Fixed data_frame_tool to handle category columns correctly

src/python/tests_extended/data_frame_tool.py

@@ -112,7 +112,7 @@ def _process_input_list(self, df, input_metas, require):
                 if (df[input_meta.name].dtype) == 'datetime64[ns]':
                     input_array = np.array([dt.timestamp() for dt in df[input_meta.name]]).astype(np.int64)
                 elif (str(df[input_meta.name].dtype)) == 'category':
-                    input_array = np.array([dt + 1 for dt in df[input_meta.name]]).astype(np.uint32) # in ONNX models trained in ML.NET "categorical columns" are 1 based, not 0 based                    
+                    input_array = np.array([key + 1 for key in df[input_meta.name].array.codes]).astype(np.uint32)  # in ONNX models trained in ML.NET input coming from "categorical columns" is 1 based indices, whereas Categorical columns save indices that are 0 based, and that need to be retrieved from .array.codes                    
                 else:
                     # With strings we must cast first to np.object then then reshape
                     # so we do it for everything