How to strip or slice the bottom items of a SplHeap (also SplMaxHeap, SplMinHeap)?

Question

I have a SplMaxHeap with 30 items. Now I want to strip the last 10 ones. How to do this in a decent and efficient way?

The only solutions I see so far is to copy the first 20 items to a new heap or to convert the heap to an array, use the array_slice and put them also in a new heap.

I even can't see a way to get something out of the heap not from the top.

I guess the reason for this is a limitation of the general design of a heap? I was also thinking to use a DoublyLinkedList, however there I am missing the automatic sorting.

@PaulCrovella I see and understand. I just hope someone finds a clever way to work around this design limitation. If there is no good answer, I know at least that I don't have to try it and that I have to use a custom data structure. — flori, Mar 26 '15 at 10:55

score 1 · Answer 1 · answered Apr 11 '15 at 13:19

One approach is to create a new class derived from an SPL heap, which upon insert only keeps the top (or bottom, depending on the heap) n items.

A super-basic example might look something like:

class Top100 extends SplMinHeap
{
    protected $limit = 100;
    public function insert($value)
    {
        parent::insert($value);
        while (count($this) > $this->limit) {
            $this->extract();
        }
    }
}

This is discussed further in Tobias Schlitt's 2010 article "Heap, heap, hooray!".

This has the problem that now you cannot extract the top anymore, unless you take them all, and reverse them. — JayTaph, Jun 10 '15 at 07:30

score 1 · Answer 2 · answered Jun 10 '15 at 07:35

The problem is that inside a SPLHeap there is no "bottom". Elements are stored in a tree-like structure. The term "bottom" becomes a bit meaningless, however, there is always a "top" of the tree, and this is exactly what SPLHeap can return.

If you have a small number of items, you might get away with alternative data structures (even simple arrays would suffice). Inserting elements might take a bit extra time (this is where trees are much efficient), but maintaining a top and bottom is cheap again (like the SplDoubleLinkedList).

Unless you are storing tens of thousands of elements, you might want to skip the SplHeap for this.

Note: it's trivial enough to create a fixedHeap structure in PHP itself, but i reckon it would not be faster. Benchmark to check what satisfies your needs.

How to strip or slice the bottom items of a SplHeap (also SplMaxHeap, SplMinHeap)?

2 Answers2