clustertools-0.1.1: src/Xcerpt.lhs

Extract sequences matching a list of labels

\begin{code}

module Main where

import Data.Set hiding (null,filter)
import System
import Data.List (foldl')
import IO

usage :: String
usage = "xcerpt <labels> <input>\n\n" ++
        "where <labels> is the name of a file containing the IDs\n" ++
        "of sequences to extract from the <input> file.\n" ++
	"The result ends up in <input>.match and <input>.rest"

main :: IO ()
main =  do
	args <- System.getArgs
	if length args /= 2 then 
	   putStrLn usage
	   else do
	     d <- readFile (args!!0)
	     let dict = mkdict d
	     xcerpt dict (args!!1)

mkdict :: String -> Set String
mkdict = foldl' (flip insert) empty . words

xcerpt :: Set String -> String -> IO ()
xcerpt dict input = do
		    m <- openFile (input++".match") WriteMode
		    r <- openFile (input++".rest") WriteMode
		    i <- readFile input
		    xtr dict m r $ filter (\l->(not.null) l && head l /= '#') 
			    $ lines i

xtr _ m r [] = do
	       hClose m
	       hClose r
xtr _ _ _ [x] = error ("Odd number of lines?\n"++x)
xtr d m r (l1:ls) = if head l1 == '>' then
		       let f = if (drop 1 $ head $ words l1) `member` d
			       then m else r
			   in do
                              let (sequence,rest) = break ((=='>').head) ls
			      hPutStr f $ unlines (l1:sequence)
			      xtr d m r rest
		       else error ("Not a FASTA header:\n"++l1)

\end{code}