Update filter_data_by_segment() function: call chunk_episodes() inside the filter function

src/features/phone_activity_recognition/rapids/main.py

@@ -3,8 +3,6 @@ import numpy as np
 
 def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_segment, *args, **kwargs):
 
-    chunk_episodes = kwargs["chunk_episodes"]
-
     ar_episodes = pd.read_csv(sensor_data_files["sensor_episodes"])
     activity_classes = provider["ACTIVITY_CLASSES"]
 
@@ -18,10 +16,6 @@ def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_seg
     if not ar_episodes.empty:
         ar_episodes = filter_data_by_segment(ar_episodes, day_segment)
 
-        if not ar_episodes.empty:
-            # chunk episodes
-            ar_episodes = chunk_episodes(ar_episodes)
-
         if not ar_episodes.empty:
             ar_features = pd.DataFrame()
 

src/features/phone_battery/rapids/main.py

@@ -4,7 +4,6 @@ from datetime import datetime, timedelta, time
 def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_segment, *args, **kwargs):
     
     battery_data = pd.read_csv(sensor_data_files["sensor_episodes"])
-    chunk_episodes = kwargs["chunk_episodes"]
 
     # name of the features this function can compute
     base_features_names = ["countdischarge", "sumdurationdischarge", "countcharge", "sumdurationcharge", "avgconsumptionrate", "maxconsumptionrate"]
@@ -16,10 +15,6 @@ def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_seg
     if not battery_data.empty:
         battery_data = filter_data_by_segment(battery_data, day_segment)
 
-        if not battery_data.empty:
-            # chunk_episodes
-            battery_data = chunk_episodes(battery_data)
-
         if not battery_data.empty:
         
             battery_data["episode_id"] = ((battery_data.battery_status != battery_data.battery_status.shift()) | (battery_data.start_timestamp - battery_data.end_timestamp.shift() > 1)).cumsum()

src/features/phone_screen/rapids/main.py

@@ -34,7 +34,6 @@ def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_seg
     requested_episode_types = provider["EPISODE_TYPES"]
     ignore_episodes_shorter_than = provider["IGNORE_EPISODES_SHORTER_THAN"]
     ignore_episodes_longer_than = provider["IGNORE_EPISODES_LONGER_THAN"]
-    chunk_episodes = kwargs["chunk_episodes"]
 
     # name of the features this function can compute
     base_features_episodes = ["countepisode", "episodepersensedminutes", "sumduration", "maxduration", "minduration", "avgduration", "stdduration", "firstuseafter"]
@@ -50,10 +49,8 @@ def rapids_features(sensor_data_files, day_segment, provider, filter_data_by_seg
     if not screen_data.empty:
 
         screen_data = filter_data_by_segment(screen_data, day_segment)
-        if not screen_data.empty:
-            # chunk_episodes
-            screen_data = chunk_episodes(screen_data)
         
+        if not screen_data.empty:
             if ignore_episodes_shorter_than > 0:
                 screen_data = screen_data.query('@ignore_episodes_shorter_than <= duration')
             if ignore_episodes_longer_than > 0:

src/features/utils/utils.py

@@ -11,7 +11,12 @@ def filter_data_by_segment(data, day_segment):
         data["timestamps_segment"] = None
     else:
         data[["local_segment","timestamps_segment"]] = data["local_segment"].str.split(pat =";",n=1, expand=True)
-    return(data)
+    
+    # chunk episodes
+    if (not data.empty) and ("start_timestamp" in data.columns) and ("end_timestamp" in data.columns):
+        data = chunk_episodes(data)
+    
+    return data
 
 # Each minute could fall into two segments.
 # Firstly, we generate two rows for each resampled minute via resample_episodes rule: