Improve clean_feed() when shape_id not available

[ethan-moss]

Description of the Feature to be Added

clean_feed() is a shallow wrapper around gtfs_kit's clean() function which implements 4 cleaning functions:

1. clean_ids(): strip whitespace from all string IDs and then replace every remaining whitespace chunk with an underscore
2. clean_times(): convert H:MM:SS time strings to HH:MM:SS time strings to make sorting by time work as expected.
3. clean_route_short_names(): In feed.routes, assign ‘n/a’ to missing route short names and strip whitespace from route short names. Then disambiguate each route short name that is duplicated by appending ‘-’ and its route ID. Note: this is the method that fixes the "Repeated pair (route_short_name, route_long_name)" warning.
4. drop_zombies(): does the following:
   • Drop stops of location type 0 or NaN with no stop times.
   • Remove undefined parent stations from the parent_station column.
   • Drop trips with no stop times.
   • Drop shapes with no trips.
   • Drop routes with no trips.
   • Drop services with no trips.

clean_feed() and hence clean() will fail if there is no shape_id column in trip.txt. However, drop_zombies() is the only one that relies on that column, the 3 other cleaning functions work fine:

(OPTIONAL) Suggested Implementations

Instead of returning without cleaning if shape_id is not present, this could be improved to instead only action the first 3 cleaning methods.

[CBROWN-ONS]

Note to self:
To replicate error, delete shape_id from feed.trips

[CBROWN-ONS]

UPDATE

This issue will not be completed before the end of sprint 4 due to it relying on other PRs that are yet to be merged. This issue should however be completed early on in sprint 5.

[r-leyshon]

Migrated to datasciencecampus/assess_gtfs#6