Updated the patch with documentation. There's an interesting side effect to be considered. Method#to_proc
returns a lambda that checks its arity. Proc#curry
takes an optional arity
argument. If this doesn't match the method's arity then the result can not be called.
def two_args(a,b)
end
curried = method(:two_args).to_proc.curry(1)
curried.(1)
# -:4:in `curry': wrong number of arguments (1 for 2) (ArgumentError)
# from -:4:in `<main>'
curried.(1, 2)
# -:4:in `curry': wrong number of arguments (1 for 2) (ArgumentError)
# from -:4:in `<main>'
curried.(1, 2, 3)
# -:4:in `curry': wrong number of arguments (1 for 2) (ArgumentError)
# from -:4:in `<main>'
In the case of a method with variable arguments the arity argument is important though, because otherwise the currying has no effect.
def varargs(*args)
args
end
curried = method(:varargs).to_proc.curry
curried.(1) # => [1]
curried.(1,2) # => [1, 2]
curried = method(:varargs).to_proc.curry(3)
curried.(1) # => #<Proc:0x007f0cb8fc6fe0 (lambda)>
curried.(1).(2).(3) # => [1, 2, 3]
I think it would make sense for a Method#curry
method to raise an error or at least give a warning when passing the wrong arity to a fixed-arity function.