Improve Parser Error Message When Using Malformed Function Argument

[tmciver]

As a PS newbie coming from Haskell, I tried to do the following:

> a = map (+1) [1, 2, 3]

not realizing that the correct syntax for the increment function is (_ + 1) . The above produces the error:

Unexpected token '=' at line 1, column 3

It would be nice to improve that error message so that the actual problem is clearer.

[natefaubion]

Note, this is only in the repl. The error from purs compile is:

Error found:
at Test3.purs:3:10 - 3:11 (line 3, column 10 - line 3, column 11)

  Unable to parse module:
  Unexpected token '+'

This is due to the parser backtracking the repl is doing. I did fix an issue with that in the polykinds branch, so I wonder if the error is better on master.

[natefaubion]

For reference, I fixed an issue with how it accumulated errors in a oneOf, which the repl parser uses.
https://github.com/purescript/purescript/blame/323df38b5fffcabd219f72cd8efd706750e2ac90/lib/purescript-cst/src/Language/PureScript/CST/Monad.hs#L151-L153

[natefaubion]

This error is not better in master for the repl.

[huntwj]

I'm guessing this is the same bug I'm running into. You get similarly weird stuff in the REPL when you forget your function arrow in a lambda function:

> foldr (\n acc show n <> acc) "" myNonEmptyData
Unexpected token 'foldr' at line 1, column 1
> foldr (\n acc -> show n <> acc) "" myNonEmptyData
"123"

I'll not create a duplicate issue for now and assume this is the same issue.

[hdgarrood]

I think in general the repl syntax errors are bad because we're assembling the parser out of the parser combinator language provided by the CST. If we write a proper Happy parser for the repl then the syntax error ought to be on par with those from the compiler, I think.