Jupyter notebooks are great, but they take me away from my nice terminal environment: tmux panels and vim buffers.

I thought Jupyter Lab would take me back to a comfortable workflow where I interact easily between terminals, text editors and python/R consoles, but it’s simply not as slick. However, two major points of notebooks are:

  1. reproducibility, something not easy to achieve when just typing commands in a terminal 1;
  2. visualisation: any plot produced is displayed for ever in the notebook.

So I’d like to add these two essential parts of a scientific workflow to my data exploration.

Some people like me seem to prefer the more markdown-oriented philosophy of rmarkdown, but with Python (or any other language possible in jupyter), and some of them developped solutions:

  1. podoc

    As far as I’m concerned, the interesting feature is the conversion from markdown to notebook.

    Then converting the executed notebook could be done with jupyter nbconvert --execute.

    About the comparison with Rmarkdown, see this issue.

  2. knitpy

  1. I’m not yet efficient enough to produce snakemake pipeline on the fly during exploratory analysis; that’s an idea to dig though: creating pipelines from notebooks? ↩