osm-codes/s2geom-poc

Proof of Concept for PostgreSQL implementation of S2geometry functions, based on AfieldTrails/s2-postgis

s2geom-poc's Issues

Validate Javascript libraries

For example https://github.com/countingCrow/simple-s2-port can be tested by an script of asserts generated by SQL.

Need for cell's GeoJSON

Each cell ID of S2 Geometry can be represented geometrically, so, this representation can be expessed as a GeoJSON polygon.

We need a funcion that return the GeoJSON of a cell, and similar function that returns the Union of a set of cells, as multipolygon.

check metric and convert area to square meters

Supposing that areas returned by S2 functions are offered in steradians... As this comment at C++ source s2/s2testing.h

the actual metric for a cell is obtained by integrating the differential quantity,
which is not constant across the cell. Therefore cells at low levels
(bigger cells) have smaller variations

and the implementation of conversion functions at s2testing.cc:

const double S2Testing::kEarthRadiusKm = 6371.01;

double S2Testing::AreaToMeters2(double steradians) {
  return 1e6 * AreaToKm2(steradians);
}

double S2Testing::AreaToKm2(double steradians) {
  return steradians * kEarthRadiusKm * kEarthRadiusKm;
}

About kEarthRadiusKm, seems the "Average distance from center to surface", that can be 6,371.230 km by wikipedia. Nominal value: 6.3781×10^6 m but in ellipsoid can range from 6,357 to 6,378 km. Ideal is the "Median distance from center to surface", so 6371010 m is near it.

The area in steradians is useful for precise corrections and integration (and/or positioning each cell in its correct altitude). For any other use the m2 is the standard, like PostGIS functions.

The sugestion is to add a flag in the *_exact_area() functions (like ST_Area) to be true for m2 and false for steradians.

error on get area from negative cellid

There are an error on assert02 line 11, all areas are equal (and same as a leaf area). Seems that it is a problem with s2_cellid_parent( s2_cellid_from_latlng(-23.55041,-46.63394), 1+i*5), but the test below, using tokens (the test on token area transform was good) show that is not:

WITH cells AS (
 SELECT 1+i*5 as level,
        s2_cellid_parent( s2_cellid_from_latlng(-23.55041,-46.63394), 1+i*5) pt1_id,
        s2_cellid_parent( s2_cellid_from_latlng(15,15), 1+i*5) pt2_id
 FROM generate_series(0,4) t(i)
) 
  SELECT level, pt1_id, pt2_id,
         s2_token_from_cellid(pt1_id) "pt1_token",
         s2_token_from_cellid(pt2_id) "pt2_token"
  FROM cells;

level	pt1_id	pt2_id	pt1_token	pt2_token
1	-7782220156096217088	1441151880758558720	94	14
6	-7723954835917111296	1230327123202277376	94cf	1113
11	-7724137629725229056	1230478580929003520	94ce59c	111389c
16	-7724137722872332288	1230478344437366784	94ce59aa5	11138988f
21	-7724137722769309696	1230478344653635584	94ce59aa5624	11138988fce4