yihua opened a new issue, #833: URL: https://github.com/apache/hudi-rs/issues/833
`HudiDataSource::scan_parquet` builds each `PartitionedFile` with the base file's `file_metadata.size`, and passes `0` when the slice carries none (a write stat without `fileSizeInBytes`, or a metadata-table entry with size 0). DataFusion's file-group repartitioner assigns a zero-length file to no partition, so the file drops out of the scan without an error once the group's total passes `datafusion.optimizer.repartition_file_min_size`. That threshold was 10 MiB on DataFusion 54 and is 1 MiB on 55, so most real tables now cross it. A slice whose base file has no known size should not reach the parquet source; `HudiScanExec` looks the size up itself. The routing has to hold whether the table persists `hoodie.table.base.file.format` (Spark-written tables do) or the format is inferred from the slices. Fixed in #698. -- 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]
