Skip to content

Commit

Permalink
prevent loss of precision for NSIDE > 2^25
Browse files Browse the repository at this point in the history
Fixes #2.
  • Loading branch information
ntessore committed Jan 13, 2023
1 parent ea509b8 commit 13f5675
Show file tree
Hide file tree
Showing 2 changed files with 36 additions and 34 deletions.
2 changes: 1 addition & 1 deletion setup.cfg
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
[metadata]
name = healpix
version = 2022.11.1
version = 2023.1.13
maintainer = Nicolas Tessore
maintainer_email = n.tessore@ucl.ac.uk
description = Python package for HEALPix discretisation of the sphere
Expand Down
68 changes: 35 additions & 33 deletions src/healpix.c
Original file line number Diff line number Diff line change
Expand Up @@ -105,15 +105,15 @@ static int64_t compress_bits(int64_t v) {


// A structure describing the discrete Healpix coordinate system.
// f takes values in [0, 11], a and b lie in [0, nside).
// f takes values in [0, 11], x and y lie in [0, nside).
typedef struct {
int64_t a, b;
int64_t x, y;
int32_t f;
} t_hpd;


static int64_t hpd2nest(int64_t nside, t_hpd hpd) {
return (hpd.f*nside*nside) + spread_bits(hpd.a) + (spread_bits(hpd.b)<<1);
return (hpd.f*nside*nside) + spread_bits(hpd.x) + (spread_bits(hpd.y)<<1);
}


Expand All @@ -125,24 +125,24 @@ static t_hpd nest2hpd(int64_t nside, int64_t pix) {

static int64_t hpd2ring(int64_t nside, t_hpd hpd) {
int64_t nl4 = 4*nside;
int64_t jr = (jrll[hpd.f]*nside) - hpd.a - hpd.b - 1;
int64_t jr = (jrll[hpd.f]*nside) - hpd.x - hpd.y - 1;

if (jr<nside)
{
int64_t jp = (jpll[hpd.f]*jr + hpd.a - hpd.b + 1) / 2;
int64_t jp = (jpll[hpd.f]*jr + hpd.x - hpd.y + 1) / 2;
jp = (jp>nl4) ? jp-nl4 : ((jp<1) ? jp+nl4 : jp);
return 2*jr*(jr-1) + jp - 1;
}
else if (jr > 3*nside)
{
jr = nl4-jr;
int64_t jp = (jpll[hpd.f]*jr + hpd.a - hpd.b + 1) / 2;
int64_t jp = (jpll[hpd.f]*jr + hpd.x - hpd.y + 1) / 2;
jp = (jp>nl4) ? jp-nl4 : ((jp<1) ? jp+nl4 : jp);
return 12*nside*nside - 2*(jr+1)*jr + jp - 1;
}
else
{
int64_t jp = (jpll[hpd.f]*nside + hpd.a - hpd.b + 1 + ((jr-nside)&1)) / 2;
int64_t jp = (jpll[hpd.f]*nside + hpd.x - hpd.y + 1 + ((jr-nside)&1)) / 2;
jp = (jp>nl4) ? jp-nl4 : ((jp<1) ? jp+nl4 : jp);
return 2*nside*(nside-1) + (jr-nside)*nl4 + jp - 1;
}
Expand Down Expand Up @@ -231,20 +231,21 @@ static t_hpd loc2hpd(int64_t nside, t_loc loc, double* u, double* v) {
double jm = temp1+temp2; // index of descending edge line, [0; 5)
int ifp = (int)jp; // in {0,4}
int ifm = (int)jm;
hpd.a = (jm-ifm)*nside;
hpd.b = (1+ifp - jp)*nside;
hpd.x = (jm-ifm)*nside;
hpd.y = (1+ifp - jp)*nside;
hpd.f = (ifp==ifm) ? (ifp|4) : ((ifp<ifm) ? ifp : (ifm+8));
if (u) {
*u = (jm-ifm)*nside - hpd.a;
*v = (1+ifp - jp)*nside - hpd.b;
*u = (jm-ifm)*nside - hpd.x;
*v = (1+ifp - jp)*nside - hpd.y;
}
} else {
// polar region
int64_t ntt = (int64_t)tt;
if (ntt >= 4)
ntt = 3;
double tp = tt - ntt; // [0;1)
double tmp = sqrt(3.*(1.-za));
/* double tmp = sqrt(3.*(1.-za)); */
double tmp = loc.s/sqrt((1.+za)*(1./3.)); // FIXME optimize!

