[Haskell-cafe] Speed of Error handling with Continuations vs. Eithers

[wren ng thornton]

wren ng thornton wrote:
> Here's one big difference:
> 
>>> newtype ErrCPS e m a = ErrCPS { runErrCPS ::
>>>     forall r . (e -> m r) --  error handler
>>>     -> (a -> m r) --  success handler
>>>     -> m r }
> 
> The analogous version I use is:
> 
>     newtype MaybeCPS a = MaybeCPS
>         (forall r. (a -> Maybe r) -> Maybe r)
> 
> While I also offer a transformer version of MaybeCPS, the transformer 
> *does* suffer from significant slowdown. Also, for MaybeCPS it's better 
> to leave the handlers inline in client code rather than to abstract them 
> out; that helps to keep things concrete. So perhaps you should first try 
> a direct CPS translation:
> 
>     newtype ErrCPS e a = ErrCPS
>         (forall r. (a -> Either e r) -> Either e r)
> 
>     runErrCPS :: ErrCPS e a -> Either e a
>     runErrCPS (ErrCPS f) = f return
> 
> I'd be curious if this version suffers the same slowdown.


With this change [1] I can't notice any difference for your 
benchmark[2]. Then again, all the runTest calls take 0 msec and I've had 
no luck making the computation take much time; perhaps your computer can 
detect a difference.

You may want to see what standard benchmarking tools like Microbench[3] 
or the magnificent Criterion[4] have to say. I'd do it myself, but I 
haven't had a chance to reinstall everything since getting my new 
computer (due to the installation issues on newer versions of OSX).


[1] 
http://community.haskell.org/~wren/wren-extras/src/Control/Monad/ErrCPS.hs

[2] 
http://community.haskell.org/~wren/wren-extras/test/Control/Monad/ErrCPS/MaxCantorBenchmark.hs

[3] http://hackage.haskell.org/package/microbench

[4] http://hackage.haskell.org/package/criterion

-- 
Live well,
~wren