Catch errors when running --pipe black

If you try to run the --pipe black command over a set of notebooks in one or more directories, and black fails for some reason, the pipe fails at that point.

Non-runnable illustrative fragment:

jupytext --sync  --pipe black  Part*/*.ipynb

---

[jupytext] Executing black -
reformatted -
All done! ✨ 🍰 ✨
1 file reformatted.
[jupytext] Updating 'Part 03 Notebooks/03.2 Selecting and projecting, sorting and limiting.ipynb'
[jupytext] Updating 'Part 03 Notebooks/.md/03.2 Selecting and projecting, sorting and limiting.md'
[jupytext] Reading Part 03 Notebooks/03.3 Combining data from multiple datasets.ipynb in format ipynb
[jupytext] Loading 'Part 03 Notebooks/.md/03.3 Combining data from multiple datasets.md'
[jupytext] Executing black -
error: cannot format -: Cannot parse: 156:9: result = %sql SELECT * FROM quickdemo WHERE value > 25
Oh no! 💥 💔 💥
1 file failed to reformat.

In the above case, black failed trying to parse a line that assigned the result of a line magic call to a variable.

It would be useful if the Jupytext pipe did not fail, but did perhaps warn that a file could not be parsed.

Another solution might be to allow a cell to be tagged as no-black or similar and be ignored somehow (--ignore-tag no-black?) (I’m not sure if you apply the formatter at the cell level?)

[mwouts]

Hi @psychemedia , I expect to have fixed both issues in the development version. Feel free to give it a try, cf. https://jupytext.readthedocs.io/en/latest/developing.html

[MarcoGorelli]

Jupytext does not identify the line myvar = %dont_format_me as a magic command, maybe we could improve that part as well

Suggestion for how to do this:

>>> from IPython.core.inputtransformer2 import TransformerManager
>>> TransformerManager().transform_cell('myvar = %dont_format_me')
"myvar = get_ipython().run_line_magic('dont_format_me', '')\n"

Then, using ast.parse, detect lines with get_ipython().run_line_magic (and other functions which magics would get transformed to) and comment them out (e.g. of what I’m referring to: https://github.com/psf/black/pull/2357 , where jupytext has been mentioned)