“Consumed-OK” and “empty-OK” continuations take 'ParseError' as argument about megaparsec HOT 2 CLOSED

I've rewritten Text.Megaparsec.Prim using special hint-system. Now it should be clear where we have errors and where we have hints (for better "expected" messages when something fails). These concepts are separated and their interaction is clearly defined. Rewritten version passes all tests: all old tests and new tests including those that I haven't pushed into the repository yet.

There shouldn't be any performance degradation, maybe performance will be even better. To find out, we should finish benchmarks and compare Megaparsec against old Parsec.

Now type of ParsecT is:

newtype ParsecT s u m a = ParsecT
  { unParser :: forall b . State s u
             -> (a -> State s u -> Hints -> m b) -- consumed-OK
             -> (ParseError -> m b)              -- consumed-error
             -> (a -> State s u -> Hints -> m b) -- empty-OK
             -> (ParseError -> m b)              -- empty-error
             -> m b }

from megaparsec.

New hint system allows to handle tricky cases intelligently:

λ> parseTest (try (string "let") <|> string "lexical") "le"
parse error at (line 1, column 1):
unexpected end of input
expecting "lexical"
-- ↑ original Parsec
λ> parseTest (try (string "let") <|> string "lexical") "le"
parse error at line 1, column 1:
unexpected "le"
expecting "let" or "lexical"
-- ↑ Megaparsec with hints

There are many more cases when error messages have higher quality compared
to Parsec.

from megaparsec.