Programmatically create markdown and jupyter notebook from a share dictionary with questions and answers #89
Conversation
SebastianoF
changed the title
|
Amazing work! Thanks. Do you think we could have a regular md file as the main source for generating others ? I'm not too fan of heaving actual contents in a Python file? |
|
Welcome and sorry for the late answer. Uhm... it is certainly possible to use markdown as source, although I do not think it would be optimal to have a markup language for data serialisation. This would imply to have the need of marshalling the .md into a structure like the one in the .py and the unmarshalling in back into the jupyter notebook, or other files that we want (restructuredText, HTML, docuwiki, latex...?). What about storing the data into a json file? It would be already serialized, and ready to become anything we like in few lines of code (including markdown). The structure would remain simple and would contain:
|
|
json might be a better option, yes. Also, it would be good to nice to keep the original file names because I suspect there are some people that link directly to this or that version. Ideally, you could only add the json file as well as the generator and left other files (more or less) untouched. |
|
It seems there is a limitation in how to the newlines are parsed in the json, which make json not ideal as a source format for python code, whose text and indent are critical. |
|
The most direct solution I could find, alternative to the json format, was to define a "text file with keys", that can be human readable (and modifiable), preserves the indentation, and can be seamlessly parsed into python dictionaries. This is then parsed into markdown, and jupyter notebook. Not sure is the best idea, and suggestions are always welcome! |
|
(also I had moved to same filenames as the master, although I would not advocate for starting names of executable files with numbers and mixing camelcase and snake conventions). |
100_Numpy_exercises.md
Outdated
| import numpy as np | ||
| ``` | ||
| If you find an error or think you've a better way to solve some of them, feel | ||
| free to open an issue at <https://github.com/rougier/numpy-100>File automatically generated. See the documentation to update questions/answers/hints programmatically. |
There was a problem hiding this comment.
There's a trailing "File automatically generated" here
|
I've merged two small PRs, that might be the reason for the conflict. Sorry for that. |
|
No problem! Hopefully this PR would make the next updates more straightforward, as only one text file will need to be changed. |
|
Thanks. I'm merging. There are some PR pending, I'll redirect them to the new format. |
|
Just forgot. Maybe we now need a |
Over the past months I had found numpy 100 very handy for learning, and as a source of information.
At the same time it is not ideally implemented. If a question or answer needs an updated, this must be done manually in all files involved.
To solve this issue I had embedded headers, questions, hints and answer in a python file as strings and dictionaries and created a method to automatically generate the jupyter notebooks and the markdown files.
Changing the source dictionaries and stings would change the output files.
E.g: if a question needs update, just update it under -
data_source.py-source/questions.ktxand then re-create all the files with the updated questions with:Also, there is no more need to have several jupyter notebooks, since hints and answers can be queried programmatically with
hint(n)andanswer(n)from within the jupyter notebook.A jupyter notebook where to query a random question (flashcard style) had been added too.
Pleaese note: binder and link in the readme may need to be updated, to keep the binder version in sync with the current repo!
After merging the repo would look like: https://github.com/SebastianoF/numpy-100/tree/dev