$client = new Client();
Should create its own Transport, using default settings.

Likewise:
$client = new Client('somehost', 7475);
Should create its own Transport, using the supplied host/port

Standardize use of namespacing throughout the wiki. If class aliases continue to be used, make mention of the namespace on the "getting started" page and anywhere aliased classes are used.

Setting a property on an entity to null should have the same the same effect as removeProperty : https://github.com/neo4j/community/issues/115

Still return null from the client when getting the property, but make sure the property is removed from the entity persistence.

I think it would be good if it is able to check if the server is up before creating the client.. I did some modifications on my own..

Placed in transport

/**
*Check if server is online
*/
public function isOnline()
{
$holder = $this->makeRequest(self::GET, $this->path);
return $holder['code'];
}

Placed in Client

public function __construct(Transport $transport)
{
if($transport->isOnline())
$this->setTransport($transport);
else
exit("Neo4j server is offline");
}

this would ensure that the server is online before allowing anything else to take place, unfortunately its now relying on the error code 0 or 404. i am not very sure of the rest interface to get a proper one. Perhaps u could tell me O o

http://docs.neo4j.org/chunked/milestone/cypher-plugin.html#rest-api-return-paths

Due to how Neo4J handles batch processing the escape sequence needed for '' is awkward. Currently, any batch commit which contains a property value that contains a '' fails with an invalid argument exception. Removing the '' from the value seems to make it work ok but is a terrible solution. The only solution I've been able to find so far is:

In Everyman\Neo4j\Transport.php

$encoded = json_encode($data);
$encoded = str_replace('\\\\', '\\\\\\\\', $encoded);
return $encoded;

The downfall is this issue only happens during a batch commit, so the encoding fix only needs to occur on batch related data. We currently only commit in batches so we are good to go, but a better work around should probably be implemented.

Otherwise, fantastic library.

This should already be the case, but test and make sure.

I installed neo4jphp on my local machine (Mac OS X 10.6 with MAMP) by adding

require("phar://neo4jphp.phar");

as described. Creating nodes and relationships works like a charme. But I am unable to create Traversals. I always run into the following error:

include(Traversal.php) [function.include]: failed to open stream: No such file or directory. I tried to include the file separately, but it does not help.

Any ideas?

404 style errors should be handled by returning null or false where appropriate.

Other errors should be handled by throwing exceptions.

Also, don't count 0 as a valid response code; should translate to 500 unknown error

Ability to create full-text indexes. Are there other types? Spatial?

The call to Client::getId() will return a Node object if the Node was found successfully or null if not.

should be changed to

The call to Client::getNode() will return a Node object if the Node was found successfully or null if not.

Create a client method/command to retrieve the server info and version information.

http://docs.neo4j.org/chunked/snapshot/rest-api-service-root.html

Allow users to configure a transport to use https and provide a username/password for http basic authentication.

http://docs.neo4j.org/chunked/stable/rest-api-indexes.html#rest-api-list-node-indexes

Check if this is possible for relationships as well.

Hi Josh,

I'm currently trying your batch branch and have encountered an error in committing operations, the following example code should work:

<?php
    use Everyman\Neo4j\Transport,
        Everyman\Neo4j\Client,
        Everyman\Neo4j\Index,
        Everyman\Neo4j\Relationship,
        Everyman\Neo4j\Node,
        Everyman\Neo4j\Batch;

    error_reporting(-1);
    ini_set('display_errors', 1);

    function loaderTestAutoloader($sClass)
    {
        $sLibPath = __DIR__.'/batch-lib/';
        $sClassFile = str_replace('\\',DIRECTORY_SEPARATOR,$sClass).'.php';
        $sClassPath = $sLibPath.$sClassFile;
        if (file_exists($sClassPath)) {
            require($sClassPath);
        }
    }
    spl_autoload_register('loaderTestAutoloader');

    $transport = new Transport();
    $client = new Client($transport);

    $batch = new Batch($client);

    $node = new Node($client);
    $node->setProperties(array('foo' => 'bar', 'baz' => 'qux'));
    $batch->save($node);

    $result = $batch->commit();
    var_dump($result);
?>

The issue lies in that $result will always return false. After some debugging I discovered that Java is returning a 400 Bad Request with the following message: java.util.LinkedHashMap cannot be cast to java.util.List in Transport->makeRequest(). The full stacktrace is available here: https://gist.github.com/25fd7f7aef3fdb77e32a

I'm currently running Neo4j 1.4.1 (upgraded from 1.4, just in case that was the problem).

