performance of CytoscapeWeb 2 versus CytoscapeWeb 1

[Naveed]

Can someone speak a little about the performance of CytoscapeWeb 2 versus CytoscapeWeb 1. Thanks.

-Naveed

[Max Franz]

Cytoscape Web 2 is approximately on par with the Flash version, apart from in IE.  We've ensured full compatibility with IE, however.

In fact, a lot of operations have much better performance because they don't require passing JSON back and forth between Flash and JS.

On Wed, Mar 28, 2012 at 14:36, Naveed Massjouni <nave...@gmail.com> wrote:

Can someone speak a little about the performance of CytoscapeWeb 2 versus CytoscapeWeb 1. Thanks.

-Naveed

--
You received this message because you are subscribed to the Google Groups "cytoscapeweb" group.
To view this discussion on the web visit https://groups.google.com/d/msg/cytoscapeweb/-/KnLtMjHcoTgJ.
To post to this group, send email to cytosc...@googlegroups.com.
To unsubscribe from this group, send email to cytoscapeweb...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/cytoscapeweb?hl=en.

[Naveed]

Max, thanks. So the upper bound for visualizing very large graphs is going to be similar to CytoscapeWeb 1?

-Naveed

On Wednesday, March 28, 2012 3:21:31 PM UTC-4, Max Franz wrote:

Cytoscape Web 2 is approximately on par with the Flash version, apart from in IE.  We've ensured full compatibility with IE, however.

In fact, a lot of operations have much better performance because they don't require passing JSON back and forth between Flash and JS.

On Wed, Mar 28, 2012 at 14:36, Naveed Massjouni <nave...@gmail.com> wrote:

Can someone speak a little about the performance of CytoscapeWeb 2 versus CytoscapeWeb 1. Thanks.

-Naveed

--
You received this message because you are subscribed to the Google Groups "cytoscapeweb" group.
To view this discussion on the web visit https://groups.google.com/d/msg/cytoscapeweb/-/KnLtMjHcoTgJ.
To post to this group, send email to cytosc...@googlegroups.com.

To unsubscribe from this group, send email to cytoscapeweb+unsubscribe@googlegroups.com.

[Max Franz]

More or less.  Some things will be faster, some will be slower.  

Remember: The Flash version didn't support that many elements when compared to desktop apps like Cytoscape.  The limiting factor is web browsers.  They need to get faster.

It's good enough such that the cost of rewriting your code later to support the new API is much greater than the cost of differences in performance.

To view this discussion on the web visit https://groups.google.com/d/msg/cytoscapeweb/-/Ck9LjqDi2n0J.

To post to this group, send email to cytosc...@googlegroups.com.

To unsubscribe from this group, send email to cytoscapeweb...@googlegroups.com.

[Anish Kejariwal]

Nice job on cytoscapeweb.  Do you have a rough feel for the maximum number of nodes/edges that cytoscapeweb can handle in the force-directed layout?  Yes, I realize the number is probably browser dependent.  Thanks!

[Max Franz]

We haven't tested that extensively yet, because that really falls under the performance of the renderer:  The renderer is the bottleneck.

We started with a SVG renderer because it offers good cross-browser compatibility (iOS too) and event handling. However, that comes at the cost of performance: SVG < canvas < canvas with WebGL.

We'll almost certainly be writing a WebGL renderer, because it will give us much better performance.  Until then, I wouldn't worry too much about performance of the arbor.js layout.

Cytoscape Web 2 is really just the core API.  Everything else is extensions and plugins.  The renderer is just an extension that you can swap out.  This means that when the WebGL renderer is available, you can use it without having to change your code at all.

It's also important to mention that now with Cytoscape Web as an entirely JS+HTML5 project on GitHub, it's now very easy for any developer to contribute to Cytoscape Web directly.  If anyone's interested on contributing to the project, please contact me at maxk...@gmail.com.

To view this discussion on the web visit https://groups.google.com/d/msg/cytoscapeweb/-/3D-Zd2ureFcJ.

To post to this group, send email to cytosc...@googlegroups.com.

To unsubscribe from this group, send email to cytoscapeweb...@googlegroups.com.

[Liran Yogev]

So, you fully support IE? what versions?

[Max Franz]

Short version: IE9 works; older versions won't since they don't support SVG or canvas

Long version:

Internet Explorer 9 and newer support SVG, though SVG was initially chosen for first renderer based on ease of implementation and the lack of mature canvas/WebGL libraries.

We expect the next renderer to have very high performance.  However, old versions of IE are nearly impossible to support.  They don't have SVG or canvas support.

IE9 still doesn't implement everything it should -- and what it has implemented is dodgy at best.  The best bet is to use something like Chrome frame or redirecting your users to download Chrome, if you're interested in old versions of IE.

If you're interested, you could try to write some kind of custom, high-performance just for old versions of IE.  However, I don't expect that to go very well, since IE isn't known for performance even with its own proprietary technologies.  Canvas will work well for IE9+.

To close: We will support IE insofar as it doesn't require rewrites just for IE; i.e. there are minor workarounds in the SVG renderer to fix IE's problems, but we won't plan on IE-only or IE-centric renderers because (1) it goes against the philosophy of having standards on the web and (2) it's too expensive to develop and maintain.

To view this discussion on the web visit https://groups.google.com/d/msg/cytoscapeweb/-/DIwvPZMCBH4J.

To post to this group, send email to cytosc...@googlegroups.com.

To unsubscribe from this group, send email to cytoscapeweb...@googlegroups.com.