DataFrame object has no attribute 'name'

Question

I currently have a list of Pandas DataFrames. I'm trying to perform an operation on each list element (i.e. each DataFrame contained in the list) and then save that DataFrame to a CSV file.

I assigned a name attribute to each DataFrame, but I realized that in some cases the program throws an error AttributeError: 'DataFrame' object has no attribute 'name'.

Here's the code that I have.

# raw_og contains the file names for each CSV file.
# df_og is the list containing the DataFrame of each file.
for idx, file in enumerate(raw_og):
    df_og.append(pd.read_csv(os.path.join(data_og_dir, 'raw', file)))
    df_og[idx].name = file

# I'm basically checking if the DataFrame is in reverse-chronological order using the
# check_reverse function. If it is then I simply reverse the order and save the file.
for df in df_og:
    if (check_reverse(df)):
        df = df[::-1]
        df.to_csv(os.path.join(data_og_dir, 'raw_new', df.name), index=False)
    else:
        continue

The program is throwing an error in the second for loop where I used df.name.

This is especially strange because when I run print(df.name) it prints out the file name. Would anybody happen to know what I'm doing wrong?

Thank you.

Answer 1

the solution is to use a loc to set the values, rather than creating a copy.

creating a copy of df loses the name:

df = df[::-1] # creates a copy

setting the value 'keeps' the original object intact, along with name

df.loc[:] = df[:, ::-1] # reversal maintaining the original object

Example code that reverses values along the column axis:

df = pd.DataFrame([[6,10]], columns=['a','b'])
df.name='t'
print(df.name)
print(df)
df.iloc[:] = df.iloc[:,::-1]
print(df)
print(df.name)

outputs:

t
   a   b
0  6  10
    a  b
0  10  6
t

Answer 2

A workaround is to set a columns.name and use it when needed.

Example:

df = pd.DataFrame()

df.columns.name = 'name'

print(df.columns.name)

name

Answer 3

I suspect it's the reversal that loses the custom .name attribute.

In [11]: df = pd.DataFrame()

In [12]: df.name = 'empty'

In [13]: df.name
Out[13]: 'empty'

In [14]: df[::-1].name
AttributeError: 'DataFrame' object has no attribute 'name'

You'll be better off storing a dict of dataframes rather than using .name:

df_og = {file: pd.read_csv(os.path.join(data_og_dir, 'raw', fn) for fn in raw_og}

Then you could iterate through this and reverse the values that need reversing...

for fn, df in df_og.items():
    if (check_reverse(df)):
        df = df[::-1]
        df.to_csv(os.path.join(data_og_dir, 'raw_new', fn), index=False)