monkey patch to ruby's JSON that escapes chars bigger than 3 bytes in UTF-8

string_to_json_escape_outside_bmp.rb

require 'active_support'
require 'active_support/concern'
require 'active_support/core_ext'
require 'active_support/json'

# String#to_json to escape any string outside BMP
module StringToJsonWithEscape4byteUtf8
  extend ActiveSupport::Concern

  included do
    alias_method_chain :to_json, :escape_4byte_utf8
  end

  private

  def to_json_with_escape_4byte_utf8(opts = {})
    # State object already set to escape
    return to_json_without_escape_4byte_utf8(opts) if opts.respond_to?(:ascii_only?) && opts.ascii_only?

    # escape all if not UTF-8
    return to_json_with_ascii_only if self.encoding.name != 'UTF-8'

    chars_need_escape = self.chars.map { |c| c.bytesize > 3 }
    if chars_need_escape.none?
      to_json_without_escape_4byte_utf8(opts)
    elsif chars_need_escape.all?
      to_json_with_ascii_only
    else
      # split the mixture and to_json recursively
      groups = []
      prev_needs_escape = nil
      chars_need_escape.each_with_index do |char_needs_escape, idx|
        groups << [] if prev_needs_escape != char_needs_escape
        groups.last << idx
        prev_needs_escape = char_needs_escape
      end
      groups.map! { |indexes| (indexes.first .. indexes.last) }
      groups.each_with_object('') do |idx_range, buff|
        if buff.empty?
          buff << self[idx_range].to_json(opts)
        else
          # join dropping the double quote
          buff.chop!
          buff << self[idx_range].to_json(opts)[1..-1]
        end
      end
    end
  end

  def to_json_with_ascii_only
    JSON.generate([self], ascii_only: true)[1..-2]
  end

end
String.send(:include, StringToJsonWithEscape4byteUtf8)