134 api returns empty array when valid trips should exist

[clukas1]

Addressing issue raised, now there is more logic to convert gtfs transfers to raptor transfers with clear hierarchies.

• Additional Transfers generated by a Transfer Generator have lowest priority.
• Child/Parent Transfers are derived on a stop-stop basis in raptor. E.g. if parent stop holds transfers, stop A1 (child) will get all transfers from stop A (parent). See test cases to see all potential cases.
• Explicit Transfers: Even if parent transfers are defined in GTFS, if there is a direct transfer (e.g. A1-A2) is specified in GTFS, transfers derived from (A-A) will be overwritten.

In the process of implementation, I've also addressed a bug I found in our old transfer conversion. Where both additional transfers and gtfs transfers were all added (allowing duplicates), which in the end resulted in always using the shortest transfer, regardless where it was defined e062dc8.

[clukas1]

Thanks, @clukas1, really nice work!

I do have a few style and documentation suggestions.

Please improve the documentation, especially clarifying what "derive" transfers means in the context of parent and child stops. It would be helpful to describe the general idea, similar to what you outlined in the issue response, somewhere in the Javadoc of the method that encapsulates this logic.

Thanks!

Thank you for your review @munterfi!

I've addressed all of your comments, especially re-writing the docstrings. A thing I've noticed while writing, is that we don't cover all potential cases (some parent/sibling transfers) but this might actually also be better this way and this is something the transfer generators are a good alternative for. To clarify some examples:

Stops: (note in this example parent stops also have departures) A with Children A1, A2 B with Children B1, B2 C with Children C1, C2

Defined Transfers: A-A A1-A1 A-B A1-C1

Resulting A1 transfers: A1-A A1-A1 A1-A2 (from A-A) A1-B (from A-B) A1-B1 (from A-B) A1-B2 (from A-B) A1-C1

Not Covered but potentially possible transfers: A1-C (from A1-C1) A1-C2 (from A1-C2)

Hope you agree that adding more sibling parent logic would kind of be over interpreting transfers defined (or not defined) in GTFS.