autoruby.rb

#!/usr/bin/env ruby
# klg, Dec 2019
require 'MeCab'
require 'erb'

def to_hira(str)
  str.gsub(/[ァ-ヴヽヾ]/) do |x|
    (x.ord - 96).chr x.encoding
  end
end

# Try to match lexeme-internal okurigana
# and return array of [rb, rt] pairs.
def ruby_match(rb, rt)
  bb = rb.scan(/[ぁ-ゔァ-ヴ]+|[^ぁ-ゔァ-ヴ]+/)
  rr = bb.map do |x|
    r = [Regexp.quote(x), Regexp.quote(to_hira(x))]
    r.push('.*?') if x !~ /^[ぁ-ゔァ-ヴ]/
    "(#{r.join '|'})"
  end
  cc = Regexp.new("^#{rr.join}$").match(rt)
  return [[rb, rt]] if cc.nil?
  cc.captures.zip(bb).map do |t, b|
    to_hira(b) == to_hira(t) ? [b] : [b, t]
  end
end

def ruby_html(rbrt)
  if rbrt.length == 1 && rbrt[0].length == 1
    return ERB::Util.html_escape(rbrt[0][0])
  end
  a = rbrt.map do |x|
    rb, rt = x.map(&ERB::Util.method(:html_escape))
    "<rb>#{rb}</rb><rt>#{rt}</rt>"
  end
  "<ruby>#{a.join}</ruby>"
end

@mw = MeCab::Tagger.new('-Owakati')
@my = MeCab::Tagger.new('-Oyomi')

ARGV.each do |arg|
  sw = @mw.parse(arg)
  sy = @my.parse(sw)
  sw = sw.split
  sy = sy.split
  raise unless sw.length == sy.length
  ar = sw.zip(sy).map do |w, y|
    ruby_html(ruby_match(w, to_hira(y)))
  end
  puts(ar.join)
end