star-search

star-schema

This package can perform star-schema like operations on a set of JavaScript objects. Star schema optimisation is a data warehousing terminology where a set of measurements are held in a fact table which typically contains a large volume of data. Each of the measurements will have a set of attributes that can be used to filter results and aggregate data. The attributes are held in dimension tables which each have a one-to-many relationship to the fact table. The set of relationship values for any given entry in the fact table is stored as a bitmap index enabling complex searches on combinations of attribute values to be performed with a single scan of the index.

Usage

Dimensions

Define a number of attributes or dimensions. Each attribute value needs to be given a numeric identifier preferably from a sequence

const { Fact, Dimension } = require('star-search')
const colours = new Dimension('id', [
  { id: 1, name: 'Red' },
  { id: 2, name: 'White' },
  { id: 3, name: 'Blue' }
])
const sizes = new Dimension('id', [
  { id: 1, name: 'Small' },
  { id: 2, name: 'Medium' },
  { id: 3, name: 'Large' }
])
const styles = new Dimension('id', [
  { id: 1, name: "Men's" },
  { id: 2, name: "Women's" }
])

Facts

Define a potentially large array for example sales data

const salesArr = [
  { colour: 2, size: 2, style: 1, price: 45 },
  { colour: 1, size: 3, style: 2, price: 43 },
  { colour: 3, size: 2, style: 1, price: 15 },
  { colour: 3, size: 2, style: 1, price: 17 },
  { colour: 3, size: 4, style: 1, price: 19 }
]

Create a fact table using the array mapped to the dimensions

const fact = new Fact(salesArr, {
  colour: colours,
  size: sizes,
  style: styles
})

Searches

Apply searches to the results

const results = fact.search({ colour: [1, 3], style: [1, 2] })
[ { colour: 3, size: 2, style: 1, price: 15 },
  { colour: 3, size: 2, style: 1, price: 17 } ]
const value = results.reduce((a, c) => a + c.price, 0)
...
32

See examples for further information.

Caution Take care when running these algorithms in single-threaded, asynchronous context such as a Node.js web server. In such scenarios it would be preferable to prepare the index on the server and have the browser request the index for repeated searches. The Fact.index getters and setters may be used for this.