# Usecases for Jupyter kernels and notebooks

## Use cases

* Reproducible logbooks (aka lab notebooks) documenting a computer experimentation.
Ideally eventualy published along the research paper.
* Reproducible live slides [example](https://mybinder.org/v2/gh/nthiery/harmonic-modules/master?filepath=talk.ipynb)
* Interactive software [demos](https://sebasguts.github.io/oscar-website/example/)
* Interactive documentation (either as notebooks or live html page with ThebeLab)
* Teaching documents, possibly with auto-grading, ...
* Ease of software deployment (via the cloud)
* Widgets:
- data visualization (e.g. subgroup lattices, which then would supersede XGAP and make it usable on Windows)
- profiling / debugging / ... ?
- https://medium.com/ibm-watson-data-lab/the-visual-python-debugger-for-jupyter-notebooks-youve-always-wanted-761713babc62
- Explorer (guided introspection): type `explore(o)` and get an lmfdb-style page providing information about `o` (trivial prototype: [sage-explorer](https://github.com/jbandlow/sage-explorer); upcoming work on this by Odile in Orsay)
* "Lab notebook" for conducting research; keep notes, computations

## Benefits of Jupyter

- reducing entry barriers for new users by:
* providing "standard" interface which they may know from using other Jupyter kernels (Python, R, etc.)
* allowing them to easily organise their code and documentation into Jupyter notebooks
- ability to use full GAP installation on Windows by connecting to a Docker container (or a remote GAP?)
- Using JupyterKernel as a backend for other IDE's
- potentially using an equivalent of ipyparallel for parallel distributed calculations
- Benefits for reproducible research

## Inconveniences of using Jupyter

Missing features?
- input/output/error similar/identical to the GAP command line interface
- A notebook extension to force edition at the end of the notebook (it seems that what's exist already is changing the colour of cells below as soon as a cell in the middle is edited - this is already fairly good)
- break loops in Jupyter? How to do interactive debugging?

## Further directions in Jupyter developments from which we may benefit
* JupyterLab (a kind of an IDE)

## Technical questions sidetracking discussion - list them here to return to them later
---

* A standard UI for GAP
Problems to solve: Why is the current Jupyter interface so bad for "veteran users", which features do we need for "new users", what are new features which we want that "veteran" and "pro" users will miss?

* A distribution/deployment mechanism

* Teaching [GAP in Software Math course](https://github.com/pedritomelenas/Software-Matematicas-GAP), [Commutative Algebra](https://github.com/pedritomelenas/Algebra-conmutativa-y-computacional), [Logic and Discrete Math](https://github.com/pedritomelenas/LMD), [Graph and group theory](https://github.com/pedritomelenas/Algebra-II), [ThebeLab for high school study of functions](https://cursos-0-fc-ugr.github.io/Matematicas/derivadas-app-estudio-func-thebelab.html)

* Research [snip visualizer](https://github.com/mikecroucher/jupyter-interactions/blob/master/snp-view.ipynb) to browse DNA chains sequences

* Make some webpage with some GAP examples/tutorials like:
* [OSCARExamples](https://sebasguts.github.io/oscar-website/example/)
* [Jupyter-Interactions](https://mikecroucher.github.io/jupyter-interactions/)
* [Jupyter/Try](http://jupyter.org/try)

* Setup a gap-specific binder site people can run their notebooks on (Like Azure?)

* Blog posts

# Guides to Jupyter

* "Quick start" guide / best practices

# References

- The [Jupyter-education](https://groups.google.com/forum/#!overview) mailing list