--- A light implementation of Binary heaps data structure.
-- While running a search, some search algorithms (Astar, Dijkstra, Jump Point Search) have to maintains
-- a list of nodes called __open list__. Retrieve from this list the lowest cost node can be quite slow,
-- as it normally requires to skim through the full set of nodes stored in this list. This becomes a real
-- problem especially when dozens of nodes are being processed (on large maps).
--
-- The current module implements a binary heap
-- data structure, from which the search algorithm will instantiate an open list, and cache the nodes being
-- examined during a search. As such, retrieving the lower-cost node is faster and globally makes the search end
-- up quickly.
--
-- This module is internally used by the library on purpose.
-- It should normally not be used explicitely, yet it remains fully accessible.
--
--[[
Notes:
This lighter implementation of binary heaps, based on :
https://github.com/Yonaba/Binary-Heaps
--]]
if (...) then
-- Dependency
local Utils = require((...):gsub('%.bheap$','.utils'))
-- Local reference
local floor = math.floor
-- Default comparison function
local function f_min(a,b) return a < b end
-- Percolates up
local function percolate_up(heap, index)
if index == 1 then return end
local pIndex
if index <= 1 then return end
if index%2 == 0 then
pIndex = index/2
else pIndex = (index-1)/2
end
if not heap._sort(heap._heap[pIndex], heap._heap[index]) then
heap._heap[pIndex], heap._heap[index] =
heap._heap[index], heap._heap[pIndex]
percolate_up(heap, pIndex)
end
end
-- Percolates down
local function percolate_down(heap,index)
local lfIndex,rtIndex,minIndex
lfIndex = 2*index
rtIndex = lfIndex + 1
if rtIndex > heap._size then
if lfIndex > heap._size then return
else minIndex = lfIndex end
else
if heap._sort(heap._heap[lfIndex],heap._heap[rtIndex]) then
minIndex = lfIndex
else
minIndex = rtIndex
end
end
if not heap._sort(heap._heap[index],heap._heap[minIndex]) then
heap._heap[index],heap._heap[minIndex] = heap._heap[minIndex],heap._heap[index]
percolate_down(heap,minIndex)
end
end
-- Produces a new heap
local function newHeap(template,comp)
return setmetatable({_heap = {},
_sort = comp or f_min, _size = 0},
template)
end
--- The `heap` class.
-- This class is callable.
-- _Therefore,_ heap(...)
_is used to instantiate new heaps_.
-- @type heap
local heap = setmetatable({},
{__call = function(self,...)
return newHeap(self,...)
end})
heap.__index = heap
--- Checks if a `heap` is empty
-- @class function
-- @treturn bool __true__ of no item is queued in the heap, __false__ otherwise
-- @usage
-- if myHeap:empty() then
-- print('Heap is empty!')
-- end
function heap:empty()
return (self._size==0)
end
--- Clears the `heap` (removes all items queued in the heap)
-- @class function
-- @treturn heap self (the calling `heap` itself, can be chained)
-- @usage myHeap:clear()
function heap:clear()
self._heap = {}
self._size = 0
self._sort = self._sort or f_min
return self
end
--- Adds a new item in the `heap`
-- @class function
-- @tparam value item a new value to be queued in the heap
-- @treturn heap self (the calling `heap` itself, can be chained)
-- @usage
-- myHeap:push(1)
-- -- or, with chaining
-- myHeap:push(1):push(2):push(4)
function heap:push(item)
if item then
self._size = self._size + 1
self._heap[self._size] = item
percolate_up(self, self._size)
end
return self
end
--- Pops from the `heap`.
-- Removes and returns the lowest cost item (with respect to the comparison function being used) from the `heap`.
-- @class function
-- @treturn value a value previously pushed into the heap
-- @usage
-- while not myHeap:empty() do
-- local lowestValue = myHeap:pop()
-- ...
-- end
function heap:pop()
local root
if self._size > 0 then
root = self._heap[1]
self._heap[1] = self._heap[self._size]
self._heap[self._size] = nil
self._size = self._size-1
if self._size>1 then
percolate_down(self, 1)
end
end
return root
end
--- Restores the `heap` property.
-- Reorders the `heap` with respect to the comparison function being used.
-- When given argument __item__ (a value existing in the `heap`), will sort from that very item in the `heap`.
-- Otherwise, the whole `heap` will be cheacked.
-- @class function
-- @tparam[opt] value item the modified value
-- @treturn heap self (the calling `heap` itself, can be chained)
-- @usage myHeap:heapify()
function heap:heapify(item)
if self._size == 0 then return end
if item then
local i = Utils.indexOf(self._heap,item)
if i then
percolate_down(self, i)
percolate_up(self, i)
end
return
end
for i = floor(self._size/2),1,-1 do
percolate_down(self,i)
end
return self
end
return heap
end