Match columns and append to data frame, Python 3.6

6
votes

I have around 50 excel files & I want to import to dataframe and merge all files into single dataframe. But some file has 3 some are 4 columns. Every file as different columns in different order.

Total distinct column from all the files: 5 i.e col1, col2, col3, col4, col5

I know how to import but while appending facing issue.

Script:

dfAll = pd.DataFrame(columns=['col1', 'col2', 'col3', 'col4', 'col5')]
df= pd.read_excel('FilePath', sheetname='data1') # contains 3 columns i.e col1, col2, col5
columnsOFdf = df.columns
dfAll[columnsOFdf] = dfAll.append(df)

but its giving error "ValueError: Columns must be same length as key"

I want to append df['col1','col2','col5'] data to dfAll['col1','col2','col5']

Please help on this issue.

3
pythonpython-3.xpandasdataframeappend
You have are trying to append a dataframe of size 3 to a dataframe of size 5, that is not going to work with untype datasets – Sentinel
@Sentinel, thanks for response, any alternate solution? – Learnings
I'm not well referenced in using python in terms of dataFrames, you will most likely need to make a new dataframe including only the columns you want, than append the other dataframe – Sentinel

3 Answers

6
votes

Concatenation will match your columns

dfs = []
files = [...]
for file_name in files:
    dfs.append(pd.read_excel(file_name, sheetname='data1'))
df = pd.concat(dfs)

df1 = pd.DataFrame(np.random.randn(3, 3), columns=list('ABC'))
df2 = pd.DataFrame(np.random.randn(3, 3), columns=list('BCD'))
>>> pd.concat([df1, df2])
          A         B         C         D
0 -2.329280  0.644155 -0.835137       NaN
1  0.666496 -1.299048  0.111579       NaN
2  1.855494 -0.085850 -0.541890       NaN
0       NaN -1.131514  1.023610 -0.514384
1       NaN  0.670063  1.403143 -0.978611
2       NaN -0.314741 -0.727200 -0.620511

In addition, each time you append a dataframe to an existing one, it returns a copy. This will seriously degrade performance and is referred to as a quadratic copy. You are best of creating a list of all dataframes and then concatenating the result.

2
votes

One solution is to add empty columns to the dataframes you load from Excel files:

columns = ['col1', 'col2', 'col3', 'col4', 'col5']
dfAll = pd.DataFrame(columns=columns)
df= pd.read_excel('FilePath', sheetname='data1') # contains 3 columns i.e             col1, col2, col5
columnsOFdf = df.columns
for column in columns:
    if column not in columnsOFdf:
        df[column] = [""] * df.shape[0]
dfAll.append(df)
2
votes

try this:

[dfAll.append(i) for i in df]

I hope this help you.