Pandas 中的布尔索引结合可变数量的列

Question

Is there a (pythonic) way to loop over columns in dataframe for boolean-indexing in pandas ?

I'm not sure you can do this using list comprehension...

The only way I've found would be :

df = pd.DataFrame(
    [
    ["a", "b", "c"],
    ["d", "e", "f"],
    ["g", "h", "i"],
    ], 
    columns=["A", "B", "C"],
    )

my_filter = {
    "A":{"a", "g"},
    "B":{"e"},
    }
    
series = [
        df[f].isin(v)==True for k,v in my_filter.items()
        ]
s = pd.DataFrame(series, columns=series[0].index).T.any(axis=1)
ix = s[s==True].index
df.loc[ix]

This seems to do the job... But I'm not sure this is memory efficient and (IMHO) this is neither straightforward nor easy to understand.

Is there a way in pandas to make some kind of serialisation of the "|"or "&" operators ?

Answer 1

让我们试试

out = df[np.any([df[k].isin(v) for k,v in my_filter.items()], axis=0)]