FurkhanShaikh commented on issue #4079: URL: https://github.com/apache/iceberg-python/issues/4079#issuecomment-7055347730
A correction to my comment above. I wrote that "there are currently no geo bounds to misread". That's too broad: it only holds for data written by Spark with Iceberg 1.12.0, the version in the integration image. - Iceberg 1.12.0 already includes apache/iceberg#17161. That PR makes Iceberg's generic Parquet writer (`BaseParquetWriter`) emit XY bounds for **geometry** columns, serialized as `GeospatialBound` points (`x:y`). - The Spark 4.1 and 4.2 writers skipped those bounds in 1.12.0. apache/iceberg#18501, merged on Oct 28 after the 1.12.0 release, adds them there too. - In both cases, geography still gets size-only metrics. How this affects the two points above: - **Metrics evaluators:** the conclusion holds. Comparison predicates on geo columns fail at bind, so the evaluators never read these bounds. The bounds will matter once PyIceberg adds spatial predicates. On the Java side, bounding-box types are in apache/iceberg#12667, and spatial predicates are in the open apache/iceberg#14101. - **`inspect.files()` / `entries()`:** the representation question matters now rather than later. Geometry files from the generic writer can already carry real `x:y` bounds, which shouldn't be typed as WKB in `readable_metrics`. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
