TNTSearch is a full-featured full text search (FTS) engine written entirely in PHP. A simple configuration allows you to add an amazing search experience in just minutes. Features include:
- Fuzzy search
- Search as you type
- Geo-search
- Text classification
- Stemming
- Custom tokenizers
- Bm25 ranking algorithm
- Boolean search
- Result highlighting
- Dynamic index updates (no need to reindex each time)
- Easily deployable via Packagist.org
We also created some demo pages that show tolerant retrieval with n-grams in action. The package has a bunch of helper functions like Jaro-Winkler and Cosine similarity for distance calculations. It supports stemming for English, Croatian, Arabic, Italian, Russian, Portuguese and Ukrainian. If the built-in stemmers aren't enough, the engine lets you easily plugin any compatible snowball stemmer. Some forks of the package even support Chinese. And please contribute other languages!
Unlike many other engines, the index can be easily updated without doing a reindex or using deltas.
View online demo | Follow us on Twitter, or Facebook | Visit our sponsors:
- Solving the search problem with Laravel and TNTSearch
- Searching for Bobby Fisher with Laravel 5
- Did you mean functionality with Laravel Scout
The easiest way to install TNTSearch is via composer:
composer require teamtnt/tntsearch
Before you proceed, make sure your server meets the following requirements:
- PHP >= 7.1
- PDO PHP Extension
- SQLite PHP Extension
- mbstring PHP Extension
In order to be able to make full text search queries, you have to create an index.
Usage:
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig([
'driver' => 'mysql',
'host' => 'localhost',
'database' => 'dbname',
'username' => 'user',
'password' => 'pass',
'storage' => '/var/www/tntsearch/examples/',
'stemmer' => \TeamTNT\TNTSearch\Stemmer\PorterStemmer::class//optional
]);
$indexer = $tnt->createIndex('name.index');
$indexer->query('SELECT id, article FROM articles;');
//$indexer->setLanguage('german');
$indexer->run();
Important: "storage" settings marks the folder where all of your indexes will be saved so make sure to have permission to write to this folder otherwise you might expect the following exception thrown:
- [PDOException] SQLSTATE[HY000] [14] unable to open database file *
Note: If your primary key is different than id
set it like:
$indexer->setPrimaryKey('article_id');
By default the primary key is not searchable. If you wanna make it searchable, simply run:
$indexer->includePrimaryKey();
Searching for a phrase or keyword is trivial
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig($config);
$tnt->selectIndex("name.index");
$res = $tnt->search("This is a test search", 12);
print_r($res); //returns an array of 12 document ids that best match your query
// to display the results you need an additional query against your application database
// SELECT * FROM articles WHERE id IN $res ORDER BY FIELD(id, $res);
The ORDER BY FIELD clause is important, otherwise the database engine will not return the results in the required order.
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig($config);
$tnt->selectIndex("name.index");
//this will return all documents that have romeo in it but not juliet
$res = $tnt->searchBoolean("romeo -juliet");
//returns all documents that have romeo or hamlet in it
$res = $tnt->searchBoolean("romeo or hamlet");
//returns all documents that have either romeo AND juliet or prince AND hamlet
$res = $tnt->searchBoolean("(romeo juliet) or (prince hamlet)");
The fuzziness can be tweaked by setting the following member variables:
public $fuzzy_prefix_length = 2;
public $fuzzy_max_expansions = 50;
public $fuzzy_distance = 2; //represents the Levenshtein distance;
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig($config);
$tnt->selectIndex("name.index");
$tnt->fuzziness = true;
//when the fuzziness flag is set to true, the keyword juleit will return
//documents that match the word juliet, the default Levenshtein distance is 2
$res = $tnt->search("juleit");
Once you created an index, you don't need to reindex it each time you make some changes to your document collection. TNTSearch supports dynamic index updates.
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig($config);
$tnt->selectIndex("name.index");
$index = $tnt->getIndex();
//to insert a new document to the index
$index->insert(['id' => '11', 'title' => 'new title', 'article' => 'new article']);
//to update an existing document
$index->update(11, ['id' => '11', 'title' => 'updated title', 'article' => 'updated article']);
//to delete the document from index
$index->delete(12);
First, create your own Tokenizer class. It should extend AbstractTokenizer class, define word split $pattern value and must implement TokenizerInterface:
use TeamTNT\TNTSearch\Support\AbstractTokenizer;
use TeamTNT\TNTSearch\Support\TokenizerInterface;
class SomeTokenizer extends AbstractTokenizer implements TokenizerInterface
{
static protected $pattern = '/[\s,\.]+/';
public function tokenize($text) {
return preg_split($this->getPattern(), strtolower($text), -1, PREG_SPLIT_NO_EMPTY);
}
}
This tokenizer will split words using spaces, commas and periods.
After you have the tokenizer ready, you should pass it to TNTIndexer
via setTokenizer
method.
$someTokenizer = new SomeTokenizer;
$indexer = new TNTIndexer;
$indexer->setTokenizer($someTokenizer);
Another way would be to pass the tokenizer via config:
use TeamTNT\TNTSearch\TNTSearch;
$tnt = new TNTSearch;
$tnt->loadConfig([
'driver' => 'mysql',
'host' => 'localhost',
'database' => 'dbname',
'username' => 'user',
'password' => 'pass',
'storage' => '/var/www/tntsearch/examples/',
'stemmer' => \TeamTNT\TNTSearch\Stemmer\PorterStemmer::class//optional,
'tokenizer' => \TeamTNT\TNTSearch\Support\SomeTokenizer::class
]);
$indexer = $tnt->createIndex('name.index');
$indexer->query('SELECT id, article FROM articles;');
$indexer->run();
$candyShopIndexer = new TNTGeoIndexer;
$candyShopIndexer->loadConfig($config);
$candyShopIndexer->createIndex('candyShops.index');
$candyShopIndexer->query('SELECT id, longitude, latitude FROM candy_shops;');
$candyShopIndexer->run();
$currentLocation = [
'longitude' => 11.576124,
'latitude' => 48.137154
];
$distance = 2; //km
$candyShopIndex = new TNTGeoSearch();
$candyShopIndex->loadConfig($config);
$candyShopIndex->selectIndex('candyShops.index');
$candyShops = $candyShopIndex->findNearest($currentLocation, $distance, 10);
use TeamTNT\TNTSearch\Classifier\TNTClassifier;
$classifier = new TNTClassifier();
$classifier->learn("A great game", "Sports");
$classifier->learn("The election was over", "Not sports");
$classifier->learn("Very clean match", "Sports");
$classifier->learn("A clean but forgettable game", "Sports");
$guess = $classifier->predict("It was a close election");
var_dump($guess['label']); //returns "Not sports"
$classifier->save('sports.cls');
$classifier = new TNTClassifier();
$classifier->load('sports.cls');
You're free to use this package, but if it makes it to your production environment, we would highly appreciate you sending us a PS4 game of your choice. This way you support us to further develop and add new features.
Our address is: TNT Studio, Sv. Mateja 19, 10010 Zagreb, Croatia.
We'll publish all received games here
Support us with a monthly donation and help us continue our activities. [Become a backer]
Become a sponsor and get your logo on our README on Github with a link to your site. [Become a sponsor]
The MIT License (MIT). Please see License File for more information.
From Croatia with ♥ by TNT Studio (@tntstudiohr, blog)