在Pandas 0.21.0版之后，如何将Python dict转换为DataFrame列？

Question

I am trying to run the same script over two computers, and have my dict structure, data, convert to Pandas DataFrame.

df = pd.DataFrame(data, columns=[column_label])
df.to_csv('./result.csv', mode='w', index=False)

It works perfectly on the computer with Pandas version less than 0.21.0. However, when I execute the same code over to my server, suddenly it would generate a csv file with just the column labels, and none of the data.

I tried to print out the values of df, and on the server it's all NAN.

When I remove the columns part like this:

df = pd.DataFrame(data)
df.to_csv('./result.csv', mode='w', index=False)

Suddenly the data are back, albeit the columns are missing and the data are not in order.

If I do

df = pd.DataFrame(data)
df.columns = column_label
df.to_csv('./result.csv', mode='w', index=False)

The column labels are back, but the column labels are out of order, and the data is also doesn't match the order of the column labels.

If I do

df = pd.DataFrame(columns=[column_label])
df = pd.DataFrame(data)
df.to_csv('./result.csv', mode='w', index=False)

The the data matches the order of the column labels, but the column labels themselves are out of order...

I have since upgrade my pandas library on the computer that was working from 0.17.0 to 0.22.0, and it has also stopped working.

So for some reason, assigning data and columns=[column_label] to a DataFrame in one line seems to break the dict to DataFrame conversion after Pandas version 0.21.0.

How should I do columns assignment with the newer versions of Pandas?

Answer 1

I am assuming column_label is a list.The columns parameter of pandas just requires a list of column names. What you are doing by passing [column_label] to the columns parameter is passing a list of lists. Try without [] .

df = pd.DataFrame(data, columns=column_label)