This Neo4j PHP Client is a fork of the GraphAware Neo4j PHP Client. Since this package was not maintained since 2017 and was archived to a readonly repository, LongitudeOne decided to fork it and to create a really new package. A simple fork was a problem because issues and pull requests were still centralized on graphaware/neo4j-php-client.
Originally created by GraphAware, Neo4j-PHP-Client is the most advanced and flexible Neo4j Client for PHP. LongitudeOne is now maintaining this project.
Neo4j is a transactional, open-source graph database. A graph database manages data in a connected data structure, capable of representing any kind of data in a very accessible way. Information is stored in nodes and relationships connecting them, both of which can have arbitrary properties. To learn more visit What is a Graph Database?
- Supports multiple connections
- Support for Bolt binary protocol
- Built-in and automatic support for Neo4j Enterprise HA Master-Slave Mode with auto slaves fallback
Version | Tested |
---|---|
3.0 | Yes |
3.5 | Yes |
This project cannot be used with the version 4.0+.
Feature | Supported? |
---|---|
Auth | Yes |
Remote Cypher | Yes |
Transactions | Yes |
High Availability | Yes |
Embedded JVM support | No |
Binary Protocol | Yes |
As PHP7.2 and less are no more maintained, neo4j-php-client is now only tested with PHP7.3+ .
- PHP >= 7.3
- ext-bcmath
- ext-json
- ext-mbstring
- A Neo4j database (minimum version 3.0, maximum version 4.0 excluded)
To get help, you should:
- Ask a question on StackOverflow
- For bugs, please feel free to create a new issue on GitHub
Add the library to your composer dependencies :
composer require "longitude-one/neo4j-php-client:^1.0"
Require the composer autoloader, configure your connection by providing a connection alias and your connection settings :
<?php
require_once 'vendor/autoload.php';
use GraphAware\Neo4j\Client\ClientBuilder;
$client = ClientBuilder::create()
->addConnection('default', 'http://neo4j:password@localhost:7474') // Example for HTTP connection configuration (port is optional)
->addConnection('bolt', 'bolt://neo4j:password@localhost:7687') // Example for BOLT connection configuration (port is optional)
->build();
You're now ready to connect to your database.
NB: The build method will process configuration settings and return you a Client
instance.
$client->run('CREATE (n:Person)');
$client->run('CREATE (n:Person) SET n += {infos}', ['infos' => ['name' => 'Ales', 'age' => 34]]);
$result = $client->run('MATCH (n:Person) RETURN n');
// a result always contains a collection (array) of Record objects
// get all records
$records = $result->getRecords();
// get the first or (if expected only one) the only record
$record = $result->getRecord();
A Record
object contains the values of one record from your Cypher query :
$query = 'MATCH (n:Person)-[:FOLLOWS]->(friend) RETURN n.name, collect(friend) as friends';
$result = $client->run($query);
foreach ($result->getRecords() as $record) {
echo sprintf('Person name is : %s and has %d number of friends', $record->value('name'), count($record->value('friends')));
}
Ideally, you would stack your statements and issue them all at once in order to improve performance.
You can create Cypher statement stacks that act as a Bag and run this stack with the client, example :
$stack = $client->stack();
$stack->push('CREATE (n:Person {uuid: {uuid} })', ['uuid' => '123-fff']);
$stack->push('MATCH (n:Person {uuid: {uuid1} }), (n2:Person {uuid: {uuid2} }) MERGE (n)-[:FOLLOWS]->(n2)', ['uuid1' => '123-fff', 'uuid2' => '456-ddd']);
$results = $client->runStack($stack);
Sometimes, you may want to retrieve a specific result from a Stack, an easy way to do this is to tag your Cypher statements.
The tag is passed via the 3rd argument of the run
or push
methods :
$stack = $client->stack();
$stack->push('CREATE (n:Person {uuid: {uuid} })', ['uuid' => '123-fff'], 'user_create');
$stack->push('MATCH (n:Person {uuid: {uuid1} }), (n2:Person {uuid: {uuid2} }) MERGE (n)-[r:FOLLOWS]->(n2) RETURN id(r) as relId', ['uuid1' => '123-fff', 'uuid2' => '456-ddd'], 'user_follows');
$results = $client->runStack($stack);
$followResult = $results->get('user_follows');
$followRelationshipId = $followResult->getRecord()->value('relId');
The run
method returns you a single Result
object. Other methods where you can expect multiple results returns a ResultCollection
object which is Traversable.
