is it possible to do the following (dict is data for the first row index):
dict={'col1':1,'col2':[1,2],'col3':'str'}
nm=pd.DataFrame(dict,index=['new line'])
I expect to receive a data frame looking like:
nm
col1 col2 col3
new line 1 [1,2] str
with the values in col2 of type list, but instead I receive the following error:
ValueError: could not broadcast input array from shape (2) into shape (1)
Thanks!
You are miss one []:
dict={'col1':1,'col2':[[1,2]],'col3':'str'}
nm=pd.DataFrame(dict,index=['new line'])
print nm
col1 col2 col3
new line 1 [1, 2] str
Or:
dict={'col1':[1],'col2':[[1,2]],'col3':['str']}
nm=pd.DataFrame(dict,index=['new line'])
print nm
col1 col2 col3
new line 1 [1, 2] str
If you have more values, you need use list to each column as Series, so if you need list in column, you have to use list of lists:
dict={'col1':[1, 2],'col2':[[1,2], [3,5]],'col3':['str', 'str1']}
nm=pd.DataFrame(dict,index=['new line', 'new line1'])
print nm
col1 col2 col3
new line 1 [1, 2] str
new line1 2 [3, 5] str1
Storing non-scalar values as data elements is ill-advised and typically you lose access to vectorised methods as np and pandas has no vectorised methods for appending to a list in a vectorised manner. link
