Is 0 .< A .< 1 going away?

Internals & Design
1

Is 0 .< A .< 1 (where A is an array) going away in 0.6 and beyond? Would be good to know.

Reason for asking: version 0.6.0-dev.1924 (Windows 64) warns that this is deprecated.

/Paul S

2

I suspect the answer is “as far you’re concerned, no, it’s not going away.”

I say it that way because I’m interpreting your question as “can I write this expression and have it evaluate to the same thing”, but the deprecation message you’re getting is likely related to deprecation of .< as a generic function that will be replaced with dot-broadcast syntax. But that’s a change in how that expression gets evaluated, not in what it evaluates to.

2 Likes
3

I think you’ve found a bug.

julia> 0 .< eye(4) .< 1
WARNING: A::AbstractArray & B::AbstractArray is deprecated, use A .& B instead.
Stacktrace:
 [1] depwarn(::String, ::Symbol) at .\deprecated.jl:64
 [2] &(::BitArray{2}, ::BitArray{2}) at .\deprecated.jl:50

So it’s not the .< that is deprecated. But it looks like that chained comparison is getting expanded to (0 .< A) & (A .< 1), and the & there is now deprecated in favor of .&. Whatever is responsible for that expansion (I think it’s lowering but I could be wrong) wasn’t updated appropriately, and we must not have any chained comparisons of arrays in our tests otherwise we would have noticed by now.

4 Likes