#Introduction
Extremely fast double-ended queue implementation. Double-ended queue can also be used as a:
The implementation is GC and CPU cache friendly circular buffer. It will run circles around any "linked list" implementation.
Every queue operation is done in constant O(1)
- including random access from .get()
.
#Topics
#Quick start
npm install double-ended-queue
var Deque = require("double-ended-queue");
var deque = new Deque([1,2,3,4]);
deque.shift();
deque.pop();
#Why not use an Array?
Arrays take linear O(N)
time to do shift
and unshift
operations. That means in theory that an array with 1000 items is 1000x slower to do those operations than a deque with 1000 items. 10000x slower with 10000 items and so on.
V8 implements a trick for small arrays where these operations are done in constant time, however even with this trick deque is still 4x faster.
But arrays use "native" methods, they must be faster!
In V8, there is almost no advantage for a method to be a built-in. In fact many times built-ins are at a severe disadvantage of having to implement far more complex semantics than is actually needed in practice. For example, sparse array handling punishes almost every built-in array method even though nobody uses sparse arrays as is evidenced by the popularity of the underscore library which doesn't handle sparse arrays in the same way across different browsers.
#Using double-ended queue as a normal queue
Queue is a more commonly needed data structure however a separate implementation does not provide any advantage in terms of performance. Aliases are provided specifically for the queue use-case. You may use .enqueue(items...)
to enqueue item(s) and .dequeue()
to dequeue an item.
#API
#####new Deque()
-> Deque
Creates an empty double-ended queue with initial capacity of 16. If you know the optimal size before-hand, use new Deque(int capacity)
.
var deque = new Deque();
deque.push(1, 2, 3);
deque.shift();
deque.pop();
#####new Deque(Array items)
-> Deque
Creates a double-ended queue from items
.
var deque = new Deque([1,2,3,4]);
deque.shift();
deque.pop();
#####new Deque(int capacity)
-> Deque
Creates an empty double-ended queue with the given capacity
. Capacity
should be the maximum amount of items the queue will hold at a given time.
The reason to give an initial capacity is to avoid potentially expensive resizing operations at runtime.
var deque = new Deque(100);
deque.push(1, 2, 3);
deque.shift();
deque.pop();
#####push(dynamic items...)
-> int
Push items to the back of this queue. Returns the amount of items currently in the queue after the operation.
var deque = new Deque();
deque.push(1);
deque.pop();
deque.push(1, 2, 3);
deque.shift();
deque.shift();
deque.shift();
Aliases: enqueue
, insertBack
#####unshift(dynamic items...)
-> int
Unshift items to the front of this queue. Returns the amount of items currently in the queue after the operation.
var deque = new Deque([2,3]);
deque.unshift(1);
deque.toString();
deque.unshift(-2, -1, 0);
deque.toString();
Aliases: insertFront
#####pop()
-> dynamic
Pop off the item at the back of this queue.
Note: The item will be removed from the queue. If you simply want to see what's at the back of the queue use peekBack()
or .get(-1)
.
If the queue is empty, undefined
is returned. If you need to differentiate between undefined
values in the queue and pop()
return value -
check the queue .length
before popping.
var deque = new Deque([1,2,3]);
deque.pop();
deque.pop();
deque.pop();
deque.pop();
Aliases: removeBack
#####shift()
-> dynamic
Shifts off the item at the front of this queue.
Note: The item will be removed from the queue. If you simply want to see what's at the front of the queue use peekFront()
or .get(0)
.
If the queue is empty, undefined
is returned. If you need to differentiate between undefined
values in the queue and shift()
return value -
check the queue .length
before shifting.
var deque = new Deque([1,2,3]);
deque.shift();
deque.shift();
deque.shift();
deque.shift();
Aliases: removeFront
, dequeue
#####toArray()
-> Array
Returns the items in the queue as an array. Starting from the item in the front of the queue and ending to the item at the back of the queue.
var deque = new Deque([1,2,3]);
deque.push(4);
deque.unshift(0);
deque.toArray();
Aliases: toJSON
#####peekBack()
-> dynamic
Returns the item that is at the back of this queue without removing it.
If the queue is empty, undefined
is returned.
var deque = new Deque([1,2,3]);
deque.push(4);
deque.peekBack();
#####peekFront()
-> dynamic
Returns the item that is at the front of this queue without removing it.
If the queue is empty, undefined
is returned.
var deque = new Deque([1,2,3]);
deque.push(4);
deque.peekFront();
#####get(int index)
-> dynamic
Returns the item that is at the given index
of this queue without removing it.
The index is zero-based, so .get(0)
will return the item that is at the front, .get(1)
will return
the item that comes after and so on.
The index can be negative to read items at the back of the queue. .get(-1)
returns the item that is at the back of the queue,
.get(-2)
will return the item that comes before and so on.
Returns undefined
if index
is not a valid index into the queue.
var deque = new Deque([1,2,3]);
deque.get(0);
deque.get(1);
deque.get(2);
deque.get(-1);
deque.get(-2);
deque.get(-3);
Note: Even though indexed accessor (e.g. queue[0]
) could appear to return a correct value sometimes, this is completely unreliable. The numeric slots
of the deque object are internally used as an optimization and have no meaningful order or meaning to outside. Always use .get()
.
Note: The implementation has O(1) random access using .get()
.
#####isEmpty()
-> boolean
Return true
if this queue is empty, false
otherwise.
var deque = new Deque();
deque.isEmpty();
deque.push(1);
deque.isEmpty();
#####clear()
-> void
Remove all items from this queue. Does not change the queue's capacity.
var deque = new Deque([1,2,3]);
deque.toString();
deque.clear();
deque.toString();
#Performance
Clone the repo and npm install
. Then run the bench
script.
##1000 items in the queue
double-ended-queue x 15,532,714 ops/sec ±0.19% (96 runs sampled)
built-in array x 6,501,398 ops/sec ±0.87% (95 runs sampled)
node-deque x 2,938,068 ops/sec ±3.50% (68 runs sampled)
##2 million items in the queue
double-ended-queue x 14,425,547 ops/sec ±0.17% (94 runs sampled)
node-deque x 2,815,628 ops/sec ±10.56% (76 runs sampled)
built-in array x 19.23 ops/sec ±0.35% (51 runs sampled)
Noteworthy is just how bad the degradation can be for built-in array when V8 cannot use the trick.