Gephi on the web

[mbastian]

Hi all, the ability to bring network visualization to the web is a key feature and we want to work in this direction. The problem is complex however and we may try different solution.

We observed the following technology limitations

- Flash is good because interactivity can be easily done but performance are limited over 2000 nodes
- Java applets and Processing has integration issues due to Java in browser which is not the best. Processing performance are also limited, though better than Flash.
- Web canvas performance needs to be tested

New solutions appears and we are eager to try them:
- Using WebGL may solve display performance issues easily but the development time is slighty more important. We proposed it this year for the Google Summer of Code but finally decided to wait one more year to profit from the library improvements and support.
- Seadragon, the best solution I see so far. There is no network size limitations with this, but it's harder to add interactivity. With the Javascript layer it would however be possible to propose search and click on nodes.

Please join the discussion!

[jbilcke]

Hi everybody!

I've had indeed various integration issues with my Java application, despite the fact I use Processing so I can switch to the 100% software implementation (to avoid issues regarding native libraries and OpenGL). Typically, I've had, and still have issues on Macs, because of the variety of JVM versions on this platform.

I tried a couple of pure-web solutions:

- Processing.js (pixel-canvas, great for shading, Processing compatibility)
- RaphaelJS (SVG-canvas, great for curved edges and flash-like animations, by example)
- WebWorker (safe multi-threading for JavaScript) for data processing (like spatialization or filtering)

- First, about the WebWorkers, this is an interesting way to achieve good processing performances, without blocking the interface. I've used them a bit and it works as expected (these real threads that do not share states but instead do exchange JSON messages, catched asynchronously by JS callbacks), no performance issue here.

- Now, about Processing.js and Raphael : they have a cool "web-oriented" design. You can control them from your favorite JS framework (like JQuery), which means great flexibility (ajax, DOM event-based programming). But unfortunately, canvas rendering is slow. It just don't scale (to the graph size, to the screen resolution..).

- On the other side, as Mathieu pointed it out, there is WebGL. Although it is still in its infancy, is an interesting emergent technology, that could be used for a cutting edge web-based Gephi. Creating such WebGL-based graph visualization library could be a great strategical investment.

- Finally, about SeaDragon AJAX : I agree with Mathieu on it. I used it too in the past. It is a lightweight solution, the same kind of technology used by Google Maps. Yes, it is lead by Microsoft, but it is still alive, and works on every platform. It can show very large graph, and could be extended with layers and JavaScript animations or events. This is probably the best short-term solution.

[elijah]

I use Flash to deliver anything that requires performance and support across multiple platforms but I think WebGL is the best bet for displaying high quantities of information (which is, in my opinion, a much more pressing problem than presenting beautiful but data shallow visualizations).

[jbilcke]

I've been looking for the past weeks for WebGL stuff.

As expected, we are starting to see a lot of libraries for doing 3D stuff, working on top of the WebGL API.

Some are high-levels "3D engines", other are low-level.
Some are dedicated to video games, or "load one 3DS model, view it, rotate. done." but aren't very adapted for advanced usage (eg. you are not supposed to a hack or add features inside).
Some can abstract the rendering to use either WebGL or the more-compatible software canvas (remember: some browser accelerate canvas on the GPU too).
Some have a nice "JS-like" API (JSON, asynchronous..), some don't (this is not necessarily a good idea, since this can limit feature support, or add overhead with useless abstractions layers. remember: we are doing javascript!)

3D PaperVision for JS/WebGL
http://code.google.com/p/pvjswebgl/

Canvas 3D JS Library (C3DL)
http://www.c3dl.org/

CopperLicht
http://www.ambiera.com/copperlicht/

GLGE
http://www.glge.org/tag/html5/

Processing.JS "WebGL branch" by asalga
http://asalga.wordpress.com/category/programming/webgl/

(and it's personal project: XB PointStream: http://asalga.wordpress.com/2010/06/15/ ... lease-0-1/ )

SceneJS
http://scenejs.org/

SpiderGL
http://spidergl.org/

Three.js
http://github.com/mrdoob/three.js

WebGLU
http://github.com/OneGeek/WebGLU

[slarson]

Hi,

Looks like FireFox's new beta has WebGL support:

http://blog.mozilla.com/blog/2010/11/10 ... e-browser/

Best,
Stephen

[jbilcke]

Some others libraries for WebGL stuff:

EnergizeGL
http://energize.cc/


X3DOM
http://www.x3dom.org/
Comment: Only to read static files / scenes? more a viewer lib than a framework to builds apps?


GWTGL - Google Web Toolkit + WebGL
http://code.google.com/p/gwtgl/
Description: "GwtGL enables you to develop 3D websites and web applications using the Java programming language instead of JavaScript. Type safety and the Java Toolset help you to write your apps in less time compared to plain JavaScript development. The GWT compiler generates optimized JavaScript code, so you will achieve higher frame rates."


Webglet
http://www.oampo.co.uk/2010/10/webglet/
Comment: Apparently, good for basic applications / prototypes?

[hodoman]

Hello,

Does anyone know of any web implementation of Gephi as a Java applet?

Thanks

[admin]

Hi,

It is no longer possible.

[carolinemeyer]

Hi

I have read about the google summer of code project to bring Gephi to the web. I was wondering what the status of the web implementation currently is.

[admin]

An alpha version is available here:
https://github.com/uskudnik/GraphGL

Unfortunately the development is stopped.