local heap = require'heap'

Priority queues implemented as binary heaps. A binary heap is a binary tree that maintains the lowest (or highest) value at the root. The tree is laid as an implicit data structure over an array. Pushing and popping values from the heap is O(log n). Removal is O(n) by default unless a key is reserved on the elements (assuming they’re indexable) to store the element index which makes removal O(log n) too.

API

API Notes:

• values that compare equally are popped in random order.

heap.heap(push, pop, swap, len, cmp) -> push, pop, rebalance

Create a heap API:

push(v) -> i         drop a value into the heap and return its index
pop(i)               remove the value at index i (root is at index 1)
rebalance(i)         rebalance the heap after the value at i has been changed

from a stack API:

push(v)              add a value to the top of the stack
pop()                remove the value at the top of the stack
swap(i, j)           swap two values (indices start at 1)
len() -> n           number of elements in stack
cmp(i, j) -> bool    compare elements

The heap can be a min-heap or max-heap depending on the comparison function. If cmp(i, j) returns a[i] < a[j] then it’s a min-heap. Stack indices are assumed to be consecutive.

heap.cdataheap(h) -> h

Create a cdata heap over table h which must contain:

• ctype: element type (required).
• min_capacity: heap starting capacity (optional, defaults to 0).
• cmp: a comparison function (optional).
• index_key: enables O(1) h:find(v) and thus O(log n) h:remove(v) at the price of setting e[index_key] on all elements of the heap, otherwise h:find(v) is O(n) and h:remove(v) is O(n).
• dynarray: alternative glue.dynarray implementation (optional).

Note: cdata heaps are 1-indexed just like value heaps.

Example:

local h = heap.cdataheap{
   ctype = [[
      struct {
         int priority;
         int order;
      }
   ]],
   cmp = function(a, b)
      if a.priority == b.priority then
         return a.order > b.order
      end
      return a.priority < b.priority
   end}
h:push{priority = 20, order = 1}
h:push{priority = 10, order = 2}
h:push{priority = 10, order = 3}
h:push{priority = 20, order = 4}
assert(h:pop().order == 3)
assert(h:pop().order == 2)
assert(h:pop().order == 4)
assert(h:pop().order == 1)

Note: the order field in this example is used to stabilize the order in which elements with the same priority are popped.

heap.valueheap([h]) -> h

Create a value heap from table h, which can contain:

• cmp: a comparison function (optional).
• index_key: enables O(1) h:find(v) and thus O(log n) h:remove(v) at the price of setting e[index_key] on all elements of the heap, otherwise h:find(v) is O(n) and h:remove(v) is O(n).
• a pre-allocated heap in the array part of the table (optional).

Note: trying to push nil into a value heap raises an error.

Example:

local h = heap.valueheap{cmp = function(a, b)
      return a.priority < b.priority
   end}
h:push{priority = 20, etc = 'bar'}
h:push{priority = 10, etc = 'foo'}
assert(h:pop().priority == 10)
assert(h:pop().priority == 20)

TODO

• heapifying the initial array
• merge(h), meld(h)

Last updated: 3 years ago | Edit on GitHub