[Haskell-cafe] Takusen and strictness

[Bryan O'Sullivan]

Paul Moore wrote:

> ... ie, there's deep dark magic involved in the seemingly simple
> getContents, which isn't easily available to mere mortals (or even
> semi-immortal library designers).

That's really not true.  getContents looks simple from the outside, and 
it *can* be simple underneath, too.

You can write a getContents on an arbitrary source of data with just one 
built-in action and two other primitives.  The built-in is 
unsafeInterleaveIO, and the primitives from your data source are "get me 
the next item" and "am I done yet?".

getContents :: MyDataItem a => DataSource a -> IO [a]

getContents myDataSource = unsafeInterleaveIO $ do
   empty <- amIDoneYet myDataSource
   if empty
     then return []
     else do x <- getMeTheNextItem myDataSource
             xs <- getContents myDataSource
             return (x:xs)

If you're providing this in a library, you have to consider what you 
should do for your consumer if amIDoneYet or getMeTheNextItem does "the 
wrong thing", but that's hardly unusual.

You can make the implementation complex, so that it prefetches data, 
handles exceptions, and whatnot, but the basic idea isn't too terribly 
scary.  You'll see this pattern in a number of Haskell libraries 
(System.IO, Data.ByteString.Lazy, Control.Concurrent.Chan, and more).

	<b