Hi Todd,

I'm not sure I understood your data completely, but I wonder what you will get 
if you change this line:


complex_t** iqData = new complex_t*[nelems];


to:


complex_t* iqData = new complex_t[nelems];


and read without the for loop...


Binh-Minh


________________________________
From: Hdf-forum <[email protected]> on behalf of Todd 
Dobmeyer <[email protected]>
Sent: Monday, July 11, 2016 3:10 PM
To: [email protected]
Subject: [Hdf-forum] Reading Nested Compound objects in C++

All

I have an HDF5 file I need to read in a C++ application. I performed an h5dump 
on the data and have the following structure (with many items removed that I am 
not interested in). I have successfully read in all 7500 values of "num_samps" 
using the C++ HDF5 library. What I am struggling to figure out how to read is 
the array of compound "iq_data". I know in the datatype it says there are 1000 
compound objects. I don't know if I can get this value out of the DataType, but 
I do know the num_samps points to this array size as well, which is fine since 
I can read this. Can any of you help me figure out how to read the "iq_data" 
array. Below is the sample data followed by my code I have so far. In my code 
"numSamps" is good, but "iqData" is not holding anything useful. I am sure I am 
missing something.

Thanks for your help!
Todd Dobmeyer

--------------SAMPLE DATA-------------------------

HDF5 "hdfFile.be01" {
GROUP "/" {
   DATATYPE "DRdata" H5T_COMPOUND {
      H5T_ARRAY { [5] H5T_STD_U32LE } "guid";
      H5T_STD_I32LE "version";
      ...
      H5T_STD_U32LE "num_samps";
      ...
      H5T_STD_U32LE "sequence_dummy";
      H5T_ARRAY { [1000] H5T_COMPOUND {
         H5T_IEEE_F32LE "f_i";
         H5T_IEEE_F32LE "f_q";
      } } "iq_data";
   }
   DATASET "DRx1data" {
      DATATYPE  "/DRx1"
      DATASPACE  SIMPLE { ( 7500 ) / ( H5S_UNLIMITED ) }
      DATA {
      (0): {
            [ 352430272, 25602, 2314, 1442149219, 43629 ],
            1,
            1.44215e+09,
            1442149219,
            0,
            1,
            1,
            0,
            1,
            0,
            0,
            0,
            3,
            0,
            0,
            0,
            0,
            1,
            31095000,
            20000,
            0,
            1000,
            1000, // (this is num_samps)
            1,
            1000,
            [ {
                  1.09068e+09,
                  4.53469e+08
               }, {
                  4.03303e+08,
                  4.11846e+08
               }, {
                  -1.35295e+08,
                  -1.70107e+08
               }, ...

--------------------CODE------------------



typedef struct {

    float f_i;

    float f_q;

} complex_t;


int main(int argc, char *argv[])

{

    std::string path = "/data/hdfFile.be01";

    // Open the file, dataset, and dataspace

    H5::H5File file(path, H5F_ACC_RDONLY);

    H5::DataSet dataset = file.openDataSet("DRx1data");

    H5::DataSpace dataspace = dataset.getSpace();



    // Get the number of elements and number of samples

    int nelems = dataspace.getSimpleExtentNpoints();

    int numSamps[nelems];



    // Create number of samples type

    H5::CompType sampsType(sizeof(uint));

    sampsType.insertMember("num_samps", 0, H5::PredType::STD_U32LE);



    // Read the number of samples

    dataset.read(numSamps, sampsType);



    // Now try to read the I/Q data

    H5::CompType iqDataType(sizeof(complex_t));

    iqDataType.insertMember("f_i", HOFFSET(complex_t, f_i), 
H5::PredType::NATIVE_FLOAT);

    iqDataType.insertMember("f_q", HOFFSET(complex_t, f_q), 
H5::PredType::NATIVE_FLOAT);

    complex_t** iqData = new complex_t*[nelems];

    for(int i = 0 ; i < nelems ; ++i)

    {

        iqData[i] = new complex_t[numSamps[i]];

    }

    dataset.read(iqData, iqDataType);
_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5

Reply via email to