Check for pd.merge about pandas-vet HOT 4 CLOSED

Any thoughts on how to implement this check?

The AST node will look the same for both (ast.Call), so this needs to include a separate semantic check to determine the type of the base object:

type(pd)     # <class 'module'>
type(df)     # <class 'pandas.core.frame.DataFrame'>

We can still use the AST to confirm that node.func.attr == 'merge'

from pandas-vet.

Without exploring this, my first thought is maybe to catch all pd. calls and then check for merge. Anything besides pd. at the start of .merge is fine. If someone calls their DataFrame 'pd', they have bigger problems.

from pandas-vet.

Please ignore my closing of the issue - I mis-clicked.

from pandas-vet.

So to be clear, the AST does not retain any of the pandas semantics, so it doesn't have any idea what type of object the .merge() method is applied to. Without implementing some other way to determine the object type, we can infer based on the name of the object.

We can actually add one additional check to this -- if the object name is pd and the method has at least two arguments (left, right, ... ). All of the other arguments have default values, so only df.merge() could have a single argument.

from pandas-vet.