Takes care of issue #279 and also helps to support autoflake #313 .

• [ ] Test current changes on external notebooks

This PR is ready for review. But before merging I would like to run on some external notebooks.

I have tried to look around, but I am not sure if one of the recent fixed addresses this problem. I am using nbQA for its isort functionality and have the following problem.

A notebook which looks like this (MWE)

%load_ext autoreload
%autoreload 2

import sys
import pathlib

is converted by nbqa isort into

import pathlib
import sys

%load_ext autoreload
%autoreload 2

which actually invalidates the autoreload extension that must be loaded before any imports. I have tried to use # isort:skip or put it in a different cell, but to no avail. Do you know if there's a fix for this?

Otherwise, thanks for the great hooks!

Thanks for your work! I have a question, that might convert into an issue, so I'm trying to use nbqa-mypy, my config looks like this:

  - repo: https://github.com/nbQA-dev/nbQA
    rev: 0.5.7
    hooks:
      - id: nbqa-mypy
        args: ["--nbqa-ignore-cells=%%R", "--strict", "--show-error-codes"]

But I'm getting error about missing modules that are part of the repository, example:

pipelines/notebooks/foo.ipynb:cell_2:1: error: Cannot find implementation or library stub for module named 'my_utils'  [import

Where the repo contains {pipelines, my_utils} in the root, and import from my_utils doesn't raise issues in py modules and vanilla mypy checks. Am I doing sth wrong? Is there a blessed way to use nbqa-mypy and give access to the main repo? Maybe nbqa-mypy is missing some step to add repo to the pre-commit's nbqa-mypy virtualenv?

closes #203

Removed separate github actions for running pylint.

Changes to be committed: deleted: .github/workflows/pylint.yml modified: .pre-commit-config.yaml

e.g.

if True:
    %time True
pass

This will get transformed to

if True:
# commented out line
pass

which isn't valid syntax.

I'm tempted to say that the fix here would be to replace magics with

pass  # nbqamagic

and then, in the future, have a less hacky solution

… is not able to run the command and raises ValueError: Command not found.

Changes to be committed: modified: nbqa/main.py modified: nbqa/cmdline.py new file: nbqa/troubleshoot.py modified: tests/test_runtime_errors.py new file: tests/test_troubleshoot.py

Just after I reinstalled pre-commit (after getting the latest changes) and then I ran precommit run --all or even tried to git commit my changes I get this error:

An error has occurred: InvalidManifestError:
==> File /Users/user/.cache/pre-commit/repo8m9u7jls/.pre-commit-hooks.yaml
==> At Hook(id='nbqa')
==> At key: types
==> At index 0
=====> Type tag 'jupyter' is not recognized.  Try upgrading identify and pre-commit?
Check the log at /Users/user/.cache/pre-commit/pre-commit.log

Workaround: uninstall precommit and we can continue to use git commit

On the back of the issue #320 (and original issue #89) amending the following GitHub actions:

• [x] tox.yml
• [x] autoupdate-pre-commit-config.yml
• [x] pre-commit.yml

Adding the cache action to these github actions via the github action cache@v2.

This is required to support pylint. Using pass causes pylint to raise unnecessary pass warnings. Instead if we use builtin function type we are clean of any flake8 and pylint warnings.

I'm raising this just to discuss if there is a way to measure complexity of Python (or other language supported by a notebook)?

There's https://en.wikipedia.org/wiki/Cyclomatic_complexity that many languages have libraries for, I'm thinking there is one for Python as well.

In case your other analysis already covers this then please close this. Otherwise, it would be a helpful feature to have.

Maybe related to #259

Hi

Running nbqA flake8 over notebooks in a set of directories, I encountered a few errors of the form:

KeyError(1) while parsing output from applying flake8 to ...

I will try to dig into the notebooks to see what the problems appear to be and raise a separate issues as appropriate, but more generally it would be useful if nbQA could trap these errors rather than erroring and then publishing a notice or error file capturing the errors.

In passing, would it be worth adding to docs a simple example recipe for excluding multiple files with different filename patterns for folk who aren't overly familiar with regex? eg I am using a pattern of the form nbqa flake8 --nbqa-exclude "(.*02.2.*)|(.*07.1.*)|(.*07.2.*)" Part\ *\ Notebooks/*.ipynb with | separated bracketed expressions.

The following works with nbQA (1.5.2) and docformatter (1.5.0).

nbqa docformatter -i example.ipynb

The -i option is needed to tell docformatter to modify the notebook in-place.

For those using pre-commit, the following configuration can be used:

repos:
- repo: https://github.com/nbQA-dev/nbQA
  rev: "1.5.2"
  hooks:
    - id: nbqa
      entry: nbqa docformatter -i
      name: nbqa-docformatter
      alias: nbqa-docformatter
      additional_dependencies: [docformatter==1.5.0]

It might be useful to support docformatter out of the box in nbQA, or to show this example in the documentation. (The documentation already explains how to get custom tools to work, but a working example is always nice to have.)

Here is an example notebook to try out the above setup: example_nbqa_docformatter.zip