サロゲートペアなどを途中で切り詰めない様にしたい

bench-grapheme-length.rb

require 'benchmark'

N = 10

str = "がぎぐげご" * 100
gra = "か\u3099き\u3099く\u3099け\u3099こ\u3099" * 100

Benchmark.bm(64) do |x|
  x.report('String#size (1)') do |x|
    N.times { str.size }
  end

  x.report('String#size (2)') do |x|
    N.times { gra.size }
  end

  x.report('ActiveSupport::Multibyte::Unicode.unpack_graphemes (1)') do
    N.times { ActiveSupport::Multibyte::Unicode.unpack_graphemes(str) }
  end

  x.report('ActiveSupport::Multibyte::Unicode.unpack_graphemes (2)') do
    N.times { ActiveSupport::Multibyte::Unicode.unpack_graphemes(gra) }
  end

  x.report('String#scan.size (1)') do
    N.times { str.scan(/\X/).size }
  end

  x.report('String#scan.size (2)') do
    N.times { gra.scan(/\X/).size }
  end

  x.report('String#grapheme_clusters.size (1)') do
    N.times { str.grapheme_clusters.size }
  end

  x.report('String#grapheme_clusters.size (2)') do
    N.times { gra.grapheme_clusters.size }
  end

  x.report('String#each_grapheme_cluster.count (1)') do
    N.times { str.each_grapheme_cluster.count }
  end

  x.report('String#each_grapheme_cluster.count (2)') do
    N.times { gra.each_grapheme_cluster.count }
  end

  x.report('String#each_grapheme_cluster.size (1)') do
    N.times { str.each_grapheme_cluster.size }
  end

  x.report('String#each_grapheme_cluster.size (2)') do
    N.times { gra.each_grapheme_cluster.size }
  end
end

__END__
                                                                       user     system      total        real
String#size (1)                                                    0.000018   0.000008   0.000026 (  0.000022)
String#size (2)                                                    0.000012   0.000002   0.000014 (  0.000012)
ActiveSupport::Multibyte::Unicode.unpack_graphemes (1)             2.384050   0.043771   2.427821 (  2.440785)
ActiveSupport::Multibyte::Unicode.unpack_graphemes (2)             4.380538   0.015965   4.396503 (  4.457205)
String#scan.size (1)                                               0.005429   0.000144   0.005573 (  0.005697)
String#scan.size (2)                                               0.006267   0.000076   0.006343 (  0.006353)
String#grapheme_clusters.size (1)                                  0.004079   0.000104   0.004183 (  0.004222)
String#grapheme_clusters.size (2)                                  0.003742   0.000058   0.003800 (  0.003826)
String#each_grapheme_cluster.count (1)                             0.004119   0.000023   0.004142 (  0.004185)
String#each_grapheme_cluster.count (2)                             0.003672   0.000006   0.003678 (  0.003688)
String#each_grapheme_cluster.size (1)                              0.003329   0.000010   0.003339 (  0.003425)
String#each_grapheme_cluster.size (2)                              0.003561   0.000017   0.003578 (  0.003624)

bench-truncate.rb

require 'benchmark'

N = 1_000

str = "がぎぐげご" * 100
gra = "か\u3099き\u3099く\u3099け\u3099こ\u3099" * 100

Benchmark.bm(80) do |x|
  x.report('String#truncate (1)') do
    N.times { str.truncate(100, omission: "...") }
  end

  x.report('String#truncate (2)') do
    N.times { str.truncate(100, omission: "...", separator: "ぐ") }
  end

  x.report('String#truncate (3)') do
    N.times { gra.truncate(100, omission: "\u{2702 fe0f}") }
  end

