Merge pull request #1363 from cal-itp/vp-hackathon

Create dataset package as zipped folder for hackathon

Author: tiffanychu90

gtfs_funnel/create_hackathon_datasets.py

@@ -0,0 +1,262 @@
+"""
+Grab a set of Cal-ITP tables for Posit hackathon in Feb 2025.
+
+Grab GTFS schedule and vp tables from GCS bucket and 
+subset to the operators. Concatenate across dates.
+
+Our GTFS analytics data catalog: 
+https://github.com/cal-itp/data-analyses/blob/main/_shared_utils/shared_utils/gtfs_analytics_data.yml
+"""
+import geopandas as gpd
+import glob
+import pandas as pd
+import os
+import shutil
+
+from segment_speed_utils import time_series_utils
+from update_vars import GTFS_DATA_DICT
+from shared_utils import rt_dates
+
+OUTPUT_FOLDER = "./hackathon/"
+CRS = "EPSG:4326"
+
+def merge_crosswalk(
+    df: pd.DataFrame, 
+    crosswalk: pd.DataFrame
+) -> pd.DataFrame:
+    """
+    Drop feed_key after merging in schedule_gtfs_dataset_key.
+    Use this identifier when merging in vp.
+    """
+    df2 = pd.merge(
+        df,
+        crosswalk,
+        on = "feed_key",
+        how = "inner"
+    ).drop(columns = "feed_key") 
+    
+    return df2
+
+
+def export_gtfs_schedule_tables(
+    operator_list: list,
+    date_list: list
+):
+    """
+    Export subset of downloaded schedule tables (minimally processed in our warehouse),
+    by inputting a list of operator names.
+    """
+    schedule_tables = GTFS_DATA_DICT.schedule_downloads
+    
+    trips = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = schedule_tables.dir,
+        dataset_name = schedule_tables.trips,
+        date_list = date_list,
+        data_type = "df",
+        get_pandas = True,
+        filters = [[("name", "in", operator_list)]], 
+        columns = ["feed_key", "gtfs_dataset_key", "name",
+                   "service_date", "trip_id", "trip_instance_key",
+                   "route_key", "route_id", "route_type",
+                   "route_short_name", "route_long_name", "route_desc",
+                   "direction_id", 
+                   "shape_array_key", "shape_id", 
+                   "trip_first_departure_datetime_pacific", 
+                   "trip_last_arrival_datetime_pacific",
+                   "service_hours"
+                  ]
+    ).rename(
+        columns = {
+            "gtfs_dataset_key": "schedule_gtfs_dataset_key",
+    })
+    
+    subset_feeds = trips.feed_key.unique().tolist()
+    crosswalk = trips[["feed_key", "schedule_gtfs_dataset_key"]].drop_duplicates()
+    
+    shapes = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = schedule_tables.dir,
+        dataset_name = schedule_tables.shapes,
+        date_list = date_list,
+        data_type = "gdf",
+        get_pandas = True,
+        filters = [[("feed_key", "in", subset_feeds)]], 
+        columns = ["feed_key",
+                   "shape_array_key", "shape_id", 
+                   "n_trips", "geometry"
+                  ]
+    ).pipe(
+        merge_crosswalk, crosswalk
+    ).to_crs(CRS)  
+    
+    stops = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = schedule_tables.dir,
+        dataset_name = schedule_tables.stops,
+        date_list = date_list,
+        data_type = "gdf",
+        get_pandas = True,
+        filters = [[("feed_key", "in", subset_feeds)]], 
+        columns = [
+            "feed_key", "service_date",
+            "stop_id", "stop_key", "stop_name", 
+            "geometry"
+        ]           
+    ).pipe(
+        merge_crosswalk, crosswalk
+    ).to_crs(CRS) 
+    
+    stop_times = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = schedule_tables.dir,
+        dataset_name = schedule_tables.stop_times,
+        date_list = date_list,
+        data_type = "df",
+        get_pandas = True,
+        filters = [[("feed_key", "in", subset_feeds)]], 
+        columns = ["feed_key", 
+                   "trip_id", "stop_id",
+                   "stop_sequence", "arrival_sec", 
+                   "timepoint"
+                  ]     
+    ).pipe(merge_crosswalk, crosswalk)  
+    
+    trips.to_parquet(f"{OUTPUT_FOLDER}trips.parquet")
+    shapes.to_parquet(f"{OUTPUT_FOLDER}shapes.parquet")
+    stops.to_parquet(f"{OUTPUT_FOLDER}stops.parquet")
+    stop_times.to_parquet(f"{OUTPUT_FOLDER}stop_times.parquet")
+
+    return 
+
+
+def export_gtfs_vp_table(
+    schedule_gtfs_dataset_key_list: list,
+    date_list: list
+):
+    """
+    Export subset of downloaded vp (deduped in our warehouse),
+    by inputting a list of schedule_gtfs_dataset_keys.
+    """
