Support for combining OSM point and area elements in GOQL queries (out center)

[sfkeller]

A very typical use case for querying POIs (e.g. museum), is, that you like to get not only the OSM node elements as points and OSM area elements as polygons, but both aggregated into one single resultset of geometry type point.

This requires that 1. a "center point" function (Centroid or even better a PointOnSurface) is applied to the area geometry of the OSM Area elements, and 2. that this resultset is being combined ("union") in to one resultset, consisting of the OSM Node elements and the OSM Area elements with "center point" geometries.

The GOQL docs are unfortunately very rudimentary about describing and controlling the resultset https://docs.geodesk.com/goql .

Looking at the only GOQL example "Museums in Paris" in https://docs.geodesk.com/python/examples/museums, this is how an Overpass QL query would look like:

/* Museums in Paris
*/ 
[out:json];
{{geocodeArea:Paris}}->.searchArea;
//area[boundary=administrative][admin_level=8][name="Paris"]->.searchArea; // not-unique
( 
  nwr[tourism=museum](area.searchArea);
); 
out center;

=> How can this implemented in Geodesk?

[clarisma]

Hi Stefan,

GOQL only describes the filter for type and tags; it doesn't control output formatting. However, in your case you can use the features' centroid attribute, like this:

for museum in museums(paris):
    c = museum.centroid
    print(f"{museum.name}: {c.lon}, {c.lat}")

If you prefer point-on-surface, there's a Shapely method for that (Shapely is automatically installed with GeoDesk for Python, you only need to import shapely in your script):

    p = geodesk.from_mercator(shapely.point_on_surface(museum.shape))
    print(f"  Point-on-surface: {p.x}, {p.y}")

Keep in mind that unlike centroid, which returns a GeoDesk Coordinate, point_on_surface returns a Shapely Point, which has no concept of longitude and latitude -- hence, you have to un-project it using from_mercator and then access its x and y.

GeoDesk doesn't yet support OSM JSON as a native output format, so to obtain the same output as [out:json] in Overpass, you would need to assemble a JSON object from Python lists and dicts, then json.dump it.

[sfkeller]

Many thanks @clarisma for the quick answer. You wrote:

GeoDesk doesn't yet support OSM JSON as a native output format

OSM JSON is not required here - GeoJSON is better anyway. And centroid is also good-enough here.

What I'm still struggling is with the GOQL documentation: Does the resultset / output format contain also the id, version, changeset, timestamp and uid of the OSM elements?

[clarisma]

Does the resultset / output format contain also the id, version, changeset, timestamp and uid of the OSM elements?

GOLs store the feature's ID and OSM type, but no other metadata (so no information about the last edit).