Bulbflow: difference between neo4jserver Graph and neo4jserver Neo4jclient

Question

I am now trying to learn how to connect to Neo4j server and run Cypher queries on it using Bulbflow from Python. And the thing I do not understand is the difference between two possibilities to connect to the neo4j server:

1) Graph

from bulbs.neo4jserver import Graph
g = Graph()

2) Neo4jClient

from bulbs.neo4jserver import Neo4jClient
client = Neo4jClient()

Could anyone please explain the conceptual difference here? And which way is it better to choose if I then want to execute (quite a lot of) Cypher queries against the server and ultimately in parallel?

PS: I do not have enough reputation to create a tag "bulbflow" for this question :)

Answer 1

Bulbs supports three different graph database servers -- Neo4j Server, Rexster, and now Titan.

Code specific to each backend server is contained within its own Python package (directory). You should see directories for: neo4jserver, rexster, titan:

• Bulbs Source Code: https://github.com/espeed/bulbs/tree/master/bulbs

Neo4jClient is the low-level adapter for Neo4j Server -- you usually don't need to use this directly unless you are doing custom stuff -- use the high-level Graph class instead.

See the Bulbs docs for...

• Neo4j Server Client: http://bulbflow.com/docs/api/bulbs/neo4jserver/client/
• Neo4j Server Graph: http://bulbflow.com/docs/api/bulbs/neo4jserver/graph/

The Bulbs Quickstart guide provides examples on using the Graph interface:

• Bulbs Quickstart: http://bulbflow.com/quickstart/

However, your Bulbs objects always have access to the low-level client when you need it via the _client var.

Lightbulb is an example app I created to show how to use and customize Bulbs models -- it's a Python blog engine that uses Git for source control and a graph database for persistence.

• Lightbulb: https://github.com/espeed/lightbulb

Lightbulb was originally designed for use with the free Neo4j Heroku Add On, but both Bulbs and Lightbulb make heavy use of Gremlin, and the Neo4j Heroko Add On no longer offers Gremlin in the free edition.

The Lightbulb model file contains a heavily customized Entry model and a custom Graph class -- the Entry model makes use of the low-level client:

• Lightbulb Model: https://github.com/espeed/lightbulb/blob/master/lightbulb/model.py

As you can see in the Entry model, I have access to the low-level client via the _client var, and I use it to get a Gremlin script from the scripts library and then again to execute the Gremlin script.

Here's the code for the save_blog_entry Gremlin script used by the Entry model:

• Lightbulb Gremlin Scripts: https://github.com/espeed/lightbulb/blob/master/lightbulb/gremlin.groovy

NOTE: There is only one Gremlin script in the file, but it is large, contains several operations, and everything is wrapped in a transaction. Putting all the operations into a single Gremlin script allows you to do everything in one transactional request, rather than having the overhead of sending multiple requests to the server.

Unless you are doing something like customizing a model, you would normally use the scripts object and the gremlin object stored on the graph object:

>>> from bulbs.neo4jserver import Graph
>>> g = Graph() 
>>> script = g.scripts.get('some_script')
>>> params = dict(name="James", city="Dallas")
>>> g.gremlin.execute(script, params)

See the Bulbs Neo4j Gremlin docs...

• Neo4j Gremlin: http://bulbflow.com/docs/api/bulbs/neo4jserver/gremlin/

Likewise, when you want to execute a Neo4j Cypher query, use the cypher object stored on the graph object.

There are three Cypher methods (unfortunately these aren't documented on the website yet):

1. g.cypher.query(): Used when returning a list of nodes/relationships -- it will initialize them to objects.
2. g.cypher.table(): Used when returning Cypher table data.
3. g.cypher.exectue(): Used when returning arbitrary data -- it returns a generic Response object.

You can look at the source code to see how they work...

• Cypher Source: https://github.com/espeed/bulbs/blob/master/bulbs/neo4jserver/cypher.py

Here are some examples of using the Cypher query() method (the query simply returns a relationship):

>>> from bulbs.neo4jserver import Graph
>>> g = Graph() 
>>> query = "start a = relationship({eid}) return a"
>>> params = dict(eid=123) 
>>> edges = g.cypher.query(query, params)

The query method automatically initializes elements to their type. If you created the element as a custom model, Bulbs will try to initialize it to the specific type, otherwise it will default to a generic Vertex or Edge.

Note that the Bulbs Cypher query() method returns an iterator.

You can loop over the iterator...

>>> from bulbs.neo4jserver import Graph
>>> g = Graph() 
>>> query = "start a = relationship({eid}) return a"
>>> params = dict(eid=123) 
>>> edges = g.cypher.query(query, params)
>>> for edge in edges: print edge

...or convert it to a list...

>>> from bulbs.neo4jserver import Graph
>>> g = Graph() 
>>> query = "start a = relationship({eid}) return a"
>>> params = dict(eid=123) 
>>> edges = g.cypher.query(query, params)
>>> list(edges)

...or get the next item...

>>> from bulbs.neo4jserver import Graph
>>> g = Graph() 
>>> query = "start a = relationship({eid}) return a"
>>> params = dict(eid=123) 
>>> edges = g.cypher.query(query, params)
>>> edges.next()

Please let me know if you have any more questions.