Reduced memory burden by iterating over the normalization test input
file directly (instead of sucking it all into a list of lines first).
2 files changed