Closed dmcclean closed 8 years ago
I think the appropriate semantics are:
nroot (-n) x = nroot n (recip x)
which is consistent with (**)
(by definition of nroot
) and dimensionally sound, no? What is the problem you see with this?
At first I was concerned that it allowed a zeroth root, but I was wrong about that, it doesn't.
Then I left the question anyway, because I couldn't find any sources anywhere talking about negative roots of things (hard to find with all the noise about roots of negative things).
That is consistent with (**)
and with the implementation we have. I'm not sure if it is dimensionally sound or not because I can't think of anything to compare it to.
I think we can safely keep those semantics, but add a documentation note explaining them?
OK, this agrees with what wolfram alpha does.
What are the intended semantics of
I'd suggest that there probably are none, and that the
Root
type family may need to be changed to not define negative roots ofDimension
s.