Open callummole opened 1 year ago
Work on this going on in https://github.com/alan-turing-institute/rds-course/pull/126.
Have initial proof of concept - generated slides are embedded and can be full-screened. I also tested whether you can toggle a cell to be present in the book build only, the slides only, or both (you can)
To do:
Brain dump of thoughts:
We have both code-based and text-based material.
For the text-based material, I agree (properly formatted/prepared) slides would be better to present, but we need the verbose material for the course web-site and people following asynchronously. The main problem I see is the potential added cost of maintaining both.
For the code-based material, I see more challenges, particularly for:
And would this require a more explicit separation between "code-based" sections/files and "text-based" sections/files?
Instructors running the code, maybe modifying/adding cells to answer questions etc. - would that work in slides?
This is part of the beauty of having rendered slides - we can have Slide_0 be the code and Slide_1 show output without risking the inevitable embarrassments that come with live code executions.
Sorry, I misread this initially, it is a good point and probably comes down to the person delivering it and their style of teaching. Having the embedded Binder button would most likely be good enough for this, though?
Some participants like to follow along running the code with the instructor, which may be more difficult if the slides don't map onto the notebooks exactly.
I entirely agree, I don't think we should have slides which do not exactly follow the order of code blocks, and do not omit any.
And would this require a more explicit separation between "code-based" sections/files and "text-based" sections/files?
From reading through the course I think there is a degree of this already, where a lot of the text is superfluous and doesn't add much for the majority of learners who (In my biased opinion) are likely to move from src block to src block without reading unless something breaks. - My current working branch is experimenting with streamlining the amount of text by ensuring it's directly relevant and adding to the code it surrounds.
There are two competing requirements for this course that don't play well with each other:
We have investigated embedding slides using reveal.js, for examples:
The idea here is that you would have an iframe that links to some static html output, generated by, for example, pandoc or nbconvert (
jupyter nbconvert "4.1_What_and_Why.ipynb" --to slides
). You could set up a CI that executes these in a loop then the iframe picks up the resulting html files.A few obstacles: