Repository navigation
Invalid promotion of numpy scalar to float64 (Trac #346) #944
Description
Original ticket http://projects.scipy.org/numpy/ticket/346 on 2006-10-14 by @charris, assigned to unknown.
In [95]: a = np.float32(1)
In [96]: a.dtype
Out[96]: dtype('float32')
In [97]: a /= 2
In [98]: a.dtype
Out[98]: dtype('float64')
Result type of a should still be float32. This also holds for +=, -=, and *=.
Activity
numpy-gitbot commentedon Oct 19, 2012
@teoliphant wrote on 2006-10-16
This is not necessarily true. Array scalars are not considered special. They are just scalars. When two scalars are involved in an operation, the normal rules of coercion apply. Different rules would require a change to NumPy that needs to wait until 1.1
This is intended behavior.