Tantivy-cli is command line interface for tantivy search engine.

Tutorial: Indexing Wikipedia with Tantivy CLI

Introduction

In this tutorial, we will create a brand new index with the articles of English wikipedia in it.

Install

There are two ways to get tantivy. If you are a rust programmer, you probably have cargo installed and you can just run cargo install tantivy-cli.

Alternatively, if you are on Linux 64bits, you can directly try and download a static binary: binaries/linux_x86_64/

Creating the index: new

Let's create a directory in which your index will be stored.

    # create the directory
    mkdir wikipedia-index

We will now initialize the index and create its schema. The schema defines the list of your fields, and for each field :

• its name
• its type, currently u32 or str
• how it should be indexed.

You can find more information about the latter on [tantivy's schema documentation page](http://fulmicoton.com/tantivy/tantivy/schema/index.html

In our case, our documents will contain

• a title
• a body
• a url

We want the title and the body to be tokenized and index. We want to also add the term frequency and term positions to our index. (To be honest, phrase queries are not yet implemented in tantivy, so the positions won't be really useful in this tutorial.)

Running tantivy new will start a wizard that will help you go through the definition of the schema of our new index.

Like all the other commands of tantivy, you will have to pass it your index directory via the -i or --index parameter as follows.

    tantivy new -i wikipedia-index

When asked answer to the question, answer as follows:

    Creating new index 
    Let's define it's schema! 



    New field name  ? title
    Text or unsigned 32-bit Integer (T/I) ? T
    Should the field be stored (Y/N) ? Y
    Should the field be indexed (Y/N) ? Y
    Should the field be tokenized (Y/N) ? Y
    Should the term frequencies (per doc) be in the index (Y/N) ? Y
    Should the term positions (per doc) be in the index (Y/N) ? Y
    Add another field (Y/N) ? Y



    New field name  ? body
    Text or unsigned 32-bit Integer (T/I) ? T
    Should the field be stored (Y/N) ? Y
    Should the field be indexed (Y/N) ? Y
    Should the field be tokenized (Y/N) ? Y
    Should the term frequencies (per doc) be in the index (Y/N) ? Y
    Should the term positions (per doc) be in the index (Y/N) ? Y
    Add another field (Y/N) ? Y



    New field name  ? url
    Text or unsigned 32-bit Integer (T/I) ? T
    Should the field be stored (Y/N) ? Y
    Should the field be indexed (Y/N) ? N
    Add another field (Y/N) ? N

    [
    {
        "name": "title",
        "type": "text",
        "options": {
            "indexing": "position",
            "stored": true
        }
    },
    {
        "name": "body",
        "type": "text",
        "options": {
            "indexing": "position",
            "stored": true
        }
    },
    {
        "name": "url",
        "type": "text",
        "options": {
            "indexing": "unindexed",
            "stored": true
        }
    }
    ]

After the wizard has finished, a meta.json has been written in wikipedia-index/meta.json. It is a fairly human readable JSON, so you may check its content.

It contains two sections :

• segments (currently empty, but we will change that soon)
• schema

Indexing the document : index

Tantivy's index command offers a way to index a json file. More accurately, the file must contain one document per line, in a json format. The structure of this JSON object must match that of our schema definition.

    {"body": "some text", "title": "some title", "url": "http://somedomain.com"}

For this tutorial, you can download a corpus with the 5 millions+ English articles of wikipedia formatted in the right format here : wiki-articles.json (2.34 GB). Make sure to uncompress the file

    bunzip2 wiki-articles.json.bz2

If you are in a rush you can download 100 articles in the right format here.

The index command will index your document. By default it will use as many threads as there are cores on your machine. You can change the number of threads by passing it the -t parameter.

On my computer (8 core Xeon(R) CPU X3450 @ 2.67GHz), it will take around 6 minutes.

    cat wiki-articles.json | tantivy index -i ./wikipedia-index

While it is indexing, you can peek at the index directory to check what is happening.

    ls ./wikipedia-index

If you indexed the 5 millions articles, you should see a lot of new files, all with the following format The main file is meta.json.

Our index is in fact divided in segments. Each segment acts as an individual smaller index. Its named is simply a uuid.

Serve the search index

Tantivy's cli also embeds a search server. You can run it with the following command.

    tantivy serve -i wikipedia-index

By default, the server is serving on the port 3000.

You can search for the top 20 most relevant documents for the query Barack Obama by accessing the following url in your browser

http://localhost:3000/api/?q=barack+obama&explain=true&nhits=20