DOC: Warn about the difference between np.remainder and math.remainder #9702
Conversation
| * Python 3.7's `math.remainder` and C's ``remainder``, which | ||
| computes the IEEE remainder, which are the complement to | ||
| ``round(x1 / x2)``. | ||
| * Matlab's ``rem``, which is the complement to ``int(x1 / x2)``. |
There was a problem hiding this comment.
Does this need the (TM) appendix? Just curious because the original version had one.
There was a problem hiding this comment.
That was accidental - but doing a search seems to reveal that we don't use a TM sign anywhere else.
There was a problem hiding this comment.
There are other places, but all in the ufunc documentation. It looks like we don't need it for the function documentation, https://www.mathworks.com/brandguide/editorial/products.html, but we might want to put a copyright and trademark notice somewhere in the documentation, I don't know where.
There was a problem hiding this comment.
Note that the mathworks style guide wants it in caps and without the possessive suffix.
There was a problem hiding this comment.
Ok, I've switched to all caps
Python 3.7 adds its own `math.remainder` with different semantics to ours. We can't change the semantics now, but we can warn users that they are different.
c27e0db to
429f99e
Compare
|
Thanks Eric. |
|
Is there a reason that we didn't just name this function As an aside, should we add an |
|
The |
|
The IEEE remainder might have an interesting interaction with the rounding mode, but should have good precision on both sides of zero for NumPy round to even, unlike floor remainder. |
Python 3.7 adds its own
math.remainderwith different semantics to ours.We can't change the semantics now, but we can warn users that they are different.