nbviewer support about interact.jl HOT 8 CLOSED

I don't see how this could be possible with the current framework, because when the widget is changed (e.g. a slider is dragged), it calls Julia code to compute the new value to be displayed. On nbviewer, there is no Julia code running.

A similar limitation should apply to Python notebooks: there, it is Python code being called to compute the new object to display, and similarly nbviewer is not running Python. Unless IPython implements some kind of caching here?

from interact.jl.

Yeah, I think there would need to be a different macro for this, like @manipulatecached. It would basically call the code for all the various combinations of settings, saving the output, and then using javascript to switch between the saved outputs.

from interact.jl.

You're looking for something like Shiny in R? It would be nice to have that, and have it share a similar interface to Interact, but it's a pretty big project since it would have to be implemented either as a pretty elaborate caching scheme on the ipython side, or an entirely separate mechanism on the julia side.

from interact.jl.

@dcjones It looks like Shiny is client-server (which is pretty cool by the way), but I was thinking we could just save a sequence of outputs from writemime for the different parameter settings. Then the javascript control would just display the relevant output.

from interact.jl.

Right, for some reason I was under the impression they also had the ability to cache output like you describe. Now I'm not sure why I thought that. I looks like it does require an R process.

from interact.jl.

I'm pretty sure Shiny always runs R code on a server to do updates.

from interact.jl.

Depending on your level of patience, I would recommend waiting for static widget support in IPython (ipython/ipython#5900), which will work on nbviewer without you needing to do anything other than support vanilla widget messages.

from interact.jl.

Thanks @minrk, that sounds great. I'm going to close this issue in favor of the IPython issue.

from interact.jl.