[darcs-users] haskell_policy and hlint

[Petr Rockai]

Hi Neil,

Neil Mitchell <ndmitchell at gmail.com> writes:
> Yes, having darcs use HLint would be good - and I'm happy to make
> changes as necessary to accommodate this.
>
> I'm happy to add new darcs specific hints into the main part of hlint,
> and I was also going to add the ability to add new hints to HLint and
> use it like a library. I'd like to generalise as many of the hints as
> possible in to a way where they are useful for most people, but if
> they have to be 100% darcs specific then that's fine too.
Actually, I don't think we need many more rules.

> If you give me a list of examples which you want HLint to pick up on,
> then I'll raise bugs for each of those, and see the best way forward.
>
> For example, with your readFile hint, I think what you actually want
> is a banned function hint - and say Prelude.readFile is disallowed.
> That's perfectly reasonable to add as a new general method.
The current rule files provide everything needed to implement a reasonable
check for that, apart from scoping. Unfortunately, from a quick look at hlint
source, I'm not sure that's easy to solve. An example:

-- this should give a warning: ----
main = readFile "foo" >>= putStr
-----------------------------------

-- this should NOT give a warning: -------
import Prelude hiding( readFile )
import Data.ByteString.Char8( readFile )

main = readFile "foo" >>= putStr
------------------------------------------

I would just write the rule as:

error = Prelude.readFile ==> Data.ByteString.readFile

where Prelude.readFile would match readFile instances that come from Prelude
but not elsewhere.

The other thing is that we'd probably like hlint to return a failure code when
it detects errors. Probably a "when (err > 0) $ exitWith (ExitFailure 1)" near
the end of main would do the trick.

Thanks,
   Petr.

-- 
Petr Ročkai | http://web.mornfall.net
A physicist is an atom's way of knowing about atoms. (George Wald)