在正/负值上拆分列？

Question

So I came across this another little annoying problem, as another obstacle in my first steps of learning python.

I have outcome column which has positive/negative/zero values (winnings, loss, no commitment). I would like to split into winnings/loss based on sign, also zeroes would populate zeroes, and negative rows in the new winnings column; zeroes would populate zeroes and positive rows in the new loss columns.

DATA.

g=pd.DataFrame({'OUTCOME':[100,-100,400,-200,-200,-750,-250,1000,0,100,-100]},index=[1,1,2,2,2,3,3,3,4,4,4])

DESIRED OUTPUT.

g['WINNINGS']=[100,0,400,0,0,0,0,1000,0,100,0]
g['LOSS']=[0,100,0,200,200,750,250,0,0,0,100]

Answer 1

Theres's more than one way to do this, but essentially what you want to do is apply a function that returns 0 if the number is less than or equal to zero, and the input number otherwise. And then do the opposite thing for losses. One way is:

def winnings(value):
    return max(value, 0)

def losses(value):
    return min(value, 0)

df["winnings"] = df["outcome"].map(winnings)
df["loss"] = = df["outcome"].map(losses)

Answer 2

You can use Series.where :

df["winnings"] = df.OUTCOME.where(df.OUTCOME > 0, 0)
df["loss"] = -1 * df.OUTCOME.where(df.OUTCOME < 0, 0) 
print (df)
   OUTCOME  winnings  loss
1      100       100     0
1     -100         0   100
2      400       400     0
2     -200         0   200
2     -200         0   200
3     -750         0   750
3     -250         0   250
3     1000      1000     0
4        0         0     0
4      100       100     0
4     -100         0   100

Or faster solution with numpy.where :

df["winnings"] = np.where(df.OUTCOME > 0, df.OUTCOME, 0)
df["loss"] = np.where(df.OUTCOME < 0, - df.OUTCOME, 0) 

Timings :

In [68]: %timeit (jez1(df2))
100 loops, best of 3: 3.75 ms per loop

In [69]: %timeit (jez(df1))
100 loops, best of 3: 5.82 ms per loop

In [70]: %timeit (bat(df))
10 loops, best of 3: 134 ms per loop

Code for timings :

df=pd.DataFrame({'OUTCOME':[100,-100,400,-200,-200,-750,-250,1000,0,100,-100]},index=[1,1,2,2,2,3,3,3,4,4,4])
print (df)
##[110000 rows x 1 columns]
df = pd.concat([df]*10000).reset_index(drop=True)
df1 = df.copy()
df2 = df.copy()

def winnings(value):
    return max(value, 0)

def losses(value):
    return min(value, 0)

def bat(df):
    df["winnings"] = df.OUTCOME.apply(winnings)
    df["loss"] = - df.OUTCOME.apply(losses)
    return df

def jez(df):
    df["winnings"] = df.OUTCOME.where(df.OUTCOME > 0, 0)
    df["loss"] = -1 * df.OUTCOME.where(df.OUTCOME < 0, 0) 
    return (df)

def jez1(df):
    df["winnings"] = np.where(df.OUTCOME > 0, df.OUTCOME, 0)
    df["loss"] = np.where(df.OUTCOME < 0, - df.OUTCOME, 0) 
    return (df)

print (bat(df))   
print (jez(df1))  
print (jez1(df2))