From: Run Paint Run Run <runrun@...>
Date: 2010-12-16T03:01:23+09:00
Subject: [ruby-core:33731] Re: [feature:trunk] Enumerable#categorize

> pp ary.categorize {|e| [e[1], e[0]] }
> #=>
> {"Yukihiro Matsumoto"=>["matz"],
>   "Nobuyoshi Nakada"=>["nobu"],
>   "Tanaka Akira"=>["akr"],
>   "Usaku NAKAMURA"=>["usa"],
>   "NARUSE, Yui"=>["naruse"],
>   "SASADA Koichi"=>["ko1"]}

ary.map.with_object({}){|e,h| h[e[1]] = [e[0]]}
=> {"Yukihiro Matsumoto"=>["matz"],
      "Nobuyoshi Nakada"=>["nobu"],
      "Tanaka Akira"=>["akr"],
      "Usaku NAKAMURA"=>["usa"],
      "NARUSE, Yui"=>["naruse"],
      "SASADA Koichi"=>["ko1"]}

We currently have #chunk, #group_by--so to some extent #partition--and
#slice_before for segregating an Enumerable into an Enumerable of
Enumerables. I understand their difference on paper, but still need to
experiment with them in IRB before use. This proposal would add two
more methods to this list--#categorize and matz's alternative--for
what appear to be increasingly specialised uses. I'm unconvinced.