Add pp in stdlib

[ksss]

I added the type for the standard library pp since it was missing.

I am performing the following tasks also:

• Modifying the test code generation template.
• Fix Kernel#pp.
  • ref: Fix Kernel.p and Kernel.pp type with no argument #1342
• Fix RBS::Types::Optional#to_s with Proc

[ksss]

It seems there might be an issue with the RBS::Writer.

# original
def seplist: (untyped list, ?(^() -> void)? sep, ?interned iter_method) { (*untyped, **untyped) -> void } -> void

# RBS::Writer
def seplist: (untyped list, ?^() -> void? sep, ?interned iter_method) { (*untyped, **untyped) -> void } -> void

The following RBS is interpreted as next follows:

^() -> void?

^() -> (void | nil)

[soutaro]

The second untyped was actually to skip the first overload if one argument is given.
Not sure if removing it causes another type error or not. (It won't on Steep.)

[ksss]

Oh, I see, that's what it meant.

The original definition is as follows:

module Kernel
  def pp(*objs)
    objs.each {|obj|
      PP.pp(obj)
    }
    objs.size <= 1 ? objs.first : objs
  end
  module_function :pp
end

I became concerned because Steep reported the following:

The method parameter has different kind from the declaration `[T(353)(354)] (?untyped, ?untyped, *untyped) -> (T(353)(354) | ::Array[untyped] | nil)`(Ruby::DifferentMethodParameterKind)

[soutaro]

@ksss I mean the use-site. It depends on how the overloading is resolved. (Steep uses the first matching one, so passing one argument to pp resolves to the first overload anyways.)

[ksss]

OK. I'll revert this commit later.

[soutaro]

👍

[soutaro]

@ksss I found that pp is automatically loaded today. Moving to core would make more sense?

[ksss]

@soutaro Thank you. Certainly, that's true.
I believe pp should be treated the same way as set.

I will move it to the core soon.

[ksss]

@soutaro
Oh wait.

If we move pp to the core, then its dependency pretty_print must also be moved to the core.
Looking at https://github.com/ruby/ruby/blob/5c1b7633fcd9c7c43974b72bd36cab1f9e9bd186/prelude.rb, we can see that while Set is autoloaded, but PP and PrettyPrint are not.
This might be because PP and PrettyPrint are used less frequently than Set.

It seems that pp is not exactly treated the same as set.

Moreover, the signature for pp is already included in the core.

rbs/core/kernel.rbs

Lines 1262 to 1272 in 32cd6c9

	# <!--
	# rdoc-file=lib/pp.rb
	# - pp(*objs)
	# -->
	# prints arguments in pretty form.
	#
	# pp returns argument(s).
	#
	def self?.pp: [T] (T arg0) -> T
	| (untyped, untyped, *untyped) -> Array[untyped]
	| () -> nil

I believe it might be better not to add pp to the core.

[soutaro]

@ksss Got it.

#pp (method) can be used without explicitly loading it. PP (the library) should be loaded manually. It looks reasonable.

Clicking the merge button. 👍