Orgdown code blocks aren't exported as code blocks to Jupyter

[morgandavidson]

I'm using Pandoc 2.19.1 on Emacs 28.1 through Pandoc-mode 2.31. OS is Ubuntu 20.04.

When converting a notebook from an org file to an ipynb file the notebook is converted into a single markdown cell instead of multiple cells containing markdown and code.

For example the following org file :

* Load HealthCoach dataset
  - Parse date column as dates
  - Use data column as index
  - [[../../../Logbook/Health/Measurements/weight.org][Weight DB]]
  #+begin_src jupyter-python :session py :display plain 
  import pandas as pd 
  df = pd.read_csv('../../../Logbook/Health/Measurements/weight.csv', parse_dates=['Date'],index_col=0)
  #+end_src

Will be converted into a single markdown cell in ipynb, instead of 1 markdown cell for the text part and 1 code cell for the code part.

[tarleb]

Pandoc doesn't if the code should be part of the text cell or should be treated as a separate code cell. Currently all code blocks with class code will be converted to code cells. You can use a Lua filter like the below to add the class:

function CodeBlock (cb)
  if cb.classes[1] == 'jupyter-python' then
    cb.classes:insert 'code'
    return cb
  end
end

We could make this the default if there is any agreed-upon way to mark code cells in orgmode.

[morgandavidson]

Thanks for the tip. Indeed it would be great to make this default. Actually, the standard way of adding code blocks in org-mode is to start with #+BEGIN_SRC and to end with #+END_SRC, as I did in my example. See Working with Source Code from Org-Mode manual. Do you need more info?

[tarleb]

I guess my question is if there's a standard way of telling apart src blocks that are code cells and src blocks that are part of a text cell. Your example uses jupyter-python as the block's language. Is that a general convention that's documented somewhere?

[morgandavidson]

In my example #+BEGIN_SRC and #+END_SR indicate that what is between these two terms is a code block. Whereas jupyter-python means that the code is to be interpreted with a Jupyter kernel.

So what Pandoc should understand is that everything below a line starting with #+BEGIN_SR and under a line starting with #+END_SR is a code block to be converted into a code cell.

So none of the content starting with #+BEGIN_SR and ending with #+END_SR should be interpreted as a text cell by Pandoc.

Which explanation are you looking for that is not already mentioned in the link from my previous comment?

[tarleb]

I found the answer to my questions: it appears that you are using the emacs-ipython package. The use of jupyter-LANG is documented there. This is what I needed to know.

In essence, we should treat src blocks that use a language with a jupyter- prefix as code cells. Conversely, code blocks with class code should probably add this prefix when writing org output.

Side note: The emacs-ipython and ob-ipython packages appear to enjoy a similar level of popularity, but their interfaces are different. It might make sense to support the latter as well.

[morgandavidson]

Glad you found the information you were looking for, I'm sorry I wasn't able to properly understand your question.

Indeed I'm using emacs-jupyter package. Never tried ob-ipython, but it doesn't seem to be maintained (last update was 5 years ago).

[morgandavidson]

I just tested ipynb export with pandoc 2.19.2 and it works perfectly, bravo! Also I asked emacs-jupyter team to mention exports from org to ipynb through pandoc in their documentation.

[tarleb]

Great news, thank you!