no method for scalar divsion by Array{}

[brossetti]

It appears that there is no method for scalar division by an Array{}. Here is a quick example:

julia> x = 1.0:4.0
1.0:1.0:4.0

julia> y = [1.0, 2.0, 3.0, 4.0]
4-element Array{Float64,1}:
 1.0
 2.0
 3.0
 4.0

julia> 1+x
2.0:1.0:5.0

julia> 1+y
4-element Array{Float64,1}:
 2.0
 3.0
 4.0
 5.0

julia> 1-x
0.0:-1.0:-3.0

julia> 1-y
4-element Array{Float64,1}:
  0.0
 -1.0
 -2.0
 -3.0

julia> 1*x
1.0:1.0:4.0

julia> 1*y
4-element Array{Float64,1}:
 1.0
 2.0
 3.0
 4.0

julia> 1/x
4-element Array{Float64,1}:
 1.0
 0.5
 0.333333
 0.25

julia> 1/y
ERROR: MethodError: no method matching /(::Int64, ::Array{Float64,1})
Closest candidates are:
  /(::Union{Int128, Int16, Int32, Int64, Int8, UInt128, UInt16, UInt32, UInt64, UInt8}, ::Union{Int128, Int16, Int32, Int64, Int8, UInt128, UInt16, UInt32, UInt64, UInt8}) at int.jl:38
  /(::Union{Int16, Int32, Int64, Int8, UInt16, UInt32, UInt64, UInt8}, ::BigInt) at gmp.jl:381
  /(::T<:Integer, ::T<:Integer) where T<:Integer at int.jl:36

As you can see, / works when using StepRangeLen{}, but not for Array{}. It might be worth noting that 1./y works as expected.

I observed this behavior on:

Julia Version 0.6.2
Commit d386e40c17 (2017-12-13 18:08 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin14.5.0)
  CPU: Intel(R) Core(TM) i5-4288U CPU @ 2.60GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, haswell)

I'd be surprised if I'm the first person to notice this, so my apologies if this is a known issue.

[KristofferC]

On master:

julia> 1/x
1×4 Transpose{Float64,Array{Float64,1}}:
 0.0333333  0.0666667  0.1  0.133333

julia> 1/y
1×4 Transpose{Float64,Array{Float64,1}}:
 0.0333333  0.0666667  0.1  0.13333

Ok to close?

[brossetti]

Looks good to me. Thanks!

[mbauman]

Welcome! Just a quick note: this now means something very different from 1 ./ x. If you have more questions about this, please feel free to open a topic about this over at the Discourse discussion board.

[brossetti]

Thanks for the note, @mbauman. As a quick clarification, are you referring to the changes noted in the 0.6.0 release?

For every binary operator ⨳, a .⨳ b is now automatically equivalent to the broadcast call (⨳).(a, b). Hence, one no longer defines methods for .* etcetera. This also means that "dot operations" automatically fuse into a single loop, along with other dot calls f.(x) (#17623). Similarly for unary operators (#20249).

Operations like .+ and .* on Range objects are now generic broadcast calls (see above) and produce an Array. If you want a Range result, use + and *, etcetera (#17623).

[mbauman]

No, I mean to say that 1 / (1:4) is no longer equivalent to 1 ./ (1:4), and it returns something completely different in 0.7 than it did in 0.6. This is a new breaking change between 0.6 and 0.7.

Julia 0.6: Ranges have a special non-idiomatic behavior where 1/r is the same as 1 ./ r — it means elementwise division. Division of scalars by all other vectors is a missing method error, as you reported.

julia> 1/(1:4) # 0.6
4-element Array{Float64,1}:
 1.0
 0.5
 0.333333
 0.25

Julia 0.7: 1/x for any vector x computes the multiplicative inverse of the vector itself, akin to solving the equation 1 ≈ y*x for y. This is now consistent between ranges and all other vector types.

julia> y = 1/(1:4) # 0.7
1×4 LinearAlgebra.Transpose{Float64,Array{Float64,1}}:
 0.0333333  0.0666667  0.1  0.133333

julia> y*(1:4)
1.0