macros -> lambdas

[yakra]

macros

Every macro you write makes you less 1337.

-Some random person on the internet.

While not 100% the case across the board, there's some truthiness to it.
Lambdas will work in many cases. Advantages include:

• Safer
• Smaller program size
• Speed? Instruction cache?

Can't get rid of all of them, but here are some candidates:

• Just use a reference for this one, not a lambda, but including it anyway.
  

  DataProcessing/siteupdate/cplusplus/functions/sql_file.cpp

  Line 496 in 387c99b

  #define G GraphListEntry::entries[g]

• DataProcessing/siteupdate/cplusplus/classes/HighwaySystem/route_integrity.cpp

  Line 17 in 387c99b

  #define CSV_LINE r.system->systemname + ".csv#L" + std::to_string(r.index()+2)

• DataProcessing/siteupdate/cplusplus/classes/Waypoint/Waypoint.cpp

  Line 203 in 387c99b

  #define PYTHON_STYLE_FLOAT(F) e=sprintf(s," %.15g",F); if (F==int(F)) strcpy(s+e,".0"); nmpnmp<<s;

• Binary size down 4000 B. Gets used 8 times.
  

  DataProcessing/siteupdate/cplusplus/classes/Waypoint/Waypoint.cpp

  Line 490 in 387c99b

  #define FLAG(SUBTYPE) Datacheck::add(route, label, "", "", "LABEL_SELFREF", SUBTYPE)

• This may not work, with the continue in the lambda and not the body of the loop itself? Try it & see what happens.
  

  DataProcessing/siteupdate/cplusplus/classes/Route/read_wpt.cpp

  Line 55 in 387c99b

  #define SKIP {--points.size; if (segments.size) /*cope with all-whitespace files*/ --segments.size; continue;}

• DataProcessing/siteupdate/cplusplus/classes/Args/Args.cpp

  Line 30 in 387c99b

  #define ARG(N,S,L) ( n+N < argc && (!strcmp(argv[n],S) || !strcmp(argv[n],L)) )

• Eventually, the need for these macros will go away entirely. But that's a heavyweight operation that'll take some time to get around to. In the interim, could this maybe have an effect if it's easier on instruction cache?
  Edit: Deleted in 897c231. No need for a lambda.
  

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

  Lines 38 to 39 in 387c99b

  	#define AREA (!g->placeradius || subgraph_contains(e->vertex1, threadnum) && subgraph_contains(e->vertex2, threadnum))
  	#define REGION (!g->regions || contains(*g->regions, e->segment->route->region))

• Eventually, HGEdge::detach may go away entirely (another heavyweight opeation). If this does help in the interim; it'll be low-impact; edge compression is already pretty fast and may get faster.
  Edit: It's still here, just smaller & called less often. The lambda is invoked twice, as opposed to the macro being invoked 6 times.
  

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

  Lines 122 to 127 in 387c99b

  	#define DETACH(v) \
  	for (std::list<HGEdge*>::iterator e = v.begin(); e != v.end(); e++) \
  	if (*e == this) \
  	{ v.erase(e); \
  	break; \
  	}

• DataProcessing/siteupdate/cplusplus/classes/TravelerList/TravelerList.cpp

  Lines 126 to 129 in 387c99b

  	#define UPDATE_NOTE(R) if (R->last_update) \
  	{ updated_routes.insert(R); \
  	log << " Route updated " << R->last_update[0] << ": " << R->readable_name() << '\n'; \
  	}

---

other stuff

• datacheck boilerplate
  A couple examples are above already, but in general, look for Datacheck::add.
• sorting
  grep -n 'sort([^)]' \`find . -name \*.cpp`
  • Sorting regions.csv lines & eliminating the 1 use of sort_1st_csv_field reduces binary size 144 B.
  • sort_route_updates_oldest
  • What if a function is used multiple times, i.e. sort_root_at_label (5x)?
• splitregion.cpp