Python Driver for ArangoDB with built-in validation
Clone or download
Latest commit dd72e5f Dec 8, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples fix example for examples/ May 10, 2017
pyArango Merge branch 'master' into travis-with-docker-containers Dec 7, 2018
.gitignore Initial commit Oct 5, 2014
.travis.yml use arangodb via docker container Dec 7, 2018
CHANGELOG.rst opened 1.3.3 Sep 27, 2018
DESCRIPTION.rst new stable version Jul 21, 2016
LICENSE True license Nov 19, 2014 added manifest Jan 5, 2015
README.rst Mention python 3.5 Oct 24, 2018 opened 1.3.3 Sep 27, 2018



NoSQL is really cool, but in this harsh world it is impossible to live without field validation.

WARNING: The last versions of pyArango are only compatible with ArangoDB 3.X. For the old version checkout the branch ArangoDBV2

Key Features

pyArango, is geared toward the developer. It's here to help to you develop really cool apps using ArangoDB, really fast.

  • Light and Simple interface
  • Built-in Validation of fields on seting or on saving
  • Support for all index types
  • Supports graphs, traversals and all types of queries
  • Caching of documents with Insertions and Lookups in O(1)

Collections are treated as types that apply to the documents within. That means that you can define a Collection and then create instances of this Collection in several databases. The same goes for graphs

In other words, you can have two databases cache_db and real_db each of them with an instance of a Users Collection. You can then be assured that documents of both collections will be subjected to the same validation rules. Ain't that cool?

You can be 100% permissive or enforce schemas and validate fields, on set, on save or both.


Supports python 2.7 and 3.5.

From PyPi:

pip install pyArango

For the latest version:

git clone
cd pyArango
python develop

Full documentation

This is the quickstart guide, you can find the full documentation here.

Initialization and document saving

from pyArango.connection import *

conn = Connection()

db = self.conn["test_db"] # all databases are loaded automatically into the connection and are accessible in this fashion
collection = db.createCollection(name="users") # all collections are also loaded automatically

# collection.delete() # self explanatory

for i in xrange(100):
    doc = collection.createDocument()
    doc["name"] = "Tesla-%d" % i
    doc["number"] = i
    doc["species"] = "human"

doc = collection.createDocument()
doc["name"] = "Tesla-101"
doc["number"] = 101
doc["species"] = "human"

doc["name"] = "Simba"
# # overwrites the document
doc.patch() # updates the modified field

Queries : AQL

aql = "FOR c IN users FILTER == @name LIMIT 10 RETURN c"
bindVars = {'name': 'Tesla-3'}
# by setting rawResults to True you'll get dictionaries instead of Document objects, useful if you want to result to set of fields for example
queryResult = db.AQLQuery(aql, rawResults=False, batchSize=1, bindVars=bindVars)
document = queryResult[0]

Queries : Simple queries by example

PyArango supports all types of simple queries (see for the full list). Here's how you do a query by example:

example = {'species': "human"}
query = collection.fetchByExample(example, batchSize=20, count=True)
print query.count # print the total number or documents

Queries : Batches

for e in query :
    print e['name']

Defining a Collection and field/schema Validation

PyArango allows you to implement your own field validation. Validators are simple objects deriving from classes that inherit from Validator and implement a validate() method.

import pyArango.collection as COL
import pyArango.validation as VAL
from pyArango.theExceptions import ValidationError
import types

class String_val(VAL.Validator):
 def validate(self, value):
     if type(value) is not types.StringType :
         raise ValidationError("Field value must be a string")
     return True

class Humans(COL.Collection):

    _validation = {
        'on_save': False,
        'on_set': False,
        'allow_foreign_fields': True  # allow fields that are not part of the schema

    _fields = {
        'name': Field(validators=[VAL.NotNull(), String_val()]),
        'anything': Field(),
        'species': Field(validators=[VAL.NotNull(), VAL.Length(5, 15), String_val()])

collection = db.createCollection('Humans')

In addition you can also define collection properties (creation arguments for ArangoDB) right inside the definition

class Humans(COL.Collection):

