`local` clears all hints

[Lev135]

I was very surprised, when observed this. Maybe we should add a note for MonadReader instance, if we can't implement this better?

[mrkkrp]

Does this section of the tutorial help? https://markkarpov.com/tutorial/megaparsec.html#the-monadparsec-type-class

[Lev135]

Does this section of the tutorial help?

As far as I understood the section in the tutorial is about the influence of parsing on other effects (like backtracking state). However, here I have the converse problem. I had a simple parser

p = optional (string "a")

I expected, that local (+ 1) p should behave exactly as the p itself (we are changing environment, which is not used). However, it's not true! The call of local clears hints:

run p = case runParserT p "" "b" `runReader` 0 of
  Left peb  -> putStrLn $ errorBundlePretty peb
  Right str -> print str

run (p <* hidden eof)
1:1:
  |
1 | b
  | ^
unexpected 'b'
expecting 'a'

run (local (+ 1) p <* hidden eof)
1:1:
  |
1 | b
  | ^
unexpected 'b'

I found that this is due to the use of calling runParsecT in local, which seems to clear all hints. It's a bit strange that hints can't be preserved when the environment changes. But even if it's so, I think this should be documented in the MonadReader instance for ParsecT

[mrkkrp]

This is the only way to define local for ParsecT. You have to delegate to local on the inner monad, and to do that you have to run the ParsecT monad to get to the inner m monad. You are welcome to try to improve on this.

What happens when you try the other way around, e.g.:

p :: ReaderT Int (Parsec Void String) (Maybe String)
p = optional (chunk "a")

run :: Show a => ReaderT Int (Parsec Void String) a -> IO ()
run x = case runIdentity (runParserT (runReaderT x 0) "" "b") of
  Left peb  -> putStrLn $ errorBundlePretty peb
  Right str -> print str

I think this might allow ParsecT bind without interruption and so hints should work.

[Lev135]

It seems like we can fix this easy just by adding Hints to Result.
I've added this in my fork. local in the simplest test case works as expected. I haven't tested other transformers but I hope they will work fine. I'm not sure about MonadCont and MonadFix instances -- to be honest, I haven't ever used them and don't understand properly, how they should work.

[mrkkrp]

I quite like this idea, actually. Would you like to open a PR with this change?