voonhous commented on issue #19782: URL: https://github.com/apache/hudi/issues/19782#issuecomment-5451833594
Two ways to close this out: 1. Fix: apply the parquet-format name rule (a repeated group named `array` or `<parent>_tuple` is the element) in `HoodieAvroReadSupport` before delegating to parquet-avro, plus a single-field struct case in `TestParquetReaderCompatibility`. Small, read-side only, works on any engine parquet version. 2. Won't Fix: document the workaround and move on. Set `parquet.avro.write-old-list-structure=false` (Spark: `spark.hadoop.parquet.avro.write-old-list-structure=false`) before the first write so lists are written 3-level, or add a second field to the element struct, or run on Spark 4 where parquet-avro 1.14+ already carries the fix. Existing 2-level files with a one-field element struct stay unreadable through `HoodieAvroParquetReader` and need a rewrite. Going with 2. This has been a known limitation since 2020 with a working preventive setting, it does not affect Spark 4, and dev cycles are better spent on higher priority items. Closing as not planned. -- 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]
