sliding-window-max

Given a stream of data, this algorithm returns (for every added value) the current max value.

It uses a strategy that:

• Stores at most a number of values defined by the window size
• Performs a rolling max calculation

Example

const SlidingWindowMax = require('sliding-window-max')

const windowSize = 3

const slidingWindowMax = new SlidingWindowMax(windowSize)
slidingWindowMax.add(0) // returns null
slidingWindowMax.add(5) // returns null
slidingWindowMax.add(10)// returns 10
slidingWindowMax.add(7) // returns 10
slidingWindowMax.add(8) // returns 10
slidingWindowMax.add(5) // returns 8
slidingWindowMax.add(3) // returns 8
slidingWindowMax.add(4) // returns 5
//etc ...

VALUE	Max
[0] 5 10 7 8 5 3 4	null
[0 5] 10 7 8 5 3 4	null
[0 5 10] 7 8 5 3 4	10
0 [5 10 7] 8 5 3 4	10
0 5 [10 7 8] 5 3 4	10
0 5 10 [7 8 5] 3 4	8
0 5 10 7 [8 5 3] 4	8
0 5 10 7 8 [5 3 4]	5
...	...

Reference

new SlidingWindowMax(windowSize, options)

Param	Default	Description
windowSize	required	Defines how many values should be considered to calculate the max
options.comparator	(a, b) => b - a	Override the custom comparator function
options.waitFullRange	true	If false the functions returns the max value also if the window size hasn't been reached yet

add(value)

Evaluate a new value and returns the calculated max value

Credits

I made this algorithm starting from this code https://github.com/chihungyu1116/leetcode-javascript/blob/master/239%20Sliding%20Window%20Maximum.js. My requirements were a bit different. The leetcode algorithm requires that all the data are known before it starts. With sliding-window-max you can:

• evaluate the data as soon as they are produced
• evaluate big array using less memory

Changelog

• 0.x - Beta version
• 1.0 - First official version
• 1.1 - Migrates to gh-workflows; Upgrades deps

Contributors

• chrvadala (author)