fix(go): geo type auto-detect should not require edges:spherical for GEOGRAPHY

[jatorre]

resolveGeoType's own doc comment says "missing CRS, EPSG:4326, or unparsable CRS stays GEOGRAPHY", and docs/snowflake.md's option table described detection differently again — but the implementation required edges == "spherical" in addition to SRID 4326 before choosing GEOGRAPHY.

Most GeoArrow producers emit a crs but no edges field (DuckDB's spatial extension and GeoPandas among them), so in practice every WGS84 column ingested through auto-detect silently landed as GEOMETRY. We hit this in CARTO's integration suite when moving from the pre-merge #117 branch to go/v1.11.0: same import, GEOGRAPHY became GEOMETRY.

This aligns the code with the documented contract:

• SRID 4326 or no CRS → GEOGRAPHY
• any other SRID → GEOMETRY (SRID survives the round trip via ST_SETSRID)
• edges:spherical with a non-4326 SRID → still an error (unchanged)
• explicit adbc.snowflake.statement.ingest_geo_type → unchanged, pins all columns

Updated TestResolveGeoType and the option docs to match. The two auto-detect validation txtcases are unaffected (the GEOGRAPHY one uses 4326+spherical, the GEOMETRY one uses EPSG:2029).

https://claude.ai/code/session_01Y3xqZXWrx9qXdPNXBzNxa2

[lidavidm]

CC @paleolimbot but my understanding is that this was semantically required

[paleolimbot]

CC paleolimbot but my understanding is that this was semantically required

Yes, this is my fault 🙂 .

DuckDB's spatial extension and GeoPandas among them

Both of these only support "geometry", so "geometry" is the type that should be automatically chosen (although it's fair that you'd want to override that choice if you want to insert into a Snowflake geography, which I believe is possible). Specifically, these tools treat edges as Cartesian planar edges regardless of CRS, so they map to the Snowflake version of that.

The only tools I know about that support the union of Geography, Geometry, and GeoArrow are SedonaDB and PyArrow (after GeoArrow extension types are loaded via import geoarrow.pyarrow). PyArrow will emit "edges": "spherical" when reading a GEOGRAPHY type from a Parquet file; SedonaDB has fairly complete support and is a few functions away from BigQuery parity (ST_TessellateGeom(), ST_TessellateGeog() , ST_ToGeometry() and ST_ToGeography() can get you between them nicely).

doc comment says "missing CRS, EPSG:4326, or unparsable CRS stays GEOGRAPHY"

Probably the fix here is to fix the comment, which is outdated.