dons@cse.unsw.edu.au (Donald Bruce Stewart) writes:
I'd like to say that all I want to do is have the Word8 "bare metal" layer, and a minimal Char8 layer layer on top
This is what's currently implemented.
I've now added a Latin1 module, that works like Char8, but where packing a Char >255 is an error. This means some extra checking, packing 45M characters from [Char] to ByteString slows down from (very rougly) 6.4 to 5.6 Mb/s. Many (but probably less important) operations will be faster, checking if c >= 256 is `elem` a Latin1 ByteString will be O(1) and always False. (Char8 will need to scan the string for c `mod` 256). Feel free to grab, read, criticize, or benchmark, darcs get http://www.ii.uib.no/~ketil/src/fps -k -- If I haven't seen further, it is by standing in the footprints of giants