The Result
object contains the records
and the summary
of the statement, the following methods are available in the API :
$result->firstRecord(); // Returns the first record of the Statement Result
$result->records(); // Returns all records
$result->summarize(); // Returns the ResultSummary
The ResultSummary
contains the Statement
, the Statistics and the QueryPlan if available :
$summary = $result->summarize();
$query = $summary->statement()->text();
$stats = $summary->updateStatistics();
$nodesUpdated = $stats->nodesUpdated();
$propertiesSet = $stats->propertiesSet();
// Does the statement affected the graph ?
$affected = $stats->containsUpdates();
Each record contains one row of values returned by the Cypher query :
$query = 'MATCH (n:Person) n, n.name as name, n.age as age';
$result = $client->run($query);
foreach ($result->records() as $record) {
print_r($record->get('n')); // nodes returned are automatically hydrated to Node objects
echo $record->value('name') . PHP_EOL;
echo $record->value('age') . PHP_EOL;
}
The client takes care of the hydration of Graph objects to PHP Objects, so it is for Node, Relationship and Path :
labels()
: returns an array of labels (string)identity()
: returns the internal ID of the nodevalues()
: returns the properties of the node (array)value($key)
: returns the value for the given property keyhasValue($key)
: returns whether or not the nodes has a property with the given keykeys()
: returns you an array representing the keys of the node propertieshasLabel($label)
: returns whether or not the node has the given label (boolean)
type()
: returns the relationship typeidentity()
: returns the internal ID of the relationshipvalues()
: returns the properties of the relationship (array)value($key)
: returns the value for the given property keyhasValue($key)
: returns whether or not the relationship has a property with the given keykeys()
: returns you an array representing the keys of the relationship propertiesstartNodeIdentity
: returns the start node idendNodeIdentity
: returns the end node id
start()
: returns the start node of the pathend()
: returns the end node of the pathlength()
: returns the length of the pathnodes()
: returns all the nodes in the pathrelationships
: returns all the relationships in the path
There are 3 main concepts around this topic :
- a Result
- a Record
- a RecordValue
Let's take a look at a query we do in the browser containing multiple possibilities of types :
MATCH (n:Address)
RETURN n.address as addr, n, collect(id(n)) as ids
LIMIT 5
A Result
is a collection of Record
objects, every row you see in the browser is a Record
and contains Record Value
s.
- In blue the Result
- In orange a Record
- In green a RecordValue
In contrary to the previous versions of the client, there is no more automatic merging of all the records into one big record, so you will need to iterate all the records from the Result
:
$query = 'MATCH (n:Address)
RETURN n.address as addr, n, collect(id(n)) as ids
LIMIT 5';
$result = $client->run($query);
foreach ($result->records() as $record) {
// here we do what we want with one record (one row in the browser result)
print_r($record);
}
Every record contains a collection of Record Value
s, which are identified by a key
, the key is the identifier you give in the RETURN
clause of the Cypher query. In our example, a Record
will contain the following keys :
- addr
- n
- ids
In order to access the value, you make use of the get()
method on the Record
object :
$address = $record->get('addr');
The type of the value is depending of what you return from Neo4j, in our case the following values will be returned :
- a
string
for theaddr
value - a
Node
for then
value - an
array
ofintegers
for theids
value
Meaning that :
$record->get('addr'); // returns a string
$record->get('n'); // returns a Node object
$record->get('ids'); // returns an array
Node
, Relationship
and Path
objects have then further methods, so if you know that the node returned by the identifier n
has a countries
property on it, you can access it like this :
$addressNode = $record->get('n');
$countries = $addressNode->value('countries');
The Record
object contains three methods for IDE friendlyness, namely :
$record->nodeValue('n');
$record->relationshipValue('r');
$record->pathValue('p');
This does not offer something extra, just that the docblocks hint the IDE for autocompletion.
When you use Stack
objects for sending multiple statements at once, it will return you a ResultCollection
object containing a collection of Result
s. So you need to iterate the results before accessing the records.
The Client provides a Transaction object that ease how you would work with transactions.
$tx = $client->transaction();
At this stage, nothing has been sent to the server yet (the statement BEGIN has not been sent), this permits to stack queries or Stack objects before commiting them.
$tx->push('CREATE (n:Person) RETURN id(n)');
Again, until now nothing has been sent.
