Regex string interpolation breaks unicode properties

String interpolation into regexes appears to break unicode property names.
This script works under MRI 1.9.2.

{{# encoding: UTF-8

1. The following two scenarios work
   puts /[\p {L}]/u
   puts /[#{"
   p{L}

   "}]/u

1. this one raises an exception
2. RegexpError: (RegexpError) invalid character property name {L}: /\p{L}

   /
   b = "
   p

   {L}

   "
   puts /[#

   {b}

   ]/u}}

This is currently preventing datamapper from running on jruby 1.9, because it uses this technique in dm-validations to validate email addresses.