mirror of
				https://github.com/zenorogue/hyperrogue.git
				synced 2025-11-04 07:43:02 +00:00 
			
		
		
		
	C++20 introduces `std::format` and we `using namespace std`, so some of these would be ambiguous in C++20.
		
			
				
	
	
		
			80 lines
		
	
	
		
			2.3 KiB
		
	
	
	
		
			C++
		
	
	
	
	
	
			
		
		
	
	
			80 lines
		
	
	
		
			2.3 KiB
		
	
	
	
		
			C++
		
	
	
	
	
	
// test the conjecture comparing triangulation distances and hyperbolic distances
 | 
						|
 | 
						|
namespace dhrg {
 | 
						|
 | 
						|
ld stats[32][3], wstats[32][3];
 | 
						|
  
 | 
						|
hyperpoint celltopoint(cell *c) {
 | 
						|
  return tC0(calc_relative_matrix(c, croot(), C0));
 | 
						|
  }
 | 
						|
 | 
						|
void do_analyze_grid(int maxv) {
 | 
						|
  cell *root = croot();
 | 
						|
  celllister cl(root, 32, maxv, NULL);
 | 
						|
  // if this works too slow, use a smaller number
 | 
						|
  // (you can also use a larger number if you have time of course)
 | 
						|
 | 
						|
  // int rot = 0;
 | 
						|
  
 | 
						|
  vector<ld> distances[128];
 | 
						|
  
 | 
						|
  for(cell *c: cl.lst) {
 | 
						|
    hyperpoint h = celltopoint(c);
 | 
						|
    ld dd = hdist0(h);
 | 
						|
    int d = celldist(c);
 | 
						|
    stats[d][0] ++;
 | 
						|
    stats[d][1] += dd;
 | 
						|
    stats[d][2] += dd*dd;
 | 
						|
    distances[d].push_back(dd);
 | 
						|
 | 
						|
    if(d>0) {
 | 
						|
      ld alpha[2];
 | 
						|
      int qalpha = 0;
 | 
						|
      forCellCM(c2, c) if(celldist(c2) == d) {
 | 
						|
        hyperpoint h1 = celltopoint(c2);
 | 
						|
        alpha[qalpha++] = atan2(h1[0], h1[1]);
 | 
						|
        }
 | 
						|
      if(qalpha != 2) printf("Error: qalpha = %d\n", qalpha);
 | 
						|
      ld df = raddif(alpha[0], alpha[1]);
 | 
						|
      df /= 720._deg;
 | 
						|
      wstats[d][0] += df;
 | 
						|
      if(d==2) printf("df == %" PLDF " dd = %" PLDF "\n", df, dd);
 | 
						|
      wstats[d][1] += df*dd;
 | 
						|
      wstats[d][2] += df*dd*dd;
 | 
						|
      }
 | 
						|
    }
 | 
						|
  
 | 
						|
  println(hlog, "log(gamma) = ", log(cgi.expansion->get_growth()));
 | 
						|
  
 | 
						|
  ld lE, dif, lwE;
 | 
						|
  for(int d=0; d<32; d++) if(stats[d][0]) {
 | 
						|
    int q = stats[d][0];
 | 
						|
    if(q != cgi.expansion->get_descendants(d).approx_int()) continue; 
 | 
						|
    ld E = stats[d][1] / q;
 | 
						|
    ld E2 = stats[d][2] / q;
 | 
						|
    ld Vr = E2 - E * E;
 | 
						|
    if(Vr < 0) Vr = 0;
 | 
						|
    dif = E- lE; lE = E;
 | 
						|
    ld Vd = d > 1 ? Vr/(d-1) : 0;
 | 
						|
    
 | 
						|
    ld wE = wstats[d][1];
 | 
						|
    ld wE2 = wstats[d][2];
 | 
						|
    ld wVr = wE2 - wE * wE;
 | 
						|
 | 
						|
    print(hlog, hr::format("d=%2d: q = %8d E = %12.8" PLDF " dif = %12.8" PLDF " Vr = %12.8" PLDF " Vr/(d-1)=%12.8" PLDF,
 | 
						|
      d, q, E, dif, Vr, Vd));
 | 
						|
 | 
						|
    if(0) print(hlog, hr::format(" | <%" PLDF "> ex = %12.8" PLDF " d.ex = %12.8" PLDF " Vr = %12.8" PLDF, wstats[d][0], wE, wE - lwE, wVr));
 | 
						|
    
 | 
						|
    ld Sigma = sqrt(Vr);
 | 
						|
    sort(distances[d].begin(), distances[d].end());
 | 
						|
    if(Sigma) for(int u=1; u<8; u++)
 | 
						|
      print(hlog, hr::format(" %8.5" PLDF, (distances[d][u * isize(distances[d]) / 8] - E) / Sigma));
 | 
						|
    
 | 
						|
    println(hlog);
 | 
						|
    lwE = wE;
 | 
						|
    }
 | 
						|
  }
 | 
						|
 | 
						|
}
 |