[FEATURE] more flexible parsing of geometries for /snapshot

[spwoodcock]

This is low priority

Problem

• /snapshot accepts a JSON, with a geometry key.
• The geometry must be of type Polygon or MultiPolygon.
• Passing a Feature or FeatureCollection fails.
• I have not testing passing a GeometryCollection, but I assume that fails too.
• Upstream services that use raw-data-api must parse GeoJSON and extract only a Polygon or MultiPolygon first.

Solution

• As raw-data-api is the lowest level, it would be nice to have the handling here, to prevent duplicating geometry extraction code.
• It's not too tricky to identify the type of GeoJSON passed in by the user, using various libraries available: python-geojson, geojson-pydantic, etc.

Multiple Geometries

• This also raises the question for what to do if the user includes multiple geometries in their FeatureCollection or GeometryCollection.
• If all nested geometries are of type Polygon, then they could be merged together.
• But if different types are contained together, this causes issues.
• If handling multiple geometries is required, then this becomes more complex.

Additional context

• We mostly parse everything into a standard FeatureCollection for FMTM, for flexibility of parsing many types of files.

[spwoodcock]

One option for supporting multiple geometries could be to simply convex hull them:

import json
from shapely import to_geojson
from shapely.geometry import shape
from shapely.ops import unary_union

if (geom_type := aoi_geojson.get("type")) == "FeatureCollection":
    # Convert each feature into a Shapely geometry
    geometries = [
        shape(feature.get("geometry")) for feature in aoi_geojson.get("features", [])
    ]
    merged_geom = unary_union(geometries)
elif geom_type == "Feature":
    merged_geom = shape(aoi_geojson.get("geometry"))
else:
    merged_geom = shape(aoi_geojson)

# Convex hull with enclose all geometries in a single polygon
merged_aoi_polygon = json.loads(to_geojson(merged_geom.convex_hull))     

• It could also be possible to generate extracts based on multipolygons / multiple features, instead of enclosing in a convex hull, however I am sure this is more complex to implement.
• If the user provides a feature collection with geoms very far apart (e.g. in different cities), then the merged AOI would be far too large. Extract generation would presumably fail.