An implementation of Coromode and Muthukrishnan's Count-Min sketch data structure for JavaScript. The count-min sketch is basically a high powered generalization of the bloom filter. While a bloom filter gives an efficient way to approximate membership of a set, a count-min sketch can give approximate data about the relative frequency of items in the set.
For more information see the following references:
- Count-Min sketch: https://sites.google.com/site/countminsketch/
- next big thing syndrome: http://lkozma.net/blog/sketching-data-structures/
- G. Cormode, S. Muthukrishnan. "Approximating Data with the Count-Min Data Structure". IEEE Trans. on Software (2012)
//Import library
var createCountMinSketch = require("count-min-sketch")
//Create data structure
var sketch = createCountMinSketch()
//Increment counters
sketch.update("foo", 1)
sketch.update(1515, 104)
//Query results
console.log(sketch.query(1515)) //Prints 104
console.log(sketch.query("foo")) //Prints 1
npm install count-min-sketch
module.exports
is a constructor for the data structure, and you import it like so:
var createCountMinSketch = require("count-min-sketch")
Creates a count-min sketch data structure.
-
epsilon
is the accuracy of the data structure (ie the size of bins that we are computing frequencies of) -
probError
is the probability of incorrectly computing a value -
hashFunc(key, hashes)
is a hash function for the data structure. (optional) the parameters to this function are as follows:key
is the item that is being hashedhashes
is an array ofk
hashes which are required to be pairwise independent.
Returns A count-min sketch data structure
Adds v
to key
key
is the item in the table to increment.v
is the amount to add to it
Returns the frequency of the item key
key
is the item whose frequency we are counting
Returns An estimate of the frequency of key
Returns a serializable JSON representation of the table.
Converts a JSON object into a deserialized sketch. The hash function is reused from the current sketch.
Note In order for this to be successful both the serialized hash table and the current hash table have to have the same hash function.
(c) 2013 Mikola Lysenko. MIT License