Rule C416 broke some pandas code

[wlach]

This snippit:

import pandas as pd
obj = { k: v for k,v in pd.DataFrame.from_records([], columns=["foo", "bar", "baz"]).groupby("foo")}

Is rewritten to the following using the C416 rule (no unnecessary comprehensions):

import pandas pd
obj = dict(pd.DataFrame.from_records([], columns=["foo", "bar", "baz"]).groupby("foo"))

However that generates a TypeError when I run it. I filed a bug about this upstream (see pandas-dev/pandas#53287) but it makes me wonder if this rule is safe to apply. Current using ruff 0.0.262.

[charliermarsh]

Huh, interesting. It's not immediately clear to me why / how those are different.

[edgarrmondragon]

FWIW it's fixed by wrapping the DataFrameGroupBy object with iter:

import pandas as pd
obj = dict(iter(pd.DataFrame.from_records([], columns=["foo", "bar", "baz"]).groupby("foo")))

or using the splat * operator

import pandas as pd
obj = dict(*pd.DataFrame.from_records([], columns=["foo", "bar", "baz"]).groupby("foo"))

[Charlie-XIAO]

The splat * operator does not work in general (only works when empty). iter() works because DataFrameGroupBy.__iter__() is a public method that returns a generator yielding sequence of (name, subsetted object) for each group.

[zanieb]

Interesting...

Thanks for reporting this issue. I looked online and apparently Python's dict() method checks for existence of the .keys() method.

https://stackoverflow.com/questions/40667093/what-is-a-mapping-object-according-to-dict-type

Unfortunately, in this case obj.keys returns a str, which is not callable.

ref pandas-dev/pandas#53287 (comment)