Unexpected type when using comprehensions

[ghost]

Usually I get a type from comprehensions which I expected, except for the last example below. I find it weird that y becomes an array of Any. I was expecting an array of Float64. I was able to get an array of Float64 by defining x as a const. However that will not be possible in all use cases. How can I avoid this 'weird' behaviour (without using const x or const x2 = copy(x)? Or is this a bug perhaps?

julia> y = [i for i = 1:3]
3-element Array{Int64,1}:
 1
 2
 3

julia> y = [rand() for i = 1:3]
3-element Array{Float64,1}:
 0.263477 
 0.0517146
 0.678627 

julia> x = rand(3)
3-element Array{Float64,1}:
 0.679907
 0.559466
 0.874267

julia> y = [x[i] for i = 1:3]
3-element Array{Any,1}:
 0.679907
 0.559466
 0.874267

[kmsquire]

Inferring types in array comprehensions has frequently been brittle, but
Jeff seems to be improving things when he can. I suggest filing an issue.

In the mean time, assuming the type is always the same, you can get around
it by declaring the type of the array:

julia> x = rand(3)
3-element Array{Float64,1}:
0.205122
0.0720471
0.664416

julia> y = Float64[x[i] for i = 1:3]
3-element Array{Float64,1}:
0.205122
0.0720471
0.664416

Kevin

On Thu, Oct 3, 2013 at 3:23 PM, Carlos Baptista [email protected]:

Usually I get a type from comprehensions which I expected, accept for the
last example below. I find it weird that y becomes an array of Any. I was
expecting an array of Float64. I was able to get an array of Float64 by
defining x as a const. However that will not be possible in all use
cases. How can I avoid this 'weird' behaviour (without using const x or const
x2 = copy(x)? Or is this a bug perhaps?

julia> y = [i for i = 1:3]
3-element Array{Int64,1}:
1
2
3

julia> y = [rand() for i = 1:3]
3-element Array{Float64,1}:
0.263477
0.0517146
0.678627

julia> x = rand(3)
3-element Array{Float64,1}:
0.679907
0.559466
0.874267

julia> y = [x[i] for i = 1:3]
3-element Array{Any,1}:
0.679907
0.559466
0.874267

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4430
.

[johnmyleswhite]

Interesting point: this problem is related to being in the global scope and doesn't happen inside of functions.

julia> function foo()
       x = [1, 2, 3]
       return [x[i]^2 for i in 1:3]
       end
foo (generic function with 1 method)

julia> foo()
3-element Array{Int64,1}:
 1
 4
 9

I think the problem is that, in the global scope, the compiler doesn't feel certain of the type of x.

[kmsquire]

Thanks, John--I've forgotten this point more than once!

On Thu, Oct 3, 2013 at 3:36 PM, John Myles White
[email protected]:

Interesting point: this problem is related to being in the global scope
and doesn't happen inside of functions.

julia> function foo()
x = [1, 2, 3]
return [x[i]^2 for i in 1:3]
end
foo (generic function with 1 method)

julia> foo()
3-element Array{Int64,1}:
1
4
9

I think the problem is that, in the global scope, the compiler doesn't
feel certain of the type of x.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4430#issuecomment-25664147
.

[ghost]

Thanks Kevin. That did the trick. And likewise I keep on forgetting to check things out in functions. With a background in MATLAB and Python, I still have to get used to that.

[JeffBezanson]

related: #524, #964