Exception With np.frompyfunc

[craustin]

from aqr.core.frame import DataFrame
import numpy as np
def func(a, b):
return a + b
df = DataFrame({'US':[1,2,3]})
f = np.frompyfunc(func, 2, 1)
print f.accumulate(df)

In pre-0.4.0 pandas, this printed:
US
0 1
1 3
2 6

Now, this:
ValueError: could not find a matching type for func (vectorized).accumulate, requested type has type code 'q'

[wesm]

I can't find a code revision where this works-- tried 0.4.0 and mid-august commit too, any help here? np.add.accumulate(df) works fine however. Are you using the same version of numpy for both?

[wesm]

The vectorized function above doesn't work on regular NumPy arrays so this may be a NumPy bug. I am using 1.6.1

[craustin]

Yes, this works in 1.4.0. I'll close this issue as it's external.

[wesm]

Weird. i can create an issue on numpy/numpy on github

[lamblin]

I think you can work around the problem by specifying "dtype='object'":
print f.accumulate(df, dtype='object')