Spark-only BerlinMOD benchmark harness consuming the canonical suite#23

Open

estebanzimanyi wants to merge 31 commits into

MobilityDB:mainfrom

estebanzimanyi:feat/berlinmod-benchmark

Commits on Jun 13, 2026

Add the catalog-driven Spark UDF generator
estebanzimanyi
committed
Drive the full JMEOS surface from the catalog, organized by doxygen group
estebanzimanyi
committed
Generate the portable bare-name dispatch from the contract families
estebanzimanyi
committed

Commits on Jun 14, 2026

Make the generated-dispatch surface build cleanly
estebanzimanyi
committed
Wire build-time generation of the UDF surface (generate-sources)
estebanzimanyi
committed
Add runtime verification of the generated UDF surface
estebanzimanyi
committed
Generate the *_in parse functions (const char* -> Java String)
estebanzimanyi
committed
Broaden generated-surface verification to cbuffer + npoint families
estebanzimanyi
committed
Generate uint64_t functions + default codegen output to target/
estebanzimanyi
committed
Generate the full portable bare-name operator dispatch surface
estebanzimanyi
committed
Generate the *_as_hexwkb family (swallow size_out, map unsigned char)
estebanzimanyi
committed
Generate the @sqlfn canonical MobilityDB SQL surface with overload dispatch
estebanzimanyi
committed
Re-vendor the catalog from the consolidated MEOS-API at pin 14h
estebanzimanyi
committed
Run the generated-dispatch surface on Java 21 / Spark 3.5 under CI
estebanzimanyi
committed
Rebuild the BerlinMOD benchmark on the generated UDF surface
estebanzimanyi
committed
Re-vendor the generator with the wrong-type-WKB crash + SQL-arity fixes
estebanzimanyi
committed
Bump the CI libmeos pin to ecosystem-pin-2026-06-14i
estebanzimanyi
committed
Re-vendor the generator with the atTime time-restrict + ecosystem-TZ fix
estebanzimanyi
committed
Wire the bench loader to the canonical load.sql H3 build (ready for the tags)
estebanzimanyi
committed
Delete the committed BerlinMOD data copy; read the canonical data instead
estebanzimanyi
committed
Drop the unused hand-registration helpers; the UDF surface is fully generated
estebanzimanyi
committed
Advance to ecosystem-pin-2026-06-14l (catalog + JMEOS jar + CI pin)
estebanzimanyi
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Spark-only BerlinMOD benchmark harness consuming the canonical suite#23

Spark-only BerlinMOD benchmark harness consuming the canonical suite#23
estebanzimanyi wants to merge 31 commits into
MobilityDB:mainfrom
estebanzimanyi:feat/berlinmod-benchmark

Commits on Jun 13, 2026

Add the catalog-driven Spark UDF generator

Drive the full JMEOS surface from the catalog, organized by doxygen group

Generate the portable bare-name dispatch from the contract families

Commits on Jun 14, 2026

Make the generated-dispatch surface build cleanly

Wire build-time generation of the UDF surface (generate-sources)

Add runtime verification of the generated UDF surface

Generate the _in parse functions (const char -> Java String)

Broaden generated-surface verification to cbuffer + npoint families

Generate uint64_t functions + default codegen output to target/

Generate the full portable bare-name operator dispatch surface

Generate the *_as_hexwkb family (swallow size_out, map unsigned char)

Generate the @sqlfn canonical MobilityDB SQL surface with overload dispatch

Re-vendor the catalog from the consolidated MEOS-API at pin 14h

Run the generated-dispatch surface on Java 21 / Spark 3.5 under CI

Rebuild the BerlinMOD benchmark on the generated UDF surface

Re-vendor the generator with the wrong-type-WKB crash + SQL-arity fixes

Bump the CI libmeos pin to ecosystem-pin-2026-06-14i

Re-vendor the generator with the atTime time-restrict + ecosystem-TZ fix

Wire the bench loader to the canonical load.sql H3 build (ready for the tags)

Delete the committed BerlinMOD data copy; read the canonical data instead

Drop the unused hand-registration helpers; the UDF surface is fully generated

Advance to ecosystem-pin-2026-06-14l (catalog + JMEOS jar + CI pin)

Commits on Jun 15, 2026

Advance to ecosystem-pin-2026-06-14m + generate the NxN array UDFs

Add a Spark-dialect overlay for the PG-only NxN array queries (q05/q06/q10)

Make the Spark-dialect NxN queries actually run (rows.field + bare dist literal)

Advance to ecosystem-pin-2026-06-15a (canonical operator dialect)

Advance to ecosystem-pin-2026-06-15c (catalog + JMEOS jar + CI pin)

Advance to ecosystem-pin-2026-06-15d (h3 @sqlfn fix → geoToH3IndexSet)

Advance pin to ecosystem-pin-2026-06-15e (npoint_test build fix)

Fix the th3 spatial prefilter on Spark: EWKT/SRID geo parse, eEq set overload, hex-reject guard

Advance to ecosystem-pin-2026-06-15f (thread-safe geo_from_text; geoToH3Cell)