DB API 2.0 driver for the Neo4j graph database.

pip install neo4jdb==0.0.9



Implements the Python DB API 2.0 for Neo4j, compatible with python 2.6, 2.7, 3.2 and 3.3 and Neo4j >= 2.0.0


pip install neo4jdb

Minimum viable snippet

import neo4j

connection = neo4j.connect("http://localhost:7474")

cursor = connection.cursor()
for name, age in cursor.execute("MATCH (n:User) RETURN, n.age"):
    print name, age


The library generally adheres to the DB API, please refer to the documentation for the DB API for detailed usage.

# Write statements
cursor.execute("CREATE (n:User {name:'Stevie Brook'}")
connection.commit() # Or connection.rollback()

# With positional parameters
cursor.execute("CREATE (n:User {name:{0}})", "Bob")
# or
l = ['Bob']
cursor.execute("CREATE (n:User {name:{0}})", *l)

# With named parameters
cursor.execute("CREATE (n:User {name:{name}})", name="Bob")
# or
d = {'name': 'Bob'}
cursor.execute("CREATE (n:User {name:{name}})", **d)
# or
d = {'node': {'name': 'Bob'}}
cursor.execute("CREATE (n:User {node})", **d)

If you ask Cypher to return Nodes or Relationships, these are represented as Node and Relationship types, which are dict objects with additional metadata for id, labels, type, end_id and start_id.

# Retrieve and access a node
for node, in cursor.execute("MATCH (n) RETURN n"):
    print, node.labels
    print node['a_property']

# Retrieve and access a relationship
for rel, in cursor.execute("MATCH ()-[r]->() RETURN r"):
    print, rel.type, rel.start_id, rel.end_id
    print rel['a_property']

Using the context manager. Any exception in the context will result in the exception being thrown and the transaction to be rolled back.

from neo4j.contextmanager import Neo4jDBConnectionManager
manager = contextmanager.Neo4jDBConnectionManager('http://localhost:7474')

with as r:  # r is just a cursor
    for name, age in r.execute("MATCH (n:User) RETURN, n.age"):
        print name, age
# When leaving read context the transaction will be rolled back

with manager.write as w:
    w.execute("CREATE (n:User {name:{name}})", name="Bob")
# When leaving write context the transaction will be committed

# When using transaction a new connection will be created
with manager.transaction as t:
    t.execute("CREATE (n:User {name:{name}})", name="Bob")
# When leaving transaction context the transaction will be
# committed and the connection will be closed

# Rolling back or commit in contexts
with manager.transaction as t:
    t.execute("CREATE (n:User {name:{name}})", name="Bob")
    if something == True:
        t.connection.commit()  # This will commit the transaction
        t.connection.rollback()  # This will rollback the transaction

Building & Testing

Neo4jDB uses paver as its build system. To install paver:

pip install paver

Then you can build Neo4jDB with:

paver build

And install it with:

paver install

Running tests requires a Neo4j server running on localhost. Paver can handle this for you:

paver start_server
paver nosetests
paver stop_server

Incompliance with the spec


The library delegates to Neo4j for parameter substitution, which means it does not use any of the standard parameter substitution types defined in the spec.

Instead it uses curly brackets with named and/or positional parameters:

{0} or {identifier}

Type system

Because the wire format for Neo4j is JSON, the library does not support the date or binary value types. This may change in the future as the wire format for Neo4j evolves.

In a similar vein, because Neo4j is a schema-optional database, it may return arbitrary types in each cell in the result table. As such, the description of the result table always marks each column type as neo4j.MIXED.