Skip to main content

Object-oriented Python library to interact with Neo4j standalone REST server

Project description

The first objective of Neo4j Python REST Client is to make transparent for Python programmers the use of a local database through or a remote database thanks to Neo4j REST Server. So, the syntax of this API is fully compatible with However, a new syntax is introduced in order to reach a more pythonic style.


Available throught Python Package Index:

$ pip install neo4jrestclient


$ easy_install neo4jrestclient

Getting started

The main class is GraphDatabase, exactly how in

>>> from neo4jrestclient.client import GraphDatabase

>>> gdb = GraphDatabase("http://localhost:7474/db/data/")


There some global options available:: If CACHE is ‘True’, a ‘.cache’ directory is created and the future request to the same URL will be taken from cache:

neo4jrestclient.options.CACHE = False # Default

If DEBUG is ‘True’, ‘httplib2’ is set to debuglevel = 1:

neo4jrestclient.options.DEBUG = False # Default

And SMART_ERRORS, set to ‘False’ by default. In case of ‘True’, the standard HTTP errors will be replaced by more pythonic errors (i.e. ‘KeyError’ instead of ‘NotFoundError’ in some cases):

neo4jrestclient.options.SMART_ERRORS = False # Default

Node, Relationships and Properties

Due to the syntax is fully compatible with, the next lines only show the commands added and its differences.

Creating a node:

>>> n = gdb.nodes.create()

# Equivalent to
>>> n = gdb.node()

Specify properties for new node:

>>> n = gdb.nodes.create(color="Red", widht=16, height=32)

# Or
>>> n = gdb.node(color="Red", widht=16, height=32)

Accessing node by id:

>>> n = gdb.node[14]

# Using the identifier or the URL is possible too
>>> n = gdb.nodes.get(14)

Accessing properties:

>>> value = n['key'] # Get property value

>>> n['key'] = value # Set property value

>>> del n['key']     # Remove property value

# Or the other way
>>> value = n.get('key', 'default') # Support 'default' values

>>> n.set('key', value)

>>> n.delete('key')

Besides, a Node object has other attributes:


>>> = {'name': 'John'}
{'name': 'John'}

# The URL and the identifier assigned by Neo4j are added too

>>> n.url

Create relationship:

>>> n1.Knows(n2)

# Or
>>> n1.relationships.create("Knows", n2) # Usefull when the name of
                                         # relationship is stored in a variable

Specify properties for new relationships:

>>> n1.Knows(n2, since=123456789, introduced_at="Christmas party")

# It's the same to
>>> n1.relationships.create("Knows", n2, since=123456789,
                                         introduced_at="Christmas party")

The creation returns a Relationship object, which has properties, setter and getters like a node:

>>> rel = n1.relationships.create("Knows", n2, since=123456789)

>>> rel.start
<Neo4j Node: http://localhost:7474/db/data/node/14>

>>> rel.end
<Neo4j Node: http://localhost:7474/db/data/node/32>

>>> rel.type

{'since': 123456789}

Or you can create the relationship using directly from GraphDatabse object:

>>> rel = gdb.relationships.create(n1, "Hates", n2)

>>> rel
<Neo4j Relationship: http://localhost:7474/db/data/relationship/66>

>>> rel.start
<Neo4j Node: http://localhost:7474/db/data/node/14>

>>> rel.end
<Neo4j Node: http://localhost:7474/db/data/node/32>

Others functions over ‘relationships’ attribute are possible. Like get all, incoming or outgoing relationships (typed or not):

>>> rels = n1.relationships.all()
<Neo4j Iterable: Relationship>

In order improve the performance of the ‘neo4jrestclient’, minimizing the number of HTTP requests that are made, all the functions that should return list of objects like Nodes, Relationships, Paths or Positions, they actually return an Iterable object that extends the Python ‘list’ type:

>>> rels = n1.relationships.all()[:]
[<Neo4j Relationship: http://localhost:7474/db/data/relationship/35843>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35840>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35841>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35842>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35847>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35846>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35845>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35844>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/11>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/10>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/9>]

>>> rels = n1.relationships.incoming(types=["Knows"])[:]
[<Neo4j Relationship: http://localhost:7474/db/data/relationship/35843>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35840>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/11>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/10>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/9>]

>>> rels = n1.relationships.outgoing(["Knows", "Loves"])[:]
[<Neo4j Relationship: http://localhost:7474/db/data/relationship/35842>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/35847>]

There’s a shortcut to access to the list of all relationships:

>>> rels = n1.relationships.all()[2]
<Neo4j Relationship: http://localhost:7474/db/data/relationship/47>

It’s the same to:

>>> rels = n1.relationships[2]
<Neo4j Relationship: http://localhost:7474/db/data/relationship/47>


>>> rels = n1.relationships.get(2)
<Neo4j Relationship: http://localhost:7474/db/data/relationship/47>


The traversals framework is supported too with the same syntax of, but with some added issues.

