def reset_feed_dates(daterange):
if not gtfs.has_table('feed_info'): return
gtfs_daterange = GTFSDateRange(daterange['start'], daterange['end'])
feed_info = gtfs.get_table('feed_info')
feed_info['feed_start_date'] = gtfs_daterange.start.datestring()
feed_info['feed_end_date'] = gtfs_daterange.end.datestring()
gtfs.update_table('feed_info', feed_info)
def remove_trips_with_nonexistent_calendars():
calendar = gtfs.get_table('calendar', index=False)
trips = gtfs.get_table('trips')
trips_filtered = trips[trips['service_id'].isin(calendar['service_id'])]
if (gtfs.has_table('frequencies')):
frequencies = gtfs.get_table('frequencies')
frequencies_filtered = frequencies[frequencies['trip_id'].isin(
trips_filtered.index.to_series())]
gtfs.update_table('frequencies', frequencies_filtered)
gtfs.update_table('trips', trips_filtered)
def filter_board_alight_by_daterange(daterange):
if not gtfs.has_table('board_alight'): return
board_alight = gtfs.get_table('board_alight', index=False)
if 'service_date' not in board_alight.columns: return
filter_daterange = GTFSDateRange(daterange['start'], daterange['end'])
board_alight['_inrange'] = board_alight.apply(
lambda row: filter_daterange.includes(row['service_date']), axis=1)
board_alight_filtered = board_alight[board_alight['_inrange']]
gtfs.update_table('board_alight', board_alight_filtered)
def filter_trips_by_date(date):
# removes trips that do not occur on specified date
# TODO consider replacing with filter_calendars_by_date, prune
trips_extended = get_trips_extended()
dow = GTFSDate(date).dow()
date_in_range = (trips_extended['start_date'] <=
date) & (date <= trips_extended['end_date'])
dow_in_service = trips_extended[dow] == GTFSBool.TRUE
trips_filter = date_in_range & dow_in_service
# filter calendar_dates for relevant calendar exceptions
if gtfs.has_table('calendar_dates'):
calendar_dates = gtfs.get_table('calendar_dates')
added_on_date = (calendar_dates['date'] == date) & (
calendar_dates['exception_type'] == GTFSExceptionType.ADDED)
services_added_on_date = calendar_dates[added_on_date]['service_id']
removed_on_date = (calendar_dates['date'] == date) & (
calendar_dates['exception_type'] == GTFSExceptionType.REMOVED)
services_removed_on_date = calendar_dates[removed_on_date][
'service_id']
service_added_on_date = trips_extended['service_id'].isin(
services_added_on_date)
service_removed_on_date = trips_extended['service_id'].isin(
services_removed_on_date)
if gtfs.has_table('calendar'):
trips_filter = (date_in_range & dow_in_service
& ~service_removed_on_date) | service_added_on_date
else:
trips_filter = service_added_on_date
trips_filtered_df = trips_extended[trips_filter]
gtfs.update_table('trips', trips_filtered_df)
def filter_calendar_dates_by_daterange(daterange):
if not gtfs.has_table('calendar_dates'): return
calendar_dates = gtfs.get_table('calendar_dates')
filter_daterange = GTFSDateRange(daterange['start'], daterange['end'])
calendar_dates['_gtfs_date'] = calendar_dates.apply(
lambda row: GTFSDate(row['date']), axis=1)
calendar_dates['_inrange'] = calendar_dates.apply(
lambda row: filter_daterange.includes(row['date']), axis=1)
calendar_dates_filtered = calendar_dates[calendar_dates['_inrange']]
gtfs.update_table('calendar_dates', calendar_dates_filtered)
def filter_calendars_by_daterange(daterange):
calendar = gtfs.get_table('calendar')
filter_daterange = GTFSDateRange(daterange['start'], daterange['end'])
calendar['_gtfs_daterange'] = calendar.apply(
lambda row: GTFSDateRange(row['start_date'], row['end_date']), axis=1)
calendar['_overlap'] = calendar['_gtfs_daterange'].apply(lambda dr: \
filter_daterange.get_overlap(dr) \
)
# we want to remove calendar entries that don't overlap DOWs
calendar['_dows_overlap'] = calendar.apply(lambda row: \
GTFSBool.TRUE in (row[dow] for dow in filter_daterange.days_of_week()),
axis=1
)
# we want to keep calendar entries that are used in overlapping exceptions
if gtfs.has_table('calendar_dates'):
calendar_dates = gtfs.get_table('calendar_dates')
