Running nbtutor with kernels other than a Python kernel

[psychemedia]

I notice that the original Python Tutor application can be used with with "Python, Java, JavaScript, TypeScript, Ruby, C, and C++" programmes, though I'm not sure how it actually works...

Which got me wondering - will this plugin be able to work with other kernels? If so, to what extent might the approach generalise to other language kernels too, such as R?

[lgpage]

At the moment no, it will not be able to work with other kernels. To be honest this is not something I have looked into yet, but it is something I would like to incorporate at some point.

The way this extension works at the moment is: On the kernels side:

1. Trace through the cells code line-by-line
2. Generate a json object (metadata) of stack frame data
3. Send that metadata to the front end

On the front end side:

• Once new metadata is received from the kernel, find the correspond cell
• Create the html elements representing the names and objects
  • Moving from line to line in the code is just the stepping though of this metadata and the re-creating of the html elements

I will have to get back to you regarding what exactly is need to support R. Off the top of my head:

• As far as I am aware the R kernel does not have cell magics, so code would need to be sent to a function that implements 1 to 3.
• 3 is currently done using comms, which I do not know if the R kernel has an equivalent for or not.
• For the front end the notable changes would be the need to document and standardize the metadata it receives and of course different visualizations for the different languages. I have never used R before so I don't know how its memory model looks.

[nthiery]

I just stumbled on nbtutor through an e-mail an jupyter-education. That sounds great for teaching!

My main class is introduction to programming in C++, using xeus-cling, so I'd be very much interested in support for this kernel. To be more specific about the scope, we use and would like to graphically debug programs involving:

• basic types, vectors, strings
• control structures, functions

@SylvainCorlay: any insight of what it would take to split this extension into a kernel agnostic part and primitives to be implemented in the each kernel? @gouarin: you may be interested too

Cheers,

[Naereen]

Hi there, @nthiery I have the same need as you, but for C using jupyter-c-kernel (or maybe xeus-cling too)

From what I understand, this project was not using the PythonTutor website, am I right @lgpage ? So there is no easy way to extend this jupyter nbtutor extension to non-Python kernel.

A possibility could be to see what was discussed https://github.com/kikocorreoso/tutormagic/issues/8 and use a recent open-source fork version of PythonTutor, such as @seamile/PyTutor @seamile if you're interested in joing the discussion?

Cheers.

[Naereen]

Ok, I just tried by switching offline/online, or being online and disabling just pythontutor.com from my DNS, and it stops working as soon as the pythontutor.com website is unavailable. So I was wrong when claiming

From what I understand, this project was not using the PythonTutor website, am I right @lgpage ? So there is no easy way to extend this jupyter nbtutor extension to non-Python kernel.

• Bad news, it can't be used offline!
• Good news, it means that extending it to use all the other modes of PythonTutor.com should not be too hard!

@lgpage if you are interested in trying this, see https://github.com/kikocorreoso/tutormagic/issues/10 et https://github.com/kikocorreoso/tutormagic/pull/11 for the similar project tutormagic (just one hour ago);

Cheers!

[Naereen]

Just FYI, currently PythonTutor.com accepts the following mode: python2, python3, py3anaconda, java, javascript, typescript, ruby, c, c++