Should I expect that the length of a range is preserved under arithmetic operations on the range? I ran into a bug where simply changing the sign in front of a range in an expression caused a dimension mismatch error. For example:
julia> collect((22:-1:7)-0.0030026930763747783)
15-element Array{Float64,1}:
21.997
20.997
19.997
18.997
17.997
16.997
15.997
14.997
13.997
12.997
11.997
10.997
9.997
8.997
7.997
julia> collect((22:-1:7))-0.0030026930763747783
16-element Array{Float64,1}:
21.997
20.997
19.997
18.997
17.997
16.997
15.997
14.997
13.997
12.997
11.997
10.997
9.997
8.997
7.997
6.997
After some experimentation, I discovered, at least in this example, if the range is floating point, I get the expected behavior. But then you get the weird property that the following are not the same. Also, I don’t know if the length preservation holds in general.
# Should these be the same?
((22:-1:7)*1.0)-0.0030026930763747783
(22:-1:7)-0.0030026930763747783