::Bool, ::Bool)
Closest candidates are:
  !(::Bool) at bool.jl:36
Stacktrace:
 [1] (::##7#8)(::Bool) at ./<missing>:0
 [2] broadcast(::Function, ::Bool) at ./broadcast.jl:16

Am I making a mistake, or is this a parsing issue? cc @stevengj. Best!

[Sacha0]

(Discovered from failure when writing @deprecate !(B::BitArray) .!B.)

[stevengj]

Seems like a parsing issue, since [1 .+1] parses as [1 (.+1)] I think the same should be true for other unary operators.

Actually, it's stronger than that: since ! is only a unary operator, it should parse as unary regardless of the spacing.

[Sacha0]

Actually, it's stronger than that: since ! is only a unary operator, it should parse as unary regardless of the spacing.

Agreed, I am confused why .! parses as binary in some cases. Perhaps

julia/src/julia-parser.scm

Line 19 in ddb1e1e

(append! '(|<:| |>:| |.!| in isa)

, from 814c974?

[stevengj]

I'm marking this as a bug, since .! should never parse as a binary operator.

[stevengj]

Yes, it seems like .! should be removed from prec-comparison

[stevengj]

Ah, now I see why it is there: see #17393 (comment)

This hack should have been removed when .! was parsed as a unary operator. Can you submit the one-line PR (+ test) to remove .! from prec-comparison?