Regular way:

>>> n1.relationships.create("Knows", n2, since=1970)
<Neo4j Relationship: http://localhost:7474/db/data/relationship/36009>

>>> class TraversalClass(gdb.Traversal):
   ...:     types = [
   ...:         client.All.Knows,
   ...:     ]

>>> [traversal for traversal in TraversalClass(n1)]
[<Neo4j Node: http://localhost:7474/db/data/node/15880>]

Added way (the types of relationships are ‘All’, ‘Incoming’, ‘Outgoing’):

>>> n1.relationships.create("Knows", n2, since=1970)
<Neo4j Relationship: http://localhost:7474/db/data/relationship/36009>

>>> n1.traverse(types=[client.All.Knows])[:]
[<Neo4j Node: http://localhost:7474/db/data/node/15880>]

For getting a paginated traversal is only needed one of the next parameters: ‘paginated’ to enable the pagination, ‘page_size’ to set the size of returned page, and ‘time_out’ to establish the lease time that the server will wait for. After set any of this parameters, the traversal call will return an iterable object of traversals called ‘PaginatedTraversal’:

>>> pages = n1.traverse(types=[client.All.Knows], stop=stop, page_size=5)

>>> pages
<PaginatedTraversal object at 0x25a5150>

>>> [n for n in [traversal for traversal in pages]]
[<Neo4j Node: http://localhost:7474/db/data/node/15880>]


Due to the original currently doesn’t provide support for the new index component, for nodes and for relationships, the syntax for indexing is not compliant, quite different and, hopefully, more intuitive:

>>> i1 =  gdb.nodes.indexes.create("index1")

>>> i2 =  gdb.nodes.indexes.create("index2", type="fulltext", provider="lucene")

>>> gdb.nodes.indexes
{u'index2': <Neo4j Index: http://localhost:7474/db/data/index/node/index2>,
 u'index1': <Neo4j Index: http://localhost:7474/db/data/index/node/index1>}

>>> gdb.nodes.indexes.get("index1")
<Neo4j Index: http://localhost:7474/db/data/index/node/index1>

You can query and add elements to the index like a 3-dimensional array or using the convenience methods:

>>> i1["key"]["value"]

>>> i1.get("key")["value"]

>>> i1.get("key", "value")

>>> i1["key"]["value"] = n1

>>> i1.add("key", "value", n2)

>>> i1["key"]["value"][:]
[<Neo4j Node: http://localhost:7474/db/data/node/1>,
 <Neo4j Node: http://localhost:7474/db/data/node/2>]

Advanced queries are also supported if the index is created with the type ‘fulltext’ (‘lucene’ is the default provider) by entering a Lucene query:

>>> n1 = gdb.nodes.create(name="John Doe", place="Texas")

>>> n2 = gdb.nodes.create(name="Michael Donald", place="Tijuana")

>>> i1 = gdb.nodes.indexes.create(name="do", type="fulltext")

>>> i1["surnames"]["doe"] = n1

>>> i1["places"]["Texas"] = n1

>>> i1["surnames"]["donald"] = n2

>>> i1["places"]["Tijuana"] = n2

>>> i1.query("surnames", "do*")[:]
[<Neo4j Node: http://localhost:7474/db/data/node/295>,
 <Neo4j Node: http://localhost:7474/db/data/node/296>]

…or by using the DSL described by lucene-querybuilder to support boolean operations and nested queries:

>>> i1.query(Q('surnames','do*') & Q('places','Tijuana'))[:]
[<Neo4j Node: http://localhost:7474/db/data/node/295>]

Deleting nodes from an index:

>>> i1.delete("key", "values", n1)

>>> i1.delete("key", None, n2)

And in order to work with indexes of relationships the instructions are the same:

>>> i3 =  gdb.relationships.indexes.create("index3")

For deleting an index just call ‘delete’ with no arguments:

>>> i3.delete()


The server plugins are supported as extensions of GraphDatabase, Node or Relationship objects:

>>> gdb.extensions
{u'GetAll': <Neo4j ExtensionModule: [u'get_all_nodes',
>>> gdb.extensions.GetAll
<Neo4j ExtensionModule: [u'get_all_nodes', u'getAllRelationships']>

>>> gdb.extensions.GetAll.getAllRelationships()[:]

