TravelerList dicts

[yakra]

It appears we calculate & store data here, but never retrieve & use it.

DataProcessing/siteupdate/python-teresco/siteupdate.py

Lines 3347 to 3353 in 3d56bd9

	# "traveled" dictionaries indexed by system name, then conn or regular
	# route in another dictionary with keys route, values mileage
	# "clinched" dictionaries indexed by system name, values clinch count
	t.con_routes_traveled = dict()
	t.con_routes_clinched = dict()
	t.routes_traveled = dict()
	#t.routes_clinched = dict()

DataProcessing/siteupdate/cplusplus/classes/TravelerList/TravelerList.h

Lines 19 to 24 in c15f3c2

	std::unordered_map<HighwaySystem*, std::unordered_map<ConnectedRoute*, double>> con_routes_traveled; // mileage per ConRte per system
	// TODO is this necessary?
	// ConRtes by definition exist in one system only
	std::unordered_map<Route*, double> routes_traveled; // mileage per traveled route
	std::unordered_map<HighwaySystem*, unsigned int> con_routes_clinched; // clinch count per system
	//std::unordered_map<HighwaySystem*, unsigned int> routes_clinched; // commented out in original siteupdate.py

• Getting rid of these structures & their contents should free up some RAM, and maybe speed things up a tiny bit.

---

• userlog.cpp: system_con_umap (system_con_dict in Python) can also go bye-bye.

[yakra]

Getting rid of these structures & their contents should free up some RAM, and maybe speed things up a tiny bit.

By "a tiny bit", I mean "54.7% faster Creating per-traveler stats logs and augmenting data structure on BiggaTomato, C++, 4 threads."
This calls for making a pretty graph. Haven't done one of those in a while.

• (Did 53ca19d slow us down a tiny bit?) No!
  Not on lab2 at least. Didn't bother testing anywhere else.

  threads	3d56bd9	53ca19d	diff
  1	30.98	30.6	-0.38
  2	15.24	15.02	-0.22
  3	10.29	10.18	-0.11
  4	7.74	7.64	-0.10
  5	6.29	6.24	-0.05
  6	5.29	5.24	-0.05
  7	4.67	4.59	-0.08
  8	4.1	4.09	-0.01
  9	3.83	3.76	-0.07
  10	3.62	3.56	-0.06
  11	3.43	3.35	-0.08
  12	3.22	3.2	-0.02
  13	3.09	3.05	-0.04
  14	2.94	2.89	-0.05
  15	2.81	2.78	-0.03
  16	2.7	2.68	-0.02

[yakra]

By "a tiny bit", I mean "54.7% faster Creating per-traveler stats logs and augmenting data structure on BiggaTomato, C++, 4 threads."

Heh, nope. Small sample size (1) + not too tightly controlled testing conditions led to an inflated time for the old code.
There is some improvement, but it's much smaller.

[yakra]

HighwayData @ aeaeb953348711a4265e31f66031af1d362485a8
UserData @ f5458d401b8a7856efae631fe2b89d11d56bc4b2
Saves about 11.3 MB C++ & 13.2 MB Python, plus whatever overhead may exist within the hash table structures themselves.