Avoid cumulative real-time updates

[t2gran]

Summary

Realtime updates (at least Siri ET) should in general not be treated as cumulative updates. This make the logic complicated, you need to account for all kind of updates when applying a new on top of others. In the case of a malicious update, resending a ok one will not fix the problem as well.

This PR changes the stop-pattern logic so that a update is applied on top of the planned stop-pattern and not the last version from the previous update. There are a few places where this happens, so I included a few commits witch revert the changes if needed. The plan is either to drop those commits or to squash them. (I have removed these commits now, since no one requested them).

I would like to see som tests of this with various updators:

• Siri (Entur)
• Siri (Skaanetrafiken)
• GTFS (HSL)
• StopConsolidation - IBI (Sandbox)

Issue

At Entur we applied the following update several time and OTP toggled between the expected state (Stop 1,,2,3 CANCELED) and not (Stop 1,2,3 SCHEDULED).

Netex data set - One line
Siri update

To reproduce the problem, you can start OTP with the given data set and use the following build-config.json

"updaters": [
 {
      "type": "siri-et-updater",
      "url": "file://[PATH-TO-LOCAL-FOLDER-WITH-SIRI-XML-FILE]",
      "frequency": "3s",
      "previewInterval": "3h",
      "feedId": "RB",
      "blockReadinessUntilInitialized": true,
      "fuzzyTripMatching": true
     }
  ]

Possible challanges

Existing systems may relay on cumulative updates.

Unit tests

✅ One unit tests is added, and a few updated.

Documentation

Only JavaDoc

Changelog

✅

Bumping the serialization version id

🟥 Not needed

[codecov]

Codecov Report

Attention: Patch coverage is 80.88235% with 13 lines in your changes are missing coverage. Please review.

Project coverage is 67.71%. Comparing base (e35ac67) to head (ed3d05a).
Report is 56 commits behind head on dev-2.x.

Files	Patch %	Lines
...tripplanner/transit/model/network/StopPattern.java	66.66%	6 Missing and 1 partial ⚠️
...tripplanner/transit/model/network/TripPattern.java	33.33%	2 Missing and 2 partials ⚠️
.../opentripplanner/ext/siri/ModifiedTripBuilder.java	84.61%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             dev-2.x    #5705      +/-   ##
=============================================
+ Coverage      67.69%   67.71%   +0.01%     
- Complexity     16369    16410      +41     
=============================================
  Files           1890     1896       +6     
  Lines          71812    71964     +152     
  Branches        7407     7419      +12     
=============================================
+ Hits           48611    48728     +117     
- Misses         20692    20726      +34     
- Partials        2509     2510       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[leonardehrenfried]

I'm frankly suprised that the updates are cumulative and would have expected (at least in the GTFS-RT case) it not to be the case, so I think this is the correct behaviour.

[leonardehrenfried]

I've tested this with several IBI data sets and all worked fine, including the stop consolidation.

However, I don't think anybody was ever relying on updates to stop patterns being cumulative.

[jtorin]

Found some minor spelling while reading through the PR.