  _properties = {
      "keyOptions" : {
          "allowUserKeys": False,
          "type": "autoincrement",
          "increment": 1,
          "offset": 0,

    _validation = {
        'on_save': False,
        'on_set': False,
        'allow_foreign_fields': True  # allow fields that are not part of the schema

    _fields = {
        'name': Field(validators=[VAL.NotNull(), String_val()]),
        'anything': Field(),
        'species': Field(validators=[VAL.NotNull(), VAL.Length(5, 15), String_val()])

A note on inheritence

There is no inheritence of the "_validation" and "_fields" dictionaries. If a class does not fully define it's own, the defaults will be automatically assigned to any missing value.

Creating Edges

from pyArango.collection import Edges

class Connections(Edges):

    _validation = {
        'on_save': False,
        'on_set': False,
        'allow_foreign_fields': True # allow fields that are not part of the schema

    _fields = {
        'length': Field(NotNull=True),

Linking Documents with Edges

from pyArango.collection import *

class Things(Collection):

class Connections(Edges):

a = myThings.createDocument()
b = myThings.createDocument()

conn = myConnections.createEdge()

conn.links(a, b)
conn["someField"] = 35 # once an edge links documents, save() and patch() can be used as with any other Document object

Geting Edges linked to a vertex

You can do it either from a Document or an Edges collection:

# in edges

# out edges

# both

# you can also of ask for the raw json with
myDocument.getInEdges(myConnections, rawResults=True)
# otherwise Document objects are retuned in a list

Creating a Graph

By using the graph interface you ensure for example that, whenever you delete a document, all the edges linking to that document are also deleted.

from pyArango.collection import Collection, Field
from pyArango.graph import Graph, EdgeDefinition

class Humans(Collection):
    _fields = {
        "name": Field()

class Friend(Edges): # theGraphtheGraph
    _fields = {
        "lifetime": Field()

# Here's how you define a graph
class MyGraph(Graph) :
    _edgeDefinitions = [EdgeDefinition("Friend", fromCollections=["Humans"], toCollections=["Humans"])]
    _orphanedCollections = []

# create the collections (do this only if they don't already exist in the database)
# same for the graph
theGraph = self.db.createGraph("MyGraph")

# creating some documents
h1 = theGraph.createVertex('Humans', {"name": "simba"})
h2 = theGraph.createVertex('Humans', {"name": "simba2"})

# linking them'Friend', h1, h2, {"lifetime": "eternal"})

# deleting one of them along with the edge

Document Cache

pyArango collections have a caching system for documents that performs insertions and retrievals in O(1)

# create a cache a of 1500 documents for collection humans

# disable the cache

Statsd Reporting

pyArango can optionally report query times to a statsd server for statistical evaluation.

import statsd from pyArango.connection import Connection statsdclient = statsd.StatsClient(os.environ.get('STATSD_HOST'), int(os.environ.get('STATSD_PORT'))) conn = Connection('', 'root', 'opensesame', statsdClient = statsdclient, reportFileName = '/tmp/queries.log')
Its intended to be used in a two phase way: (we assume you're using bind values - right?)
  • first run that will trigger all usecases. You create the connection by specifying statsdHost, statsdPort and reportFileName. reportFilename will be filled with your queries paired with your hash identifiers. Its reported to statsd as 'pyArango_<hash>'. you can later on use this digest to identify your queries to the gauges.
  • on subsequent runs you only specify statsdHost and statsdPort; only the request times are reported to statsd.


More examples can be found in the examples directory. To try them out change the connection strings according to your local setup.

Debian Dependency Graph

If you are on a Debian / Ubuntu you can install packages with automatic dependency resolution. In the end this is a graph. This example parses debian package files using the deb_pkg_tools, and will then create vertices and edges from packages and their relations.

Use examples/ to install it, or examples/ to browse it as an ascii tree.

ArangoDB Social Graph

You can create the ArangoDB SocialGraph using examples/ It resemples The original ArangoDB Javascript implementation in python.