I have tried both. Jupyter is fine for demonstrations of small snippets of code that fit on a screen. It is fantastic for e.g. pandas and plotting. But as soon you have to scroll back and forth in the browser to relate different part of code to each other, I get lost and so do the students. Great for experimentation but not for the development skills we try to teach.
Cheers Olav On Wed, May 10, 2017 at 9:45 PM, Eric Jankowski < [email protected]> wrote: > I've taught both ways and lean towards Jupyter for shorter-contact > workshops, and lean towards CLI for longer-contact workshops and classes. > > Jupyter notebooks lower the cognitive load associated with CLI, text > editors, edit-save-execute iterations, and are a lower barrier to providing > a standardized implementation (e.g., Anaconda on learner's machines or a > public-facing notebook server you control). With the lower load, it's > possible to go from 0 to image processing in no time, and that time saved I > think contributes substantially to the take-aways of our learners in short > workshops. "Wow! I did some real data wrangling there, and I can use this > on my work today!" I also *really* like having the code and the > presentation of data in one place, so when I work with students now it's > easy to fiddle with things in plots in the notebooks which significantly > accelerates our group understanding of what our data means. (as compared to > "plot it a different way and see you next week") > > My minor qualm is that the notebooks don't provide exposure to the CLI and > the very common workflows that combine loops, pipes, python scripts, and > other unixy tools. In the context of my semester-long courses we'll start > with CLI, python, and git, and maybe 5 weeks in show "Here are jupyter > notebooks, which are a really great compliment to what we've learned so > far!". > And in terms of keeping track of code used by, managing jupyter notebooks > with version control can be tricky. > > So, the quick version: Jupyter notebooks are amazing, but in some ways > orthogonal to the automation, reproducibility, and sharing ecosystem that > software carpentry has been thoughtful about making self-consistent. > > Talk permissions granted! > > Best, > Eric > > On Wed, May 10, 2017 at 1:03 PM, Maneesha Sane <[email protected]> > wrote: > >> Hi all, >> >> I am giving two talks next week at Bryn Mawr's Jupyter days conference ( >> http://jupyterday.blogs.brynmawr.edu/) >> >> One is running an intro to Python workshop using Jupyter and I'm piecing >> together parts of the SWC and DC curricula to teach it. >> >> The other talk is a shorter talk (~20 minutes, incuding Q&A) about the >> differences teaching Python via straight command line compared to teaching >> from the Jupyter notebook. When I've taught for the Carpentries, I've >> always used Jupyter. In other contexts, I've taught strictly from the >> CLI. I'm curious to know if others have taught in both ways (either for >> the Carpentries or in other circumstances) and what you've thought of it. >> Advantages/disadvantages? What you like and don't like about Jupyter and >> CLI environments? >> >> With your permission, I'd like to use some of your feedback in my talk. >> I also think it would be useful for Carpentries instructors to know in >> general, so please share any feedback to the list. >> >> Thanks! >> >> >> >> -- >> Maneesha Sane >> Software Carpentry: https://software-carpentry.org/ >> Data Carpentry: http://www.datacarpentry.org/ >> >> >> _______________________________________________ >> Discuss mailing list >> [email protected] >> http://lists.software-carpentry.org/listinfo/discuss >> > > > _______________________________________________ > Discuss mailing list > [email protected] > http://lists.software-carpentry.org/listinfo/discuss >
_______________________________________________ Discuss mailing list [email protected] http://lists.software-carpentry.org/listinfo/discuss
