D4.7: Full featured Jupyter interface for GAP, PARI/GP, Singular

[minrk]

• WP4: User Interfaces
• Lead Institution: Université Paris-Sud
• Due: 2017-08-31 (month 24)
• Nature: Other
• Task: T4.1 (T4.1: Uniform notebook interface for all interactive components #69)
• Proposal: p. 47
• Final report (sources)

For the basic Jupyter interface, see #93.

• GAP: http://gap-packages.github.io/JupyterKernel
  • support complete_request (tab completion)
  • support inspect_request (shift-tab in notebook)
  • provide some rich display of objects and/or plotting, as appropriate
  • reasonably well documented, installable, etc.
  • distributed with SageMath: https://trac.sagemath.org/ticket/21416
  • installed on CoCalc: jupyter GAP kernel sagemathinc/cocalc#1706
• PARI: https://github.com/jdemeyer/pari_jupyter
  • support complete_request (tab completion)
  • support inspect_request (shift-tab in notebook)
  • provide some rich display of objects and/or plotting, as appropriate
  • reasonably well documented, installable, etc.
  • distributed with SageMath: https://trac.sagemath.org/ticket/19372
  • installed on CoCalc: support Jeroen Demeyer's PARI/GP Jupyter kernel once it stabilizes sagemathinc/cocalc#160
• Singular: https://github.com/sebasguts/jupyter_kernel_singular
  • support complete_request (tab completion)
  • support inspect_request (shift-tab in notebook)
  • provide some rich display of objects and/or plotting, as appropriate
  • reasonably well documented, installable, etc.
  • distributed with SageMath: https://trac.sagemath.org/ticket/22378
  • installed on CoCalc: Singular Jupyter Kernel sagemathinc/cocalc#1553
• xeus-cling (C++)
  • installs on CoCalc as a user: https://share.cocalc.com/share/4711033a-b5d8-4fb3-9708-e9019ba1c158/Install-xeus-cling-in-CoCalc-project.md?viewer=share
  • installed on CoCalc system-wide: Install cling jupyter kernel for interactive C++ sagemathinc/cocalc#2324
• MMT; see D4.11: Notebook Import into MathHub.info (interactive display) #85.
  • widgets
• Syntax highlighting: Syntax highlighting for GP, GAP, Singular, ... #175

[defeo]

Trac ticket 20589 relates to this.

[minrk]

Month 14 as the due date for this deliverable is a typo in the proposal, and should be Month 24. This should be formally updated to M24 (@nthiery).

[jdemeyer]

Month 14 as the due date for this deliverable is a typo in the proposal, and should be Month 24. This should be formally updated to M24 (@nthiery).

ping (Nicolas?)

[bpilorget]

Ah yes... Well we will ask the Project Officer to take that into account. If Michael really does change partner we will need to to another amendment anyway...

[jdemeyer]

What should be done here? Of course 'full featured' is very ambiguous. The only thing I can immediately think of is syntax highlighting. But I know essentially nothing about Javascript, so I don't think it's a good idea for me to work on that.

[minrk]

I would say a full-featured Jupyter kernel means:

• syntax highlighting
• supports complete_request (tab completion)
• supports inspect_request (shift-tab in notebook)
• provides some rich display of objects and/or plotting, as appropriate
• is reasonably well documented, installable, etc.

omitting (with reason) any individual item, as appropriate.

[minrk]

I think the gist is: D4.4's goal is to show that it's possible and get the basics going, and D4.7 was to develop it the level of being useful, which is dependent on the expectations of the user community of each kernel. Based on the deliverable consolidation discussion, these probably shouldn't have been separate deliverables and I would merge them into "Jupyter kernels for GAP, PARI/GP, Singular" if D4.4 weren't already delivered.

[nthiery]

I would say a full-featured Jupyter kernel means: * [ ] syntax highlighting * [ ] supports complete_request (tab completion) * [ ] supports inspect_request (shift-tab in notebook) * [ ] provides some rich display of objects and/or plotting, as appropriate * [ ] is reasonably well documented, installable, etc. omitting (with reason) any individual item, as appropriate.

+1 I would add: distribution of the kernel with Sage (preferably as a standard package) and installation on CoCalc (possibly as a side effect of the former). It could be reasonable for the kernel to be included as well in the distribution of the relevant component itself. Possibly also some of the items at the end of the description of: #94 @jdemeyer: which of the above would be willing to handle? For the others (plausibly those requiring javascript work) we will need to find someone! Cheers, Nicolas

[videlec]

For the others (plausibly those requiring javascript work) we will need to find someone!

Vincent K (engineer at Bordeaux) is mostly done with the gmpy2 and pplpy integration in Sage. That might be a good continuation! I will ask him about working on the Pari kernel (starting with fixing issue 2 and issue 3). And possibly more kernels after that. Is that ok?

[embray]

I can help with all of the above if needed, though I know virtually nothing about GAP, PARI/GP, etc. beyond some basics of how they work. So if there are people with specific knowledge of what features would be useful in kernels for them (as in, the ones @minrk listed) I can work with them on the implementation.

[sebasguts]

Short update on the Singular kernel: Tab-completion is available, as well as documentation. I will give an update of the remaining points during this week.

[bpilorget]

@minrk @jdemeyer This deliverable is officially due for 31/08/2017. Can you please report on the progress you have achieved?
Thanks in advance

[sebasguts]

inspect_request will be possible in the Singular kernel in the near future. I am not sure I can make it until end of August, but I will try. Syntax-Highlighting might work until then, too.

There is the possibility to plot images, using surf. Does this count as rich output?