Any ideas?

Cheers,

• Peter

Example:

$qryObj = new \Everyman\Neo4j\Cypher\Query($client, 'START n=node(0) MATCH (n)<--(x) RETURN n, collect(x) as a');
$data = $qryObj->getResultSet();
$stuff = $data[0];

$n = $stuff['n'];
$a = $stuff['a'];

var_dump($n); //object(Everyman\Neo4j\Node)
var_dump($a[0]); //array

It seems arrays of nodes in the restult are not recognized properly. This is probably true for relations as well, but I haven't tested it.
$a[0] = $client->getEntityMapper()->getEntityFor($a[0]) can be used as a workaround

Allow batches to be opened and committed transparently. Example:

$batch = $client->startBatch();
// None of the following operations are sent to the server until...
$nodeA->save();
$nodeB->save();
$nodeA->relateTo($nodeB, 'KNOWS')->save();
$nameIndex->add($nodeB, 'name', 'foo');
// ...now
$client->commitBatch();
// or
$batch->commit();

Maybe in the future, all data operations will be wrapped in an implicit batch?

http://docs.neo4j.org/chunked/stable/rest-api-relationship-types.html

Both Cypher and Gremlin plugins support parameterized queries by passing a JSON map with the script.
This is much more efficient. See http://neo4j.org/forums/#nabble-td3569774|a3574235

Tested on examples/directions.php
Neo4J version neo4j-community-1.4.M05

string(194) "Unsupported value type class java.util.LinkedHashMap. Supported value types are all java primitives (byte, char, short, int, long, float, double) and String, as well as arrays of all those types"
["exception"]=>
string(244) "org.neo4j.server.rest.web.PropertyValueException: Unsupported value type class java.util.LinkedHashMap. Supported value types are all java primitives (byte, char, short, int, long, float, double) and String, as well as arrays of all those types"
["stacktrace"]=>
array(38) {
[0]=>
string(88) "org.neo4j.server.rest.domain.JsonHelper.assertSupportedPropertyValue(JsonHelper.java:87)"
[1]=>
string(77) "org.neo4j.server.rest.domain.JsonHelper.jsonToSingleValue(JsonHelper.java:65)"

The following script repeatedly creates nodes and relationships. Around iteration 10000 (on my machine), an exception is thrown saying:

PHP Everyman\Neo4j\Exception: No relationship start node specified in /home/kratib/www/recotype/sites/all/libraries/Neo4jPHP/lib/Everyman/Neo4j/Command/CreateRelationship.php on line 76

Dumping the parameters of Everyman\Neo4j\Command\CreateNode::handleResult() shows they're all empty.

My environment:

• Ubuntu 11.04
• PHP 5.3, memory_limit = -1
• Neo4j 1.4.1

#!/usr/bin/env php
<?php
use Everyman\Neo4j\Transport,
  Everyman\Neo4j\Client,
  Everyman\Neo4j\Index,
  Everyman\Neo4j\Relationship,
  Everyman\Neo4j\Node;
error_reporting(-1);
ini_set('display_errors', 1);

function loaderTestAutoloader($sClass)
{
  $sLibPath = __DIR__.'/../lib/';
  $sClassFile = str_replace('\\',DIRECTORY_SEPARATOR,$sClass).'.php';
  $sClassPath = $sLibPath.$sClassFile;
  if (file_exists($sClassPath)) {
    require($sClassPath);
  }
}
spl_autoload_register('loaderTestAutoloader');

$transport = new Transport();
$client = new Client($transport);

for ($i = 1; $i<=500000; ++$i) {
  $node = new Node($client);
  $node->setProperty('status_id', $i)->save();
  $user = new Node($client);
  $user->setProperty('user_id', $i)->save();
  try {
    $user->relateTo($node, 'TEST')->save();
  }
  catch (Exception $e) {
    echo $e->getMessage() . "\n";
  }
  if ($i % 100 == 0) {
    echo "$i\n";
  }
}

Doing any batch operation does not update the appropriate entities in the cache. They should.