Sometimes you want to get an immediate result of a statement inside the transaction, this can be done with the run
method :
$result = $tx->run('CREATE (n:Person) SET n.name = {name} RETURN id(n)', ['name' => 'Michal']);
echo $result->getRecord()->value("id(n)");
If the transaction has not yet begun, the BEGIN of the transaction will be done automatically.
$stack = $client->stack();
$stack->push('CREATE (n:Person {uuid: {uuid} })', ['uuid' => '123-fff']);
$stack->push('MATCH (n:Person {uuid: {uuid1} }), (n2:Person {uuid: {uuid2} }) MERGE (n)-[:FOLLOWS]->(n2)', ['uuid1' => '123-fff', 'uuid2' => '456-ddd']);
$tx->pushStack($stack);
// or
$results = $tx->runStack($stack);
if you have queued statements in your transaction (those added with the push
methods) and you have finish your job, you can commit the transaction and receive
the results :
$stack = $client->stack();
$stack->push('CREATE (n:Person {uuid: {uuid} })', ['uuid' => '123-fff']);
$stack->push('MATCH (n:Person {uuid: {uuid1} }), (n2:Person {uuid: {uuid2} }) MERGE (n)-[:FOLLOWS]->(n2)', ['uuid1' => '123-fff', 'uuid2' => '456-ddd']);
$tx->pushStack($stack);
$tx->pushQuery('MATCH (n) RETURN count(n)');
$results = $tx->commit();
After a commit, you will not be able to push
or run
statements in this transaction.
Generally speaking, you would better use HAProxy for running Neo4j in a cluster environment. However sometimes it makes sense to have full control to which instance you send your statements.
Let's assume a environment with 3 neo4j nodes :
$client = ClientBuilder::create()
->addConnection('node1', 'bolt://10.0.0.1')
->addConnection('node2', 'bolt://10.0.0.2')
->addConnection('node3', 'bolt://10.0.0.3')
->setMaster('node1')
->build();
By default, the $client->run()
command will send your Cypher statements to the first registered connection in the list.
You can specify to which connection to send the statement by specifying its alias as 4th argument to the run parameter :
$result = $client->run('CREATE (n) RETURN n', null, null, 'node1');
The client is also aware of the manually configured master connection, so sending your writes can be easier with :
$client->runWrite('CREATE (n:User {login: 123})');
$client->getLabels();
Returns an array of Label
objects.
3 types of events are dispatched during the run
methods :
PreRunEvent
: before the statement or stack is run.PostRunEvent
: after the statement or stack is run.FailureEvent
: in case of failure, you can disable the client to throw an exception with this event.
Example :
$client = ClientBuilder::create()
->addConnection('default', 'bolt://localhost')
->registerEventListener(Neo4jClientEvents::NEO4J_PRE_RUN, array($listener, 'onPreRun')
->build();
The event dispatcher is available via the client with the $client->getEventDispatcher
methods.
You can configure a global timeout for the connections :
$client = ClientBuilder::create()
->addConnection('default', 'http://localhost:7474')
->setDefaultTimeout(3)
->build();
The timeout by default is 5 seconds.
This feature is deprecated and will be removed in version 5. See Http client settings below.
You can enable TLS encryption for the Bolt Protocol by passing a Configuration
instance when building the connection, here
is a simple example :
$config = \GraphAware\Bolt\Configuration::newInstance()
->withCredentials('bolttest', 'L7n7SfTSj')
->withTLSMode(\GraphAware\Bolt\Configuration::TLSMODE_REQUIRED);
$client = ClientBuilder::create()
->addConnection('default', 'bolt://hodccomjfkgdenl.dbs.gdb.com:24786', config)
->build();
We use HTTPlug to give you full control of the HTTP client. Version 4 of the Neo4jClient comes with Guzzle6 by default to preserve backward compatibility. Version 5 will give you the option to choose whatever client you want. Read more about HTTPlug in their documentation.
To configure your client you may add it to Configuration
. Below is an example using php-http/curl-client
.
use Http\Client\Curl\Client;
$options = [
CURLOPT_CONNECTTIMEOUT => 3, // The number of seconds to wait while trying to connect.
CURLOPT_SSL_VERIFYPEER => false // Stop cURL from verifying the peer's certificate
];
$httpClient = new Client(null, null, $options);
$config = \GraphAware\Neo4j\Client\HttpDriver\Configuration::create($httpClient);
$client = ClientBuilder::create()
->addConnection('default', 'http://neo4j:password@localhost:7474', $config)
->build();
The library is released under the MIT License, refer to the LICENSE file bundled with this package.