Backport #8463
closed
Proc auto-splat bug with named arguments
Description
=begin
I'd expect a proc to either do an implicit splat or not, but right now it looks for options before doing the implicit splat. Should it not do it after doing the implicit splat?
I thought that when a proc had an argument list with more than one element, it was the same to call it with a single array argument than with the same array splatted:
Proc{|a, ...| ... }.call([...]) == Proc{|a, ...}| ... }.call(*[...]) # => Because of implicit splat
But we have currently:
Proc.new{|a, *b, **c| p a, b, c}.call(1,2, bar: 3)
=> 1, [2], {:bar=>3} : OK¶
Proc.new{|a, *b, **c| p a, b, c}.call([1,2, bar: 3])
=> 1, [2, {:bar=>3}], {}: Expected same as above¶
Proc.new{|(a, *b), **c| p a, b, c}.call([1,2], bar: 3)
=> 1, [2], {:bar=>3} : OK¶
Proc.new{|(a, *b), **c| p a, b, c}.call([[1,2], bar: 3])
=> [1, 2], [{:bar=>3}], {}: Expected same as above¶
As an additional note, this affects some methods of Enumerable when yielding multiple arguments.
For example:
def each; yield 1, 2, bar: 3; end
include Enumerable
each{|a, *b, **c| p a, b, c} # => 1, [2], {:bar => 3}: ok
detect{|a, *b, **c| p a, b, c} # => 1, [2, {:bar => 3}], {}: should be the same, no?
=end
Updated by 
Updated by