fix schema.parse_columns function

mindspore/dataset/engine/datasets.py

@@ -2464,47 +2464,53 @@ class Schema:
         Parse the columns and add it to self.
 
         Args:
-            columns (dict or list[str]): names of columns.
+            columns (dict or list[dict]): dataset attribution information, decoded from schema file.
+            if list: columns element must be dict, 'name' and 'type' must be in keys, 'shape' optional.
+            if dict: columns.keys() as name, element in columns.values() is dict, and 'type' inside, 'shape' optional.
+              example 1)
+                [{'name': 'image', 'type': 'int8', 'shape': [3, 3]},
+                 {'name': 'label', 'type': 'int8', 'shape': [1]}]
+              example 2)
+                {'image': {'shape': [3, 3], 'type': 'int8'}, 'label': {'shape': [1], 'type': 'int8'}}
 
         Raises:
-            RuntimeError: If failed to parse schema file.
+            RuntimeError: If failed to parse columns.
-            RuntimeError: If unknown items in schema file.
+            RuntimeError: If unknown items in columns.
             RuntimeError: If column's name field is missing.
             RuntimeError: If column's type field is missing.
         """
-        if columns is None:
-            raise TypeError("Expected non-empty dict or string list.")
         self.columns = []
-        for col in columns:
+        if isinstance(columns, list):
-            name = None
+            for column in columns:
-            shape = None
+                try:
-            data_type = None
+                    name = column.pop("name")
-            col_details = None
+                except KeyError:
-            if isinstance(columns, list):
+                    raise RuntimeError("Column's name is missing")
-                col_details = col
+                try:
-                if "name" in col:
+                    de_type = column.pop("type")
-                    name = col["name"]
+                except KeyError:
-            elif isinstance(columns, dict):
+                    raise RuntimeError("Column' type is missing")
-                col_details = columns[col]
+                shape = column.pop("shape", None)
-                name = col
+                column.pop("t_impl", None)
-            else:
+                column.pop("rank", None)
-                raise RuntimeError("Error parsing the schema file")
+                if column:
-
+                    raise RuntimeError("Unknown field {}".format(",".join(column.keys())))
-            for k, v in col_details.items():
+                self.add_column(name, de_type, shape)
-                if k == "shape":
+        elif isinstance(columns, dict):
-                    shape = v
+            for key, value in columns.items():
-                elif k == "type":
+                name = key
-                    data_type = v
+                try:
-                elif k in ("t_impl", "rank"):
+                    de_type = value.pop("type")
-                    pass
+                except KeyError:
-                else:
+                    raise RuntimeError("Column' type is missing")
-                    raise RuntimeError("Unknown field %s" % k)
+                shape = value.pop("shape", None)
-
+                value.pop("t_impl", None)
-            if name is None:
+                value.pop("rank", None)
-                raise RuntimeError("Column's name field is missing.")
+                if value:
-            if data_type is None:
+                    raise RuntimeError("Unknown field {}".format(",".join(value.keys())))
-                raise RuntimeError("Column's type field is missing.")
+                self.add_column(name, de_type, shape)
-            self.add_column(name, data_type, shape)
+        else:
+            raise RuntimeError("columns must be dict or list, columns contain name, type, shape(optional).")
 
     def from_json(self, json_obj):
         """