Bugfix in Horner's scheme #79
Conversation
| @@ -372,7 +372,7 @@ function polyval{T,S}(p::Poly{T}, x::S) | |||
| if lenp == 0 | |||
| return zero(R) * x | |||
| else | |||
| y = convert(R, p[end]) + zero(T)*x | |||
| y = convert(R, p[end]) | |||
There was a problem hiding this comment.
I don't understand why zero(T)*x is here. If there is any reason, perhaps changing to zero(T)*zero(x)?
There was a problem hiding this comment.
actually, I do not see any benefit in using zero(T)*zero(x). since we have convert(R, p[end]) where R = promote_type(T,S), zero(x) will be automatically promoted to R. For normal numbers, I mean standard Numbers, zero(x) is not needed at all. Then, if there is a custom type Type <: Number, it might make sense to use zero(x) and rely on the conversion rules among Numbers.
What do you think?
Do you think we can push these issues (i.e., Horner's bug, Base./ and perhaps polyder) this weekend?
There was a problem hiding this comment.
I feel the same way, but didn't write this so may be missing something. As for time, yes if no one chimes in I'll merge the first two in a few days.
Closes #78.
|
Thanks! |
Closes #78.