[<Neo4j Relationship: http://localhost:7474/db/data/relationship/0>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/1>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/2>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/3>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/4>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/5>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/6>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/7>,
 <Neo4j Relationship: http://localhost:7474/db/data/relationship/8>]

An example using extensions over nodes:

>>> n1 = gdb.nodes.get(0)

>>> n1.extensions
{u'DepthTwo': <Neo4j ExtensionModule: [u'nodesOnDepthTwo',
 u'ShortestPath': <Neo4j ExtensionModule: [u'shortestPath']>}

>>> n2 = gdb.nodes.get(1)

>>> n1.relationships.create("Kwnos", n2)
<Neo4j Relationship: http://localhost:7474/db/data/relationship/36>

>>> n1.extensions.ShortestPath
<Neo4j ExtensionModule: [u'shortestPath']>

>>> n1.extensions.ShortestPath.shortestPath.parameters

[{u'description': u'The node to find the shortest path to.',
  u'name': u'target',
  u'optional': False,
  u'type': u'node'},
 {u'description': u'The relationship types to follow when searching for ...',
  u'name': u'types',
  u'optional': True,
  u'type': u'strings'},
 {u'description': u'The maximum path length to search for, ...',
  u'name': u'depth',
  u'optional': True,
  u'type': u'integer'}]


Currently, the transaction support is not complete in this client, although a work in progress is being carried out, and hopefully the capacity to handle objects created in the same transaction will be done.

Basic usage for deletion:

>>> n = gdb.nodes.create()

>>> n["age"] = 25

>>> n["place"] = "Houston"

{'age': 25, 'place': 'Houston'}

>>> with gdb.transaction():
   ....:         n.delete("age")

{u'place': u'Houston'}

When a transaction is performed, the values of the properties of the objects are updated automatically. However, this can be controled by hand adding a parameter in the transaction:

>>> n = gdb.nodes.create()

>>> n["age"] = 25

>>> with gdb.transaction(update=False):
   ....:         n.delete("age")

{'age': 25}

>>> n.update()


Apart from update or deletion of properties, there is also creation. In this case, the object just created is returned through a ‘TransactionOperationProxy’ object, which is automatically converted in the proper object when the transaction ends. This is the second part of the commit process and a parameter in the transaction can be added to avoid the commit:

>>> n1 = gdb.nodes.create()

>>> n2 = gdb.nodes.create()

>>> with gdb.transaction(commit=False) as tx:
   .....:         for i in range(1, 11):
   .....:             n1.relationships.create("relation_%s" % i, n2)

>>> len(n1.relationships)

The ‘commit’ method of the transaction object returns ‘True’ if there’s no any fail. Otherwose, it returns ‘None’:

>>> tx.commit()

>>> len(n1.relationships)

In order to avoid the need of setting the transaction variable, ‘neo4jrestclient’ uses a global variable to handle all the transactions. The name of the variable can be changed using de options:

>>> client.options.TX_NAME = "_tx"  # Default value

And this behaviour can be disabled adding the right param in the transaction: ‘using_globals’. Even is possible (although not very recommendable) to handle different transactions in the same time and control when they are commited. There are many ways to set the transaction of a intruction (operation):

>>> n = gdb.nodes.create()

>>> n["age"] = 25

>>> n["name"] = "John"

>>> n["place"] = "Houston"

>>> with gdb.transaction(commit=False, using_globals=False) as tx1, \
   ....:      gdb.transaction(commit=False, using_globals=False) as tx2:
   ....:         n.delete("age", tx=tx1)
   ....:     n["name"] = tx2("Jonathan")
   ....:     n["place", tx2] = "Toronto"

>>> "age" in

>>> tx1.commit()

>>> "age" in

>>> n["name"] == "John"

>>> n["place"] == "Houston"

>>> tx2.commit()

>>> n["name"] == "John"

>>> n["place"] == "Houston"


1.5.0 (2011-10-31)

  • Removing the smart_quote function from indexing. It’s not needed anymore with the new way to add elements to indices.
  • Fixes #37.
  • Using JSON object to set index key and value

1.4.5 (2011-09-15)

  • Adding more testing to returns parameter in the extensions.
  • Fixes 32. It needs some more testing, maybe.
  • Updated to using lucene-querybuilder 0.1.5 (bugfixes and better wildcard support).
  • Fixed the test issue found in #34, and updated the REST client to using lucene-querybuilder 0.1.5.
  • Fixes #34. Fixing dependency of lucene-querybuilder version
  • Fixes #30. Fixing an issue deleting all index entries for a node.
  • Fixing an issue with parameters in extensions.
  • Ensure that self.result is always present on the object, even if it’s None.
  • Fixing naming glitch in exception message
  • Ensure that self.result is always present on the object, even if it’s None
  • Fixing an error retrieving relationships in paths.
  • Fixing an error in extensions, Path and Position.

