/* Copyright (c) 2019, Anthony Latorre * * This program is free software: you can redistribute it and/or modify it * under the terms of the GNU General Public License as published by the Free * Software Foundation, either version 3 of the License, or (at your option) * any later version. * This program is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for * more details. * You should have received a copy of the GNU General Public License along with * this program. If not, see . */ #include "misc.h" #include #include /* for size_t */ #include #include "vector.h" #include #include static struct { int n; double f; } ln_table[LN_MAX + 1] = { {0,-INFINITY}, {1,0}, {2,0.69314718055994529}, {3,1.0986122886681098}, {4,1.3862943611198906}, {5,1.6094379124341003}, {6,1.791759469228055}, {7,1.9459101490553132}, {8,2.0794415416798357}, {9,2.1972245773362196}, {10,2.3025850929940459}, {11,2.3978952727983707}, {12,2.4849066497880004}, {13,2.5649493574615367}, {14,2.6390573296152584}, {15,2.7080502011022101}, {16,2.7725887222397811}, {17,2.8332133440562162}, {18,2.8903717578961645}, {19,2.9444389791664403}, {20,2.9957322735539909}, {21,3.044522437723423}, {22,3.0910424533583161}, {23,3.1354942159291497}, {24,3.1780538303479458}, {25,3.2188758248682006}, {26,3.2580965380214821}, {27,3.2958368660043291}, {28,3.3322045101752038}, {29,3.3672958299864741}, {30,3.4011973816621555}, {31,3.4339872044851463}, {32,3.4657359027997265}, {33,3.4965075614664802}, {34,3.5263605246161616}, {35,3.5553480614894135}, {36,3.5835189384561099}, {37,3.6109179126442243}, {38,3.6375861597263857}, {39,3.6635616461296463}, {40,3.6888794541139363}, {41,3.713572066704308}, {42,3.7376696182833684}, {43,3.7612001156935624}, {44,3.784189633918261}, {45,3.8066624897703196}, {46,3.8286413964890951}, {47,3.8501476017100584}, {48,3.8712010109078911}, {49,3.8918202981106265}, {50,3.912023005428146}, {51,3.9318256327243257}, {52,3.9512437185814275}, {53,3.970291913552122}, {54,3.9889840465642745}, {55,4.0073331852324712}, {56,4.0253516907351496}, {57,4.0430512678345503}, {58,4.0604430105464191}, {59,4.0775374439057197}, {60,4.0943445622221004}, {61,4.1108738641733114}, {62,4.1271343850450917}, {63,4.1431347263915326}, {64,4.1588830833596715}, {65,4.1743872698956368}, {66,4.1896547420264252}, {67,4.2046926193909657}, {68,4.219507705176107}, {69,4.2341065045972597}, {70,4.2484952420493594}, {71,4.2626798770413155}, {72,4.2766661190160553}, {73,4.290459441148391}, {74,4.3040650932041702}, {75,4.3174881135363101}, {76,4.3307333402863311}, {77,4.3438054218536841}, {78,4.3567088266895917}, {79,4.3694478524670215}, {80,4.3820266346738812}, {81,4.3944491546724391}, {82,4.4067192472642533}, {83,4.4188406077965983}, {84,4.4308167988433134}, {85,4.4426512564903167}, {86,4.4543472962535073}, {87,4.4659081186545837}, {88,4.4773368144782069}, {89,4.4886363697321396}, {90,4.499809670330265}, {91,4.5108595065168497}, {92,4.5217885770490405}, {93,4.5325994931532563}, {94,4.5432947822700038}, {95,4.5538768916005408}, {96,4.5643481914678361}, {97,4.5747109785033828}, {98,4.5849674786705723}, {99,4.5951198501345898}, {100,4.6051701859880918}, }; static struct { int n; double f; } ln_fact_table[LNFACT_MAX + 