Why can't I modify DataFrame in place by selecting some columns when iterating through list of DataFrames?

dfl = [pd.DataFrame(
     {
        "A": 1.0,
        "B": pd.Timestamp("20130102"),
         "C": pd.Series(1, index=list(range(4)), dtype="float32"),
         "D": "foo",
     }
)]
for df in dfl:
  df = df[["A", "B"]]
print(dfl)

I was expecting the output has only column "A" and "B" since I was modifying the DataFrame in place (df = ...). However I got:

[     A          B    C    D
0  1.0 2013-01-02  1.0  foo
1  1.0 2013-01-02  1.0  foo
2  1.0 2013-01-02  1.0  foo
3  1.0 2013-01-02  1.0  foo]

What is the reason and how can I select (not drop) some columns from the each DataFrame in that list in place?

>Solution :

Because df is a local pointer to the loop. By doing df = df[['A','B']], you tell the pointer to point to something new, which doesn’t override the existing element of the list. Another similar example:

ll = [[1]]
for l in ll:
    l = None
print(ll)
# output: [[1]]

To override the element, you would want to do:

for i, d in enumerate(dfl):
    dfl[i] = d[['A','B']]

print(dfl)

Out:

[     A          B
0  1.0 2013-01-02
1  1.0 2013-01-02
2  1.0 2013-01-02
3  1.0 2013-01-02]