Pandas: numpy.where logic with multiple conditions

Question

New to Python Data Science

The below is my Raw_data

raw_data = {'var1': ['true','false','ture'],'var2': [10,20,50], 'var3':['eggs','milk','eggs']}
df = pd.DataFrame(raw_data, columns = ['var1','var2','var3'])`

Expermented code but not working

def my_fun (var1,var2,var3,var4):
    df[var4]= np.where((df[var1] == 'true', 
                       df[var3] == 'eggs',
                       df[var2] < 10),
                       'hello',
                       'zello')
return df

Here I like to use var1, var2 and var3 conditions and get the conditional result. Please help

"but not working" - please include the complete error message. Also, you probably should use True and False instead of "true" and "false". — DYZ
– DYZ, Commented Mar 23, 2018 at 2:21

jpp · Accepted Answer · 2019-01-26 13:55:34Z

2

First use Boolean True / False rather than strings to simplify your logic. To apply this conversion to series 'var1':

df['var1'] = df['var1'] == 'true'

You can then use the bitwise operator & to compare Boolean series:

def my_fun (var1,var2,var3,var4):
    df[var4]= np.where(df[var1] & df[var3].eq('eggs') & df[var2].lt(10),
                       'hello', 'zello')
    return df

A less efficient alternative is to use np.logical_and.reduce:

def my_fun (var1,var2,var3,var4):
    conds = (df[var1], df[var3] == 'eggs', df[var2] < 10)
    df[var4]= np.where(np.logical_and.reduce(conds), 'hello', 'zello')
    return df

edited Jan 26, 2019 at 13:55

answered Mar 23, 2018 at 2:22

jpp

166k37 gold badges301 silver badges362 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

user2357112 Over a year ago

You could also use numpy.all with an axis argument, or just & twice.

miradulo Over a year ago

@jpp For a tiny amount of data like this there's probably no difference, but if you had any significant amount, np.logical_and would waste some significant time dealing with the Series objects during reduction, (presumably because it isn't optimized for this like Pandas bitwise-ands are). You could avoid this of course using .values everywhere but that gets pretty verbose.

Murali Over a year ago

@jpp Works great. Can add additional condition to np.where? (df[var1].values == 'true') & (df[var3].values == 'eggs') & (df[var2]].values == 10), 'equal'

Murali Over a year ago

@jpp I am accepting your answer. However, I am able to add new condition by adding addtional np.where. Thank you so much.

Collectives™ on Stack Overflow

Pandas: numpy.where logic with multiple conditions

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related