I have defined a subclass of AbstractArray for which I do *not* want to use
the default AbstractArray printing machinery in show.jl . My assumption,
apparently incorrect, is that I simply needed to overload show.
Unfortunately, my overloaded version is never used.
Here is some sample code. Note two things:
- The code contained within a named module, and Base.Array is not
imported, so there is not a name conflict here.
- The function eleltype is indeed defined, and returns T. The actual
eltype of this array is Scalar{T}. But the data is not stored as a
contiguous memory buffer containing Scalar{T} objects; it has been
compressed in a certain way.
show(io::IO,x::Array) = print( io, "CVX $(eleltype(x)) $(ndims(x))D Array
..." )
show{T,N}(io::IO,::Type{Array{T,N}}) = print( io, "CVX $(T) $(N)D Array" )
show(io::IO,::Type{Array}) = print( io, "CVX Array" )
The last two lines work as expected, suggesting I'm indeed exporting if I
type CVX.Array or CVX.Array{Float64,2} into the REPL, my custom show
commands are called. But if I attempt to instantiate an instance, and show
that, it still calls the show.jl code; my overloaded version is ignored.
Here's what I get:
2x2 Array{Float64,2}:
#undef #undef
#undef #undef
And here is the output of an explicit call to show(A)
CVX Float64 2D Array ...
I tried modifying that first line as follows
show{T,N}(io::IO,x::Array{T,N}) = print( io, "CVX $(T) $(N)D Array ..." )
but that didn't make a difference. I tried different combinations of
import/export
show as well. As you probably know I do have to import it, but I don't have
to export (at least, I don't think so), since I'm overloading. I also tried
overloading print as well.
Any hints? Thanks in advance for the help. I'm using Julia 0.3rc1 on the
Mac...
