Is it possible to access a pointer to a structure within another structure via libpointer?

Hello,

I am using an external c/c++ shared library from within MATLAB.

Please see the code below:

%  c/c++ interface function
%  bool oa_get_pattern_objects ( int instance_id , oa_object_array * pReturned_objects )

    function [found object_xyrecords] = get_pattern_objects(instance_id)

    oa_object_array.m_objects = {};
    oa_object_array.m_size = 0;
    pOA_pattern_objects = libpointer('oa_object_array',oa_object_array);

    if calllib('OpenAccessCore','oa_get_pattern_objects',instance_id,pOA_pattern_objects) == true
        found = true;

        proxy_objects = get(pOA_pattern_objects,'Value');

        size  = proxy_objects.m_size;

%THIS IS ILLEGAL

        setdatatype(proxy_objects.m_objects,'oa_xyrecordPtr',1,size)

        object_xyrecords = proxy_objects.m_objects.Value;
    else
        found = false;
        object_xyrecords = [];
    end

I cannot use setdatatype to inform MATLAB that the field, m_objects, points to an array of structures of the same type.

Perhaps I could get a pointer to the first element in the array and manually convert them in MATLAB one-by-one. Is there a way to do this?

My real desire is to give MATLAB the information it needs and allow it to convert all the elements for me more efficiently and elegantly than I can. Is there a way I can do this?

For a simple c-structure like below is it possible to access the elements in the array pointed to by the pointer, sc_array, by any means in MATLAB?

struct simple_class
{
    int            my_id;
    simple_class * sc_array;
};

Any thoughts are greatly appreciated :-)

Answer 1

[shon]

I've discovered a solution that works but I don't really understand why that is.

I can create records in which a record may contain different structures. If these records are to be initialized in C/C++ then structures must precede any PODs in the record's definition.

struct file
{
   char *path_;
}

struct record
{
   file f_;
   int  file_size_;  // POD cannot be defined before any struct
}

MATLAB appears to be "reshaping" the memory for structs in MATLAB code.

If a struct is placed after a POD then it's fields are not aligned in memory as I would desire as I cannot access the fields of that struct and confirm the right data is where it should be.

If I define all the structs before PODs then for some reason everything works out okay.

This answer is incomplete as I do not fully understand what is happening or why but instead have only discovered how to circumvent the affects of what is happening so that I can work with MATLAB. It would be wonderful to understand just what IS going on.

[Walter Roberson]

Sorry, what is "POD" in this context?

[Philip Borghesani]

What release of MATLAB are you using? There have been alignment bugs with structure layout when there were imbedded structure. I believe they were all fixed by release R2010a.

It may be possible for you to work around this by specifing a structure packing in your header file start with 8 but you can try other values:

#pragma pack 8

[Philip Borghesani]

Walter POD is a c/c++term it stands for "plain old data" but has a much more specific meaning in the specifications.

[shon]

Thank you Philip. I'm using the "plain old data" definition for POD.

I've been using R2012a.

It may be possible to use #pragma pack 8 to set my alignment boundaries. I had not thought of that. Thank you. I'll try to keep this in mind going forward but I'm still a little weary.

What is MATLAB doing to my memory or should I be asking how is MATLAB interpreting my memory layout?