Difference in line-length handling between ruff and black when line ends in #

[dwyatte]

Settings: Default ruff 0.1.3

Note: there is something specific about the example line below in terms of variable string length vs actual assignment value such that reproduction becomes difficult when changing the name/value, but this example captures the behavior I'm interested in raising in this issue

Example

AAAAA_BBBBBBB_CCCC = "xxxxxxxx_xxxxxxxxxx_xxx_xxxxxxxx_xx_xxxx_xxxxxxx_xxxx_xxxx_xxxx_xxxxxxx_xxxx_xxxxx"

When the line above is suffixed with a # e.g., to disable linting / formatting, black will not break it:
https://black.vercel.app/?version=stable&state=_Td6WFoAAATm1rRGAgAhARYAAAB0L-Wj4AFEAGhdAD2IimZxl1N_WjPZmMnerqoaitTnd2a09E8Bi61KJWbnQshHGxfdai0l17K2DqAdXFD89RHKQ6jT30nA8Il4tW9g8JRaFzqeaQ-5TntovkAApVkiDMflIUXe64zjHKvCTRtNm3SZQwHmAKLERh2XSkjRAAGEAcUCAAD5YQM5scRn-wIAAAAABFla

Ruff however, will always break this line into

AAAAA_BBBBBBB_CCCC = (
    "xxxxxxxx_xxxxxxxxxx_xxx_xxxxxxxx_xx_xxxx_xxxxxxx_xxxx_xxxx_xxxx_xxxxxxx_xxxx_xxxxx"  #
)

My code has some long constant names that I'd prefer not to split and apparently # noqa: E501 is enough to exclude them from flake8 as well as black's formatting by virtue of how it handles lines ending in #

[zanieb]

In general, we're far more careful with comment handling and therefore can wrap lines like that without fear of dropping comments accidentally.

We explicitly exclude pragma comments from the line length #6772

We provide a # fmt: skip pragma to skip formatting of statements, which appears to be what you need here?

[dwyatte]

We provide a # fmt: skip pragma to skip formatting of statements, which appears to be what you need here?

Thanks, I think this will get me what I need although I suppose it's not clear how to support both disabling linting E501 and formatting on a single line -- I wonder if that's something ruff should support since it does both linting and formatting? Similar black feature that was recently implemented: psf/black#3959

Seems this is a known divergence from black as you mention though.

[zanieb]

Hm yeah I don't see a good way to do that cc @MichaReiser

[charliermarsh]

This works as expected, which seems reasonable to me?

AAAAA_BBBBBBB_CCCC = "xxxxxxxx_xxxxxxxxxx_xxx_xxxxxxxx_xx_xxxx_xxxxxxx_xxxx_xxxx_xxxx_xxxxxxx_xxxx_xxxxx"  # fmt: skip  # noqa: E501

[zanieb]

@charliermarsh that is formatted for me with

❯ ruff version
ruff 0.1.3 (3e7b92991 2023-10-26)

but it looks fixed in 0.1.4 so @dwyatte perhaps upgrade

[charliermarsh]

Ah okay, interesting. I'll close for now then. It does require that the skip comes before the noqa, but I think that's okay.

[dwyatte]

Ah okay, interesting. I'll close for now then. It does require that the skip comes before the noqa, but I think that's okay.

Perfect, thanks all!