  x.report('String#truncate (4)') do
    N.times { gra.truncate(100, omission: "\u{2702 fe0f}", separator: "く\u3099") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_count (1)') do
    N.times { str.truncate_graphemes(100, omission: "...") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_count (2)') do
    N.times { str.truncate_graphemes(100, omission: "...", separator: "ぐ") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_count (3)') do
    N.times { gra.truncate_graphemes(100, omission: "\u{2702 fe0f}") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_count (4)') do
    N.times { gra.truncate_graphemes(100, omission: "\u{2702 fe0f}", separator: "く\u3099") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_size (1)') do
    N.times { str.truncate_graphemes_with_each_grapheme_cluster_size(100, omission: "...") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_size (2)') do
    N.times { str.truncate_graphemes_with_each_grapheme_cluster_size(100, omission: "...", separator: "ぐ") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_size (3)') do
    N.times { gra.truncate_graphemes_with_each_grapheme_cluster_size(100, omission: "\u{2702 fe0f}") }
  end

  x.report('String#truncate_graphemes_with_each_grapheme_cluster_size (4)') do
    N.times { gra.truncate_graphemes_with_each_grapheme_cluster_size(100, omission: "\u{2702 fe0f}", separator: "く\u3099") }
  end

  x.report('String#truncate_graphemes_with_grapheme_clusters_size (1)') do
    N.times { str.truncate_graphemes_with_grapheme_clusters_size(100, omission: "...") }
  end

  x.report('String#truncate_graphemes_with_grapheme_clusters_size (2)') do
    N.times { str.truncate_graphemes_with_grapheme_clusters_size(100, omission: "...", separator: "ぐ") }
  end

  x.report('String#truncate_graphemes_with_grapheme_clusters_size (3)') do
    N.times { gra.truncate_graphemes_with_grapheme_clusters_size(100, omission: "\u{2702 fe0f}") }
  end

  x.report('String#truncate_graphemes_with_grapheme_clusters_size (4)') do
    N.times { gra.truncate_graphemes_with_grapheme_clusters_size(100, omission: "\u{2702 fe0f}", separator: "く\u3099") }
  end

  x.report('String#gsub (1)') do
    N.times { gra.gsub(/./, '.') }
  end

  x.report('String#gsub (2)') do
    N.times { gra.gsub(/\X/, '.') }
  end
end

__END__
                                                                                       user     system      total        real
String#truncate (1)                                                                0.001988   0.000649   0.002637 (  0.002635)
String#truncate (2)                                                                0.003430   0.000627   0.004057 (  0.004072)
String#truncate (3)                                                                0.002081   0.000568   0.002649 (  0.002652)
String#truncate (4)                                                                0.004323   0.000713   0.005036 (  0.005125)
String#truncate_graphemes_with_each_grapheme_cluster_count (1)                     0.496839   0.035412   0.532251 (  0.544658)
String#truncate_graphemes_with_each_grapheme_cluster_count (2)                     0.546470   0.007729   0.554199 (  0.568006)
String#truncate_graphemes_with_each_grapheme_cluster_count (3)                     0.431042   0.001793   0.432835 (  0.433713)
String#truncate_graphemes_with_each_grapheme_cluster_count (4)                     0.422881   0.000698   0.423579 (  0.424463)
String#truncate_graphemes_with_each_grapheme_cluster_size (1)                      0.366217   0.004378   0.370595 (  0.371266)
String#truncate_graphemes_with_each_grapheme_cluster_size (2)                      0.527039   0.013160   0.540199 (  0.550277)
String#truncate_graphemes_with_each_grapheme_cluster_size (3)                      0.402696   0.010364   0.413060 (  0.418723)
String#truncate_graphemes_with_each_grapheme_cluster_size (4)                      0.390556   0.010561   0.401117 (  0.401974)
String#truncate_graphemes_with_grapheme_clusters_size (1)                          0.420647   0.014305   0.434952 (  0.435687)
String#truncate_graphemes_with_grapheme_clusters_size (2)                          0.527312   0.010127   0.537439 (  0.544975)
String#truncate_graphemes_with_grapheme_clusters_size (3)                          0.478933   0.006017   0.484950 (  0.488538)
String#truncate_graphemes_with_grapheme_clusters_size (4)                          0.426145   0.001380   0.427525 (  0.428163)
String#gsub (1)                                                                    0.149356   0.000599   0.149955 (  0.150370)
String#gsub (2)                                                                    0.290007   0.001179   0.291186 (  0.291967)

