I am new at Parsec. Will appreciate pointers for the problem here. Say, I have a csv file with fixed number of headers. Instead of parsing each line separately, I will like to look for a token at the beginning of the line, and get all the lines until next line with non-empty token. Example below:
token,flag,values
a,1,
,,a
,,f
b,2,
Rule for a valid input is: if the token field is filled in, get all the lines until next non-empty token field. So, I will like Parsec to get multiple lines below as first input (those multiple lines can then be parsed by another rule):
a,1,
,,a
,,f
Then, the process starts again on the next line with non-empty token field (last line in the example here). What I am trying to figure out is if there a simple way to specify the rule like that in Parsec - get all the lines that meet a certain rule. They could then be handed off to another parser. Basically, it looks like some kind of lookahead rule to specify what is a valid multi-line input. Did I get it right?
We can ignore the comma separator above for now, and just say that an input begins when a character is found at the beginning of a line, and ends when a character is found at the beginning of a line.
I solved the problem with the help of @user2407038 who suggested the basic outline in a comment. Solution and explanation below (please see the comments after the function - they show how the function behaves with the input):
{-# LANGUAGE FlexibleContexts #-}
import Control.Monad
import Text.Parsec
import Control.Applicative hiding ((<|>), many)
-- | this one accepts everything until newline, and discards the newline
-- | This one is used as building block in the functions below
restOfLine :: Stream s m Char => ParsecT s u m [Char]
restOfLine = many1 (satisfy (\x -> not $ x == '\n')) <* char '\n'
-- | a line with token is "many alphanumeric characters" followed by
-- | any characters until newline
tokenLine :: Stream s m Char => ParsecT s u m [Char]
tokenLine = (++) <$> many1 alphaNum <*> restOfLine
-- | ghci test:
-- | *Main Text.Parsec> parseTest tokenLine "a,1,,\n"
-- | "a,1,,"
-- | *Main Text.Parsec> parseTest tokenLine ",1,,\n"
-- | parse error at (line 1, column 1):
-- | unexpected ","
-- |expecting letter or digit
-- | a non-token line is a line that has any number of spaces followed
-- | by ",", then any characters until newline
nonTokenLine :: Stream s m Char => ParsecT s u m [Char]
nonTokenLine = (++) <$> (many space) <*> ((:) <$> char ',' <*> restOfLine)
-- | ghci test:
-- | *Main Text.Parsec> parseTest nonTokenLine ",1,,\n"
-- | ",1,,"
-- | *Main Text.Parsec> parseTest nonTokenLine "a,1,,\n"
-- | parse error at (line 1, column 1):
-- | unexpected "a"
-- | expecting space or ","
-- | One entry is tokenLine followed by any number of nonTokenLine
oneEntry :: Stream s m Char => ParsecT s u m [[Char]]
oneEntry = (:) <$> tokenLine <*> (many nonTokenLine)
-- | ghci test - please note that it drops last line as expected
-- | *Main Text.Parsec> parseTest oneEntry "a,1,,\n,,a\n,,f\nb,2,,\n"
-- | ["a,1,,",",,a",",,f"]
-- | We add 'many' to oneEntry to parse the entire file, and get multiple match entries
multiEntries :: Stream s m Char => ParsecT s u m [[String]]
multiEntries = many oneEntry
-- | ghci test - please note that it gets two entries as expected
-- | *Main Text.Parsec> parseTest multiEntries "a,1,,\n,,a\n,,f\nb,2,,\n"
-- | [["a,1,,",",,a",",,f"],["b,2,,"]]
The parser error seen in the comments is expected on invalid inputs. This can be easily handled. The above code is just a basic building block to get started.