Graph Stream Server

Enhancement of https://github.com/panisson/pygephi_graphstreaming to make a simple Graph Stream Server.

Main change is to use Flask Framework instead of the low-level Python to manage HTTP Streams.

When the server receive a stream event , it will broadcast this event to all current Gephi Client connected.

Requierments

• Python
• See req.txt for libraries

Install & Run

Git clone the project or Download it.

Run python graphStreamServer.py.

The default configuration bind to 0.0.0.0:5000. The server should be accessible as :

• http://localhost:5000/ : Good for test
• http://<yourIP>:5000/ : Good to share across same network
• http://<hostname>:5000/ : Good to share acress internet

Adapt examples bellow to your situation.

Subscribe in Gephi

You already have the Graph Streaming Plug-in installed in Gephi.

You already started a Graph Stream Server on at http://localhost:5000/

In the Streaming tab, click-right on Client and choose Connect to Stream.

On source Url, put http://localhost:5000/stream and click Ok.

On the Streaming tab, you should see your Url with a Green Point.

Push Web Services

The server by itself isn't generating any data. It exposes Web Services that can be used to send graph data that will be broadcasted to all current listeners.

Because it's based on REST web service, there is no restriction about the language, as long as it can handle HTTP Request .

Full description of data format : https://wiki.gephi.org/index.php/Graph_Streaming

POST http://localhost:5000/<actionToPerform>

This service is a wrapper of the http://localhost:5000/action service to automatically handle the action overhead directly on the service. It makes the data to send more accessible.

• Url Parameter <actionToPerform>
  • an : Add Node
  • cn : Change Node
  • dn : Delete Node
  • ae : Add Edge
  • ce : Change Edge
  • de : Delete Edge

Request Data :

	{"<entityID>":
		{
			"<property1>":"<stringValue>",
			"<property2>":<numericValue>,
			[....]
		}
	}

Example : Add a node with id "Node A"

• Url: http://localhost:5000/an
• Data: {"Node A":{"label":"Node A"}}

POST http://localhost:5000/action

This service is more close to the original streaming behaviour.

Request Data :

{"<actionToPerform>":
	{"<entityID>":
		{
			"<property1>":"<stringValue>",
			"<property2>":<numericValue>,
			[....]
		}
	}
}

Example : Add a node with id "Node A"

• Url: http://localhost:5000/action
• Data: {"an":{"Node A":{"label":"Node A"}}}

Just thinking

Endless possibilites of advanced Pub/Sub mechanism with Redis or RabbitMQ coupled with the Graph Stream Server (tested with Redis and works perfectly).

Possibility to plug it into a Web Socket / WAMP.ws mainly to dispatch to web browser.

Security enhancement, History when (re)connected, etc ...

Did not perform any " Scaling / Stress Test ". Let's organise a worldwide test one day.