string_truncate_graphemes.rb

class String
  # 自分自身を指定された長さ(書記素クラスタ単位)で切り詰める。
  #
  # @see https://github.com/rails/rails/blob/bfd296dda797e597e8a54709d1cd331cdffaa9f7/activesupport/lib/active_support/core_ext/string/filters.rb#L48-L79
  # @see https://github.com/rails/rails/blob/9cc463ed7b7be098602b72a98f72220ea6466ba2/activesupport/lib/active_support/core_ext/string/filters.rb#L81-L120
  #
  def truncate_graphemes(truncate_at, options = {})
    return dup if each_grapheme_cluster.size <= truncate_at

    omission = options[:omission] || '...'
    length_with_room_for_omission = truncate_at - omission.each_grapheme_cluster.size

    truncated = self[/\X{#{length_with_room_for_omission}}/]
    if options[:separator] && stop = truncated.rindex(options[:separator])
      truncated = truncated[0, stop]
    end

    "#{truncated}#{omission}"
  end
end

string_truncate_graphemes_spec.rb

RSpec.describe String do
  describe '#truncate_graphemes' do
    context 'omission 未指定(default)' do
      context '切り詰め位置が切り詰め対象文字列の長さ(書記素クラスタ単位)未満' do
        context '切り詰め位置が omission (...) の長さ(書記素クラスタ単位)以下' do
          it 'omission だけを返す' do
            # ààààà
            original = "\u{0061 0300 0061 0300 0061 0300 0061 0300 0061 0300}"
            expect(original.truncate_graphemes(0)).to eq '...'
            expect(original.truncate_graphemes(3)).to eq '...'
          end
        end

        context '切り詰め位置が omission (...) の長さ(書記素クラスタ単位)より大きい' do
          it '書記素クラスタ単位で切り詰めた結果に omission を付与する' do
            # ààààà
            original = "\u{0061 0300}" * 5
            expect(original.truncate_graphemes(4)).to eq "\u{0061 0300}..."
          end
        end
      end

      context '切り詰め位置が切り詰め対象文字列の長さ(書記素クラスタ単位)以上' do
        it '切り詰めない' do
          # ààààà
          original = "\u{0061 0300}" * 5
          expect(original.truncate_graphemes(5)).to eq original
          expect(original.truncate_graphemes(6)).to eq original
        end
      end
    end

    context 'omission にサロゲートペアを含む' do
      it 'omission の長さも書記素クラスタ単位で数えた上で切り詰めを行う' do
        # ààààà
        original = "\u{0061 0300}" * 5
        # è
        omission = "\u{0065 0300}"
        expect(original.truncate_graphemes(3, omission: omission)).to eq "\u{0061 0300 0061 0300 0065 0300}"
      end
    end

    context 'separator 指定あり(サロゲートペアを含む)' do
      context '切り詰め対象文字列が separator を含む' do
        context '切り詰め位置が separator である' do
          it 'その位置で切り詰める' do
            # è
            separator = "\u{0065 0300}"
            # àààèàààèààà
            original = Array.new(3, "\u{0061 0300}" * 3).join(separator)

            truncated = original.truncate_graphemes(8, separator: separator, omission: '')
            expect(truncated).to eq "\u{0061 0300 0061 0300 0061 0300 0065 0300 0061 0300 0061 0300 0061 0300}"
          end
        end

        context '切り詰め位置が separator でない' do
          it '一つ前の separator 直前で切り詰める' do
            # è
            separator = "\u{0065 0300}"
            # àààèàààèààà
            original = Array.new(3, "\u{0061 0300}" * 3).join(separator)

            truncated = original.truncate_graphemes(7, separator: separator, omission: '')
            expect(truncated).to eq "\u{0061 0300 0061 0300 0061 0300}"
          end
        end
      end

      context '切り詰め対象文字が separator を含まない' do
        it '切り詰める' do
          # ààààà
          original = "\u{0061 0300}" * 5
          # è
          separator = "\u{0065 0300}"

          truncated = original.truncate_graphemes(3, separator: separator, omission: '')
          expect(truncated).to eq "\u{0061 0300 0061 0300 0061 0300}"
        end
      end
    end
  end
end