collapsed=traveled edge check, compressed edge ctor

[yakra]

DataProcessing/siteupdate/cplusplus/classes/GraphGeneration/HighwayGraph.cpp

Lines 132 to 135 in 0100297

	else if ((wv.second->incident_c_edges.front() == wv.second->incident_t_edges.front()
	&& wv.second->incident_c_edges.back() == wv.second->incident_t_edges.back())
	|| (wv.second->incident_c_edges.front() == wv.second->incident_t_edges.back()
	&& wv.second->incident_c_edges.back() == wv.second->incident_t_edges.front()))

DataProcessing/siteupdate/python-teresco/siteupdate.py

Lines 1633 to 1634 in 0100297

	if (v.incident_c_edges[0] == v.incident_t_edges[0] and v.incident_c_edges[1] == v.incident_t_edges[1]) \
	or (v.incident_c_edges[0] == v.incident_t_edges[1] and v.incident_c_edges[1] == v.incident_t_edges[0]):

Just check format mask, whether set for both formats, shouldn't matter whether incident_c_edges or incident_t_edges.
3 Python, 6 C++

[yakra]

Tried 2 different approaches that don't work. Time to waste a load of time figuring out why.

[yakra]

The 1st approach differed a bit from the OP's "check format mask, whether set for both formats" suggestion:

if (wv.second->visibility == 1)
	new HGEdge(wv.second, HGEdge::collapsed, numthreads);
else if (wv.second->incident_c_edges.front()->format == wv.second->incident_t_edges.back()->format)
	new HGEdge(wv.second, HGEdge::collapsed | HGEdge::traveled, numthreads);
else // yadda yadda

Failed because:
• AK2@+SuzAve gets processed first.
• It's hidden and in use. Thus it gets collapsed into a collapsed edge, but not a traveled edge.
• +X235384 is adjacent.
• It loses AK AK2 +SuzAve +X235384 from the front of its incident_c_edges list.
• AK AK2 +X235384 GolRd, still formatted SCT, becomes the new front.
• Its incident_t_edges list stays intact.
• AK AK2 +X235384 GolRd, still formatted SCT, remains at the back.
• wv.second->incident_c_edges.front() == wv.second->incident_t_edges.back(); IOW,
• wv.second->incident_c_edges.front()->format == wv.second->incident_t_edges.back()->format
• One dual-format edge is created, when we should instead create two single-format edges
• Wacky antics ensue.

[yakra]

else if ((wv.second->incident_c_edges.front()->format & ~HGEdge::simple) == (HGEdge::collapsed | HGEdge::traveled)
 && (wv.second->incident_c_edges.back()->format & ~HGEdge::simple) == (HGEdge::collapsed | HGEdge::traveled))

works, per the OP.
Even if (HGEdge::collapsed | HGEdge::traveled) is optimized to a constant by the compiler, we still have 5 logical operators + 2 -> dereference operators to get our format.
No-build uses 7 operators iff we need to use ||. Call it... 4 (including ||) otherwise?
One trial run had experimental @ 1.9s with no-build @ 1.8s. Good enough for me.
No-build it is.

[yakra]

This part of the code will need a rethink for TravelMapping/EduTools#156.

[yakra]

^^ May be the ticket if paring down to just a c_front & c_back.

• If it's slower, it's only slower during edge compression, which is relatively quick & painless.
• Any speed benefits during subgraph generation from keeping sizeof(HGVertex) down sans t_front & t_back?
  If able to lose vertex numbers, 48; else 72