style: apply ruff format

src/python/parla/common/dataflow.py

@@ -103,7 +103,7 @@ def process_crosspys(
                                 _out.append((parray, i))
                 else:
                     raise TypeError(
-                       f"Invalid Type: {type(element)}. Dataflow should be PArray, CrossPyArray, or Tuple[PArray, int]"
+                        f"Invalid Type: {type(element)}. Dataflow should be PArray, CrossPyArray, or Tuple[PArray, int]"
                     )
         return _out
 

src/python/parla/common/parray/coherence.py

@@ -27,14 +27,14 @@ class MemoryOperation:
 
     # Flag
     SWITCH_DEVICE_FLAG = (
-        101
-    )  # if the flag is set, it means dst is not the current device
+        101  # if the flag is set, it means dst is not the current device
+    )
     LOAD_SUBARRAY = (
-        102
-    )  # if the flag is set, it means a subarray of src should be loaded
+        102  # if the flag is set, it means a subarray of src should be loaded
+    )
     ENSURE_IS_COMPLETE = (
-        103
-    )  # if the flag is set, check data will also check if the data is complete
+        103  # if the flag is set, check data will also check if the data is complete
+    )
 
     def __init__(self, inst: int = NOOP, dst: int = -1, src: int = -1, flag: int = []):
         self.inst = inst
@@ -125,9 +125,7 @@ def __init__(self, init_owner: int, num_gpu: int, cyparray_state: CyPArrayState)
         self._is_complete[CPU_INDEX] = None
 
         self._local_states[init_owner] = self.MODIFIED  # initial state is MODIFIED
-        self.owner = (
-            init_owner
-        )  # the device that has the complete copy (take the role of main memory)
+        self.owner = init_owner  # the device that has the complete copy (take the role of main memory)
         self._versions[init_owner] = 0  # the first version is 0
         self._is_complete[init_owner] = True  # the copy is complete
         self._latest_version = 0  # the latest version in the system

src/python/parla/common/parray/core.py

@@ -170,13 +170,11 @@ def get_array(self, device_idx: Optional[int] = None) -> ndarray:
         """
 
         if device_idx is None:
-            device_idx = self._current_device_index 
+            device_idx = self._current_device_index
 
         if self._slices:  # so this is a sub-parray object
             # index into origin array by saved slices
-            ret = self._array.get_by_global_slices(
-                device_idx, self._slices[0]
-            )
+            ret = self._array.get_by_global_slices(device_idx, self._slices[0])
             for s in self._slices[1:]:
                 ret = ret[s]
             return ret
@@ -214,10 +212,9 @@ def _current_device_index(self) -> int:
             # to avoid import gpu context, which is slow to setup.
             return device.device.id  # device.device should be a cupy.cuda.Device object
 
-
     # Public API:
 
-    def get(self, device: Optional[PyDevice] = None) -> 'np.ndarray' | 'cp.ndarray':
+    def get(self, device: Optional[PyDevice] = None) -> "np.ndarray" | "cp.ndarray":
         if device is None:
             return self.array
         else:

src/python/parla/utility/graphs.py

@@ -251,8 +251,8 @@ class RunConfig:
     """
 
     outer_iterations: int = (
-        1
-    )  # Number of times to launch the Parla runtime and execute the task graph
+        1  # Number of times to launch the Parla runtime and execute the task graph
+    )
     # Number of times to execute the task graph within the same Parla runtime
     inner_iterations: int = 1
     inner_sync: bool = False  # Whether to synchronize after each kernel launch
@@ -504,7 +504,7 @@ def get_task_properties(line: str):
 
 
 def parse_blog(
-    filename: str = "parla.blog"
+    filename: str = "parla.blog",
 ) -> Tuple[Dict[TaskID, TaskTime], Dict[TaskID, List[TaskID]]]:
     try:
         result = subprocess.run(