Neo4j PHP Client and Driver

Control to worlds' most powerful graph database

• Pick and choose your drivers with easy configuration
• Intuitive API
• Extensible
• Designed, built and tested under close supervision with the official neo4j driver team
• Validated with testkit
• Fully typed with psalm
• Bolt, HTTP and auto routed drivers available

See the driver in action

An example project exists on the neo4j github. It uses Slim and neo4j-php-client to build an API for the classic movie's example of neo4j.

Start your driving experience in three easy steps

Step 1: install via composer

composer require laudis/neo4j-php-client

Find more details here

Step 2: create a client

use Laudis\Neo4j\Authentication\Authenticate;
use Laudis\Neo4j\ClientBuilder;

$client = ClientBuilder::create()
    ->withDriver('bolt', 'bolt+s://user:password@localhost') // creates a bolt driver
    ->withDriver('https', 'https://test.com', Authenticate::basic('user', 'password')) // creates an http driver
    ->withDriver('neo4j', 'neo4j://neo4j.test.com?database=my-database', Authenticate::kerberos('token')) // creates an auto routed driver
    ->withDefaultDriver('bolt')
    ->build();

You have now created a client with bolt, HTTPS and neo4j drivers. The default driver that the client will use is bolt.

Read more about the URLs and how to use them to configure drivers here.

Step 3: run a transaction

use Laudis\Neo4j\Contracts\TransactionInterface;

$result = $client->writeTransaction(static function (TransactionInterface $tsx) {
    $result = $tsx->run('MERGE (x {y: "z"}:X) return x');
    return $result->first()->get('x')['y'];
});

echo $result; // echos 'z'

Decide how to send your Cypher queries

You can control the driver using three different approaches:

• Transaction functions (recommended and portable)
• Auto committed queries (easiest and most intuitive)
• Unmanaged transactions (for the highest degree of control)

Transaction functions

Transaction functions are the de facto standard when using the driver. It is the most portable as it is resistant to a lot of the pitfalls when first developing with high availability solutions such as Neo4j aura or a cluster.

The driver manages transaction functions:

• It re-executes the function in case of a transient error.
• It commits the transaction on successful execution
• It rolls back the transaction in case of a timeout.
• It routes the execution to a relevant follower or leader server when the neo4j protocol is enabled.

ATTENTION: Because of the automatic retry functionality, the function should produce the same result on subsequent recalls, or in more technical terms: should be idempotent. Always remember this when designing the execution logic within the function.

Some examples:

use Laudis\Neo4j\Contracts\TransactionInterface;

// Do a simple merge and return the result
$result = $client->writeTransaction(static function (TransactionInterface $tsx) {
    $result = $tsx->run('MERGE (x {y: "z"}:X) return x');
    return $result->first()->get('x')['y'];
});

// Will result in an error
$client->readTransaction(static function (TransactionInterface $tsx) {
    $tsx->run('MERGE (x {y: "z"}:X) return x');
});

// This is a poorly designed transaction function
$client->writeTransaction(static function (TransactionInterface $tsx) use ($externalCounter) {
    $externalCounter->incrementNodesCreated();
    $tsx->run('MERGE (x {y: $id}:X) return x', ['id' => Uuid::v4()]);
});

// This achieves the same effect but is safe in case it should be retried. The function is now idempotent.
$id = Uuid::v4();
$client->writeTransaction(static function (TransactionInterface $tsx) use ($id) {
    $tsx->run('MERGE (x {y: $id}:X) return x', ['id' => $id]);
});
$externalCounter->incrementNodesCreated();

Auto committed queries

Auto committed queries are the most straightforward and most intuitive but have many drawbacks when running complex business logic or within a high availability environment.

Run a simple cypher query

$client->run(
    'MERGE (user {email: $email})', //The query is a required parameter
    ['email' => '[email protected]'],  //Requests can be optionally added
    'backup' //The default connection can be overridden
);

Run a statement object:

use Laudis\Neo4j\Databags\Statement;

$statement = new Statement('MERGE (user {email: $email})', ['email' => '[email protected]']);
$client->runStatement($statement, 'default');

Running multiple queries at once

The runStatements method will run all the statements at once. This method is an essential tool to reduce the number of database calls, especially when using the HTTP protocol.

use Laudis\Neo4j\Databags\Statement;

$results = $client->runStatements([
    Statement::create('MATCH (x) RETURN x LIMIT 100'),
    Statement::create('MERGE (x:Person {email: $email})', ['email' => '[email protected]'])
]);

Unmanaged transactions

If you need lower-level access to the drivers' capabilities, then you want unmanaged transactions. They allow for completely controllable commits and rollbacks.

Opening a transaction

The beginTransaction method will start a transaction with the relevant driver.

use Laudis\Neo4j\Databags\Statement;

$tsx = $client->beginTransaction(
    // This is an optional set of statements to execute while opening the transaction
    [Statement::create('MERGE (x:Person({email: $email})', ['email' => '[email protected]'])],
    'backup' // This is the optional connection alias
);

Note that beginTransaction only returns the transaction object, not the results of the provided statements.

Running statements within a transaction

The transaction can run statements just like the client object as long as it is still open.

$result = $tsx->run('MATCH (x) RETURN x LIMIT 100');
$result = $tsx->runStatement(Statement::create('MATCH (x) RETURN x LIMIT 100'));
$results = $tsx->runStatements([Statement::create('MATCH (x) RETURN x LIMIT 100')]);

Finish a transaction

Rollback a transaction:

$tsx->rollback();

Commit a transaction:

$tsx->commit([Statement::create('MATCH (x) RETURN x LIMIT 100')]);

Accessing the results

Results are returned in a standard format of rows and columns:

// Results are a CypherList
$results = $client->run('MATCH (node:Node) RETURN node, node.id AS id');

// A row is a CypherMap
foreach ($results as $result) {
    // Returns a Node
    $node = $result->get('node');

    echo $node->getProperty('id');
    echo $result->get('id');
}

Cypher values and types map to these php types and classes:

Cypher	Php
null	null
string	string
integer	int
float	float
boolean	bool
Map	\Laudis\Neo4j\Types\CypherMap
List	\Laudis\Neo4j\Types\CypherList
Point	\Laudis\Neo4j\Contracts\PointInterface *
Date	\Laudis\Neo4j\Types\Date
Time	\Laudis\Neo4j\Types\Time
LocalTime	\Laudis\Neo4j\Types\LocalTime
DateTime	\Laudis\Neo4j\Types\DateTime
LocalDateTime	\Laudis\Neo4j\Types\LocalDateTime
Duration	\Laudis\Neo4j\Types\Duration
Node	\Laudis\Neo4j\Types\Node
Relationship	\Laudis\Neo4j\Types\Relationship
Path	\Laudis\Neo4j\Types\Path

(*) A point can be one of four types implementing PointInterface: \Laudis\Neo4j\Types\CartesianPoint \Laudis\Neo4j\Types\Cartesian3DPoint \Laudis\Neo4j\Types\WGS84Point \Laudis\Neo4j\Types\WGS843DPoint

Diving Deeper

Differentiating between parameter type

Cypher has lists and maps. This notion can be problematic as the standard php arrays encapsulate both. When you provide an empty array as a parameter, it will be impossible to determine an empty list or map.

The ParameterHelper class is the ideal companion for this:

use Laudis\Neo4j\ParameterHelper;

$client->run('MATCH (x) WHERE x.slug in $listOrMap RETURN x', ['listOrMap' => ParameterHelper::asList([])]); // will return an empty CypherList
$client->run('MATCH (x) WHERE x.slug in $listOrMap RETURN x', ['listOrMap' => ParameterHelper::asMap([])]); // will error
$client->run('MATCH (x) WHERE x.slug in $listOrMap RETURN x', ['listOrMap' => []]); // will return an empty CypherList

Neo4j Version Support

Version	Tested
3.0 +	Yes
4.0 +	Yes

Neo4j Feature Support

Feature	Supported?
Authentication	Yes
Transactions	Yes
Http Protocol	Yes
Bolt Protocol	Yes
Cluster	Yes
Aura	Yes
Jolt Protocol	Roadmap

In-depth requirements

• PHP >= 7.4
• A Neo4j database (minimum version 3.5)
• ext-bcmath *
• ext-json **

(*) Needed to implement the bolt protocol

(**) Needed to implement the http protocol

If you plan on using the HTTP drivers, make sure you have psr-7, psr-17 and psr-18 implementations included into the project. If you don't have any, you can install them via composer:

composer require nyholm/psr7 nyholm/psr7-server kriswallsmith/buzz

Result formats/hydration

In order to make the results of the bolt protocol and the http uniform, the driver provides result formatters (aka hydrators). The client is configurable with these formatters. You can even implement your own.

The default formatter is the \Laudis\Neo4j\Formatters\OGMFormatter, which is explained extensively in the result format section.

The driver provides three formatters by default, which are all found in the Formatter namespace:

• \Laudis\Neo4j\Formatter\BasicFormatter which erases all the Cypher types and simply returns every value in the resulting map as a scalar, null or array value.
• \Laudis\Neo4j\Formatter\OGMFormatter which maps the cypher types to php types as explained here.
• \Laudis\Neo4j\Formatter\SummarizedResultFormatter which decorates any formatter and adds an extensive result summary.

