Parser::Builder::Default.check_lvar_name patch

[rubys]

Test case:

require 'parser/ruby30'

if RUBY_ENGINE == 'opal'
  require 'corelib/string/unpack'

  # https://github.com/opal/opal/blob/master/lib/opal/parser/patch.rb
  class Parser::Source::Buffer
    def source_lines
      @lines ||= begin
        lines = @source.lines.to_a
        lines << '' if @source.end_with?("\n")
        lines.map { |line| line.chomp("\n") }
      end
    end
  end

  # https://github.com/whitequark/parser/issues/784
  module Parser
    class Diagnostic
      undef :render_line

      def render_line(range, ellipsis=false, range_end=false)
        source_line    = range.source_line
        highlight_line = [' '] * source_line.length

        @highlights.each do |highlight|
          line_range = range.source_buffer.line_range(range.line)
          if highlight = highlight.intersect(line_range)
            highlight_line[highlight.column_range] = ['~'] * highlight.size
          end
        end

        if range.is?("\n")
          highlight_line << "^"
        else
          if !range_end && range.size >= 1
            highlight_line[range.column_range] = ['^'] + ['~'] * (range.size - 1)
          else
            highlight_line[range.column_range] = ['~'] * range.size
          end
        end

        highlight_line += %w(. . .) if ellipsis
        highlight_line = highlight_line.join

        [source_line, highlight_line].
          map { |line| "#{range.source_buffer.name}:#{range.line}: #{line}" }
      end
    end
  end
end

p Parser::Ruby30.parse('1 => a')

The issue is a regular expression using POSIX bracket expressions: https://github.com/whitequark/parser/blob/21684f12eb73ac416fa88ecee7198739a1847596/lib/parser/builders/default.rb#L1726-L1732

Patch:

  module Parser
    class Builders::Default
      def check_lvar_name(name, loc)
        if name =~ /^[_a-z][_\w]*$/
          # OK
        else
          diagnostic :error, :lvar_name, { name: name }, loc
        end
      end
    end
  end

[rubys]

An alternative regex:

/^\p{Ll}(\p{L}|\p{N}|_)*$/u

Seems to be nearly universally supported these days

[hmdne]

Proposed here: https://github.com/opal/opal/pull/2243/files#diff-5e29ec5ef4397f3b4d0a4d402fa6b59b22a295cafabd9e22aefb8c95fa92c2a5R42-R50