Skip to content
This repository has been archived by the owner on Oct 20, 2021. It is now read-only.
/ sqerzo Public archive

Tiny ORM for graph databases: Neo4j, RedisGraph, AWS Neptune or Gremlin

License

Notifications You must be signed in to change notification settings

BBVA/sqerzo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

29 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SQErzo logo

SQErzo Tiny ORM for Graph databases

What is SQErzo

SQErzo is simple and tiny ORM (Object-Relational Mapping) for graph databases for Python developers.

It's compatible with databases that supports Open Cypher language.

Which databases are supported

Database Status
Neo4j Supported
Redis Graph Supported
Arango DB Looking for contributor
AWS Neptune Looking for contributor
Gremlin Looking for contributor

Why use SQErzo?

SQErzo intermediates between the graph database and your application logic in a database agnostic way. As such, SQErzo abstracts the differences between the different databases. For examples:

  • RedisGraph doesn't support Date times or CONSTRAINTS, SQErzo does the magic to hide that.
  • Neo4j need different channels for writing than for read. SQErzo does the magic to hide that.
  • SQErzo integrates a in memory cache to avoid queries to Graph DB and try to improve the performance.
  • Every database uses their own Node/Edge identification system. You need to manage and understand then to realize when a node already exits in Graph DB. SQErzo do this for you. It doesn't matter the Graph DB engine you use.
  • SQErzo was made to avoid you to write useless code. You can create and manage Nodes and Edges in a few lines of code without know Graph DB internals.
  • SQErzo supports Graph DB bases on Open cypher language (a Graph databases query language). You don't need to learn them to perform day a day operations.

Project status

Project is in a very early stage. If you want to use them, have in count that.

Install

Install is easy. Only run:

> pip install sqerzo

Usage examples

Run databases uses Docker.

Start Neo4j

> docker run -d -p7474:7474 -p7687:7687 -e NEO4J_AUTH=neo4j/s3cr3t neo4j

Start RedisGraph

> docker run -p 7000:6379 -d --rm redislabs/redisgraph

Simple usage

Create some nodes and setup database in both databases:

  • Neo4j
  • RedisGraph

Without the need to change any code:

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph

class MeetEdge(GraphEdge):
    pass

@dataclass
class UserNode(GraphNode):
    name: str = None

    
def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database
    
    u1 = UserNode(name=f"UName-1")
    gh.save(u1)
    
    d1 = UserNode(name=f"DName-2")
    gh.save(d1)
        
    u1_meet_g1 = MeetEdge(
        source=u1,
        destination=d1
    )
    gh.save(u1_meet_g1)
        

if __name__ == '__main__':
    create_graph("redis://127.0.0.1:7000/?graph=email")   
    create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email")

This is the result database in Node4j:

user_meet_neo4j logo

This is the result database in RedisGrap:

user_meet_redisgraph logo

Recovering database nodes by their ID

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph
from sqerzo.exceptions import SQErzoElementExistException

@dataclass
class UserNode(GraphNode):
    name: str = None

def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    user = UserNode(name=f"UName-{n}")
    gh.save(user)

    # First argument: node ID we want to recover
    # Second argument: node class in which we want to map the result
    recovered_user = gh.get_node_by_id(user.id, UserNode)

Recovering database nodes by their properties

Getting one node:

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph

@dataclass
class UserNode(GraphNode):
    __keys__ = ["name"]

    name: str = None

def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    u1 = UserNode(name="Eustaquio")
    gh.save(u1)
    u2 = UserNode(name="Guachinche")
    gh.save(u2)
    
    # First argument: node ID we want to recover
    # Second argument: node class in which we want to map the result
    node = gh.fetch_one(UserNode, name="Eustaquio")

if __name__ == '__main__':
  create_graph("redis://127.0.0.1:7000/?graph=email")
  create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email")

Getting multiple nodes:

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph

@dataclass
class UserNode(GraphNode):
    name: str = None
    age: int = None

def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    u1 = UserNode(name="Eustaquio", age=22)
    gh.save(u1)
    u2 = UserNode(name="Guachinche", age=22)
    gh.save(u2)
    
    # First argument: node ID we want to recover
    # Second argument: node class in which we want to map the result
    for n in gh.fetch_many(UserNode, age=22):
        print(n)

if __name__ == '__main__':
  create_graph("redis://127.0.0.1:7000/?graph=email")
  create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email")
from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph
from sqerzo.exceptions import SQErzoElementExistException

@dataclass
class UserNode(GraphNode):
    name: str = None

def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    user = UserNode(name=f"UName-{n}")
    gh.save(user)

    #
    # First argument: node ID we want to recover
    # Second argument: node class in which we want to map the result
    recovered_user = gh.get_node_by_id(user.id, UserNode)

if __name__ == '__main__':
  create_graph("redis://127.0.0.1:7000/?graph=email")
  create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email")

Raw queries

SQErzo try to be simple. So, if you want to do complex queries, you'll write them in the DB Engine language.

