It's like heapq (blazingly fast) but object-oriented + more features.

Read more here for the background story.

Less code.

When you need the smallest item of a large list—fast and with no overhead.

Let's suppose you have a heap, you can use pop to get its smallest item.

from xheap import Heap

heap = Heap(['H', 'D', 'B', 'A', 'E', 'C', 'L', 'J', 'I'])
heap.pop()   # returns A
heap.pop()   # returns B
heap.pop()   # returns C
heap.pop()   # returns D

Heapsort works this way.

Indeed and it's as fast as pop. Use push for insertion.

heap = Heap(['A', 'D', 'B', 'H', 'E', 'C', 'L', 'J', 'I'])
heap.push('Z')

Yes, that's what RemovalHeap.remove is supposed to do.

from xheap import RemovalHeap

heap = RemovalHeap(['A', 'D', 'B', 'H', 'E', 'C', 'L', 'J', 'I'])
heap.remove('L')

Just imagine two heaps of the very same set of items but you need different sorting for each heap. That is what OrderHeap is designed for:

from xheap import OrderHeap

items = [date(2016, 1, 1), date(2016, 1, 2),  date(2016, 1, 3),  date(2016, 1, 4)]

day_heap = OrderHeap(items, key=lambda date: date.day)
day_heap.peek()      # returns date(2016, 1, 1)

weekday_heap = OrderHeap(items, key=lambda date: date.weekday())
weekday_heap.peek()  # returns date(2016, 1, 4)

No problem. Use XHeap.

If you wonder why there are 4 distinct heap implementations, it's a matter of speed. Each additional feature slows a heap down. Thus, you could always use XHeap but beware of the slowdown.

A heap is just a list. So, if you tinker with it, you can check whether its invariant still holds:

heap = Heap([4, 3, 7, 6, 1, 2, 9, 8, 5])
heap[3] = 10           # I know what I am doing here
heap.check_invariant() # but better check... ooops

• uses C implementation if available (i.e. fast)
• object-oriented
• no slowdown if you don't need more than a simple heap
• removal possible
• custom orders possible
• works with Python2 and Python3

• no drawbacks discovered so far ;-)
• needs fix/work:
  • item wrapper which allows duplicate items
  • decrease-key+increase-key: another missing use-case of heapq
  • merge heaps
• ideas are welcome :-)