data class DefaultType(override val dataType: DataType) : DataTypeRef.Type {
        override fun toString(): String = dataType.toString()
    }

    @Serializable
    @SerialName("dataTypeRefName")
    data class DefaultName(override val fqName: FqName) : DataTypeRef.Name {
        override fun toString(): String = fqName.simpleName
    }
}


val DataType.ref: DataTypeRef

//   * an attribute named "type" containing the TF_DataType of the output tensor
//
// Given an input tensor of shape [...], if the input DataType "T" is larger
// than the output DataType "type", then the shape changes from [...]
// to [..., sizeof(T)/sizeof(type)].
//
// If "T" is smaller than "type", the operator requires that the rightmost

  llvm::ArrayRef<llvm::StringRef> input_names;
  llvm::ArrayRef<
      std::pair<tensorflow::DataType, tensorflow::PartialTensorShape>>
      input_specs;
  llvm::ArrayRef<llvm::StringRef> input_devices;

  // The following are metadata for outputs.
  llvm::ArrayRef<llvm::StringRef> output_names;
  llvm::ArrayRef<
      std::pair<tensorflow::DataType, tensorflow::PartialTensorShape>>
      output_specs;

  bool ref_variable = 2;
  bool incompatible_variable = 3;
  bool incompatible_attribute = 4;
  bool control_flow_v1 = 5;
  string method_name = 6;

  // TODO(chky): add more checks, eg. tensor datatypes.
}

message CompatibilityAnalysisProto {
  CompatibilityAnalysisReportProto summary = 1;

  message OpInfo {
    int32 count = 1;

    CompatibilityAnalysisReportProto report = 2;
  }

}
Status TapeOperation::SetAttrTypeList(const char* attr_name,
                                      const DataType* values, int num_values) {
  forward_op_.attrs.Set(attr_name,
                        gtl::ArraySlice<const DataType>(values, num_values));
  return parent_op_->SetAttrTypeList(attr_name, values, num_values);
}
Status TapeOperation::SetAttrBoolList(const char* attr_name,

 public:
  explicit MlirTensor(Value value)
      : TracingTensorHandle(kMlir), value_(value) {}

  tensorflow::DataType DataType() const override {
    tensorflow::DataType type;
    Status s = ConvertToDataType(value_.getType(), &type);
    if (!s.ok()) {
      return tensorflow::DT_INVALID;
    }
    return type;
  }

  tensorflow::Status Shape(

  static TensorBuffer* Buffer(const Tensor& tensor) { return tensor.buf_; }
  static Tensor MakeTensor(TF_DataType type, const TensorShape& shape,
                           TensorBuffer* buf) {
    return Tensor(static_cast<DataType>(type), shape, buf);
  }
};

// Allocates tensor data buffer using specified allocator.
// `operation` is a name for this operation.
void* allocate_tensor(const char* operation, size_t len, Allocator* allocator);

  Status SetAttrFloat(const char* attr_name, float value) override;
  Status SetAttrBool(const char* attr_name, bool value) override;
  Status SetAttrType(const char* attr_name, DataType value) override;
  Status SetAttrShape(const char* attr_name, const int64_t* dims,
                      const int num_dims) override;
  Status SetAttrFunction(const char* attr_name,

func (m Migrator) DataTypeOf(field *schema.Field) string {
	fieldValue := reflect.New(field.IndirectFieldType)
	if dataTyper, ok := fieldValue.Interface().(GormDataTypeInterface); ok {
		if dataType := dataTyper.GormDBDataType(m.DB, field); dataType != "" {
			return dataType
		}
	}

	return m.Dialector.DataTypeOf(field)
}

// FullDataTypeOf returns field's db full data type

  std::vector<int64_t> dimvec(num_dims);
  for (int i = 0; i < num_dims; ++i) {
    dimvec[i] = static_cast<int64_t>(dims[i]);
  }

  Tensor ret(static_cast<tensorflow::DataType>(dtype),
             tensorflow::TensorShape(dimvec), buf);
  buf->Unref();
  size_t elem_size = TF_DataTypeSize(dtype);
  if (elem_size > 0 && len < (elem_size * ret.NumElements())) {
    return nullptr;
  }