I have a df with some binary columns (1,-1) and a list with N columnnames. i need to create a new variable like that ...
df['test'] = np.where(((df['Col1']==-1) & (df['Col2']==-1)), -1, 0)
... but dynamically. so the rule is: if all the columns from the list have the same value (1,-1) take it. otherwise value = 0. the length of the list is not fixed. can i simply iterate over the list and create that "where-String" or is there a more elegant way?
thanks! e
IIUC you can just do
df['test'] = np.where((df[list_of_col_names] == -1).all(axis=1), -1, 0)
So here you can just pass a list of cols of interest to sub-select from the orig df as all you're doing is comparing all cols of interest to a scalar value, you then do all(axis=1) to test if all row values match that value and pass the boolean mask to np.where as before.
e.g.:
list_of_col_names = ['col1','col2']
df['test'] = np.where((df[list_of_col_names] == -1).all(axis=1), -1, 0)
it's important you pass an actual list of names or iterable, if you do this it'll raise a KeyError:
df['test'] = np.where((df['col1','col2'] == -1).all(axis=1), -1, 0)
as it'll interpret this as a tuple and it's likely that this column 'col1','col2' doesn't exist
df[['col1','col2']] instead of df['col1','col2'], in the past I've had people comment that it didn't due to the latter, I'll edit and make it clearer