Adds overload for NumpyToTensorView and io::ReadMAT so that the shape…

… of the array can be obtained from numpy

include/matx/core/file_io.h

@@ -219,6 +219,44 @@ void ReadMAT(TensorType &t, const std::string fname,
   pb->NumpyToTensorView(t, v);
 }
 
+/**
+ * @brief Read a MAT file into a tensor view
+ *
+ * MAT files use SciPy's loadmat() function to read various MATLAB file
+ * types in. MAT files are supersets of HDF5 files, and are allowed to
+ * have multiple fields in them.
+ *
+ * @tparam TensorType
+ *   Data type of tensor
+ * @param t
+ *   Tensor to read data into
+ * @param fname
+ *   File name of .mat file
+ * @param var
+ *   Variable name inside of .mat to read
+ *
+ **/
+template <typename TensorType>
+auto ReadMAT(const std::string fname,
+             const std::string var)
+{
+
+  if (!std::filesystem::exists(fname)) {
+    const std::string errorMessage = "Failed to read [" + fname + "], Does not Exist";
+    MATX_THROW(matxIOError, errorMessage.c_str());
+  }
+
+  MATX_NVTX_START("", matx::MATX_NVTX_LOG_API)
+
+  auto pb = std::make_unique<detail::MatXPybind>();
+
+  auto sp = pybind11::module_::import("scipy.io");
+  auto obj = (pybind11::dict)sp.attr("loadmat")("file_name"_a = fname);
+  auto v = obj[var.c_str()];
+
+  return pb->NumpyToTensorView<TensorType>(v);
+}
+
 /**
  * @brief Write a MAT file from a tensor view
  *

include/matx/core/pybind.h

@@ -32,7 +32,7 @@
 #pragma once
 
 #include "matx/core/type_utils.h"
-
+#include "matx/core/make_tensor.h"
 
 #if MATX_ENABLE_PYBIND11
 
@@ -358,6 +358,26 @@ class MatXPybind {
     }
   }
 
+  template <typename TensorType>
+  auto NumpyToTensorView(const pybind11::object &np_ten)
+  {
+    using T = typename TensorType::scalar_type;
+    constexpr int RANK = TensorType::Rank();
+    using ntype = matx_convert_complex_type<T>;
+    auto ften = pybind11::array_t<ntype, pybind11::array::c_style | pybind11::array::forcecast>(np_ten);
+
+    auto info = ften.request();
+
+    assert(info.ndim == RANK);
+
+    std::array<matx::index_t, RANK> shape;
+    std::copy_n(info.shape.begin(), RANK, std::begin(shape));
+
+    auto ten =  make_tensor<T> (shape);
+    std::copy(ften.data(), ften.data() + ften.size(), ten.Data() );
+    return ten;
+  }
+
   template <typename TensorType>
   auto TensorViewToNumpy(const TensorType &ten)
   {

test/00_io/FileIOTests.cu

@@ -142,6 +142,37 @@ TYPED_TEST(FileIoTestsNonComplexFloatTypes, MATWriteRank5)
   // Read "myvar" from mat file
   io::WriteMAT(t, "test_write.mat", "myvar");
   io::ReadMAT(t2, "test_write.mat", "myvar");
+  for (index_t i = 0; i < t.Size(0); i++) {
+    for (index_t j = 0; j < t.Size(1); j++) {
+      for (index_t k = 0; k < t.Size(2); k++) {
+        for (index_t l = 0; l < t.Size(3); l++) {
+          for (index_t m = 0; m < t.Size(4); m++) {
+            ASSERT_EQ(t(i,j,k,l,m), t2(i,j,k,l,m));
+          }
+        }
+      }
+    }
+  }
+  MATX_EXIT_HANDLER();
+}
+
+TYPED_TEST(FileIoTestsNonComplexFloatTypes, MATWriteRank5GetShape)
+{
+  MATX_ENTER_HANDLER();
+
+  auto t = make_tensor<TypeParam>({2,3,1,2,3});
+  tensor_t<TypeParam,5> t2;
+
+  randomGenerator_t<TypeParam> gen(t.TotalSize(), 0);
+  auto random = gen.GetTensorView(t.Shape(), UNIFORM);
+  (t = random).run();
+
+  cudaDeviceSynchronize();
+
+  // Read "myvar" from mat file
+  io::WriteMAT(t, "test_write.mat", "myvar");
+  t2.Shallow(io::ReadMAT<decltype(t2)>("test_write.mat", "myvar"));
+
   for (index_t i = 0; i < t.Size(0); i++) {
     for (index_t j = 0; j < t.Size(1); j++) {
       for (index_t k = 0; k < t.Size(2); k++) {