how to change dataframe in apply functions pandas - pandas

Question

I want to use apply to dynamically modify the content of my dataframe, the table is like:
index price signal stoploss
0 0 1000 True 990.0
1 1 1010 False 990.0
2 2 1020 True 1010.0
3 3 1000 False 1010.0
4 4 990 False 1010.0
5 5 980 False 1010.0
6 6 1000 False 1010.0
7 7 1020 True 1010.0
8 8 1030 False 1010.0
9 9 1040 False 1010.0
my code is :
def test(row, dd):
if row.signal:
dd['inorder']=True
row['stoploss']=1
df = pd.DataFrame({'index':[0,1,2,3,4,5,6,7,8,9],
'price':[1000,1010,1020,1000,990,980,1000,1020,1030,1040],
'signal':[True, False, True, False, False, False, False, True, False, False]})
if __name__ == '__main__':
df['stoploss'] = df.loc[df['signal'], 'price'] - 10
df['stoploss'].ffill(inplace=True)
xx = dict(inorder=False)
df.apply(lambda row: test(row, xx), axis=1)
print(df)
When I trace into the function test, I can see that the value is indeed changed to 1, but out of the function test scope, it seems has no effect on the dataframe.
I tried to the use another way to modify the content of the dataframe,
for k, row in df.iterrows():
if row.signal:
xx['inorder'] = True
df.loc[k, 'stoploss'] = 1
this one works, but obviously it's a lot slower than apply.
The correct result I expect is :
index price signal stoploss
0 0 1000 True 1.0
1 1 1010 False 990.0
2 2 1020 True 1.0
3 3 1000 False 1010.0
4 4 990 False 1010.0
5 5 980 False 1010.0
6 6 1000 False 1010.0
7 7 1020 True 1.0
8 8 1030 False 1010.0
9 9 1040 False 1010.0
How to achieve that assignment in apply please?
Thanks

Answer 1

If you look at the docs for apply, you'll notice that apply does not change the DataFrame in place, but rather returns a new dataframe where the function has been applied.
So, in your seconds to last line, you can try
df = df.apply(lambda row: test(row, xx), axis=1)
Edit:
IMO, this isn't very well documented, but the call
df.apply(func, axis=1) will apply func to each row, and set the row to the return value of func.
As written, your example won't work because the function you're applying doesn't return anything. The following minimal example works the way you intend.
df = pd.DataFrame({'index':[0,1,2,3,4,5,6,7,8,9],
'price':[1000,1010,1020,1000,990,980,1000,1020,1030,1040],
'signal':[True, False, True, False, False, False, False, True, False, False]})
df['stoploss'] = df.loc[df['signal'], 'price'] - 10
df['stoploss'].ffill(inplace=True)
def test(row):
row.stoploss = 1 if row.signal else row.stoploss
return row
modified_df = df.apply(test, axis=1)
As an aside, I don't think you actually need to use apply to get the result you want. Have you tried something like
df.loc[df['signal'] == True, 'stoploss'] = 1
That would be a much simpler and faster way to get your target output.