I was bored, so I wrote this as a solution. I wrote the replace
string function using attoparsec. Uses bytestrings throughout so that
it will be fast.
I intially tried regexes and soon abandoned that. Then I tried
parsec, but parsec doesn't do bytestrings very well, so I went to
attoparsec and that was the best code.
Requires csv-bytestring and attoparsec libraries.
import Data.ByteString as B
import Data.ByteString.Internal (c2w)
import Text.CSV.ByteString
import Data.Attoparsec.Char8 as AC
import Control.Applicative ((<|>))
import Control.Monad (liftM2)
loadcsv = do
temp <- B.readFile "blah.temp"
fmap parseCSV (B.readFile "blah.csv") >>= parse' temp
where
parse' _ Nothing = return ()
parse' temp (Just csv) = mapM_ (B.putStrLn . replaceInStr temp) csv
replaceInStr :: ByteString -> [ByteString] -> ByteString
replaceInStr str xs = case feed (parse replacePat str) B.empty of
Done _ r -> r
otherwise -> error "Something wrong with my parser."
where
replacePat = liftM2 B.append (fmap B.concat (many replace)) rest
replace = do
beg <- takeTill (== '@')
char '@'
i <- decimal
return $ beg `B.append` (xs !! i)
rest = do
x <- takeTill (isEndOfLine . c2w)
endOfLine
return x
On Wed, Mar 16, 2011 at 8:00 PM, Sean Charles
I tried in vain using Text.CSV to write a small utility for myself that would *simply* map a CSV file across a template, specifically for mass-producing Drupal nodes from a CSV file... Here's what the Haskell code has to compete with :-
TESTDATA.CSV eric,42,"Hacker" bert,20,"Janitor" harry,15,"Web nerd"
TEST.SQL UPDATE foo SET rating = @1, role = '@2' where name = '@0';
And the PHP code that is used to produce ANY output from the template and CSV data (forget injection attacks and all that for now, this is proof-of-concept remember)...
<?php $template = file_get_contents($argv[1]); if ($fh=fopen($argv[2],"r")) { while($line = fgetcsv($fh)) { foreach ($line as $k => $v ) $line['@'.$k] = $v; echo strtr($template,$line)."\n"; } fclose($fh); } ?>
Note that I have had to remap the index keys to @n in order not to accidentally hit any numerical data that may be in the template. This is my test, the *real* template creates a Drupal node dynamically and has lots of [0]['value'] bits in it so I had to do this, I will allow any shortcuts to be used though as I would have liked to have coded the utility in Haskell to get better at it but tonight it defeats me!
Here then is my unsatisfactory code and no, it doesn't compile or work very well as I am still trying hard with it! I want to learn!
Still my biggest headache is type matching and deciphering the sometimes complete gibberish output from the type inference system when I screw up... I am finding it very hard to work out what I did wrong at times. Sitting here with Real World Haskell is proving fruitless tonight. :(
Here's the rubbish code...
import Text.CSV import System.Environment import System.IO import Text.ParserCombinators.Parsec.Error
main = do args <- getArgs case length args of 2 -> do csvdata <- parseCSVFromFile (last args) either (\error -> print error) (\csvdata -> processData (head args) csvdata) csvdata _ -> print "Usage: csvsql template datafile.csv (options later!)"
{- csvdata is [CSV] a CSV is [Record] a Record is [Field] a Field is String -} processData :: String -> CSV -> IO () processData filename csvdata = do tin <- openFile filename ReadMode template <- hGetContents tin processRow template csvdata hClose tin return ()
processRow :: String -> [Record] -> String processRow template row = "eric" ++ template
_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners