dry-initializer: `to_s': wrong number of arguments (1 for 0)

[deepj]

This failed under truffleruby-1.0.0.rc9. No issue under MRI.

require 'dry-initializer'

class Test
  extend Dry::Initializer

  param :foo, proc(&:to_s)
end

Test.new(:FOO)

Error:

dry-initializer.rb:6:in `to_s': wrong number of arguments (1 for 0) (ArgumentError)
	from dry-initializer.rb:6:in `to_s'
	from (eval):5:in `call'
	from (eval):5:in `__dry_initializer_initialize__'
	from ~/.gem/truffleruby/2.4.4/gems/dry-initializer-2.5.0/lib/dry/initializer/mixin/root.rb:7:in `initialize'
	from dry-initializer.rb:9:in `new'
	from dry-initializer.rb:9:in `<main>'

[chrisseaton]

Thanks - we'll have to dig into the dry stack a bit for this.

[deepj]

@eregon Is possible to look at this issue, please? This issue affects the whole dry-rb/ROM ecosystem. It's some ugly issue with metaprogramming,

[eregon]

@deepj This variant seems to work fine:

require 'dry-initializer'

class Test
  extend Dry::Initializer

  # param :foo, proc(&:to_s)
  param :foo, &:to_s
end

p Test.new(:FOO)

[eregon]

Mmh, but that's different as it passes a block argument to param.

require 'dry-initializer'

class Test
  extend Dry::Initializer

  # param :foo, proc(&:to_s)
  param :foo, :to_s.to_proc
end

p Test.new(:FOO)

also fails.

[eregon]

This works:

require 'dry-initializer'

class Test
  extend Dry::Initializer

  param :foo, -> obj { obj.to_s }
end

p Test.new(:FOO)
p Test.new(:FOO).foo # => "FOO"

The code generated by dry-initializer here differs based on the Proc#arity of a Symbol#to_proc Proc. TruffleRuby currently returns -2, MRI -1.

[eregon]

@deepj Fix coming.
FWIW, an arity of -2 would better express the number of required arguments: Symbol#to_proc needs at least one argument (the receiver).
An arity of -1 means "0 or more arguments".
:to_s.to_proc.call gives ArgumentError: no receiver given on MRI, showing at least one argument is indeed required.

So it might be a MRI bug, on which dry-initializer depends. I'll report it to them.
In any case, I'll follow MRI semantics for now for compatibility.

[eregon]

I filed dry-rb/dry-initializer#52.

[eregon]

@deepj Fixed, it should be in the next release. Thank you for the report and the ping!

[deepj]

@eregon thank you! it's great this has been fixed. I can't wait on running dry-rb/ROM on TruffleRuby. If you would have another time, this is another issue in that ecosystems. #1469

I'm not sure how this can be relative to this issue.