Hi,

I want to constraint some matrix that is given as a linear functions of the variables to be semi-definite positive. However, I found something strange :

```
function get_hankels(s,a,b)
m = length(s)-1
if iseven(m)
# m = 2n
n = Int(m//2)
H_low = [s[i+j-1] for i in 1:n+1,j in 1:n+1]
H_up = [(a+b)*s[i+j] - s[i+j+1] - a*b*s[i+j-1] for i in 1:n,j in 1:n]
else
# m = 2n+1
n = Int((m-1)//2)
H_low = [s[i+j] - a*s[i+j-1] for i in 1:n+1,j in 1:n+1]
H_up = [b*s[i+j-1] - s[i+j] for i in 1:n+1,j in 1:n+1]
end
return H_low,H_up
end
# To see the matrices I want to use :
using DynamicPolynomials
m=6
@polyvar x[1:m+1]
H_low,H_up = get_hankels(x,0,1)
display(H_low)
display(H_up)
# Now with convex variables, I want to build the constraints :
using Convex
m=6
s = Variable(m+1)
H_low, H_up = get_hankels(s,0,1) # works
cstr1 = (H_low âȘ° 0) # does not work.
cstr1 = ([s[1] s[2] s[3] s[4]; s[2] s[3] s[4] s[5]; s[3] s[4] s[5] s[6]; s[4] s[5] s[6] s[7]] âȘ° 0) # works
# Same for H_up
cstr2 = (H_up âȘ° 0)
```

I have the follwoing error message :

```
julia> cstr1 = (H_low âȘ° 0) # does not work.
ERROR: MethodError: no method matching âȘ°(::Matrix{Convex.IndexAtom}, ::Int64)
Closest candidates are:
âȘ°(::Convex.AbstractExpr, ::Union{Number, AbstractArray}) at C:\Users\u009192\.julia\packages\Convex\ubaUN\src\constraints\sdp_constraints.jl:115
âȘ°(::Union{Number, AbstractArray}, ::Convex.AbstractExpr) at C:\Users\u009192\.julia\packages\Convex\ubaUN\src\constraints\sdp_constraints.jl:124
Stacktrace:
[1] top-level scope
@ REPL[58]:1
julia>
```

Indeed, the type is not the same :

```
julia> typeof([s[1] s[2] s[3] s[4]; s[2] s[3] s[4] s[5]; s[3] s[4] s[5] s[6]; s[4] s[5] s[6] s[7]])
Convex.TransposeAtom
julia> typeof(H_low)
Matrix{Convex.IndexAtom} (alias for Array{Convex.IndexAtom, 2})
julia>
```

Is there a way to build this `H_low`

matrix automatically in the right type ? The size of the problem might change from a run to the otherâŠ

@ericphanson Do you now the right syntax ?