Change equational constraints on IsString (Parser a) to a ~ ()

[sol]

I think I just got this wrong when I originally implemented this (888b7d7, 0265427).

When you parse a constant string it is not really useful to inspect the parse result. For that reason I think it is justified to change the type from Parser Text to Parser ().

This doesn't make a difference for applicative-style code, but it helps with monadic code, making it -Wall sane, e.g. you can now say

parser = do
  ...
  "foo"
  ...

instead of

parser = do
  ...
  _ <- "foo"
  ...

Technically this is a breaking change, but I have a hard time to imagine non-contrived code that would break from this.

[Anteproperispomenon]

I know this is kinda old now, but I'd recommend against this. While it's very rare to see use of the form txt <- "static text", there are definitely cases of the form txt <- "text1" <|> "text2" <|> "text3". e.g.

parsePhrase :: Parser Text
parsePhrase = do
  txt1 <- "Not" <|> "Already"
  skipSpace
  txt2 <- "enough" <|> "too many"
  skipSpace
  txt3 <- "fish" <|> "batteries"
  void "."
  return (txt1 <> " " <> txt2 <> " " <> txt3 <> ".")

...unless this counts as a contrived example. I suppose a lot of the time, you'd be more likely to use
("foo" $> Foo) <|> ("bar" $> Bar). Still, it kinda feels like taking away functionality for not much gain.

If it's more efficient to return () instead of the text, I guess you could write a function version of that, called stxt (static text) or something short like that. Otherwise, it's not that much of an imposition to write _ <- "abc" or void "abc".

[sol]

there are definitely cases of the form txt <- "text1" <|> "text2" <|> "text3"

Have you written code like this yourself, or where have you seen this?

I suppose a lot of the time, you'd be more likely to use ("foo" $> Foo) <|> ("bar" $> Bar).

Yes, exactly.