1] = { {0,0}, {1,0}, {2,0.69314718055994529}, {3,1.791759469228055}, {4,3.1780538303479458}, {5,4.7874917427820458}, {6,6.5792512120101012}, {7,8.5251613610654147}, {8,10.604602902745251}, {9,12.801827480081469}, {10,15.104412573075516}, {11,17.502307845873887}, {12,19.987214495661885}, {13,22.552163853123425}, {14,25.19122118273868}, {15,27.89927138384089}, {16,30.671860106080672}, {17,33.505073450136891}, {18,36.395445208033053}, {19,39.339884187199495}, {20,42.335616460753485}, {21,45.380138898476908}, {22,48.471181351835227}, {23,51.606675567764377}, {24,54.784729398112319}, {25,58.003605222980518}, {26,61.261701761002001}, {27,64.557538627006338}, {28,67.88974313718154}, {29,71.257038967168015}, {30,74.658236348830158}, {31,78.092223553315307}, {32,81.557959456115043}, {33,85.054467017581516}, {34,88.580827542197682}, {35,92.136175603687093}, {36,95.719694542143202}, {37,99.330612454787428}, {38,102.96819861451381}, {39,106.63176026064346}, {40,110.32063971475739}, {41,114.03421178146171}, {42,117.77188139974507}, {43,121.53308151543864}, {44,125.3172711493569}, {45,129.12393363912722}, {46,132.95257503561632}, {47,136.80272263732635}, {48,140.67392364823425}, {49,144.5657439463449}, {50,148.47776695177302}, {51,152.40959258449735}, {52,156.3608363030788}, {53,160.3311282166309}, {54,164.32011226319517}, {55,168.32744544842765}, {56,172.35279713916279}, {57,176.39584840699735}, {58,180.45629141754378}, {59,184.53382886144948}, {60,188.6281734236716}, {61,192.7390472878449}, {62,196.86618167289001}, {63,201.00931639928152}, {64,205.1681994826412}, {65,209.34258675253685}, {66,213.53224149456327}, {67,217.73693411395422}, {68,221.95644181913033}, {69,226.1905483237276}, {70,230.43904356577696}, {71,234.70172344281826}, {72,238.97838956183432}, {73,243.26884900298271}, {74,247.57291409618688}, {75,251.89040220972319}, {76,256.22113555000954}, {77,260.56494097186322}, {78,264.92164979855278}, {79,269.29109765101981}, {80,273.67312428569369}, {81,278.06757344036612}, {82,282.4742926876304}, {83,286.89313329542699}, {84,291.32395009427029}, {85,295.76660135076065}, {86,300.22094864701415}, {87,304.68685676566872}, {88,309.1641935801469}, {89,313.65282994987905}, {90,318.1526396202093}, {91,322.66349912672615}, {92,327.1852877037752}, {93,331.71788719692847}, {94,336.26118197919845}, {95,340.81505887079902}, {96,345.37940706226686}, {97,349.95411804077025}, {98,354.53908551944079}, {99,359.1342053695754}, {100,363.73937555556347}, }; double trapz(const double *y, double dx, size_t n) { /* Returns the integral of `y` using the trapezoidal rule assuming a * constant grid spacing `dx`. * * See https://en.wikipedia.org/wiki/Trapezoidal_rule. */ size_t i; double sum = 0.0; if (n < 2) return 0.0; sum = y[0]; for (i = 1; i < n-1; i++) { sum += 2*y[i]; } sum += y[n-1]; return sum*dx/2.0; } /* Compute the first intersection of a ray starting at `pos` with direction * `dir` and a sphere centered at the origin with radius `R`. The distance to * the intersection is stored in `l`. * * Returns 1 if the ray intersects the sphere, and 0 if it doesn't. * * Example: * * double l; * double pos[0] = {0,0,0}; * double dir[3] = {1,0,0}; * * if (intersect_sphere(pos,dir,PSUP_RADIUS,&l)) { * hit[0] = pos[0] + l*dir[0]; * hit[1] = pos[1] + l*dir[1]; * hit[2] = pos[2] + l*dir[2]; * printf("ray intersects sphere at %.2f %.2f %.2f\n", hit[0], hit[1], hit[2]); * } else { * printf("ray didn't intersect sphere\n"); * } * */ int intersect_sphere(double *pos, double *dir, double R, double *l) { double b, c; b = 2*DOT(dir,pos); c = DOT(pos,pos) - R*R; if (b*b - 4*c <= 0) { /* Ray doesn't intersect the sphere. */ *l = 0.0; return 0; } /* First, check the shorter solution. */ *l = (-b - sqrt(b*b - 4*c))/2; /* If the shorter solution is less than 0, check the second solution. */ if (*l < 0) *l = (-b + sqrt(b*b - 4*c))/2; /* If the distance is still negative, we didn't intersect the sphere. */ if (*l < 0) return 0; return 1; } void get_path_length(double *pos1, double *pos2, double R, double *l1, double *l2) { /* Returns the path length inside and outside a circle of radius `R` for a * ray starting at position `pos1` and ending at position `pos2`. * * The path length inside the sphere is stored in `l1` and the path length * outside the sphere