an exercise of Text.ParserCombinators.ReadP

csv-parser.hs

-- CSV file parser (as an exercise of Text.ParserCombinators.ReadP)
-- 2013-09-21 @nfunato
import Text.ParserCombinators.ReadP
import Control.Applicative ((<$>), (<*>), (<*), (*>))

-- NOTE:
--   The code here is baesd on d.hatena.ne.jp/kazu-yamamoto/20100104/1262597082
--   which shows code for Parsec2, not ReadP

-- BUGS:
--   Apparently it cannot handle an escaped dquote. Someday it might be fixed :)

-- the top-level API
parseCSV :: ReadS [[String]]
parseCSV = (parse' eatAll) csvFile

-- definitions for parsec2 compatibility
eatAll   = filter (\(a,str)->str=="")
parse' f = (f .) . readP_to_S
parse    = parse' id                -- for interactive tests
noneOf   = satisfy . flip notElem

-- CSV specification (RFC4180)
csvFile  = record `endBy` eol      -- assuming the last eol existence
record   = field `sepBy1` comma
field    = nonEscF +++ escF
nonEscF  = many txData
escF     = dquote *> many quotedCh <* dquote
quotedCh = (dquote>>dquote) <++ comma +++ eol +++ txData
dquote   = char '"'
comma    = char ','
txData   = noneOf ",\"\r\n"
eol      = (cr>>lf) <++ cr <++ lf   -- only crlf in RFC4180, but it's practical
cr       = char '\r'
lf       = char '\n'


{-
--
-- test data (which might include Japanese text in utf-8 encoding at some time)
--

-- Set1 (RWH Ch.16)
""
"hi"
"hi\n"
"line1\nline2\nline3\n"
"cell1,cell2,cell3\n"
"l1c1,l1c2\nl2c1,l2c2\n"
"Hi,\n\n,Hello\n"

-- Set2 (escaping)
"\"cell1\",cell2,\"\"\n"
"\"Product\",\"Price\"\n"
"\"O'Reilly Socks\",10\n"
"\"Shirt with \"\"Haskell\"\" text\",20\n"
"\"Shirt, \"\"O'Reilly\"\" version\",20\n"
"\"Haskell Caps\",15\n"
-}