unary operators eat function call parentheses when argument is a tuple

[eschnett]

This is Julia 0.4-dev in the Jupyter REPL.

This calls f with two arguments: f(1,2)
This calls f with one argument: f((1,2)) (the argument is a tuple).
This calls + with two arguments: +(1,2)
But this doesn't work: +((1,2)) -- it still calls + with two arguments instead of a tuple. This is very counter-intuitive, and I argue it is wrong.

Work-arounds are this: (+)((1,2)) (put operator into parenthesis) or this: +((1,2),) (enforce a single-argument function call via a trailing comma).

(This trailing-comma syntax is documented for tuples; is it also intended and/or documented for function calls?)

The same issue exists for -, !, and ~.

[eschnett]

Here is a test case: Compare the values of

:( (+)((1,2)) ) |> string

and

:( (+)((1,2)) ) |> string |> parse |> string

which differ, i.e. string and parse disagree about the meaning of this construct.

[JeffBezanson]

Yes, I think this can/should be fixed.

[mschauer]

@timholy What was the intention of the revert, should this be reopened?

[KristofferC]

@mschauer #12771

[mschauer]

Thanks.

[stevengj]

Seems be fixed nowadays:

julia> +((1,2))
ERROR: MethodError: no method matching +(::Tuple{Int64,Int64})
Closest candidates are:
  +(::Any, ::Any, ::Any, ::Any...) at operators.jl:538
  +(::Pkg.Resolve.FieldValue, ::Pkg.Resolve.FieldValue) at /Users/julia/buildbot/worker/package_macos64/build/usr/share/julia/stdlib/v1.5/Pkg/src/Resolve/fieldvalues.jl:43
  +(::Float32, ::Float32) at float.jl:400
  ...
Stacktrace:
 [1] top-level scope at REPL[13]:1

[mschauer]

Reference: #26154