Global compare FCI should be saved and restored (allowing nested comparators)

[rtheunissen]

All structures share the same global FCI, so if while sorting something you try to sort something else, things will break. The best way to handle this is to save the current FCI, use it, then restore it.

[nikic]

Why does this need a global at all?

[rtheunissen]

Because we need to call a user function inside a comparator (C function pointer).

[nikic]

Why can't it be passed along as a context pointer? As you aren't using zend_hash_sort, you shouldn't be limited by a context-free comparator.

[rtheunissen]

I don't believe qsort supports a context parameter.

[nikic]

Huh, you're right. I could have sworn there was a qsort_ex variant, but looks like there isn't :(

[rtheunissen]

There is, but it's not standard. I wonder how plausible it is to implement quick sort manually.

[morrisonlevi]

There are two variants commonly found: qsort_r and qsort_s. Sadly the APIs aren't quite the same even though they are supporting the same idea: add a context parameter.

Perhaps we could use this: https://github.com/noporpoise/sort_r. It is in the public domain, so no issue there.

[rtheunissen]

@morrisonlevi interesting idea, would be very easy to just nab that header file and include it. I like.

[rtheunissen]

Coming back to this 2 years later, do we think that a third party sort is still the way to go?