ENH: Add datatime64 support in np.interp #14990
Conversation
|
For other's reviewing - previously this raised |
| yp = np.array([1, 3]) | ||
| x = np.datetime64('2019-01-02') | ||
| assert_equal(np.interp(x, xp, yp), 2.) | ||
| assert_raises(TypeError, np.interp(1., xp, yp)) |
There was a problem hiding this comment.
I was expecting this test to fail, but it seems to pass.. Any idea which line of your code causes the TypeError?
There was a problem hiding this comment.
It should fail.
np.interp(1., xp, yp)
doesn't raise this error. It might be a problem with assert_raises and it's similar to #15008
But do we want to check if types in x and xp are the same? Maybe it's enough that both can be cast to np.float64?
There was a problem hiding this comment.
My bad, it should have been
assert_raises(TypeError, np.interp, 1., xp, yp)
But again, is this a better approach?
|
This PR would make np.interp support any type that can be converter to float64. In my opinion that's a better approach. Happy to receive any feedback.. |
ENH: Add support for types convertible to float64 in np.interp
Closes #14941