In the example

$traversal = new Traversal($client);
$traversal->setPruneEvaluator(Traversal::PruneNone)
    ->setReturnFilter('javascript', 'return true;')
    ->setOrder(Traversal::OrderBreadthFirst);

chaining will not work because setPruneEvaluator() doesn't return the object. Moreover the properties of prune are "language" and "name" not "language" and "body" so in this case setPruneEvaluator should set
"prune_evaluator" : {"name" : "none","language" : "builtin"}
instead of
"prune_evaluator" : {"body" : "none","language" : "builtin"}

http://php-java-bridge.sourceforge.net/pjb/

May be useful to have a second type of Transport that talks to a native Java backend.

The most likely cause of this is my total ignorance of PHP Namespaces, but when I followed the instructions and did the following:

include("neo4jphp.phar");

...

I got the following error:

Fatal error: Class 'Everyman\Neo4j\Client' not found in XXX.php on line YYY

After this, I tried to stick the source-code in a folder called lib in the root of my project, and the following code works perfectly:

<?php

    spl_autoload_register("autoload");

    use Everyman\Neo4j\Client,
        Everyman\Neo4j\Transport;

    $client = new Client(new Transport('localhost', 7474));
    print_r($client->getServerInfo());

    function autoload($file)
    {
        $base = dirname(__FILE__)."/lib";
        $filename = str_replace('\\', DIRECTORY_SEPARATOR, $file) . '.php';
        if(file_exists($base.DIRECTORY_SEPARATOR.$filename))
            require_once($base.DIRECTORY_SEPARATOR.$filename);
    }

?>

I think it might be a far smaller barrier to entry if you were to include an example like this in your wiki - that is, if I didn't totally miss something fundamental with the phar example.

Export includes rel_id

Import includes rel_id and index

Hello,

Neo4JPHP is working properly on my terminal (Ubuntu 12.04) but it seems I cannot achieve to run the same php file on my browser. Everything else is also working properly on my browser but when any connection via Neo4JPHP returns me a HTTP Error 500.

I'm using phar package and even the code below crashes on browser:

require( "phar://neo4jphp.phar");
$client = new Everyman\Neo4j\Client();
$client->getTransport();

I cannot give any more details, what should it be caused from?

Batch operations do not handle removing unsaved nodes from an index. This is appropriate to do if the node is saved in a prior batch operation. Helps ensure uniqueness in an index. Should work the same as adding an unsaved node to an index.

https://github.com/neo4j/community/issues/25

Gremlin scripts may return other types than array.

For instance, executing g.v(1).name will return

Warning: array_key_exists() expects parameter 2 to be array, string given in /Users/Pierre/Sites/EC2/survey/lib/Everyman/Neo4j/Command/ExecuteGremlinQuery.php on line 89

Warning: Invalid argument supplied for foreach() in /Users/Pierre/Sites/EC2/survey/lib/Everyman/Neo4j/Command/ExecuteGremlinQuery.php on line 92

To prevent those warnings, a check may be added on top of Command/ExecuteGremlinQuery.php:

if (!is_array($data)) {
    $data = array($data);
}

When iterating through ResultSet, you may use f.i. is_object($row[0]) to check what Gremlin returns.

Hope that makes sense for you,
Pierre

http://py2neo.org/geoff/

Also, the GEOFF extension: https://github.com/neo4j/neo4j-geoff

Consider deprecating or removing the client-side GEOFF implementation.

server v1.5 <= uses old /db/data/ext/CypherPlugin endpoint
server 1.6 > uses new /db/data/cypher endpoint

https://github.com/neo4j/community/commit/f011c51fcbd5f12236fc633f57dcec9fb8cade83

http://docs.neo4j.org/chunked/1.6.M03/rest-api-unique-indexes.html

See title.

This gets rid of the need to install the cUrl extension. Make sure the same stream context can be reused.

The line ./bacon.php "Keanu Reeves" "Kevin Bacon" should be ./bacon.php path "Keanu Reeves" "Kevin Bacon".

After two years of using Github daily, I still don't know how I can contribute to wikis on repos :-)

As Neo4j server has difficulty responding with large data sets, it might make sense to automatically and transparently paginate the results of Cypher (and Gremlin?) queries if explicit SKIP/LIMIT pagination is not set.

