sepEndBy' about replace-megaparsec HOT 8 CLOSED

or

sepEndBy' :: m a -> m (Either String a)
sepEndBy'  sep =  sep  <|> (fmap mconcat $ manyTill anyChar sep)

inefficient because sep has to match twice for each pattern, but still maybe something useful here?

from replace-megaparsec.

how about skipManyTill?

https://www.stackage.org/haddock/nightly-2019-09-19/parser-combinators/Control-Monad-Combinators.html#v:skipManyTill

sepEndBy' :: m a -> m (String, a)
sepEndBy' sep = do
  getOffset <- off1
  skipManyTill $ do
    off2 <- getOffset
    x <- sep
    return (stringSectionBetween off1 off2, x)
    // and what about when we come to the end of the stream?

from replace-megaparsec.

https://www.stackage.org/haddock/nightly-2019-09-19/megaparsec-7.0.5/src/Text.Megaparsec.Internal.html#pTakeWhileP

from replace-megaparsec.

See also @chshersh 's issue:

mrkkrp/megaparsec#367

from replace-megaparsec.

So here's the thing. Parsers such as takeWhileP are based on span, which has type

span :: (Char -> Bool) -> Text -> (Text, Text) 

The problem is the predicate only takes one char.

There is also breakOn which is predicated on a constant string.

What we need to write a sepCap which is fast in the sparse case is a function which doesn't seems to exist in the world of Text or ByteString, which is like breakOn, but predicated on the beginning of the second string in the break.

breakPred :: (Text -> Bool) -> Text -> (Text, Text)

Actually, what we want is not a predicate, because we need information about how much of the string to break on. So we need this function.

breakSatisfy :: (Text -> Maybe Int) -> Text -> (Text, Text, Text)

Let's call (Text -> Maybe Int) the "satisfaction" function. breakSatisfy will call the satisfaction function on each tail of the input string until the satisfaction function returns a span. Then breakSatisfy will return the substring up until satisfaction began, the satisfying substring, and the remaining unsatisfying substring.

from replace-megaparsec.

untilCap :: Parser a -> Parser (Text, a)

Parser fails when no match. Prefix or suffix may be zero length. Zero-width match allowed?

streamSplit :: Parser a -> s -> Maybe (Text, a, Text)

Nothing when untilCap fails.

from replace-megaparsec.

@mrkkrp mostly solved this problem this summer in v1.2.0, see manyTill_ and someTill_.

from replace-megaparsec.

Added anyTill for untilCap

Added splitCap for streamSplit

from replace-megaparsec.