QT or html5?

[tbates]

Hi, this is a bit off base. But given the increasing difficulty of installing QT and its widgets, I wonder if it's worth thinking of implementing cranvas in html5 with libraries like D3 or Raphaël.

The clickme package is a nice example of the kinds of fluid processing and display graphics that is doable there.

The benefit is that everyone has a super-fast and uptodate cross platform browser and JScript implementation, so no additional installations. The audience is vastly larger than people willing to install QT via XCode etc. and then maintain that.

Plus all the CSS and other goodness of html5 for making things look good.

[yihui]

There are at least two reasons to stay in R/Qt:

1. JS is indeed becoming faster (e.g. the V8 engine is impressively fast), but at the moment I do not see it is ready to visualize millions of points in the browser (try a D3 scatterplot with ten thousand points and my Chrome will freeze);
2. there are too many things in R to be reinvented in JS such as the binning algorithms for histograms and the grand tour, etc

You are right that the installation is still a big problem to us. We are waiting for Qt 5.1. Hopefully by then someone will pre-build the binary packages for Mac and Windows.

[tbates]

OK, I grok not re-writing R in JS... I wonder, then if an html-based window reading and writing to R via server(), with all the display work done by R, displayed by writing double-buffered updates to or , and then with an overlay of GUI elements/mouse tracking/event handling in css/html5, which feed back into the data pipeline/plumbr?

i.e., just replace the QT widgets with html5, and open a pathway to R for responding to UI events, and requesting image refreshes for updated output. Sorry to bother if this is off topic...

[yihui]

This is not entirely off-topic. It is certainly useful if we can port cranvas to the web (and it has been asked for a few times), but I really do not know how. It is a question for our Qt master Michael :)