[nthiery]

Dear Jupyter kernel writers, Of course it's always good to look over each other shoulders for inspiration of what could be useful features. Yet each component (GAP/Singular/...) is different in its use cases. Rather than trying to tick boxes from a uniform list of should-have features, let's use this deliverable as an occasion to look at the feedback from users of each basic kernel, and devote some time to implement the most requested features. If this requires a couple more weeks, that's ok. By the way: it would be good to include in the report some quantitative and/or qualitative metrics of actual use of each kernel. Example of good qualitative metric: a brief mention of Wolfram's use of Jupyter for teaching Singular at AIMS, with a link to a blog post on the topic.

[jdemeyer]

let's use this deliverable as an occasion to look at the feedback from users of each basic kernel

I'm sorry to say this, but I do not know anybody actually using one of these kernels.

[jdemeyer]

There is the possibility to plot images, using surf

Well, modulo the installation issues regarding surf. For example, it was refused as optional Sage package because it is too hard to build on OS X.

The problem is that surf-in-Jupyter only needs a small subset of the full surf program, and it's precisely the parts which are not needed for Jupyter that contribute to the build difficulties.

Anyway, I don't know how relevant this is to this deliverable, but I thought that it should be mentioned.

[jdemeyer]

Somebody added xeus-cling on this page, but can we really consider that as part of ODK?

[slel]

I think it's ok, especially since

• Loïc Gouarin is part of the team behind xeus-cling, and conceivably
  devoted part of his OpenDreamKit funding to work on xeus-cling.

• I used OpenDreamKit funded time to work on integration in CoCalc.

[nthiery]

Somebody added xeus-cling on this page, but can we really consider that as part of ODK?

ODK indeed did not initiate it, but Loïc Gouarin made heavy contributions, and I contributed by early adoption (beta testing, bug reports, feature requests, ...).

[jdemeyer]

I just pushed a major update to the report. Should I add the generated PDF also in the git repo? I don't remember how this is supposed to work.

At the XFEL meeting, somebody mentioned the MMT kernel https://github.com/UniFormal/mmt_jupyter_kernel

I really don't know much about MMT or its user interface, so I don't think that I could write something about it.

[nthiery]

Thanks Jeroen! Don't worry about the PDF; we will add it at the last minute.

@florian-rabe: would you get a chance to write a few paragraphs describing the MMT kernel?

[florian-rabe]

You can copy-paste from https://github.com/KWARC/OpenDreamKit/tree/master/WP4/D4.11
I wrote that this week.

[jdemeyer]

Nicolas, you mentioned that I should add widgets but this is already covered by D4.5. It seems strange to me to add that feature to two deliverables.

[nthiery]

Nicolas, you mentioned that I should add widgets but this is already covered by D4.5. It seems strange to me to add that feature to two deliverables, especially in the same reporting period.

Ah, right, good point. Nevertheless, it's still worthwhile mentioning with a cross-reference "implemented as part of D4.5", to give a bit of the big picture to the reader. Btw: on #94 the item for interacts/widgets is not checked; any reason? Otherwise go ahead an check it.

[nthiery]

I added a brief section about MMT (@florian-rabe: you may want to double-check), and double checked that the work was widgets work was indeed referenced as implemented as part of D4.5.
@jdemeyer: do you see anything else you'd like to change, or shall I submit the report as is?

[florian-rabe]

I made minor changes to the MMT paragraph. OK otherwise.

[nthiery]

Thanks @florian-rabe!
@jdmeyer: should I submit the report as is?

[jdemeyer]

I thought you wanted me to add something about widgets.

[nthiery]

Yeah; in the mean time, I realized that it was already there in Section 4.7; sorry for that noise:

- ...
- Improvements to Jupyter’s widgets and interact functionality to be as feature rich as in
the old Sage Notebook.
The last item was implemented by Jeroen Demeyer under OpenDreamKit funding as part of
D4.5: “Sage notebook / JUPYTER notebook convergence”.

Unless of course you see something more specific to say!

[olexandr-konovalov]

I've a collection of the some GAP Jupyter notebooks at https://github.com/alex-konovalov/gap-teaching/ which I used at the GAP tutorial at PGTC 2018. One could also run on binder. Is this interesting for this report?

[nthiery]

I've a collection of the some GAP Jupyter notebooks at [1]https://github.com/alex-konovalov/gap-teaching/ which I used at the GAP tutorial at [2]PGTC 2018. One could also run on binder. Is this interesting for this report?

Thanks for the suggestion Alex! Could you add a few lines about it around: https://github.com/OpenDreamKit/OpenDreamKit/blob/master/WP4/D4.7/report.tex#L111 We should also have a use case around authoring interactive course notes on our web page, with links to instances including the above: https://opendreamkit.org/tag/use-case Volunteers welcome :-)

[nthiery]

I integrated Alex's suggestion, and submitted.

Jeroen: thank you so much for taking the lead on this deliverable; thanks @sebasguts for contributing!
And of course thanks to all our kernel writers: we now have a pretty wide array of computational math software that are all using the same user interface (plus all the rest that Jupyter brings). Would you have just imagined this five years back in time? That's really really nice!

[slel]

For greater visibility, it would make sense to add binder-enabled examples
of using the GAP, MMT, PARI/GP, SageMath, Singular kernels for Jupyter at
https://github.com/jupyter/jupyter.github.io/blob/master/_data/try.yml
so they can appear at
http://jupyter.org/try

[slel]

Making everything conda-installable from the conda-forge channel
would also facilitate adoption.

See #278.

[IzabelaFaguet]

Accepted by the EU on December 5th.