+    vp_tables = GTFS_DATA_DICT.speeds_tables
+    
+    vp = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = vp_tables.dir,
+        dataset_name = vp_tables.raw_vp,
+        date_list = analysis_date_list,
+        data_type = "gdf",
+        get_pandas = True,
+        filters = [[("schedule_gtfs_dataset_key", "in", schedule_gtfs_dataset_key_list)]], 
+        columns = [
+            "gtfs_dataset_name", "gtfs_dataset_key", "schedule_gtfs_dataset_key", 
+            "trip_id", "trip_instance_key",
+            "location_timestamp_local",
+            "geometry"]
+    ).rename(
+        columns = {"gtfs_dataset_key": "vp_gtfs_dataset_key"}
+    ).to_crs(CRS)
+    
+    vp.to_parquet(f"{OUTPUT_FOLDER}vp.parquet")
+    
+    return
+
+    
+def export_segments_table(
+    schedule_gtfs_dataset_key_list: list,
+    date_list: list
+):
+    """
+    Export subset of cut segments from gtfs_segments.create_segments,
+    by inputting a list of schedule_gtfs_dataset_keys.
+    """
+    vp_tables = GTFS_DATA_DICT.rt_stop_times
+    
+    segments = time_series_utils.concatenate_datasets_across_dates(
+        gcs_bucket = vp_tables.dir,
+        dataset_name = vp_tables.segments_file,
+        date_list = analysis_date_list,
+        data_type = "gdf",
+        get_pandas = True,
+        filters = [[("schedule_gtfs_dataset_key", "in", schedule_gtfs_dataset_key_list)]],
+        columns = ["trip_instance_key", 
+                   "schedule_gtfs_dataset_key", "route_id", "direction_id",
+                   "stop_pair", "stop_id1", "stop_id2", 
+                   "stop_sequence", "geometry"
+                  ]
+    ).to_crs(CRS)
+    
+    segments.to_parquet(f"{OUTPUT_FOLDER}segments.parquet")
+
+    return
+
+
+def export_roads_for_district(district: int = 7):
+    """
+    Export subset roads in District 7, which is where
+    all the operators are.
+    """
+    SHARED_GCS = GTFS_DATA_DICT.gcs_paths.SHARED_GCS
+    
+    roads = gpd.read_parquet(
+        f"{SHARED_GCS}all_roads_2020_state06.parquet"
+    ).to_crs(CRS)
+    
+    caltrans_districts = gpd.read_parquet(
+        f"{SHARED_GCS}caltrans_districts.parquet",
+        filters = [[("DISTRICT", "==", district)]]
+    ).to_crs(CRS)
+    
+    roads_in_district = gpd.sjoin(
+        roads, 
+        caltrans_districts,
+        how = "inner",
+        predicate = "within"
+    ).drop(columns = ["DISTRICT", "index_right"])
+    
+    roads_in_district.to_parquet(f"{OUTPUT_FOLDER}district7_roads.parquet")
+    
+    return
+    
+    
+
+if __name__ == "__main__":
+    
+    analysis_date_list = [
+        rt_dates.DATES[f"{m}2024"] for m in ["oct", "nov"]
+    ]
+    
+    operators = [
+        "Big Blue Bus Schedule",
+        "Culver City Schedule",
+        "BruinBus Schedule",
+        "Santa Clarita Schedule", 
+        "LA DOT Schedule",
+    ]
+    
+    # Schedule tables - trips, shapes, stops, stop_times
+    export_gtfs_schedule_tables(
+        operators, analysis_date_list
+    )
+    
+    schedule_keys = pd.read_parquet(
+        f"{OUTPUT_FOLDER}trips.parquet"
+    ).schedule_gtfs_dataset_key.unique().tolist()
+           
+    # vehicle positions table
+    export_gtfs_vp_table(
+        schedule_keys, analysis_date_list
+    )
+    
+    # segments for mapping
+    export_segments_table(
+        schedule_keys, analysis_date_list
+    )    
+    
+    # roads for D7, which is where these operators are
+    export_roads_for_district(district = 7)
+    
+    # Zip up folder of parquets within OUTPUT_FOLDER -- download this and send
+    # do not check into GitHub
+    shutil.make_archive("gtfs_assembled_hackathon", "zip", OUTPUT_FOLDER)
+    
+    # Delete local parquets, just keep zipped folder to download and send
+    for f in glob.glob(f"{OUTPUT_FOLDER}*.parquet"):
+        os.remove(f)
+    
+    

rt_segment_speeds/segment_speed_utils/time_series_utils.py

@@ -18,6 +18,7 @@
 import gcsfs
 fs = gcsfs.GCSFileSystem()
 
+'''
 def concatenate_datasets_across_dates(
     gcs_bucket: str,
     dataset_name: str,
@@ -56,8 +57,8 @@ def concatenate_datasets_across_dates(
         df = compute(df)[0]    
     
     return df
+'''
 
-"""
 def concatenate_datasets_across_dates(
     gcs_bucket: str,
     dataset_name: str,
@@ -75,11 +76,12 @@ def concatenate_datasets_across_dates(
         date_list, 
         data_type = data_type,
         get_pandas=get_pandas,
-        add_date = True
+        add_date = True, 
+        **kwargs
     )  
 
     return df
-"""
+
 
 def clean_standardized_route_names(
     df: pd.DataFrame,