How to bind a 2D vector with a fixed dimension with Pybind11?

[YuqiHu]

Hi there! I am trying to write some test codes to see if pybind11 can fulfill all our needs to bind our C++ program.

The goal is to bind a struct to include the following data:

#define MACROSIZE 5

struct MyStruct {
    // arr[MACROSIZE][dynamic]
    std::vector<int> arr[MACROSIZE];
}

I tried to bind a fixed-size C-array by array_t. I tried to bind the vector by the bind_vector. Both worked. Following the same logic, I tried to bind it as a 2D vector but it seems not to work. I tried to use a combination of array_t and bind_vector and it still reports me error. I found it works if I include get and set functions within the structure like the following

py::array_t<int> getDataElement(int index) {
    std::vector<int>& vec = arr[index];
    py::array_t<int> result(vec.size());
    auto buffer = result.request();
    int* ptr = static_cast<int*>(buffer.ptr);
    std::copy(vec.begin(), vec.end(), ptr);
    return result;
}

void setDataElement(int index, const py::array_t<int>& newData) {
    py::buffer_info bufInfo = newData.request();
    int* newDataPtr = static_cast<int*>(bufInfo.ptr);

    size_t newDataSize = bufInfo.size;
    arr[index].clear();
    arr[index].reserve(newDataSize);

    for (size_t i = 0; i < newDataSize; ++i) {
        arr[index].push_back(newDataPtr[i]);
}

Since the model has a large amount of data and can be edited quite often so we would like to avoid copying the data as what we have now in this process. We also want to fix one of the dimensions to reduce workload. I have been stuck on this problem for a few days since I figured out how to bind all out types we need. I appreciate it if you could me some insights or suggestions!