Compute areasize() for multipolygons

[naoliv]

Shouldn't areasize() also return the area size of relations?

The area selector matches "any area regardless of whether the area border is only modelled with a single way or with a set of ways glued together with a relation" while areasize() only works with a closed way.

[naoliv]

Using the attached example file and this example mapcss rule it should match all the 3 areas.

area:closed[foo = yes][eval(areasize()) > 1] {
        throwWarning: tr("foo");
}

[simon04]

An implementation could be taken from or be inspired by the measurement plugin, MeasurementDialog.java

[bastiK]

Or this:

Multipolygon multipolygon = MultipolygonCache.getInstance().get(nc, r);
double area = 0;
for (PolyData pd : multipolygon.getCombinedPolygons()) {
    area += pd.getAreaAndPerimeter().getArea();
}

[simon04]

Replying to bastiK:

org.openstreetmap.josm.tools.Geometry#getAreaAndPerimeter returns the area wrt to the current projection. Thus, the value cannot be directly represented in the user's SystemOfMeasurement. Can we somehow handle this?

[bastiK]

To get a meaningful area in square meters, the coordinates need to be projected using an equal-area projection. Measurement plugin uses ​Sinusoidal projection. ​Cylindrical equal-area projection is even simpler.

The way this is done in Geometry (calcX/calcY) is incorrect: The author invented a funny projection that calculates the easting of a point (lat/lon) as great circle distance to the point (lat/0.0) and northing as great circle distance to the point (0.0/lon). This is more complicated and less correct than the projections mentioned above. Plus it is broken for areas crossing the prime meridian. I committed the code myself in [4085], so unfortunately, I have no one to blame. ;-)

Replying to simon04:

org.openstreetmap.josm.tools.Geometry#getAreaAndPerimeter returns the area wrt to the current projection. Thus, the value cannot be directly represented in the user's SystemOfMeasurement. Can we somehow handle this?

Right, the values are needed in units of the current projection for map rendering. Not sure if this can be merged somehow, but it may be simpler to add an extra method to Geometry which copies and adapts the code in PolyData.getAreaAndPerimeter.

[simon04]

In 9878/josm:

see #11516 - Add Geometry unit tests

[simon04]

For the outer ring of source:trunk/test/data/create_multipolygon.osm​,

1. the current implementation gives 5721923.660644531
2. the implementation from the measurement​ plugin gives 5736957.869628906
3. my attempt​ to implement ​https://en.wikipedia.org/wiki/Sinusoidal_projection gives 5757045.313964844

[bastiK]

Replying to simon04:

For the outer ring of source:trunk/test/data/create_multipolygon.osm​,

1. the current implementation gives 5721923.660644531
2. the implementation from the measurement​ plugin gives 5736957.869628906
3. my attempt​ to implement ​https://en.wikipedia.org/wiki/Sinusoidal_projection gives 5757045.313964844

The difference of the last two comes from different earth radii used as input. Regarding the patch: The EastNorth coordinates in the current projection are cached and should be used in this calculation (runs during rendering). Implementation of sinusoidal should either include ellipsoidal case (see ​geotools implementation) or throw an error when ellipsoid is not spherical.

[simon04]

bastiK, thank you for the review.

• _11516___Sinusoidal_projection-v2.patch​ computes 5760015.7353515625

What do you think? Is the order phi/lambda and east/north correct in the (inv-)project methods?

[bastiK]

Without testing, I would say

  return new double[]{east / cos(north), north};

should be

  return new double[]{north, east / cos(north)};

Otherwise +1.

As you may be aware, with this patch it uses ellipsoidal case of the projection, which is even more precise than the method of measurement plugin. The efficiency is fine too - the forward projections is fairly simple and overall it should be faster that what we have at the moment.

[simon04]

In 9950/josm:

see #11516 - Add Sinusoidal projection

[simon04]

In 9951/josm:

see #11516 - Compute closed way area using Sinusoidal projection

[simon04]

In 9952/josm:

see #11516 - Compute multipolygon area, include in MapCSS, JOSM search

[Don-vip]

In 9955/josm:

see #11516 - fix javadoc warning

[bastiK]

In 9956/josm:

update reference data for Sinusoidal projection (see #11516)

[Don-vip]

Milestone renamed