Piecewise fix for 0d #4792
Piecewise fix for 0d #4792
Conversation
| if n == n2-1: # compute the "otherwise" condition. | ||
| totlist = condlist[0] | ||
| for k in range(1, n): | ||
| totlist |= condlist[k] |
There was a problem hiding this comment.
this could be a np.logical_or.reduce(condlist, axis=0) I think
There was a problem hiding this comment.
I'm not used to functional thinking but definitely seems legit.
|
thanks this should probably still go into 1.9, can probably also be done post beta. please squash the commits together and write a commit message like this (with BUG: in the title): |
|
I'm making the changes you suggested. After that, do I have to locally squash all the commits and do a |
|
yes e.g. |
|
Ready, waiting for Travis to give the green light. |
When `x` has more than one element the condlist `[True, False]` is being made equivalent to `[[True, False]]`, which is correct. However, when `x` is zero dimensional the expected condlist is `[[True], [False]]`: this commit addresses the issue. Besides, the documentation stated that there could be undefined values but actually these are 0 by default: using `nan` would be desirable, but for the moment the docs were corrected. Closes numpy#331.
|
There, seems to be an issue with scalar domains and three or more conditions: >>> np.piecewise(3, [True, False, False],[4,2,0])
ValueError: function list and condition list must be the same Funny enough it works fine with two conditions (this is what has been tested): >>> np.piecewise(3, [True, False],[4,2])
array(4)The three conditions version works fine with select: >>> np.select([True, False, False],[4,2,0])
array(4) |
|
I won't be able to fix this for some time, I fear :( |
|
The changes applied from this pull request don't seem to work well with multidimensional arrays that have an extra "otherwise" condition: >>> x = array([[-2.5, -1.5, -0.5],
[ 0.5, 1.5, 2.5]])
>>> np.piecewise(x, [x < 0, x >= 2], [-1, 1, 3])
ValueError: all the input arrays must have same number of dimensionsThe same code produces the following result in numpy 1.8.2: >>> array([[-1., -1., -1.],
[ 3., 3., 1.]])If you reshape the array to a single dimension, it works as intended. Also, I noticed there aren't any multidimensional test cases, so maybe this can be one of them? |
|
@jakobgager, @bastula Can you convert your comments here into issues? It is the only way to not lose track of them. |
| @@ -1505,6 +1506,15 @@ def test_0d(self): | |||
| assert_(y.ndim == 0) | |||
| assert_(y == 0) | |||
|
|
|||
| x = 5 | |||
| y = piecewise(x, [[True], [False]], [1, 0]) | |||
There was a problem hiding this comment.
This doesn't make any sense. If x is 0d, the conditions should be 0d too, as [True, False]
…h a single boolean is ok Also fix the test added in numpygh-4792, which didn't make sense, but passed anyway
I provide a fix for #331. Actually little changes were needed, but I had to carefully think about the assumptions being made by the function. Some additional tests were added and the behavior remains backwards compatible.
Besides, I corrected the documentation because there was an inconsistent bit, see 0d1f7d3.