alitrack commented on issue #445: URL: https://github.com/apache/hudi-rs/issues/445#issuecomment-6282470798
Adding a local-path repro for this, observed on Windows through the DuckDB C++ extension (hudi 0.5.0, `default-features=false`, `x86_64-pc-windows-msvc`, invoked over the C FFI): **Two-level hive-style partitioned COW table, local `file://` path -> 0 rows / 0 partitions.** - table partitioned by `(byteField, shortField)` (complex keygen), hive-style dirs `byteField=10/shortField=300/...` - unfiltered read returns **0 rows**; `EXPLAIN` shows `Partitions: 0 / 0`, `File Slices: 0 / 0` - the base `.parquet` files are intact (present, non-empty; `.hoodie_partition_metadata` present) **Controls (same build, same host):** - a single-level *non-hive* partitioned table (`10/ 20/ 30/`) reads 4 rows; `WHERE byteField = 10` prunes to `Partitions: 1 / 3` correctly - the same two-level hive table on Linux (same commit): 4 rows So partition *discovery* (not reading) fails on Windows for nested hive-style partition directories, while flat partition directories work. This is consistent with the `PathBuf` backslash handling in this issue affecting the local directory walk. Happy to provide the exact fixture and repro steps if useful. -- 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]
