ignore jupyter magics

[basnijholt]

With jupytext one can keep notebooks as Markdown files.

Often one uses Jupyter magics (e.g., %%time), when running blacken-docs on the .md files I get many warnings like:

FILENAME_HERE.md:30: code block parse error Cannot parse: 1:0: %load_ext autoreload

Would it be possible to easily ignore those magics?

[asottile]

blacken-docs is just a thin wrapper around black -- if black can't parse it I don't think we can either

and even if we did put some weird substitution in place, I'm not sure how it could survive a round of formatting

so I think the answer here is "probably not" -- thanks for the issue though!

[grst]

I would also be interested in this.
Since the cell magics can only occur at the first line of a code chunk, I think it should be rather easy to exclude them.

In fact, adding a line to your regex parser that skips a line starting with % should already solve the problem.

[MarcoGorelli]

In fact, adding a line to your regex parser that skips a line starting with % should already solve the problem.

Unfortunately, this would break:

foo = (
'%s'
% 'bar'
)

if black can't parse it I don't think we can either
and even if we did put some weird substitution in place, I'm not sure how it could survive a round of formatting

Just FYI, this might soon be happening (psf/black#2357)

In the meantime, there's my little project nbQA which, like the above PR to black, transform cells using IPython's TransformerManager and ast.parse to preserve/round-trip magics