segment name mismatch in HighwayGraphCollapsedEdgeInfo error report

[yakra]

DataProcessing/siteupdate/python-teresco/siteupdate.py

Lines 1349 to 1350 in 791e6be

	if edge1.segment_name != edge2.segment_name:
	print("ERROR: segment name mismatch in HighwayGraphCollapsedEdgeInfo: edge1 named " + edge1.segment_name + " edge2 named " + edge2.segment_name + "\n")

Example output:
ERROR: segment name mismatch in HighwayGraphCollapsedEdgeInfo: edge1 named OR18,OR99W edge2 named OR99W

This could be added as a fatal error to the ErrorList.
To more easily find the highway data that needs to be fixed, we can print out a list of all waypoints colocated at that vertex.

[yakra]

Test with HighwayData @ a46d6d9b7ca2f2f452f7e3ebe7123e109e51ced2

[yakra]

...or would #360 make this all moot?

[jteresco]

Haven't seen/heard about this in a long time. Anyone else?

[yakra]

There's some overlap with #91 here. These are caused by HIDDEN_TERMINUS errors.
b642e06 (unhide hidden termini in graphs) should prevent the name mismatch from actually occurring I believe, but I'd like to checkout the commit above atsome point and take a closer look at how things behave.
Low priority though.

[yakra]

b642e06 (unhide hidden termini in graphs) should prevent the name mismatch from actually occurring I believe

Nope LOL what was I thinking. That only unhides at vertices with <2 incident edges.
This error happens during compression, i.e. at vertices with exactly 2 incident edges.

[yakra]

Playing around with some concurrency detection optimizations produced:
ERROR: segment name mismatch in HGEdge collapse constructor: edge1 named AH70 edge2 named AH70,E40.
This could be anywhere on their concurrency in RUS or KAZ except maybe from Astrakhan to the KAZ border, as the bug probably requres >2 concurrent routes at a location.
Kind of surprised there was just the one error and not loads more of these.

To more easily find the highway data ... we can print out a list of all waypoints colocated at that vertex.

A couple years after the OP, 315f834 made that harder to do.
A few options:

1. Bring back first_waypoint.
   This would increase sizeof(HGVertex). This could be offset by nixing lat & lng and replacing them with functions. Slightly slower, unless the decreased size is a win now that vertices are stored sequentially.
   The easiest option to implement.
2. Pass the quadtree to the collapse ctor and find colocated points via lat & lng.
   Possibly the most efficient option; all the happy path has to do is put the QT in a register. Heck, it may already be there?
3. Throw an exception and handle this in the HighwayGraph ctor where we have access to the QT.
   Not a good option if I wanna make any of this permanent; it's guaranteed to break the edge collapse process.