Skip to content

Latest commit

 

History

History
321 lines (267 loc) · 10.3 KB

README.md

File metadata and controls

321 lines (267 loc) · 10.3 KB
Raw

Tarantool Mapper

License: MIT Testing Latest Version Total Downloads Telegram

Getting started

Supported versions are php 8+ and tarantool 2+.
The recommended way to install the library is through Composer:

composer require tarantool/mapper

Usually, you manage dependencies in your service provider.
To get started you should create client instance and pass it to mapper constructor.
In this example we use PurePacker and StreamConnection.
To see other implementations please check client documentation

use Tarantool\Client\Client;
use Tarantool\Mapper\Mapper;

$client = Client::fromDefaults();
$mapper = new Mapper($client);

// internaly mapper wraps client with special middleware
assert($mapper->client !== $client);

Schema management

To get started you should describe your spaces, their format and indexes.

$person = $mapper->createSpace('person', [
    'engine' => 'memtx',
    'if_not_exists' => true,
]);

// add properties - name, type and options
$person->addProperty('id', 'unsigned');
$person->addProperty('name', 'string');
$person->addProperty('birthday', 'unsigned');
$person->addProperty('gender', 'string', [
    'default' => 'male'
]);

// indexes are created using fields array and optional index configuration
$person->addIndex(['name']);
$person->addIndex(['birthday'], ['unique' => true]);

// index name is fields based, but you can specify any preffered one
$person->addIndex(['name', 'birthday'], [
    'type' => 'hash',
    'name' => 'name_with_birthday',
]);

/**
 * define format using properties
 */
class Tracker
{
    public int $id;
    public int $reference;
    public string $status;

    public static function initSchema(\Tarantool\Mapper\Space $space)
    {
        $space->addIndex(['reference']);
    }
}

$tracker = $mapper->createSpace('tracker');
$tracker->setClass(Tracker::class);
$tracker->migrate();

/**
 * define format using constructor promotion
 */
class Policy
{
    public function __construct(
        public int $id,
        public string $nick,
        public string $status,
    ) {
    }

    public static function initialize(\Tarantool\Mapper\Space $space)
    {
        $space->addIndex(['nick'], ['unique' => true]);
    }
}

$policy = $mapper->createSpace('policy', ['if_not_exists' => true]);
$policy->setClass(Policy::class, 'initialize'); // use custom initialize method
$policy->migrate();

/**
 * mapper access using class name
 */
class Login
{
    public function __construct(
        public int $id,
        public string $username,
        public string $password,
    ) {
    }

    public static function initSchema(\Tarantool\Mapper\Space $space)
    {
        $space->addIndex(['username'], ['unique' => true]);
    }
}

$nekufa = $mapper->findOrCreate(Login::class, ['username' = 'nekufa']);
$mapper->update(Login::class, $nekufa, ['password' => password_hash("secret")]);

Working with the data

Now you can store and retreive data from tarantool storage using mapper instance.

// get space instance
$persons = $mapper->getSpace('person');

// create new entity
$dmitry = $persons->create([
    'id' => 1,
    'name' => 'Dmitry'
]);

// create entities using mapper wrapper.
// this way entity will be created and saved in the tarantool
$vasily = $mapper->create('person', [
    'id' => 2,
    'name' => 'Vasily'
]);

// retreive entites by id using space
$helloWorld = $mapper->getSpace('post')->findOne(['id' => 3]);

// or using mapper wrapper
$helloWorld = $mapper->findOne('post', ['id' => 3]);

// pass client criteria object as well
$criteria = Criteria::index('age')->andKey([18])->andGeIterator();
$adults = $mapper->find('user', $criteria);

// updates are easy
$posts = $mapper->getSpace('post');
$helloWorld = $posts->update($helloWorld, [
    'title' => 'Hello world'
]);

// if you use instance classes, instance would be updated
$policy = $mapper->findOrFail('policy', ['id' => 3]);
$policy = $mapper->get('policy', 3); // getter shortcut
$mapper->update('policy', $policy, [
    'title' => 'updated title',
]);
echo $policy->title; // updated title

// use client operations as well
use Tarantool\Client\Schema\Operations;
$mapper->getSpace('policy')->update($policy, Operations::add('counter', 1));
var_dump($policy->counter); // actual value

Schema Cache

Any new mapper instance will fetch schema from the tarantool, this requests can takes a bit of database load.
Use your favorite psr/cache implementation to persist schema on the application side.
For example, we use apcu adapter from symfony/cache package.
If new schema version is not persisted in cache, mapper will fetch it

use Symfony\Component\Cache\Adapter\ApcuAdapter;
$cache = new ApcuAdapter();