I just finished making a memcache caching plugin. By make I mean I changed line 64 of the Memcached plugin to

return $this->memcached->set($key, $value, 0, $expire);

That's the only difference as far as I can tell and it has been working fine for me so far.

I'm using a Windows server so this small addition saved me the from trying to compile the memcached extension myself.

Josh,

I'm opening a separate ticket for this as it's not related to #4. I'm running the batch operations using the following example script:
https://gist.github.com/9ca60c4b021168af33aa

This yields the following benchmarks:

Created 10000 nodes via batch took: 5.60 minutes.
Created 10000 nodes sequentially took: 27 seconds.

I thought that perhaps it's related to my memory configuration, so I increased them based on what Neo4j recommend (http://wiki.neo4j.org/content/Batch_Insert), but it had little to no effect on the above stats.

This may not be a Neo4PHP issue, as Neo4j do state that batch insertions are experimental, but thought it'd be worth flagging!

Cheers,

• Peter

Get rid of QueryAssembler/custome parameterization and use the proper mapping capability of the Cypher plugin: http://docs.neo4j.org/chunked/milestone/cypher-plugin.html#rest-api-send-queries-with-parameters

Is the spatial plugin API stable enough?

Example code:

$client = new Everyman\Neo4j\Client();
$client->startBatch();
// ..
// some code which not executed (if condition for example)
// ..
$client->commitBatch();

throws "PHP Fatal error: Uncaught exception 'Everyman\Neo4j\Exception' with message 'Unable to commit batch [500]:"

I think it is a good idea to allow to make empty batch which does nothing.

The GetIndexes command also retrieves the index data like "provider" and "type", which is currently ignored. Set these as the Index config during construction.

Neo4j rest api actually support multiple relation search.

http://components.neo4j.org/neo4j-server/milestone/rest.html#Traverse

"relationships": [
{ "type": "KNOWS", "direction": "out" },
{ "type": "LOVES" }
]

About a month ago whenever I searched for an edge node [ getRelationship( $id ) ], both the start and end nodes, along with the edge node properties were loaded. Now, after the last changes, only the edge node properties are there, the start and end nodes must be explicitly loaded [ load() ] to access their properties.

I also noticed that the "lazyLoad" protected property is on for start and end nodes, but off for the edge node: I am new to Neo4j-PHP so I don't really know how and if this property is used, but I suspect it may have to do with this issue.

Is there a fix for this or what is the default workflow when getting nodes and edges?

Thanks!

Milko

PHP Notice: Undefined offset: 2 in phar:///neo4jphp.phar/lib/Everyman/Neo4j/Command/GetServerInfo.php on line 72

This is GetServerInfo::parseVersion method and $fullVersion variable is "1.5".

// php code
$client = new Client(new Transport());
$batch = new Batch($client);
$index = new Index($client, Index::TypeNode, 'index_test');

$name = 'string \ and "test"';
// escape part
//$name = str_replace('\', '\\', $name);
//$name = str_replace('"', '"', $name);

$node = new Node($client);
$node->setProperty('NAME', $name);
$batch->save($node);

//$batch->addToIndex($index, $node, 'NAME', $node->getProperty('NAME'));
$batch->commit();
//end of code

if there are \ or " string in property of node or value of index, i got exception.
the request of commit will be like this below and it get error.
post /db/data/batch[{"method":"POST","to":"/node","body":{"NAME":"string\ and "test""},"id":0}]

then i uncomment the escape part of code, the result of request will be like this:
post /db/data/batch[{"method":"POST","to":"/node","body":{"NAME":"string\\ and \"test\""},"id":0}]
this runs well.

But it runs correctly when i use $node->save() what ever what string it is.
Only get error with batch.
i'm not sure if this is a problem of php or server of neo4j.

i'm using Neo4j 1.5.M02, and tested on both of 0.0.4-beta and 0.0.5-beta.

Handle this: https://github.com/neo4j/community/issues/349

Would it be possible to make this lib async using CURL multi ?

This lib might be of help https://github.com/ericmuyser/php-class-curl/blob/master/src/curl.php but there is still a need to run a ->update() in a while loop to gather returns... Which is faked async imo.

Any insight ?

Thanks !

Start looking at annotations, possibly following the Spring Data Graph syntax. Look to Doctrine2 for inspiration.