遍历 pandas dataframe 的列并为循环中的每个选定列创建一个新的 dataframe

Question

I have a pandas dataframe with multiple columns and I am trying to iterate through the dataframe by selecting one column at a time, create a new dataframe with that one column, perform some functions. Then select the next column in the dataframe, perform functions and continue the process until I reach the last column in the dataframe.

Currently, I am doing it with only one column. I am stuck on how to do this in a loop and run the functions inside a loop. Could someone please help on how I can iterate through the columns in a loop, create a new dataframe for each selected column and run the functions inside that loop.

df:

date                   Col1      Col2       Col3      Col4           
1990-01-02 12:00:00     24        24        24.8      24.8           
1990-01-02 01:00:00     59        58        60        60.3   
1990-01-02 02:00:00     43.7      43.9      48        49

Code

df_new = pd.DataFrame(df['Col1'])
df.reset_index(inplace=True)

def function1(df_new):
    line 1
    line 2

def function2():
    line 1
    line 2

The answer I am looking for is something like below, where I just have to iterate over the columns and perform the same set of functions for each. IS there a better way to do this?

for col in df.columns:
        col_df = df_full[[col]]
        col_df.reset_index(inplace=True)
        col_df
         
        #perform functions on col_df

Answer 1

If you insist on iterating through columns then you'll have a series for every column, in which case I don't see the added value of converting it to a DataFrame first.

Instead, perform the functions on each series:

def Add(col):
    return col+1

def Minus(col):
    return col-1

def Double(col):
    return col*2

for col in df.columns:
    print(Add(df[col]))
    Minus(df[col])
    Double(df[col])

Be sure to save the results if you want to do further manipulations with them when the loop is finished.

However, I advise instead looking at other possibilities, for example using apply() and lambda :

df.apply(lambda x: x+1 , axis=0)

This is much more efficient.