jatorre opened a new issue, #3639: URL: https://github.com/apache/iceberg-rust/issues/3639
### Is your feature request related to a problem or challenge? PyIceberg's `add_files` registers existing Parquet files without rewriting them, but PyIceberg cannot write format-version 3 snapshots yet (apache/iceberg-python#3620–#3622), so there is no way from Python to add files to a v3 table, for example one with a `geometry` column. iceberg-rust already does the hard part: its fast append writes v3 manifest lists with `first-row-id` and the snapshot's row range. Two pieces are missing to use it from Python: 1. `ParquetWriter::parquet_files_to_data_files` (data files from existing Parquet footers) is `pub(crate)` and `#[allow(dead_code)]`. 2. `Transaction::commit` needs a Rust `Catalog`. A Python caller already has a catalog client (PyIceberg's), so it only needs the commit itself: the requirements and updates of a `CommitTableRequest`. ### Describe the solution you'd like - Make `parquet_files_to_data_files` public. - Add `Transaction::build_commit(self) -> Result<TableCommit>`: apply the actions (writing the manifests) and return the commit without sending it. No refresh and no retry: the caller rebuilds on a conflict. - In `pyiceberg-core`, add `pyiceberg_core.transaction.add_files(identifier, metadata_location, table_metadata_json, file_paths, io_properties, snapshot_properties=None) -> str`, returning the commit as JSON for PyIceberg to send with `catalog.commit_table`. We tried this against a REST catalog (our own, SeaweedFS-based): v3 table with a geometry column created by DuckDB 2.0, two Parquet files added through the binding, commit sent by PyIceberg 0.12. DuckDB and PyIceberg then read 3,000 rows with consistent row ids, and a later DuckDB insert continued them (`next-row-id` 3000 → 3010). ### Willingness to contribute I can contribute this feature independently (PR to follow). -- 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]
