Run make to compile the program with GHC.
Specify a dictionary file with the -f option; the default is /usr/share/dict/words. All other command-line arguments are strung together to form the input phrase.
The algorithm turns out to be slightly similar to that of the an
program. An attempt as been made to be efficient, by using Haskell’s
unboxed vectors and limiting the contents to 8-bit integers. This
means you can’t have more that 255 of any given character in a
word/anagram string, but this is fairly trivially modifiable in the
source (look for Int8 and replace with Int32 or whatever).
When reading in the dictionary, some pre-processing is done to cut it
down to size. Some of this will be perhaps configurable on the
command-line in the future (all the filters are composed together into
a pipeline; see filterAll in Main.hs).
To do: cut down the dictionary size dynamically during the algorithm.