Fall 2018 Wishlist

[milancurcic]

Topics requested by the attendees:

• Python for complete beginners
• Python + Dask for parallel data analysis
• Getting started with cloud computing (AWS or DigitalOcean)

[tibbben]

I can do python for absolute beginners (as the software carpentry style introduction) and help with python parallel using dask and concurrent futures ...

[raybellwaves]

@tibbben thanks. I did some basic python today. If you could do dask that would be awesome! I think there is plenty of material in https://github.com/dask/dask-tutorial, https://github.com/dask/dask-examples and https://github.com/martindurant/dask-tutorial-scipy-2018 which should be good to go out of the box. I didn't have time to do dask-jobqueue today so you are also welcome to present that (https://github.com/milancurcic/lunch-bytes/blob/master/Fall_2018/LB14/py_intro_lb_nb_parallel.ipynb)

[raybellwaves]

Very excited for the cloud computing one

[milancurcic]

@tibben That would be fantastic. @raybellwaves did good job today with getting started with Python on Pegasus, but I think the bar was pretty high and mostly an overview. I think we'd still need a dedicated seminar with focus on programming and the language itself.

I like the idea of @tibben doing Python for absolute beginners, and @raybellwaves and @tibben teaming up on Dask.

Now that there seems to be demand for cloud computing, I will be happy to do that one, though later in the semester. Earlier, I will probably do one on git + Github, and perhaps one on Markdown.

A strong recommendation when designing a seminar: If you want to teach a topic and people to follow, please stick to one topic! If you want to do an overview of 5 topics, that's OK, but don't try to teach them in detail and expect people to follow.

[milancurcic]

There was also a question and interest in a Python-specific course. I mentioned Software Carpentry, but I think that happens twice a year. Perhaps we can start talking about designing a mini-course as part of Lunch Bytes, with something like:

• Python
• numpy
• matplotlib
• notebooks
• pandas, xarray, dask, and similar

[milancurcic]

We got a request for a LaTeX seminar.

[raybellwaves]

Bash/shell scripting? Awk, sed, find, grep. Maybe covering CDO and NCO as well? Software carpentry have a nice lesson on this http://swcarpentry.github.io/shell-novice/

[raybellwaves]

Some git material here: https://github.com/AllenDowney/amgit

[raybellwaves]

Probably on the periphery but testing/continuous integration/developing software. e.g. pytest, flake8, Travis CI, conda forge.

[milancurcic]

@raybellwaves Are these topics you'd like to learn or a suggestion for a seminar to teach? On their own they seem very specific and perhaps advanced.

Scientific DevOps (I saw some people call it SciDevOps) is something I haven't seen a seminar about and I think most people are not aware is a thing. Maybe a good topic for an overview, with examples in Github or GitLab.

[milancurcic]

I really like that idea (testing/continuous integration/developing software) if we can somehow frame it in a format that does not depend on specific tools, but can use specific examples

[milancurcic]

Bash/shell scripting? Awk, sed, find, grep. Maybe covering CDO and NCO as well?

These are good and important topics. I wonder how much people use stuff like bash+awk+sed nowadays. Myself, as soon as I have to write a loop or if-block in shell I switch to Python. It may be redundant.

I would really like to hear a seminar on NCO. Not much material out there.

Dian did a seminar on CDO: https://github.com/milancurcic/lunch-bytes/tree/master/Fall_2015/LB02

If you can find a good speaker, I'm all for doing seminars on some topics that we already had in the past. We are now catering to a new generation of students anyway.

[raybellwaves]

@raybellwaves Are these topics you'd like to learn or a suggestion for a seminar to teach? On their own they seem very specific and perhaps advanced.

I would like to learn. I have experience with python software development but it's all self taught along the way. SciDevOps would be useful. It could take a while to setup but you could have an exercise where people work out how to contribute code to the same repo and work on different parts (I will see if some kind of lesson exists like this; may overlap with the git lesson).

I wonder how much people use stuff like bash+awk+sed nowadays.

Very true. xarray and other packages in https://github.com/pydata/xarray/blob/master/doc/related-projects.rst have the capability of CDO and NCO. Whilst the performance may not quite be quite as good, the python implementation is more user friendly.

[tibbben]

Bash/shell scripting? Awk, sed, find, grep. Maybe covering CDO and NCO as well?

These are good and important topics. I wonder how much people use stuff like bash+awk+sed nowadays. Myself, as soon as I have to write a loop or if-block in shell I switch to Python. It may be redundant.

perhaps still useful when on remote machines ...

As a total aside, perhaps we should have a virtual meeting sometime just to coordinate a little bit? Especially if I am to help out with some of these.

[milancurcic]

I'll set up a Doodle thing..

[tibbben]

I have used this for a software carpentry style introduction to parallel computing with python, https://github.com/zonca/swcarpentry_inflammation_hpc

It includes a little dask and a cool module called concurrent_futures

[milancurcic]

We got another request for a seminar on xarray.

I'm starting to think if we should design a survey. A topic for our meeting tomorrow.

[tibbben]

Possible seminars: Paul Clough (library) - virtual environments

[tibbben]

Abraham Parrish - GIS library Cameron Riopelle - Stats and data Library

[raybellwaves]

some good stuff here https://carpentrieslab.github.io/python-aos-lesson/