Skip to content


Subversion checkout URL

You can clone with
Download ZIP
A simple inverted index for javascript.
CoffeeScript JavaScript Shell
Branch: master


A simple inverted index for javascript. An Index is used to store and retrieve objects by one or more of the terms in the object.

Indexing arbitrarilly formatted objects in 3 steps

Use these steps to index an object, an xml document, a web page, or whatever else you can put in an array.

  1. Build a document with DocumentBuilder
  2. Invert the document - Build a term vector with DocumentInverter
  3. Index the object - Add the object with its term vector to the Index.

Building Documents

For our purposes, a document is an object where the key is the field name and the value is a string ready for tokenization and filtering, or a pre-tokenized term vector, like this:

document = {name:"Red delicious", color:["Red"]}

Documents can be built with the DocumentBuilder and inverted (turned into a token vector) with DocumentInverter.


The DocumentBuilder builds a dictionary object of field to value pairs, where the value is a string that is ready to be inverted.

DocumentBuilder Example

# Objects to put in index
apples = [
        variety: "Golden Delicious"
        identified: 1914
        color: "Yellow"
        description: "The Golden Delicious is a cultivar of apple with a yellow color..."
        variety: "Red Delicious"
        identified: 1880
        color: "Red"
        description: "The Red Delicious is a clone of apple cultigen..."

# This converter defines the fields and where to get them from the object.
converter =
  name:  (d) -> d.variety
  body:  (d) -> d.description
  year:  (d) -> d.identified.toString()
  color: (d) -> [d.color] # a vector is treated as pre-tokenized terms

# Builds a document object - a simple dictionary of field=value
# (where value is the string to be inverted).
db = new DocumentBuilder converter
documents = [ a for a in {apples}]


The DocumentInverter takes a document object or string and converts it to a term vector. By default, DocumentInverter will use Filters to normalize terms into lower case and remove duplicate terms.

DocumentInverter Example

docInv = new DocumentInverter new DedupFilter new LowerCaseFilter()
apple = variety: "Red Delicious", identified: 1880, color: "Red"
terms = docInv.invertSync apple
# terms = ["name:red", "name:delicious", "year:1880", "color:Red"]

Indexing an Object

Now that your object has been described with a term vector, it is ready to be added to the index.


An Index is used to store and retrieve objects by one or more of the terms representing the object.

Indexing Example

index = new Index()
apple = variety: "Red Delicious", identified: 1880, color: "Red"
index.addSync apple, ["name:red", "name:delicious", "year:1880", "color:Red"]


Using Filters

Filters transform a term stream to prepare it for indexing. Filters have a .filter method, which accepts and returns an array or array-like object.

Standard Filters

These filters ought to get you started.

DedupFilter - Removes duplicate terms from the term stream

new DedupFilter()
new DedupFilter(subfilter)

LowerCaseFilter - Yields terms converted to lowercase

new LowercaseFilter()
new LowercaseFilter(subfilter)

StopWordFilter - Yields terms that are not in the configurable list of stopwords

new StopWordFilter(stopwordsArray)
new StopWordFilter(stopwordsArray, subfilter)

PrefixFilter - Yields terms prepended with a string

new PrefixFilter(prefix)
new PrefixFilter(prefix, subfilter)

# Example:
new PrefixFilter("tag:").filter(['salad', 'breakfast'])
# yields ['tag:salad', 'tag:breakfast']

Filter Chaining

Most filters can be chained together so that the output of one is the input of the next, thus working inside-out.

For example, this combination converts each term to lower, then removes duplicates:

new DedupFilter(new LowerCaseFilter()).filter(["APPLE","apple", "Orange"])
# yields ["apple", "orange"]


Searching an index with an IndexSearcher and a Query

An IndexSearcher lets you query an index. A query finds all the matches in an index and returns a BitArray representing the matching doctors.

lunchButNotSaladQuery = new Query (index) ->
  hits = index.getIndexesForTermSync 'tag:salad'
  hits = hits.copy()  # don't edit original
  hits.and index.getIndexesForTermSync 'tag:lunch'
  return hits

searcher = new IndexSearcher index
hits = lunchButNotSaladQuery
documents = index.getItemsSync hits
Something went wrong with that request. Please try again.