calendar_dates['_date_overlap'] = calendar_dates.apply(
lambda row: filter_daterange.includes(row['date']), axis=1)
calendar_dates = calendar_dates[calendar_dates['_date_overlap']]
calendar['_exception_overlap'] = calendar.index.to_series().isin(
calendar_dates['service_id'])
else:
calendar['_exception_overlap'] = False
calendar = calendar[(calendar['_overlap'].notnull()
& calendar['_dows_overlap'])
| calendar['_exception_overlap']]
# trim bounds to fit within daterange
calendar['start_date'] = calendar['_overlap'].apply(
lambda dr: dr.start.datestring())
calendar['end_date'] = calendar['_overlap'].apply(
lambda dr: dr.end.datestring())
gtfs.update_table('calendar', calendar)
def get_feed_start_end_daterange():
if not gtfs.has_table('feed_info'): return None
feed_info = gtfs.get_table('feed_info')
return GTFSDateRange(feed_info.loc[0, 'feed_start_date'],
feed_info.loc[0, 'feed_end_date'])
def interpolate_stop_times():
# returns false if interpolation not possible
stop_times = gtfs.get_table('stop_times')
shapes = gtfs.get_table('shapes')
no_shape_dist_traveled = 'shape_dist_traveled' not in stop_times.columns \
or stop_times['shape_dist_traveled'].isna().all()
no_shapes_txt = not gtfs.has_table('shapes') or shapes.empty
if (no_shape_dist_traveled or no_shapes_txt):
return False
# build table with chunk information
df = stop_times.copy()
df['has_arrival'] = df['arrival_time'].notna()
df['has_departure'] = df['departure_time'].notna()
df = df[df['has_arrival'] | df['has_departure']]
timepoints_only = df[df['has_arrival'] | df['has_departure']]
# https://stackoverflow.com/questions/50411098/how-to-do-forward-rolling-sum-in-pandas
df['next_stop_sequence'] = timepoints_only.sort_values(by=['trip_id', 'stop_sequence']) \
.iloc[::-1] \
.groupby('trip_id')['stop_sequence'].transform(lambda x: x.rolling(2).max()) \
.iloc[::-1] \
# cleanup
df['next_stop_sequence'] = df['next_stop_sequence'].fillna(df['stop_sequence']).astype('int64')
df['stop_sequence_list'] = df.apply(lambda row: \
list(range(row['stop_sequence'], row['next_stop_sequence']) \
if row['stop_sequence'] != row['next_stop_sequence'] \
else [row['stop_sequence']] \
), axis=1)
df = df.explode('stop_sequence_list')
df = df.rename(columns={'stop_sequence': 'start_seq', 'next_stop_sequence': 'end_seq', 'stop_sequence_list': 'stop_sequence'})
chunks = df.set_index(['trip_id', 'stop_sequence']) \
[['start_seq', 'end_seq']]
stop_times = stop_times.set_index(['trip_id', 'stop_sequence'])
stop_times = stop_times.merge(chunks, \
how='left',
right_index=True,
left_index=True,
)
start_time = stop_times['departure_time'].rename('start_time')
end_time = stop_times['arrival_time'].rename('end_time')
start_sdt = stop_times['shape_dist_traveled'].rename('start_sdt')
end_sdt = stop_times['shape_dist_traveled'].rename('end_sdt')
stop_times = stop_times.merge(start_time, \
left_on=['trip_id', 'start_seq'],
right_index=True
)
stop_times = stop_times.merge(end_time, \
left_on=['trip_id', 'end_seq'],
right_index=True
)
stop_times = stop_times.merge(start_sdt, \
left_on=['trip_id', 'start_seq'],
right_index=True
)
stop_times = stop_times.merge(end_sdt, \
left_on=['trip_id', 'end_seq'],
right_index=True
)
def interpolate_row(row):
# happens if last stop or on 1-stop chunks (consecutive timepoints)
if (row['start_time'] == row['end_time']):
return row['start_time']
return seconds_to_military( \
seconds_since_zero(row['start_time']) + \
int(round( \
( \
(row['shape_dist_traveled'] - row['start_sdt']) / (row['end_sdt'] - row['start_sdt']) \
) * ( \
seconds_since_zero(row['end_time']) - seconds_since_zero(row['start_time']) \
) \
))
)
stop_times['interp'] = stop_times.apply(lambda row: interpolate_row(row), axis=1)
stop_times['arrival_time'] = stop_times['arrival_time'].fillna(stop_times['interp'])
stop_times['departure_time'] = stop_times['departure_time'].fillna(stop_times['interp'])
gtfs.update_table('stop_times', stop_times.reset_index(), cascade=False)
return True