The client builder provides an easy way to change the formatter:

$client = \Laudis\Neo4j\ClientBuilder::create()
    ->withFormatter(\Laudis\Neo4j\Formatter\SummarizedResultFormatter::create())
    ->build();

/**
 * The client will now return a result, decorated with a summary.
 *
 * @var \Laudis\Neo4j\Databags\SummarizedResult $results
 */
$summarisedResult = $client->run('MATCH (x) RETURN x');

// The summary contains extensive information such as counters for changed values in the database,
// information on the database, potential notifications, timing, a (profiled) plan, the type of query
// and information on the server itself.
$summary = $summarisedResult->getSummary();
// The result is exactly the same as the default.
$result = $summarisedResult->getResult();

In order to use a custom formatter, implement the Laudis\Neo4j\Contracts\FormatterInterface and provide it when using the client builder.

Concepts

The driver API described here is the main target of the driver. Because of this, the client is nothing more than a driver manager. The driver creates sessions. A session runs queries through a transaction.

Because of this behaviour, you can access each concept starting from the client like this:

use Laudis\Neo4j\ClientBuilder;

// A builder is responsible for configuring the client on a high level.
$builder = ClientBuilder::create();
// A client manages the drivers as configured by the builder.
$client = $builder->build();
// A driver manages connections and sessions.
$driver = $client->getDriver('default');
// A session manages transactions.
$session = $driver->createSession();
// A transaction is the atomic unit of the driver where are the cypher queries are chained.
$transaction = $session->beginTransaction();
// A transaction runs the actual queries
$transaction->run('MATCH (x) RETURN count(x)');

If you need complete control, you can control each object with custom configuration objects.

Client

A client manages drivers and routes the queries to the correct drivers based on preconfigured aliases.

Driver

The driver object is the thread-safe backbone that gives access to Neo4j. It owns a connection pool and can spawn sessions for carrying out work.

Session

Sessions are lightweight containers for causally chained sequences of transactions. They borrow connections from the connection pool as required and chain transactions using bookmarks.

Transaction

Transactions are atomic units of work that may contain one or more query. Each transaction is bound to a single connection and is represented in the causal chain by a bookmark.

Statement

Queries are executable units within transactions and consist of a Cypher string and a keyed parameter set. Each query outputs a result that may contain zero or more records.

Result

A result contains the output from a query, made up of header metadata, content records and summary metadata. In Neo4j 4.0 and above, applications have control over the flow of result data.

In-depth configuration

Url Schemes

The URL scheme is the easiest way to configure the driver.

Configuration format:

'<scheme>://<user>:<password>@<host>:<port>?database=<database>'

Default configuration:

bolt://localhost:7687?database=neo4j

Scheme configuration matrix

This library supports three drivers: bolt, HTTP and neo4j. The scheme part of the url determines the driver.

driver	scheme	valid certificate	self-signed certificate	function
neo4j	neo4j	neo4j+s	neo4j+ssc	Client side routing over bolt
bolt	bolt	bolt+s	bolt+ssc	Single server over bolt
http	http	https	configured through PSR Client implementation	Single server over HTTP

Configuration objects

A driver, session and transaction can be configured using configuration objects. An overview of the configuration options can be found here:

name	concept	description	class
user agent	driver	The user agent used to identify the client to the neo4j server.	DriverConfiguration
Http PSR Bindings	driver	The relevant PSR implementation used by the driver when using the HTTP protocol.	DriverConfiguration
database	session	The database to connect to.	SessionConfiguration
fetch size	session	The amount of rows to fetch at once. (experimental)	SessionConfiguration
access mode	session	The default mode when accessing the server.	SessionConfiguration
bookmarks	session	The bookmarks used in the session. (experimental)	SessionConfiguration
metadata	transaction	The metadata used during the transaction. (experimental)	TransactionConfiguration
timeout	transaction	The maximum amount of time before timing out.	TransactionConfiguration

Code Example:

use \Laudis\Neo4j\Databags\DriverConfiguration;
use Laudis\Neo4j\Databags\SessionConfiguration;
use Laudis\Neo4j\Databags\TransactionConfiguration;

$client = \Laudis\Neo4j\ClientBuilder::create()
    ->withDefaultDriverConfiguration(DriverConfiguration::default()->withUserAgent('MyApp/1.0.0'))
    ->withDefaultSessionConfiguration(SessionConfiguration::default()->withDatabase('app-database'))
    ->withDefaultTransactionConfiguration(TransactionConfiguration::default()->withTimeout(5.0))
    ->build();

// The client will run the query on a driver with the provided config,
// which spawns a session with the provided session config
// and runs the query in a transaction with the provided transaction config
$client->run('MATCH (x) RETURN count(x) AS count');

// More granular control can be achieved by requesting the concepts yourself:
$tsx = $client->getDriver('default')
    ->createSession(SessionConfiguration::default()->withDatabase('management-database'))
    ->beginTransaction(null, TransactionConfiguration::default()->withTimeout(200));

$tsx->run('SOME REALLY LONG MANAGEMENT QUERY');

$tsx->commit();