C++ implementation of the Jupyter kernel protocol • Hacker News Board

Re: C++ implementation of the Jupyter kernel protocol

#11

Post by pjmlp » Tue, Feb 13, 2018, 6:01 PM UTC

eesmith wrote: post #10

pjmlp wrote: post #9

Earlier quoted context omitted.

Maybe you also would like to see the Energize C++ environment created by Lucid, after they pivoted away from Lisp Machines. https://www.youtube.com/watch?v=pQQTScuApWk http://www.dreamsongs.com/Files/Energize.pdf

I did, but while it looked like a nice IDE (as an aside, I was a Lucid Emacs user), what amazed me about the Jupyter example was the interactive C++ interface with GUI display of objects.

Yes, that is pretty neat.

Re: C++ implementation of the Jupyter kernel protocol

#12

Post by jmabille » Thu, Feb 15, 2018, 4:24 AM UTC

bfelds wrote: post #5

d33 wrote: post #3

Interesting. Are there any examples of this library being used so far? Can anybody familiar with the topic elaborate on how much can it actually simplify things and what's left to figure out?

This is an initiative backed by Bloomberg. It was pretty trivial to make a C++ ( https://github.com/QuantStack/xeus-cling ) and R ( https://github.com/JuniperKernel/JuniperKernel ) kernel around it. It's pretty easy to use, the biggest work pending is in the widgets system for charting.

We're working on it: widgets (https://github.com/QuantStack/xwidgets) and bqplot (https://github.com/QuantStack/xplot)

Also in the roadmap, a C+ backend for threejs and ipyvolume.

Re: C++ implementation of the Jupyter kernel protocol

#13

Post by alphaIuGN59 » Fri, Feb 16, 2018, 6:50 PM UTC

tavert wrote: post #7

alphaIuGN59 wrote: post #6

I wonder what was the reason this was written in C++, not in Python or other interpreted languages? Is speed a concern here?

Embed-ability, distribution. Using Jupyter from R, Julia, etc generally has an awkward dependency on python for the notebook server.

thanks