[ruby-core:49505] [ruby-trunk - Feature #6758] Object#sequence

Issue #6758 has been updated by boris_stitnicky (Boris Stitnicky).


The method seems useful, but adding it to Object or Kernel would be feature creep. I still remember how I was memorizing basic Ruby classes. The number of features in them is already such, that adding new features raises the bar for novices more than linearly.

I suggest to make this a public class method of Enumerator, which it constructs. As for the name, I do not think that "sequence" is optimal. Better name would be "produce", similar sounding but not same as "reduce". Like this:

Enumerator.produce( 1 ) { |a| a + 2 }

I would definitely leave it up to the user alone to add the shortcut method to Object class, if ze wants.
----------------------------------------
Feature #6758: Object#sequence
https://bugs.ruby-lang.org/issues/6758#change-33037

Author: merborne (kyo endo)
Status: Open
Priority: Normal
Assignee: 
Category: core
Target version: next minor


=begin

== Object#sequence

Let me propose a new method ((*Object#sequence*)).

次のような実装の((*Object#sequence*))を提案します。

    class Object
      def sequence(init=true, &blk)
        x = self
        Enumerator.new do |y|
          y << x if init
          loop { y << (x = yield x) }
        end
      end
    end

((*sequence*)) generate a sequence by applying a block recursively to the receiver object. The result is wrapped with a Enumerator object, thus it is set under lazy evaluation.

((*sequence*))は、そのレシーバオブジェクトを初期値として、渡されたブロック（漸化式）を繰り返し適用してシーケンスを生成します。適用の結果はEnumeratorオブジェクトでラップされているので、遅延評価されます。

== Usage;

使い方を示します。

    1.sequence { |x| x + 2 }.take(10) # => [1, 3, 5, 7, 9, 11, 13, 15, 17, 19]
    
    3.sequence { |x| x * 2 }.take(10) # => [3, 6, 12, 24, 48, 96, 192, 384, 768, 1536]
    
    [0, 1].sequence { |a, b| [b, a + b] }.take(10).map(&:first) # => [0, 1, 1, 2, 3, 5, 8, 13, 21, 34]
    
    [0, 1, 1].sequence { |a, b, c| [b, c, a + b + c] }.take(10).map(&:first) # => [0, 1, 1, 2, 4, 7, 13, 24, 44, 81]
    
    # square root 5
    a = 5
    eps = 0.0001
    1.0.sequence { |x| (x + a/x) / 2.0 }
       .each_cons(2)
       .detect { |a, b| (a - b).abs < eps }[1] # => 2.236067977499978
    
    # Excel label
    'A'.sequence { |x| x.succ }.take(30) # => ["A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z", "AA", "AB", "AC", "AD"]
    
    # random boolean(avoid true-true sequence)
    true.sequence { |prev| prev ? false : [true, false].sample }.take(20) # => [true, false, true, false, true, false, true, false, true, false, true, false, true, false, true, false, false, false, true, false]
    
== Some background

== 提案の経緯

Let me explain some background of this request.

本件に関しては、ここに至る若干の経緯がありますので、併せて説明します。

1. I introduced this method as ((*Object#repeat*)) on my Japanese blog.

1. 私のブログにおいて、本メソッドを((*Object#repeat*))として紹介する記事を公開。

((<URL:http://melborne.github.com/2012/07/12/object-repeat-take-care-of-sequence/>))

2. Matz tweeted to the article.

    > "um.. the feature is attractive, but the name is.." 

2. 記事に対してMatzがつぶやく。

    > 「うーん、昨日としては魅力的だけど、名前がなあ。」

((<URL:https://twitter.com/yukihiro_matz/status/223790181113806848>))


3. I updated the article to propose ((*Object#repeat_apply*)) or ((*Object#repeat_call*)).

3. Object#repeat_apply, Object#repeat_callを提案するべく記事を更新。

4. Matz tweeted to the article.

    > @merborne more clear. but combining two verbs is wrong. I understand naming is difficult.

4. 記事に対してMatzがつぶやく。

    > @merborne なるほど「repeat_apply」か「repeat_call」ですか。repeat単体よりは誤解を受けにくいとは思いますが、repeatって動詞とapply/callって動詞の組み合わせは感心しませんね。名前って難しい。 

((<URL:https://twitter.com/yukihiro_matz/status/224105896110866432>))

5. Matz tweeted to the article.

    > @merborne I suggest some clue lies around a word "series"..

5. 記事に対してMatzがつぶやく。

    > @merborne 「級数/series」あたりに名前のヒントがありそうな。

((<URL:https://twitter.com/yukihiro_matz/status/224106160591081472>))

6. I tweeted to Matz.

6. 私もつぶやく

    > @yukihiro_matz you are right.. `repeated_apply` `repeated_call`?.. 


    > @yukihiro_matz たしかに.. repeated_apply repeated_callかな.. 

((<URL:https://twitter.com/merborne/status/224108387653259264>))

    > @yukihiro_matz clue! `series_by` ? 


    > @yukihiro_matz ヒント! series_by ? 

((<URL:https://twitter.com/merborne/status/224108809948377088>))

    > @yukihiro_matz `repeated` is adjective..^^; but I don't like `repeatedly_apply`. I thought once `series_by` is good, but it would be better a method is named based on its behavior, not on its function, I think. How about `repeat_by`?


    > @yukihiro_matz repeatedは形容詞でしたね^^; するとrepeatedly_applyですが、今ひとつです。series_byもいいと思ったんですが、できれば機能ではなく動作を言いたい思いがあります。そこでrepeat_byというのはどうでしょうか。

((<URL:https://twitter.com/merborne/status/224324670764220416>))


the conversation closed..

会話終了..

7. Ideas from other Rubyists

Some Japanese Rubyists tweeted or commented me on the name. Candidates are..

7. 他のRubyistの意見

名前に関し何人かのRubyistからアイディアをもらっています。リストアップします。

    iterate
    recur
    recurrence
    recur_with
    unfold
    sequence
    seq



Haskell and Scala have the same functionality with a name of ((*iterate*)).


なお、HaskellとScalaには同種の機能が、((*iterate*))という名前で実装されているそうです。


Also, @makotokuwata-san tweeted that ((*Kernel#seq*))(function style) or ((*Enumerator.seq()*))(a Class method) is better.


また、@makotokuwata氏よりKernel#seqのような関数形式か、Enumerator.seq()のようなクラスメソッドのほうがいいとの意見も頂いています。

    > @yukihiro_matz @merborne 初期値と漸化式から順列を作る機能なので、"sequence"か"seq"に1票。あとKernel#seq(initial,&blk)のほうが好きです。RT 「級数/series」あたりに名前のヒントがありそうな。 

((<URL:https://twitter.com/makotokuwata/status/225806204390227968>))



Thank you for your consideration.

以上、ご検討のほどよろしくお願い致します。


=end



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