在 Julia 中重新解释指针

Question

I have some C++ code that has existing python bindings, and I am trying to graft it on to Julia using PyCall. One of the functions that gets called generates a pointer to a 2D array in memory to which I would like to wrap a Julia array around so that I can add/subtract/multiply by scalars, etc. I know the size of the array, and it's currently represented as a PyObject to which I can do x_ptr[1] , x_ptr[2] and get correct values out. But I'd like to have an array, x .

Answer 1

Here's a simple example using numpy array interface . By default, PyCall converts numpy arrays to Julia's arrays, but we can prevent that with @pycall and the ::Any annotation to show how you'd do it manually. You need to dig into the object to find that pointer.

julia> obj = @pycall numpy.reshape(numpy.arange(20), (4,5))::Any
PyObject array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]])

julia> array_interface = obj.__array_interface__
Dict{Any,Any} with 6 entries:
  "shape"   => (4, 5)
  "strides" => nothing
  "typestr" => "<i8"
  "data"    => (4705395216, false)
  "descr"   => Tuple{String,String}[("", "<i8")]
  "version" => 3

julia> array_interface["data"] # This is the actual pointer!
(4705395216, false)

julia> unsafe_wrap(Matrix{Int}, Ptr{Int}(array_interface["data"][1]), reverse(array_interface["shape"]))
5×4 Array{Int64,2}:
 0  5  10  15
 1  6  11  16
 2  7  12  17
 3  8  13  18
 4  9  14  19

Of course, I don't know how your python object stores its pointer. You'll have to dig into your Python object to find that pointer yourself.