[Haskell-beginners] Lazy vs Strict ponderings...

[Heinrich Apfelmus]

Sean Charles wrote:
> I *think* I understand  about lazy evaluation and its effects on I/O and 
> that it can internally create thunks, promises, continuations etc. 
> whatever you want to call them, and then at some point there will be a 
> sudden spike in memory and CPU loads as something triggers a whole chain 
> of promises to be fulfilled but what I don't have a feel for is just how 
> serious a problem that might be today in say, a simple desktop application.
> 
> [...]
> 
> Is it something one needs to worry about at all or should one just code 
> away and write the application and then worry!?
> 
> I think that having a clearer understanding of what 'types' of problems 
> and their implementations have on CPU/RAM would be a good one to have!

The practitioner's approach to laziness is usually the following 
sequence of approximations, which could be called "the lazy approach to 
laziness":

0) Don't worry about performance. Write programs that are correct 
instead. Laziness helps a lot with writing clean and compositional code. 
Example:

     zipWith (,) [1..] xs

1) Look out for common laziness gotchas. There are essentially only two 
of them:

    i)  Unevaluated expressions that take way more space
        than their evaluated forms. Prototypical example

         foldl (+) [1..n]

    ii) Memory leaks due to sharing. Prototypical example

         average xs = sum xs / length xs

2) Don't worry about memory leaks until they actually appear. If they 
do, use a profiling tool to find out what's going on. Most likely, one 
of the two things above happened in an interesting way.


Note that you might want to be a bit more careful when writing a real 
time system. In that case, would suggest to only use abstractions whose 
performance is easy to reason about and to keep the core code fairly 
small, so that you have a clear picture of what is going on. Other than 
that, there is no need to worry about garbage collection or laziness.


Regards,
Heinrich Apfelmus

--
http://apfelmus.nfshost.com