[Haskell-cafe] Haskell maximum stack depth

[Neil Mitchell]

Hi Adrian,

> The "bug" is in ghc stack management. Why is it so important that the
> stack size is arbitrarily limited?

It's not, but it makes some things easier and faster. A better
question is why is it important for the stack to grow dynamically. The
answer is that its not.

> It's just an intermediate data
> structure, no different from any other intermediate data structure
> you may build on the heap (well apart from it's efficiency). But I guess
> we would be in danger of having our programs run too fast if folk were
> silly enough to make use of the stack.

In C putting something on the stack is massively more efficient than
putting it on the heap. In Haskell, there is nearly no difference, and
I can imagine some situations where the heap is actually faster. I
guess your comment about speed relates to that assumption?

> So perhaps the current ghc defaults are too generous. What limit do you
> think should be placed on the stack size that a non buggy program can
> use?

The current limits are fine for virtually all cases. They abort on
buggy programs, but its rare that a non-buggy program will need to
change them. i.e. years of experience has ended up with good defaults.

Thanks

Neil