$mapper = new Mapper(Client::fromDefaults());
$mapper->cache = $cache;
$mapper->getSpace('_vspace'); // schema is fetched now

$mapper = new Mapper(Client::fromDefaults());
$mapper->cache = $cache;
$mapper->getSpace('_vspace'); // no new requests are made

Query Cache

If you don't want to repeat select queries you can inject cache interface implementation to space.
Use your favorite psr/cache implementation to persist schema on the application side.
For example, we use array adapter from symfony/cache package.

use Symfony\Component\Cache\Adapter\ArrayAdapter;

$mapper = new Mapper(Client::fromDefaults());
$mapper->getSpace('_vspace')->cache = new ArrayAdapter(); // feel free to set default ttl

$mapper->find('_vspace'); // query is executed
$mapper->find('_vspace'); // results are fetched from cache
$mapper->find('_vspace'); // results are fetched from cache

Changes registration

In some cases you want to get all changes that were made during current session.
By default spy configuration is set to false, this improves performance a bit.

$mapper->spy = true;

$nekufa = $mapper->create('user', ['login' => 'nekufa']);
$firstPost = $mapper->create('post', [
    'user_id' => $nekufa->id,
    'title' => 'hello world',
]);
$mapper->update('post', $firstPost, ['title' => 'Final title']);

// now there are two changes
[$first, $second] = $mapper->getChanges();
echo $first->type; // insert
echo $first->space; // user
echo $first->data; // ['login' => 'nekufa']

// all changes would be merged by space and key
// this reduces changes duplicates
echo $second->type; // insert
echo $second->space; // post
echo $second->data; // ['user_id' => 1, 'title' => 'Final title']

// of course you can flush all changes and start registration from scratch
$mapper->flushChanges();

Multiple connections

If you split your data across multiple tarantool instances you can use prefix based data api.
Api is the same but you prefix space name with a connection prefix.

$pool = new Pool(function (string $prefix) {
    return new Mapper(Client::fromDsn('tcp://' . $prefix));
});

// connect to tarantool instance `volume` and find all timelines.
$trackers = $pool->findOne('volume.timeline');

$nekufa = $pool->findOrCreate('guard.login', ['username' => 'nekufa']);
$pool->update('guard.login', $nekufa, ['locked_at' => time()]);

// pool also wraps changes with the prefixes
echo $pool->getChanges()[0]->space; // guard.login

// all expressions do the same behind the scenes
$pool->find('flow.tracker', ['status' => 'active']);
$pool->getMapper('flow')->find('tracker', ['status' => 'active']);
$pool->getMapper('flow')->getSpace('tracker')->find(['status' => 'active']);

Lua code delivery

Iproto usage is very powerful but sometimes is not enough.
You can easily execute lua code and pass local variables using associative array.

In addition, if you don't want to deliver it every request, use magic call method.
When you use call method, mapper generates unique function name and creates it if it's not exist.

// this method will always deliver and parse lua code on the tarantool side
$mapper->evaluate('return a + b', ['a' => 2, 'b' => 7]); // 9

// first call a function would be created with name evaluate_{BODYHASH}
// there would be two requests - create function and call it
$mapper->call('return a + b', ['a' => 2, 'b' => 7]); // 9

// second call will produce single request with function name and arguments
$mapper->call('return a + b', ['a' => 2, 'b' => 7]); // 9

Migrations

Use basic migration class to implement some logic before or after schema creation.
Pass migrations to mapper migrate method and that's all.


use Tarantool\Mapper\Migration;
use Tarantool\Mapper\Space;

class DropLegacySpaces extends Migration
{
    public function beforeSchema(Mapper $mapper)
    {
        $mapper->call(<<<LUA
            if box.space.legacy then
                box.space.legacy:drop()
                box.space.legacy_detail:drop()
            end
        LUA);
    }
}
class InitializeData extends Migration
{
    public function afterSchema(Mapper $mapper)
}
$mapper = $container->get(Mapper::class);

// also migrate accepts migration instance, or migration class arrays
$mapper->migrate(DropLegacySpaces::class, InitializeData::class);

Performance

We can calculate mapper overhead using getInstance method that is called per each instance.
If you don't use cache, there is single schema fetch on connection and each time schema is upgraded.
Perfomance test was made on (AMD Ryzen 5 3600X), Ubuntu 23.10 using PHP 8.3.6

Instance type Instances per second
constructor 4 664 172
properties 4 328 442
simple array 11 983 040