shifting operators are being interpreted as pointless statements

[feluelle]

Question

Hey,

I was wondering why the >> operator is interpreted as pointless statement.

For example when I do something like a >> b, the __rshift__ function from a will be called and if this function does change some kind of external state it is not a pointless statement.
Is it a bad practice to change an external state in __rshift__ or why is it being interpreted as W0104: Statement seems to have no effect (pointless-statement) ?

I am asking this because I am currently using apache-airflow and in Airflow it is recommended to use the >> operator to declare graph dependencies.
I do not want to put # pylint: disable=pointless-statement before each of those kind of statements.

It would be nice if I could somehow tell my pylintrc to only disable pointless statement for the bitshifting operator. Is this possible? Or what would you recommend me to do?

[PCManticore]

@feluelle Thanks for creating a question. RIght, that is definitely suggesting to me a code smell, special properties such as __rshift__ should either return a new object or a mutated object, but they shouldn't mutate a global state. pylint tries to figure out if any statement has its value used or not, and given that in your example you mutate a global state, pylint cannot figure (and definitely shouldn't) figure that out.
I think a solution for your use case would definitely be to disable the error locally, given that it's not possible to disable pointless-statement for one operator or another.

[feluelle]

Thank you for your opinion on this. :)

[martimlobao]

Resurfacing an old issue, but what's the recommended approach here? We're also trying to use pylint in our airflow deployment, but it's very cumbersome to put # pylint: disable=pointless-statement in every DAG. Is there a better way of doing this?

@feluelle what did you end up doing here? Is there pylint extension that we can use?

@PCManticore Or perhaps a way to disable this specific warning in a specific directory, or only for Airflow task objects or something?

[DanielNoord]

I don't think there is much more you can do then disable this for projects that use airflow.

[martimlobao]

Unfortunate, since we use a monorepo which means this is an all-or-nothing situation, and the pointless-statement warning is generally valuable. 😞

I think the #618 Add different configuration for different sub directories ticket might provide the best solution.

[feluelle]

Hey @martimlobao, sorry but I also don't have a better idea. We moved away from using pylint at some point.