My experience making an on-disk merge sort using ranges

[deadalnix]

On Wednesday, 27 February 2013 at 07:32:58 UTC, Jonathan M Davis 
wrote:
> In a way, all of that power and flexibility is exactly the 
> problem. When
> dealing with generic code like Phobos does, it's fairly 
> difficult to test it
> thoroughly enough to make sure that it works with all of the 
> various inputs
> that it can be given - especially when you start combining all 
> kinds of stuff.
> Both the code and the tests are continually improving, but for 
> a lot of it, if
> you poke it hard enough, you'll be able to find corner cases 
> that don't work
> quite right yet. But the more that gets tried, and the more 
> problems that get
> found,  the more of them that will get fixed, and the less 
> frequently you'll
> run into bugs when dealing with std.algorithm and its ilk.
>

I think this is mostly due to things having corners cases, or not 
being defined.

For instance, it is unknown what passing a range by value does 
(and it in fact does different things on different ranges). The 
same goes for many many others things.