double jp = tp*tmp; // increasing edge line index
double jm = (1.0-tp)*tmp; // decreasing edge line index
Expand All @@ -259,35 +260,36 @@ static t_hpd loc2hpd(int64_t nside, t_loc loc, double* u, double* v) {
} else {
ntt += 8;
}
hpd.a = jp*nside;
hpd.b = jm*nside;
hpd.x = jp*nside;
hpd.y = jm*nside;
hpd.f = ntt;
if (u) {
*u = jp*nside - hpd.a;
*v = jm*nside - hpd.b;
*u = jp*nside - hpd.x;
*v = jm*nside - hpd.y;
}
}
return hpd;
}


static t_loc hpd2loc(int64_t nside, t_hpd hpd, double dx, double dy) {
static t_loc hpd2loc(int64_t nside, t_hpd hpd, double u, double v) {
double z, s, phi;
const double x = (hpd.a - hpd.b + dx)/nside;
const double y = (hpd.a + hpd.b - nside + dy)/nside;
const double x = (hpd.x+u)/nside;
const double y = (hpd.y+v)/nside;
const int32_t r = 1 - hpd.f/4;
const double m = 1 - r*y;
if (m < 1) {
const double h = r-1 + (x+y);
double m = 2 - r*h;
if (m < 1.) {
// polar cap
const double tmp = m*m*(1./3.);
double tmp = m*m*(1./3.);
z = r*(1. - tmp);
s = sqrt(tmp*(2.-tmp));
phi = (PI/4.)*(jpll[hpd.f] + x/m);
phi = (PI/4)*(jpll[hpd.f] + (x-y)/m);
} else {
// equatorial region
z = (y+r)*(2./3.);
z = h*(2./3.);
s = sqrt((1.+z)*(1.-z));
phi = (PI/4.)*(jpll[hpd.f] + x);
phi = (PI/4)*(jpll[hpd.f] + (x-y));
}
return (t_loc){z, s, phi};
}
Expand Down Expand Up @@ -337,22 +339,22 @@ int64_t vec2nest(int64_t nside, t_vec vec) {


t_ang ring2ang(int64_t nside, int64_t ipix) {
return loc2ang(hpd2loc(nside, ring2hpd(nside, ipix), 0, 1));
return loc2ang(hpd2loc(nside, ring2hpd(nside, ipix), 0.5, 0.5));
}


t_ang nest2ang(int64_t nside, int64_t ipix) {
return loc2ang(hpd2loc(nside, nest2hpd(nside, ipix), 0, 1));
return loc2ang(hpd2loc(nside, nest2hpd(nside, ipix), 0.5, 0.5));
}


t_vec ring2vec(int64_t nside, int64_t ipix) {
return loc2vec(hpd2loc(nside, ring2hpd(nside, ipix), 0, 1));
return loc2vec(hpd2loc(nside, ring2hpd(nside, ipix), 0.5, 0.5));
}


t_vec nest2vec(int64_t nside, int64_t ipix) {
return loc2vec(hpd2loc(nside, nest2hpd(nside, ipix), 0, 1));
return loc2vec(hpd2loc(nside, nest2hpd(nside, ipix), 0.5, 0.5));
}


Expand Down Expand Up @@ -380,20 +382,20 @@ int64_t vec2nest_uv(int64_t nside, t_vec vec, double* u, double* v) {


t_ang ring2ang_uv(int64_t nside, int64_t ipix, double u, double v) {
return loc2ang(hpd2loc(nside, ring2hpd(nside, ipix), u-v, u+v));
return loc2ang(hpd2loc(nside, ring2hpd(nside, ipix), u, v));
}


t_ang nest2ang_uv(int64_t nside, int64_t ipix, double u, double v) {
return loc2ang(hpd2loc(nside, nest2hpd(nside, ipix), u-v, u+v));
return loc2ang(hpd2loc(nside, nest2hpd(nside, ipix), u, v));
}


t_vec ring2vec_uv(int64_t nside, int64_t ipix, double u, double v) {
return loc2vec(hpd2loc(nside, ring2hpd(nside, ipix), u-v, u+v));
return loc2vec(hpd2loc(nside, ring2hpd(nside, ipix), u, v));
}


t_vec nest2vec_uv(int64_t nside, int64_t ipix, double u, double v) {
return loc2vec(hpd2loc(nside, nest2hpd(nside, ipix), u-v, u+v));
return loc2vec(hpd2loc(nside, nest2hpd(nside, ipix), u, v));
}

0 comments on commit 13f5675

Please sign in to comment.