Missing operations for LinSpace. Fix #11049

[yuyichao]

Make it as closed to FloatRange as possible. Hopefully I'm using the right algorithm....

julia> linspace(0, 10, 20)
linspace(0.0,10.0,20)

julia> 0 + linspace(0, 10, 20) + 0
linspace(0.0,10.0,20)

julia> 0 - linspace(0, 10, 20) - 0
linspace(0.0,-10.0,20)

julia> 1 * linspace(0, 10, 20) * 1
linspace(0.0,10.0,20)

julia> linspace(0, 10, 20) / 1
linspace(0.0,10.0,20)

julia> -linspace(0, 10, 20)
linspace(-0.0,-10.0,20)

The divided by 0 case is a little different from FloatRange

julia> linspace(0, 10, 21) / 0
linspace(NaN,Inf,21)

julia> 0.0:0.5:10.0 / 0
0.0:0.5:Inf

julia> [linspace(0, 10, 21) / 0;]
21-element Array{Float64,1}:
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN
 NaN

julia> [0.0:0.5:10.0 / 0;]
ERROR: InexactError()
 in vcat at ./range.jl:650

IMHO both are wrong because

julia> [linspace(0, 10, 21);] / 0
21-element Array{Float64,1}:
 NaN
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf

julia> [0.0:0.5:10.0;] / 0
21-element Array{Float64,1}:
 NaN
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf
 Inf

But I'm waiting to see what ppl think...

[yuyichao]

Hmmm? The CI fails far before it reaches anything I changed when trying to get a version number??

[yuyichao]

Actually I think I need to handle the case when length() is 0 or 1...

[yuyichao]

• Fixed some type and the getindex for ranges.
• Added more tests.
• Rebased on latest master
• (Edit:) Also test the return type of getindex

@StefanKarpinski

[tkelman]

LGTM. I'm a little confused why both LinSpace and FloatRange exist now and aren't quite the same thing, but if we're going to have both, may as well have things tested and functioning.

[IainNZ]

Wooah yeah, I missed when LinSpace was a type, thats kinda funky

[yuyichao]

Well, yeah, that's the same question I asked in #11049.

It seems that they are using slightly different algorithms so in corner cases they may produce slightly different results (and IMHO the linspace one is better for small steps).

[timholy]

Background is #9666, #9637, but I didn't follow this in detail. @StefanKarpinski, since this is your domain it would be best if you reviewed this.

[timholy]

(But it all LGTM.)

[yuyichao]

@StefanKarpinski

[StefanKarpinski]

Yeah, seems right. Sorry about missing these. I'm still not fully convinced about this change, but since we currently have it, may as well have the more complete version.