Fix Parquet loading with columns (#6160)

* Fix Parquet loading with `columns`

* Style

Author: mariosasko

src/datasets/metric.py

@@ -65,7 +65,7 @@ def _release(self):
 # lists - summarize long lists similarly to NumPy
 # arrays/tensors - let the frameworks control formatting
 def summarize_if_long_list(obj):
-    if not type(obj) == list or len(obj) <= 6:
+    if not type(obj) == list or len(obj) <= 6:  # noqa: E721
         return f"{obj}"
 
     def format_chunk(chunk):

src/datasets/packaged_modules/parquet/parquet.py

@@ -51,7 +51,7 @@ def _split_generators(self, dl_manager):
                     with open(file, "rb") as f:
                         features = datasets.Features.from_arrow_schema(pq.read_schema(f))
                         if self.config.columns is not None:
-                            self.info.features = datasets.Features(
+                            features = datasets.Features(
                                 {col: feat for col, feat in features.items() if col in self.config.columns}
                             )
                         self.info.features = features

src/datasets/utils/py_utils.py

@@ -1141,7 +1141,7 @@ def save_function(pickler, obj):
                     state_dict["__module__"] = obj.__module__
 
                 state = obj.__dict__
-                if type(state) is not dict:
+                if type(state) is not dict:  # noqa: E721
                     state_dict["__dict__"] = state
                     state = None
                 if state_dict:
@@ -1303,7 +1303,7 @@ def save_function(pickler, obj):
                 state_dict["__module__"] = obj.__module__
 
             state = obj.__dict__
-            if type(state) is not dict:
+            if type(state) is not dict:  # noqa: E721
                 state_dict["__dict__"] = state
                 state = None
             if state_dict: