Better noqa interpretation

[sbrunner]

Description

• To completely ignore a line the # noqa should be alone
• To ignore a code we can use # noqa: <code> ad ruff do
• To ignore a code from a certain source use # noqa: <source>.<code>, this will also create an error if the ignored is not used.

Alternative of #714

Related Issue

fix #715

Motivation and Context

• Be compatible with Ruff
• Have more control on message to be ignored
• Be notified on used ignore

How Has This Been Tested?

Test added

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

[carlio]

So as I see it, it'll maintain the same behaviour (<-- I use British spelling because I am, but I won't moan about the changes in the docs 😛) for the most part.

Tools with their own syntax will do what they used to do, #noqa is still a blanket "ignore this line".

There is new syntax which is prospector-specific of #noqa flake8.F345 or similar. Prospector I think should be nothing more than a convenience wrapper and shouldn't introduce its own stuff too much, but equally there are time when flake8 or bandit might not have a way to ignore only a single line or only a specific error, so this seems like a good and valid addition to me.

[Pierre-Sassoulas]

#noqa atool.F345 is not the obvious way atool would implement noqa of a specific message so I would use #noqa: F345 so prospector does not require specific change to be adopted from a codebase where the noqa are already here. There might be collision if errors from various tools are in the same namespace so multiple errors are noqa'ed (flake8 F345 / ruff F345 bandit F345), but I think it's probably okay. In the case of ruff/flake8 it's actually the same message re-implemented, and for bandit/(ruff/flake8) how often is this actually going to happen that there's both a collision in message name between tooling AND one error is mistakenly silenced because this line is rewritten after the noqa was added and those two errors happens on the same line ?

[sbrunner]

@Pierre-Sassoulas what you say is probably true, but I think that having a test that the ignored is really use is an added value, And I can do it only I'm sure that I receive the message and with Ruff it's not the case.

If you still be against that, I will remove it :-)

[Pierre-Sassoulas]

It's possible to do both Ruff style or '# Noqa: atool: F408' style, so if you're going to use the atool: style it you might as well add it :)