On 07/10/11 15:37, Felipe Almeida Lessa wrote:

You don't need to do it, it is already done =). See the pool package by Michael Snoyman on Hackage [1]. More specifically, see createPoolCheckAlive [2].

Cheers,

[1] http://hackage.haskell.org/package/pool [2] http://hackage.haskell.org/packages/archive/pool/0.1.0.2/doc/html/Data-Pool....

How do people use this stuff? The README is empty, there's no documentation, and none of the functions are commented. I'm forced to conclude that there are people capable of looking at this, createPoolCheckAlive :: MonadControlIO m => IO a -> (a -> IO ()) -> Int -> (Pool a -> m b) -> (a -> IO Bool) -> m b and knowing immediately what it does. I'm still a beginner, and I don't think I have a use for this package (what does it do?), but I've run into similar situations with other packages and would genuinely like to know (as opposed to just complain about it).

On Mon, Jul 11, 2011 at 12:27 PM, Michael Orlitzky wrote:

How do people use this stuff? The README is empty, there's no documentation, and none of the functions are commented. I'm forced to conclude that there are people capable of looking at this,

 createPoolCheckAlive :: MonadControlIO m                          => IO a                          -> (a -> IO ())                          -> Int                          -> (Pool a -> m b)                          -> (a -> IO Bool)                          -> m b

and knowing immediately what it does.

Yes, there's a documentation problem. But you can guess by the types =). IO a: opens a new resource a. a -> IO (): disposes the given resource. Int: maximum number of open resources. Pool a -> m b: the action you want to execute; after it executes, the pool is destroyed. a -> IO Bool: check if the resource is still alive. m b: the result of running your given action with a new pool. The key to understanding all of this is looking backwards from the result. 1) The result is m b. Where did this come from? 2) Well, there's a 'Pool a -> m b'. So it runs the action for me. So this is like a 'withFile :: FilePath -> (Handle -> IO a) -> IO a' function. 3) So if it is a 'Pool a', then the resource is of type 'a'. 4) The only function returning 'a' is 'IO a', so this creates the resource. 5) By the same reason, 'a -> IO ()' disposes it. 6) 'a -> IO Bool' is confusing, but there's a doc for that. 7) 'Int' is also confusing, but since this is a pool, this must be the maximum number of open resources. HTH, -- Felipe.

On Mon, Jul 11, 2011 at 6:27 PM, Michael Orlitzky wrote:

On 07/10/11 15:37, Felipe Almeida Lessa wrote:

...

You don't need to do it, it is already done =).  See the pool package by Michael Snoyman on Hackage [1].  More specifically, see createPoolCheckAlive [2].

Cheers,

[1] http://hackage.haskell.org/package/pool [2] http://hackage.haskell.org/packages/archive/pool/0.1.0.2/doc/html/Data-Pool....

How do people use this stuff? The README is empty, there's no documentation, and none of the functions are commented. I'm forced to conclude that there are people capable of looking at this,

 createPoolCheckAlive :: MonadControlIO m                          => IO a                          -> (a -> IO ())                          -> Int                          -> (Pool a -> m b)                          -> (a -> IO Bool)                          -> m b

and knowing immediately what it does.

I'm still a beginner, and I don't think I have a use for this package (what does it do?), but I've run into similar situations with other packages and would genuinely like to know (as opposed to just complain about it).

Author of the package speaking. I agree, that's a problem :). I've just uploaded a new version[1] that is properly documented, though it will take a bit for Hackage to generates the Haddocks. As Brandon said, the reason this wasn't documented in the first place is that it was an internally used module for Persistent that I decided after the fact to release separately. As for Bryan's resource-pool: currently I would strongly recommend *against* using it for any purpose. It is based on MonadCatchIO-transformers[2], which is a subtly broken package. In particular, when I tried using it for pool/persistent in the first place, I ended up with double-free bugs from SQLite. As an abridged history, I ended up writing a replacement called MonadInvertIO, which later was superceded by MonadPeelIO[3] and MonadControlIO[4]. The latter two packages are both much more correct that MonadCatchIO, and either one should be used in its place. I did email Bryan about this a bit ago, but he didn't get back (he *is* a very busy guy). Ideally, I think that we don't need to have both pool and resource-pool, but such is the situation right now. Michael [1] http://hackage.haskell.org/package/pool-0.1.0.3 [2] http://hackage.haskell.org/package/MonadCatchIO-transformers-0.2.2.2 [3] http://hackage.haskell.org/package/monad-peel [4] http://hackage.haskell.org/package/monad-control-0.2.0.1

On Mon, Jul 11, 2011 at 10:49 AM, Michael Snoyman wrote:

I did email Bryan about this a bit ago, but he didn't get back [...]

Thanks for jogging my memory. I've released an updated version of resource-pool that drops the dependency on that package. http://hackage.haskell.org/package/resource-pool

On Jul 11, 2011, at 10:48 PM, Alistair Bayley wrote:

12 July 2011 05:49, Michael Snoyman wrote:

...

As for Bryan's resource-pool: currently I would strongly recommend *against* using it for any purpose. It is based on MonadCatchIO-transformers[2], which is a subtly broken package. In particular, when I tried using it for pool/persistent in the first place, I ended up with double-free bugs from SQLite.

Do you have a reference explaining this brokenness? e.g. a mailing list message? I wasn't aware of this. Are the other MonadCatchIO-* packages also broken?

http://www.haskell.org/pipermail/haskell-cafe/2010-October/084890.html