Search

Hex.pm VersionHex DocsHex.pm License

⚡ Fast full-text search for Elixir

This library provides simple, fast, in-memory full-text search functionality for Elixir applications.

Features

Installation

The package can be installed by adding search to your list of dependencies in mix.exs:

def deps do
  [
    {:search, "~> 0.3"}
  ]
end

Usage

Creating an Index

To create a new index, use the new/1 function with a list of fields to be indexed:

index = Search.new(fields: [:title, :content])

Adding Documents

To add a document to the index, use the add/2 function with the index and the document:

document = %{id: 1, title: "Elixir", content: "Elixir is a dynamic, functional language."}
index = Search.add!(index, document)

You can also add multiple documents at once:

documents = [
  %{id: 2, title: "Phoenix", content: "Phoenix is a web framework for Elixir."},
  %{id: 3, title: "Nerves", content: "Nerves is a framework for embedded systems."}
]
index = Search.add!(index, documents)

Removing Documents

To remove a document from the index, use the remove/2 function with the index and the document:

index = Search.remove!(index, document)

You can also remove multiple documents at once:

index = Search.remove!(index, documents)

Searching

To search the index, use the search/3 function with the index and the query string:

Search.search(index, "web famewrk", prefix?: true, fuzzy?: true)
[
  %{
    id: 2,
    matches: %{"framework" => [:content], "web" => [:content]},
    fields: %{},
    score: 1.6965399945163802,
    terms: ["web", "framework"]
  },
  %{
    id: 3,
    matches: %{"framework" => [:content]},
    fields: %{},
    score: 0.24367025800793077,
    terms: ["framework"]
  }
]

Internals

The library uses a Radix tree for efficient indexing and retrieval of terms. It also implements the BM25 algorithm for relevance scoring and the Levenstein distance algorithm for calculating edit distances.

License

Search is released under the Apache License 2.0.

About

This package was written by Elliot Jackson.