is stored in `l2`. */ double dir[3], l, b, c, d1, d2; /* Calculate the vector from `pos1` to `pos2`. */ SUB(dir,pos2,pos1); l = NORM(dir); DIV(dir,l); b = 2*DOT(dir,pos1); c = DOT(pos1,pos1) - R*R; if (b*b - 4*c <= 0) { /* Ray doesn't intersect the sphere. */ *l1 = 0.0; *l2 = l; return; } d1 = (-b + sqrt(b*b - 4*c))/2; d2 = (-b - sqrt(b*b - 4*c))/2; if (d1 < 0) { /* Ray also doesn't intersect sphere. */ *l1 = 0.0; *l2 = l; } else if (d1 >= l && d2 < 0) { /* Ray also doesn't intersect sphere. */ *l1 = l; *l2 = 0.0; } else if (d2 < 0) { /* Ray intersects sphere once. */ *l1 = d1; *l2 = l-d1; } else if (d1 >= l && d2 >= l) { /* Ray doesn't intersect the sphere. */ *l1 = 0.0; *l2 = l; } else if (d1 >= l && d2 < l) { /* Ray intersects the sphere once. */ *l2 = d1; *l1 = l-d1; } else if (d1 < l && d2 < l) { /* Ray intersects the sphere twice. */ *l1 = d1-d2; *l2 = l-(d1-d2); } } double ln(unsigned int n) { /* Returns the logarithm of n. * * Uses a lookup table to return results for n < 100. */ if (n <= LN_MAX) return ln_table[n].f; return log(n); } double lnfact(unsigned int n) { /* Returns the logarithm of n!. * * Uses a lookup table to return results for n < 100. */ if (n <= LNFACT_MAX) return ln_fact_table[n].f; return gsl_sf_lnfact(n); } double kahan_sum(double *x, size_t n) { /* Returns the sum of the elements of `x` using the Kahan summation algorithm. * * See https://en.wikipedia.org/wiki/Kahan_summation_algorithm. */ size_t i; double sum, c, y, t; sum = 0.0; c = 0.0; for (i = 0; i < n; i++) { y = x[i] - c; t = sum + y; c = (t - sum) - y; sum = t; } return sum; } double interp2d(double x, double y, double *xp, double *yp, double *zp, size_t n1, size_t n2) { /* A fast bilinear interpolation routine which assumes that the values in * `xp` and `yp` are evenly spaced. * * `zp` should be a 2D array indexed as zp[i,j] = zp[i*n2 + j]. * * If x < xp[0], x > xp[n-1], y < yp[0], or y > yp[n-1] prints an error and * exits. * * See https://en.wikipedia.org/wiki/Bilinear_interpolation. */ size_t i, j; double q11, q12, q21, q22; double idx, idy; if (x < xp[0]) { fprintf(stderr, "x < xp[0]!\n"); exit(1); } if (y < yp[0]) { fprintf(stderr, "y < yp[0]!\n"); exit(1); } idx = 1.0/(xp[1]-xp[0]); i = (x-xp[0])*idx; if (i > n1-2) { fprintf(stderr, "i > n1 - 2!\n"); exit(1); } idy = 1.0/(yp[1]-yp[0]); j = (y-yp[0])*idy; if (j > n2-2) { fprintf(stderr, "j > n2 - 2!\n"); exit(1); } q11 = zp[i*n2+j]; q12 = zp[i*n2+j+1]; q21 = zp[(i+1)*n2+j]; q22 = zp[(i+1)*n2+j+1]; return (q11*(xp[i+1]-x)*(yp[j+1]-y) + q21*(x-xp[i])*(yp[j+1]-y) + q12*(xp[i+1]-x)*(y-yp[j]) + q22*(x-xp[i])*(y-yp[j]))*idx*idy; } double interp1d(double x, double *xp, double *yp, size_t n) { /* A fast interpolation routine which assumes that the values in `xp` are * evenly spaced. * * If x < xp[0] returns yp[0] and if x > xp[n-1] returns yp[n-1]. */ size_t i; double idx; if (x <= xp[0]) return yp[0]; idx = 1.0/(xp[1]-xp[0]); i = (x-xp[0])*idx; if (i > n-2) return yp[n-1]; return yp[i] + (yp[i+1]-yp[i])*(x-xp[i])*idx; } int isclose(double a, double b, double rel_tol, double abs_tol) { /* Returns 1 if a and b are "close". This algorithm is taken from Python's * math.isclose() function. * * See https://www.python.org/dev/peps/pep-0485/. */ return fabs(a-b) <= fmax(rel_tol*fmax(fabs(a),fabs(b)),abs_tol); } int allclose(double *a, double *b, size_t n, double rel_tol, double abs_tol) { /* Returns 1 if all the elements of a and b are "close". This algorithm is * taken from Python's math.isclose() function. * * See https://www.python.org/dev/peps/pep-0485/. */ size_t i; for (i = 0; i < n; i++) { if (!isclose(a[i],b[i],rel_tol,abs_tol)) return 0; } return 1; } double logsumexp(double *a, size_t n) { /* Returns the log of the sum of the