This example explains how to perform a query in Open Cypher language and map the results to Python Classes:

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph

class MeetEdge(GraphEdge):
    pass

class WorksWithEdge(GraphEdge):
    pass

@dataclass
class UserNode(GraphNode):
    __keys__ = "email"

    name: str = None
    email: str = None


def create_graph(connection_string: str, nodes_count = 500):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    #
    # Add some data and relations: User1 -[meet]-> User 2
    #
    with gh.transaction() as tx:

        for n in range(nodes_count):
            u1_name = f"uname{n}"
            d1_name = f"dname{n}"

            u1 = UserNode(name=u1_name, email=f"{u1_name}@{u1_name}.com")
            d1 = UserNode(name=d1_name, email=f"{d1_name}@{d1_name}.com")

            tx.add(u1)
            tx.add(d1)

            u2_meet_u1 = MeetEdge(
                source=u1,
                destination=d1
            )
            u1_meet_u2 = MeetEdge(
                source=d1,
                destination=u1
            )
            tx.add(u1_meet_u2)
            tx.add(u2_meet_u1)

    #
    # HERE STARTS THE QUERY
    #

    # Execute will return a list of lists: [ 
    #   [UserNode("u1"), UserNode("u2")], 
    #   [UserNode("u1"), UserNode("u2")], 
    #   ... 
    # ]
    q = gh.Query.raw(
        "match (u1:User)-[:Meet]->(u2:User) return u1, u2"
    ).execute(map_to={"u1": UserNode, "u2": UserNode})
    
    print(q)

if __name__ == '__main__':
  count = 1000
  create_graph("redis://127.0.0.1:7000/?graph=email", nodes_count=count)
  create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email", nodes_count=count)

Transactions

Transactions are useful if you need add a lot of data. You add nodes and edges to a transaction. When they finish then perform the insertions to the database in a very efficient way:

from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph

class MeetEdge(GraphEdge):
    pass

@dataclass
class UserNode(GraphNode):
    __keys__ = ["name"]

    name: str = None

def create_graph(connection_string: str):
    gh = SQErzoGraph(connection_string)
    gh.truncate()  # Drop database

    with gh.transaction() as tx:  # Transaction starts here

        for n in range(500):  # Inserts 1000 nodes (500 * 2) and 500 relations
            u1 = UserNode(name=f"UName-{n}")
            d1 = UserNode(name=f"DName-{n}")

            tx.add(u1)
            tx.add(d1)

            u1_meet_g1 = MeetEdge(
                source=u1,
                destination=d1
            )
            tx.add(u1_meet_g1)


if __name__ == '__main__':
    print("Redis...")
    create_graph("redis://127.0.0.1:7000/?graph=email")
    print("Neo4j...")
    create_graph("neo4j://neo4j:s3cr3t@127.0.0.1:7687/?graph=email")

More complex example: Load mails to a Graph

If you need a more complex example, you can find in it examples/email_graph.py.

At this example we load a random generated mail inbox (generation script is also available) into a Graph Database following this Neo4j Blog Post suggestions.

Fraud graph db

ChangeLog

Release 0.1.2

Core

  • fixed - Node/Edge id generation when not explicit identity field was provided.
  • fixed - get_node_by_id(...) methods that raises execution.
  • fixed - Improved error control.
  • fixed - fetch_nodes(...) method, that raises when a query returns more than 1 result.
  • Improved - fetch_many(...) and fetch_one(...).

Other

  • Added new examples in examples folder.
  • Added new examples in README.
  • Updated examples for new SQErzo API.
  • Updated docker-compose with some fixes.

Release 0.1.1

  • Added queries support for raw queries in DB engine language

Release 0.1.0

  • Improved speed at insertion by 100x
  • Add support for UNIQUE create_constraints_nodes
  • Add support for INDEXES create_constraints_nodes
  • Add support for raw Cypher query
  • Errors, issues, new features and something else
  • Complete refactor to easy add new backends
  • Complete refactor to easy add new backends
  • Add new methods: fetch_many, fetch_one, raw_query, save, update & transaction
  • Add new examples
  • Improved the way to build the Node to avoid waste memory.

TODO

  • Implement update operations
  • Improve documentation
  • Improve cypher query to avoid query raises when a transaction insert a duplicate node
  • Add support for Arango DB
  • Add support for AWS Neptune
  • Add support for Gremlin
  • Add support for dates to RedisGraph using transformation of dates to numbers
  • Implementation of Query builder. Add some method to Query builder class. Here some possible examples:
from dataclasses import dataclass

from sqerzo import GraphEdge, GraphNode, SQErzoGraph


class MeetEdge(GraphEdge):
    pass

class WorksWithEdge(GraphEdge):
    pass

@dataclass
class UserNode(GraphNode):
    __keys__ = "email"

    name: str = None
    email: str = None

@dataclass
class OtherUserNode(GraphNode):
    __keys__ = "email"

    name: str = None
    email: str = None


gh = SQErzoGraph("redis://")
gh.Q().from(Node1).to(node2).execute()
gh.Q().from(name="me").to(UserNode).execute()
gh.Q().from(name="me", email="me@me.com").to((UserNode, "User")).execute()
gh.Q().from(name="me").across((WorksWithEdge, "WorksWith")).to((UserNode, "OtherUser")).execute()
gh.Q().to((UserNode, "OtherUser")).execute()
gh.Q().from(OtherUserNode).execute()
gh.Q().from(UserNode).execute()

References

I tried to use good practices for building SQErzo. Some references I used:

Authors

SQErzo is being developed by BBVA-Labs Security team members.

Contributions

Contributions are of course welcome. See CONTRIBUTING or skim existing tickets to see where you could help out.

License

SQErzo is Open Source Software and available under the Apache 2 license