Bug #12104: Procs keyword arguments affect value of previous argument

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: Tested in 2.3.0 and 2.2.3
• Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN, 2.3: UNKNOWN

[cheerfulstoic (Brian Underwood)]

This seems like a bug to me:

2.3.0 :001 > p = Proc.new {|nodes, match_array: false| puts nodes.inspect }
 => #<Proc:0x007fa52c0659e0@(irb):1>
2.3.0 :002 > p.call([])
nil
 => nil
2.3.0 :003 >  p.call([], match_array: true)
[]
 => nil

When I try the same thing in a method I get the behavior I would expect:

2.3.0 :004 > def foo(nodes, match_array: false)
2.3.0 :005?>   puts nodes.inspect
2.3.0 :006?>   end
 => :foo
2.3.0 :007 > foo([])
[]
 => nil
2.3.0 :008 > foo([], match_array: true)
[]

[cheerfulstoic (Brian Underwood)]

This seems to also happen when trying to use a options Hash as the second argument:

2.3.0 :009 > p = Proc.new {|nodes, options = {}| puts nodes.inspect }
 => #<Proc:0x007fa52b13b348@(irb):9>
2.3.0 :010 > p.call([])
nil
 => nil
2.3.0 :011 > p.call([], match_array: true)
[]

[cheerfulstoic (Brian Underwood)]

Looking deeper now it seems to be that this is because the first argument is an Array and that when that happens it's interpreting that as the elements of the array being the arguments of the Proc. Is that supposed to happen?

[ko1 (Koichi Sasada)]

• Status changed from Open to Rejected

You are right.
It is intentional.

[znz (Kazuhiro NISHIYAMA)]

How about using lambda instead of Proc.new?

>> p = lambda {|nodes, match_array: false| puts nodes.inspect }
=> #<Proc:0x0000000183a0a8@(irb):1 (lambda)>
>> p.call([])
[]
=> nil
>> p = lambda {|nodes, options = {}| puts nodes.inspect }
=> #<Proc:0x00000001850c68@(irb):3 (lambda)>
>> p.call([])
[]
=> nil

[hsbt (Hiroshi SHIBATA)]

• Related to Feature #14183: "Real" keyword argument added

[marcandre (Marc-Andre Lafortune)]

• Related to deleted (Feature #14183: "Real" keyword argument)