exponentials of the array `a`. * * This function is designed to reduce underflow when the exponentials of * `a` are very small, for example when computing probabilities. */ size_t i; double amax, sum; amax = a[0]; for (i = 0; i < n; i++) { if (a[i] > amax) amax = a[i]; } sum = 0.0; for (i = 0; i < n; i++) { sum += exp(a[i]-amax); } sum = log(sum); return amax + sum; } double log_norm(double x, double mu, double sigma) { /* Returns the log of the PDF for a gaussian random variable with mean `mu` * and standard deviation `sigma`. */ return -pow(x-mu,2)/(2*pow(sigma,2)) - log(sqrt(2*M_PI)*sigma); } double norm(double x, double mu, double sigma) { /* Returns the PDF for a gaussian random variable with mean `mu` and * standard deviation `sigma`. */ return exp(-pow(x-mu,2)/(2*pow(sigma,2)))/(sqrt(2*M_PI)*sigma); } double norm_cdf(double x, double mu, double sigma) { /* Returns the CDF for a gaussian random variable with mean `mu` and * standard deviation `sigma`. */ return erfc(-(x-mu)/(sqrt(2)*sigma))/2.0; } double mean(const double *x, size_t n) { /* Returns the mean of the array `x`. */ size_t i; double sum = 0.0; for (i = 0; i < n; i++) sum += x[i]; return sum/n; } double std(const double *x, size_t n) { /* Returns the standard deviation of the array `x`. */ size_t i; double sum, mu; mu = mean(x,n); sum = 0.0; for (i = 0; i < n; i++) sum += pow(x[i]-mu,2); return sqrt(sum/n); } double gamma_pdf(double x, double k, double theta) { /* Returns the PDF for the gamma distribution. * * See https://en.wikipedia.org/wiki/Gamma_distribution. */ return pow(x,k-1)*exp(-x/theta)/(gsl_sf_gamma(k)*pow(theta,k)); } size_t ipow(size_t base, size_t exp) { /* Returns base^exp for positive integers `base` and `exp`. * * See https://stackoverflow.com/questions/101439. */ size_t result = 1; for (;;) { if (exp & 1) result *= base; exp >>= 1; if (!exp) break; base *= base; } return result; } void product(size_t n, size_t r, size_t *result) { /* Returns the cartesian product of [1..n] with itself `r` times. * * The resulting array can be indexed as: * * result[i*r + j] * * where i is the ith product and j is the jth element in the product. For example: * * size_t result[4]; * product(2,2,result); * for (i = 0; i < 2; i++) * print("%zu %zu\n", result[i*2], result[i*2+1]); * * will print * * 0 0 * 0 1 * 1 0 * 1 1 * * `result` should be an array with at least `r`*`n`^`r` elements. */ size_t i, j; for (i = 0; i < r; i++) { for (j = 0; j < ipow(n,r); j++) { result[j*r+i] = (j/ipow(n,r-i-1)) % n; } } } typedef struct direction { int id; size_t index; } direction; static int direction_compare(const void *a, const void *b) { const direction *da = (direction *) a; const direction *db = (direction *) b; if (da->id > db->id) return 1; else if (da->id < db->id) return -1; else if (da->index > db->index) return 1; else if (da->index < db->index) return -1; return 0; } void unique_vertices(int *id, size_t n, size_t npeaks, size_t *result, size_t *nvertices) { /* Returns the set of all unique vertices for `n` particles with particle * ids `id` for `npeaks` possible direction vectors. For example, the * unique vertices for two electrons and one muon given 2 possible * directions 1 and 2 would be: * * 1 1 1 * 1 1 2 * 1 2 1 * 1 2 2 * 2 2 1 * 2 2 2 * * where the first column represents the direction for the first electron, * the second for the second electron, and the third for the muon. This is * different from the cartesian product since rows like * * 2 1 1 * * and * * 2 1 2 * * are duplicates. * * The resulting array can be indexed as: * * result[i*n + j] * * where i is the ith product and j is the jth direction in the product. * For example: * * int id[3] = {IDP_E_MINUS, IDP_E_MINUS, IDP_MU_MINUS}; * * size_t