Automatically format the import statements to adhere to the Google coding style guidelines

[yiliu30]

Can Ruff be used to format code in alignment with the guidelines outlined in https://google.github.io/styleguide/pyguide.html#22-imports?

Use import statements for packages and modules only, not for individual types, classes, or functions.

For instance:

# Original code
from module_name.util import helper_func1, helper_func2

helper_func1(a, b, ...)
helper_func2(a, ...)

# Expected code
from module_name import util

util.helper_func1(a, b, ...)
util.helper_func2(a, ...)

[tjkuson]

No, I don't think it can. If I understand the problem correctly, it would need multi-file analysis and better type inference to determine modules from other objects. Ruff can't do that yet (#3893, #7447).

You might be able to use banned-import-from (ICN003) to approximate the desired behaviour. For example, to lint against importing from functools, add

[tool.ruff.lint.flake8-import-conventions]
banned-from = ["functools"]

to your pyproject.toml file. This should report on from functools import wraps but not import functools.

[yiliu30]

@tjkuson Thanks for your reply. I get that it's up to the users to specify which imports are banned.

A follow-up question, if I use banned-import-from (ICN003) to identify the banned imports, is possible to create a rule that can automatically correct these imports? For example, the logic of the rule may looks as below:

For example:

# pyproject.toml
[tool.ruff.lint.flake8-import-conventions]
banned-from = ["transformers"]
# auto-fix rule = xxxx

If found :

from transformers import AutoModel, AutoTokenizer

model = AutoModel(...)
tokenizer = AutoTokenizer(...)

Replace these codes with:

import transformers

model = transformers.AutoModel(...)
tokenizer = transformers.AutoTokenizer(...)

[tjkuson]

It's an open issue: #6372

I might work on it if I find the time later this week.