[ENH] Add fAssumedShapeDT to support the allocatable array of derived types

[vincent-skywalker]

Description of the new feature
Allows fortran function to return an allocatable array of derived types.

Proposed Fortran code
An example of what the Fortran code looks like

    function func_return_allocatable_s_struct_nested_2() result(s)
        type(s_struct_nested_2),allocatable::s(:)
        allocate(s(2))
        s(1)%a_int = 123
        s(1)%f_nested%a_int = 234
        s(1)%f_nested%f_struct%a_int = 345
        s(2)%a_int = 456
        s(2)%f_nested%a_int = 567
        s(2)%f_nested%f_struct%a_int = 678
    end function func_return_allocatable_s_struct_nested_2

Proposed Python code
Sorry, I am not familiar.

[vincent-skywalker]

I tried and give up. It seems too complex. It is on my fork's branch fAssumedShapeDT.

Is there a plan to add this feature? @rjfarmer

[rjfarmer]

Yes I agree they are very complicated :)

There are plans but I don't know when I'll have time. They probably first need the assumed shape array code to be re written to more agnostic about what it's an array of. Before then trying to fold derived types in.

For now your beat bet would be a fixed sized array of derived types, and just size it to be as large as the largest size you need.