result[24]; * size_t nvertices; * unique_vertices(id, LEN(id), 2, result, &nvertices); * for (i = 0; i < nvertices; i++) * print("%zu %zu %zu\n", result[i*3], result[i*3+1], result[i*3+2); * * `result` should be an array with at least `n`*`npeaks`^`n` elements. */ size_t i, j, k; direction *ordered_results; size_t *results2 = malloc(sizeof(size_t)*n*ipow(npeaks,n)); int unique, equal; product(npeaks,n,results2); ordered_results = malloc(sizeof(direction)*n*ipow(npeaks,n)); direction *unique_results = malloc(sizeof(direction)*n*ipow(npeaks,n)); for (i = 0; i < ipow(npeaks,n); i++) { for (j = 0; j < n; j++) { ordered_results[i*n+j].id = id[j]; ordered_results[i*n+j].index = results2[i*n+j]; } /* Sort the vertices in each row so we can compare them. */ qsort(ordered_results+i*n,n,sizeof(direction),direction_compare); } *nvertices = 0; for (i = 0; i < ipow(npeaks,n); i++) { unique = 1; for (j = 0; j < *nvertices; j++) { equal = 1; for (k = 0; k < n; k++) { if ((unique_results[j*n+k].id != ordered_results[i*n+k].id) || (unique_results[j*n+k].index != ordered_results[i*n+k].index)) { equal = 0; break; } } if (equal) { unique = 0; break; } } if (unique) { for (j = 0; j < n; j++) { unique_results[(*nvertices)*n+j].id = ordered_results[i*n+j].id; unique_results[(*nvertices)*n+j].index = ordered_results[i*n+j].index; } *nvertices += 1; } } for (i = 0; i < *nvertices; i++) { for (j = 0; j < n; j++) { result[i*n+j] = unique_results[i*n+j].index; } } free(results2); free(ordered_results); free(unique_results); } static int is_sorted(size_t *a, size_t n) { size_t i; for (i = 1; i < n; i++) if (a[i] < a[i-1]) return 0; return 1; } void combinations_with_replacement(size_t n, size_t r, size_t *result, size_t *len) { /* Returns the set of all unique combinations of length `r` from `n` unique * elements. * * The result array can be indexed as: * * result[i*r + j] * * where i is the ith combination and j is the jth element in the * combination. For example: * * size_t result[12]; * size_t len; * product(3,2,result,&len); * for (i = 0; i < len; i++) * print("%zu %zu\n", result[i*2], result[i*2+1]); * * will print * * 0 0 * 0 1 * 0 2 * 1 1 * 1 2 * 2 2 * * `result` should be an array with at least (n+r-1)!/r!/(n-1)! elements. */ size_t i, j; size_t *tmp; tmp = malloc(sizeof(size_t)*r*ipow(n,r)); product(n,r,tmp); *len = 0; for (i = 0; i < ipow(n,r); i++) { if (is_sorted(tmp+i*r,r)) { for (j = 0; j < r; j++) { result[(*len)*r+j] = tmp[i*r+j]; } *len += 1; } } free(tmp); } size_t argmax(double *a, size_t n) { /* Returns the index of the maximum of the array `a`. */ size_t i, index; double max; if (n < 2) return 0; index = 0; max = a[0]; for (i = 1; i < n; i++) { if (a[i] > max) { max = a[i]; index = i; } } return index; } size_t argmin(double *a, size_t n) { /* Returns the index of the minimum of the array `a`. */ size_t i, index; double min; if (n < 2) return 0; index = 0; min = a[0]; for (i = 1; i < n; i++) { if (a[i] < min) { min = a[i]; index = i; } } return index; } void get_dir(double *dir, double theta, double phi) { /* Compute the 3-dimensional unit vector `dir` from the polar angle `theta` * and azimuthal angle `phi`. */ double sin_theta, cos_theta, sin_phi, cos_phi; cos_theta = cos(theta); sin_theta = sin(theta); cos_phi = cos(phi); sin_phi = sin(phi); dir[0] = sin_theta*cos_phi; dir[1] = sin_theta*sin_phi; dir[2] = cos_theta; } /* Fast version of acos() which uses a lookup table computed on the first call. */ double fast_acos(double x) { size_t i; static int initialized = 0; static double xs[N_ACOS]; static double ys[N_ACOS]; if (!initialized) { for (i = 0; i < LEN(xs); i++) { xs[i] = -1.0 + 2.0*i/(LEN(xs)-1); ys[i] = acos(xs[i]); } initialized = 1; } return interp1d(x,xs,ys,LEN(xs)); }