mchains.js

Simple library to generate string markov chains

install

npm i mchains --save

usage

supports UMD

usage with commonjs:

const Chain = require('mchains');

usage in the browser without any loader

<!-- In the html-->
<script src="node_modules/mchains/mchains.js"></script>
<script src="yourCode.js"></script>

//inside your javascript source
const Chain = window.mchains;

documentation

constructor

new Chain(parameters);

Parameters:

• Configuration object(optional)
• Training data(optional)

Configuration object(optional):

order:

Defines how many items are grouped together to form a state, in the case of strings that is either words or characters If you want to use this for word generation a lower order will produce more random-like patterns than higher order values

type:

Determines how the strings are parsed and generated:

• "character" will make the strings be parsed as sequences of characters
• "word" will make the strings be parsed as sequences of words (strings delimited by " ", "\n" and "\r" )

Training data:

Optional training data for the chain to process, more information about training data below.

methods

train()

Takes data to train from it generating new states for the chain.

Valid data:

• A string
• An array of strings
• An object with string values in its properties

Examples:

//option 1:
const training = ' just a simple string';

//option 2:
const training = ['first string', 'second string', ['string inside an array', 'etc...']];

//option 3: 
const training = {
    firstSampleData: ['first string', 'second string', ['string inside an array', 'etc...']],
    secondSampleData: {
        whateverKey: ['first string', 'second string', ['string inside an array', 'etc...']],
        anotherKey: 'just a simple string'
    },
    thirdSampleData: 'just a simple string'
}

doesStateExist(String)

returns true if the string matches one of the chain states, false otherwise

configOutput(config)

Takes a configuration object to configure the output of the generate method

Configuration object properties:

• minLength -> defaults to chain order
• maxLength -> defaults to chain's order * 2
• amount -> defaults to 1
• capitalizeFirst -> defaults to false
• cropToLength -> defaults to false

getNgrams()

returns an array of strings, representing the unique ngrams found in the provided training data

generate(config)

returns an array of strings generated randomly from the internal states of the chain

• The config param matches the structure of the parameter of configOutput method

Properties

Note: public properties that are ment to be readonly, they are initialized with the cosntructor:

Chain.order

returns the order of the chain, that is, how many items are grouped together to form each state

let chain = new Chain();
chain.order 

Chain.type

returns the method used to parse strings, if Chain.type is not String this property will not be used by the algorithm

let chain = new Chain();
chain.type 

live examples

The following page uses this algorithm in order to generate some of the names demo page