From: boris@...
Date: 2014-02-21T18:54:22+00:00
Subject: [ruby-core:60952] [ruby-trunk - Feature #7939] Alternative curry	function creation

Issue #7939 has been updated by Boris Stitnicky.


Thanks to Yui Naruse for reopening this issue. As a matter of fact, I hate the current behavior of `#curry` method with passion. In line with the OP suggestion, I propose to define `#curry` as follows:

```ruby
class Proc
  def curry *args
    hash = args[0]
    ordered_curry, named_curry = hash.each_with_object [{}, {}] do |(k, v), m|
      case k
      when Integer then m[0][k] = v
      when Symbol then m[1][k] = v
      else fail ArgumentError end
    end
    ordered_min_size = ordered_curry.keys.max
    o = -> *args {
      loop.with_index.with_object [] do |(_, i), ary|
        break ary if args.empty? and i > ordered_min_size
        ary << ( ordered_curry[ i ] || args.shift )
      end
    }
    n = -> **named_args { named_curry.update named_args }
    -> *ordered_args, **named_args {
      self.( *o.( *ordered_args ), **n.( **named_args ) )
    }
  end
end
```

And then, the usage would be:

```ruby
meal = -> main_course, soup, appetizer: nil, dessert: nil, drink: "beer" do
  { main_course: main_course,
    soup: soup,
    appetizer: appetizer,
    dessert: dessert,
    drink: drink }
end

a_la_carte = meal.curry( 1 => "chicken broth",
                         dessert: "creme brule",
                         drink: "mineral water" )

a_la_carte.( "coq au vin" )
#=> {:main_course=>"coq au vin", :soup=>"chicken broth", :appetizer=>nil, :dessert=>"creme brule", :drink=>"mineral water"}

a_la_carte.( "coq au vin", dessert: "blueberry cake" )
#=> {:main_course=>"coq au vin", :soup=>"chicken broth", :appetizer=>nil, :dessert=>"blueberry cake", :drink=>"mineral water"}

```

There are some unresolved issues lurking in the above prototype code (such as how to curry methods and what to do with blocks), but since I took the pain to write it, I'm gonna use it in my personal library. But seriously, this is the only form of `#curry` which I can understand and memorize. Current `#curry` with its `n` argument is absolutely awful. Phew. I came out and actually said what I think this time.

----------------------------------------
Feature #7939: Alternative curry function creation
https://bugs.ruby-lang.org/issues/7939#change-45352

* Author: Kresimir Bojcic
* Status: Assigned
* Priority: Normal
* Assignee: Yukihiro Matsumoto
* Category: core
* Target version: next minor
----------------------------------------
=begin
I really like the new "(({assuming}))" method used for currying in Perl 6.

For example if I loose my mind and implement ((%fizzbuzz%)) via currying I can do it like this:

  fb = ->(modulo_number, message, x) { message if x % modulo_number == 0 }
  fizzbuzz = fb.curry[15,"FizzBuzz"]
  fizz = fb.curry[3, "Fizz"]
  buzz = fb.curry[5, "Buzz"]
  (1..100).each { |i| puts fizzbuzz[i] || fizz[i] || buzz[i] || i }

Here the first hurdle is that curry is somewhat mathematical, and the secons is that you need to use (({[]})) for function invoking...
If we had something similar to this:

  class Proc
    def assuming(*args)
      curry.call *args
    end
  end

It could be written more naturally IMO:

  fb = ->(modulo_number, message, x) { message if x % modulo_number == 0 }
  fizzbuzz = fb.assuming(15,"FizzBuzz")
  buzz = fb.assuming(5, "Buzz")
  fizz = fb.assuming(3,"Fizz")

  (1..100).each { |i| puts fizzbuzz[i] || fizz[i] || buzz[i] || i }
=end




-- 
http://bugs.ruby-lang.org/