[NumPy-Tickets] [NumPy] #1890: piecewise throws TypeError for int input and sin

[NumPy Trac]

#1890: piecewise throws TypeError for int input and sin
--------------------+-------------------------------------------------------
 Reporter:  deil    |       Owner:  somebody   
     Type:  defect  |      Status:  new        
 Priority:  normal  |   Milestone:  Unscheduled
Component:  Other   |     Version:  1.6.0      
 Keywords:          |  
--------------------+-------------------------------------------------------

Comment(by deil):

 Replying to [comment:2 charris]:
 > If you truly want integer output, then you should use
 np.sin(x).astype(x.dtype).

 I want float output, I now use this, which works fine:
 {{{
 >>> np.piecewise(x.astype(float), [x < 5, x >= 5], [lambda x: -x, lambda
 x: np.sin(x)])
 }}}

 I didn't know that numpy automatically casts if the rhs is a list, but not
 if it is an array:
 {{{
 a = np.zeros(10, dtype=int)
 a[[1,4]] = [1.1, 4] # Casts to int
 a[[1,4]] = np.array([1.1, 4]) # TypeError
 }}}

 I imagine there are good reasons for not automatically casting the array,
 but throwing a TypeError?!
 I just had never run into this 'problem' of having to cast myself  and
 couldn't figure out why piecewise was throwing the error.

 Btw., this is the second ticket I filed for numpy which turned out to be
 missing understanding on my part.
 In the future, should I ask on the mailing list instead?

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1890#comment:3>
NumPy <http://projects.scipy.org/numpy>
My example project