1.4.4 (2011-08-17)

  • Merge pull request #28 from mhluongo/master
  • Made the DeprecationWarnings a bit more specific.
  • Nodes can now be used in set and as dict keys, differentiated by id.
  • Added a test for node hashing on id.
  • Removed the ‘Undirected’ reference from tests to avoid a DepreactionWarning.
  • Moved the relationship creation DeprecationWarning so creating a relationship the preferred way won’t raise it.
  • Got rid of the DeprecationWarning on import- moved in to whenever using Undirected.*.
  • Fixed traversal return filters.
  • Enabled return filters, including those with custom javascript bodies. Eventually a more elegant (Python instead of string based) solution for return filter bodies is in order.
  • Fixed a mispelling in the test_traversal_return_filter case.
  • Added a test for builtin and custom traversal return filters.
  • Small bug fix for traversal
  • Fixed bug in traverse method for POSITION and PATH return types.

1.4.3 (2011-07-28)

  • Added some deprecation warnings.
  • Added support for pickling ans some tests.
  • Fixed an error deleting nodes and relationships on transactions.
  • Finishied and refactored the full unicode support.

1.4.2 (2011-07-18)

  • Updated the documentation and version.
  • Added support for indices deletion.
  • Improved Unicode support in properties keys and values and relationships types. Adding some tests.

1.4.1 (2011-07-12)

  • Fixed an error retrieving relationships by id.
  • Added control to handle exceptions raised by Request objects.
  • Updated changes, manifest and readme files.

1.4.0 (2011-07-11)

  • Updated version number for the new release.
  • Updated documentation.
  • Updated develpment requirements.
  • Added support for paginated traversals.
  • Passed pyflakes and PEP8 on tests.
  • Added weight to Path class.
  • Index values now quoted_plus.
  • Changed quote to quote_plus for index values.
  • Added two tests for unicode and url chars in index values.
  • Added initial documentacion for transactions.
  • Added the transaction support and several tests.
  • Fixed the implementation of __contains__ in Iterable class for evaluation of ‘in’ and ‘not in’ expressions.
  • Added documentation for Iterable objects.
  • Added more transactions features.
  • Added requirements file for virtual environments in development.
  • Improved number of queries slicing the returned objects in a Iterable wrapper class.
  • Added Q syntax for more complicated queries.
  • Added support for the Q query syntax for indexes using the DSL at
  • Fixed an error in the test_query_index case (forgot to include an ‘or’. between queries).
  • Added lucene-querybuilder to the test requirements in
  • Added a test case for Q-based queries.

1.3.4 (2011-06-22)

  • Fixed the and httplib2 import error during installing.
  • Reordered the options variables in an file. Allows index.query() to be called with or without a key
  • Fixed issue #15 regarding dependency to httplib2
  • Patched index.query() so it can take a query without a key (to support, say, mutli-field Lucene queries). Ultimately, query so probably be refactored to Index (instead of IndexKey) because IndexKey doesn’t actually help with full-text queries.
  • Fixed for issue #19 (missed that urllib.quote).
  • Altered the test_query_index case to reflect how I think indexing should work.
  • Using assertTrue instead of failUnless in, failUnless is deprecated in 2.7 and up, so I figured we might as well switch.
  • Added SMART_ERRORS (aka “Django mode”). If you set SMART_ERROR to True it will make the client throw KeyError instead of NotFoundError when a key is missing.

1.3.3 (2011-06-14)

  • Fixed an introspection when the results list of a traverse is empty.
  • Merge pull request #17 from mhluongo/master
  • Resolved the STOP_AT_END_OF_GRAPH traversal test case. Calling .traverse(stop=STOP_AT_END_OF_GRAPH) will now traverse the graph without a max depth (and without 500 errors).
  • Added a failing test case for traverse(stop=STOP_AT_END_OF_GRAPH).

1.3.2 (2011-05-30)

  • Added a test for deleting relationships.
  • Fixing an Index compatibility issue with Python 2.6.1.
  • Fixing an error in extensions support with named params.

1.3.1 (2011-04-16)

  • Fixing

1.3.0 (2011-04-15)

  • First Python Index Package release with full support for Neo4j 1.3.

Project details

Release history Release notifications

History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


This version
History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


History Node


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
neo4jrestclient-1.5.0.tar.gz (41.5 kB) Copy SHA256 hash SHA256 Source None Oct 31, 2011

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page