Add Geography type support to SedonaFlink (parity with Spark)

[jiayuasu]

Goal

Bring Geography type support to SedonaFlink so that the Flink SQL/Table API reaches parity with the Spark module. Geography (geodesic, S2-backed) is already implemented in the engine-agnostic common module and fully exposed in Spark, but the Flink module exposes none of it — Flink users currently have no GEOGRAPHY type and no ST_Geog* functions.

Background

The building blocks already exist and are engine-agnostic:

• common module — Geography type and S2 backing:

  • common/src/main/java/org/apache/sedona/common/S2Geography/ — type hierarchy (Geography, PointGeography, PolylineGeography, PolygonGeography, MultiPolygonGeography, GeographyCollection, …), GeographySerializer, WKT/WKB readers & writers, Accessors, Predicates, Distance, Projection.
  • common/src/main/java/org/apache/sedona/common/geography/Constructors.java — geogFromWKT, geogFromEWKT, geogCollFromText, geogFromWKB, geogFromGeoHash, geogToGeometry, geomToGeography.
  • common/src/main/java/org/apache/sedona/common/geography/Functions.java — getEnvelope, asEWKT.

• Spark module — how it is exposed (the parity target):

  • spark/common/src/main/scala/org/apache/spark/sql/sedona_sql/UDT/GeographyUDT.scala — the GeographyUDT, wrapping GeographySerializer.
  • spark/common/src/main/scala/org/apache/spark/sql/sedona_sql/expressions/geography/Constructors.scala — 9 expressions: ST_GeogFromWKT, ST_GeogFromEWKT, ST_GeogFromText, ST_GeogCollFromText, ST_GeogFromWKB, ST_GeogFromEWKB, ST_GeogFromGeoHash, ST_GeogToGeometry, ST_GeomToGeography.
  • Registered in spark/common/src/main/scala/org/apache/sedona/sql/UDF/Catalog.scala.

• Flink module — what geography work will parallel (geometry-only today):

  • flink/src/main/java/org/apache/sedona/flink/GeometryTypeSerializer.java, GeometryArrayTypeSerializer.java — the serializer pattern to mirror.
  • flink/src/main/java/org/apache/sedona/flink/expressions/Constructors.java, Functions.java, Predicates.java — where ScalarFunction wrappers live.
  • flink/src/main/java/org/apache/sedona/flink/Catalog.java — function registration. No geography functions registered today.

Approach

Mirror the Spark integration in Flink. Each Flink ScalarFunction wraps the existing common-module geography functions, using a @DataTypeHint(value = "RAW", rawSerializer = GeographyTypeSerializer.class, bridgedTo = Geography.class) annotation — the same pattern the geometry constructors already use.

Scope

Parity with the Spark geography surface on master. Spark exposes, all backed by the engine-agnostic common/geography/:

• Type + 9 constructors/conversions — GeographyUDT, ST_GeogFromWKT, ST_GeogFromText, ST_GeogFromWKB, ST_GeogFromEWKB, ST_GeogFromEWKT, ST_GeogFromGeoHash, ST_GeogCollFromText, ST_GeogToGeometry, ST_GeomToGeography.
• 11 measurement/output functions that accept Geography — ST_Area, ST_Length, ST_Distance, ST_Buffer, ST_Centroid, ST_Envelope, ST_NPoints, ST_NumGeometries, ST_GeometryType, ST_AsText, ST_AsEWKT.
• 5 predicates that accept Geography — ST_Contains, ST_Intersects, ST_Within, ST_Equals, ST_DWithin.

In Spark these are Geography overloads on the existing geometry expressions (inferrableFunction accepting Geometry or Geography). Flink must expose the same set so geography columns are usable end-to-end, not just constructible.

Sub-tasks

• Type serializer — GeographyTypeSerializer wrapping org.apache.sedona.common.S2Geography.GeographyWKBSerializer, mirroring GeometryTypeSerializer — SedonaFlink: add GeographyTypeSerializer for the Geography type #3058 ✅
• Constructors + Catalog + tests — 9 ScalarFunction wrappers in expressions/geography/GeographyConstructors.java, registered in Catalog.java, with round-trip tests — SedonaFlink: add geography constructor functions (ST_Geog*) #3061 ✅
• Measurement/output functions — geography support for the 11 functions above, wired to common.geography.Functions, with tests — SedonaFlink: add geography support to measurement and output functions #3073 ✅
• Predicates — geography support for the 5 predicates above, with tests — SedonaFlink: add geography support to spatial predicates #3075 ✅
• Docs — SedonaFlink API docs covering geography constructors, functions, and predicates — SedonaFlink: document geography constructor functions #3069 ✅

All sub-tasks complete. SedonaFlink now has full geography parity with Spark: the Geography type serializer, 9 constructors/conversions, 11 measurement/output functions, 5 predicates, and complete API docs.

Notes

• All common-module geography infrastructure is already implemented and engine-agnostic; this epic is purely about wiring it into Flink — no new geodesic/S2 logic is required.
• No GeographyArrayTypeSerializer for now — no geography function returns an array yet.
• As Spark adds geography predicates/measurements/accessors, follow-up issues should track Flink parity for those.

[jiayuasu]

This has been fully resolved.