B018 doesn't trigger for useless expressions involving multiple variables

[swierh]

B018 doesn't trigger for useless expressions involving multiple variables

[cooperlees]

Agree this would be nice to add to B018. Thanks for reporting.

[JelleZijlstra]

I remembered that internally in our codebase we have a similar lint rule. It triggers on these types:

ast.Attribute, ast.Lambda, ast.Dict, ast.Set, ast.Compare, ast.Num, ast.Subscript, ast.List, ast.Tuple, ast.Name

So that's missing a few that I mentioned above. We also flag ast.Call nodes if the called function is one of sorted, reversed, map, and filter. If you write sorted(x) on a line by itself, you probably meant x.sort().

[JelleZijlstra]

"multiple variables" isn't a great description of what's going on here. If you look at the code for B018, it triggers if a statement consists only of an expression that is a set/list/tuple/dict literal or a non-str constant. (The original post claims that B018 triggers for +1 and -1, but I don't think it does.)

We could expand the rule to some other kinds of nodes:

• UnaryOp, like +1, not x, ~3. I don't know of any code patterns that would trigger false positives.
• BinOp, like 1 + 1 or 2 << 3. This is what the OP is asking for. This could have false positives with Airflow code, where you do dag << task to add a task to a dag.
• Compare, like 1 < 2. Can't think of any problems with this one.
• BoolOp, like a and b. People could be using this for conditional execution.
• GeneratorExp, like (x for x in y). That is definitely useless.
• JoinedStr, which is an f-string. Having an f-string in a line by itself doesn't have any use.

So we could definitely expand B018 to a few more kinds of AST nodes, but there's some room for false positives.