laskoviymishka opened a new pull request, #2986:
URL: https://github.com/apache/iceberg-rust/pull/2986
## Which issue does this PR close?
Closes #340.
## What changes are included in this PR?
Adds a `PositionDeleteFileWriter` (and its builder) under
`crates/iceberg/src/writer/base_writer/`, filling the slot the module doc
already named. It writes a spec-conforming position delete file — the two
required columns `file_path` (`string`, reserved field id 2147483546) and
`pos`
(`long`, 2147483545) — and tags the resulting `DataFile` with
`DataContentType::PositionDeletes` on close.
- `position_delete_schema()` / `position_delete_arrow_schema()` expose the
canonical (memoized) schema, built from the existing `metadata_columns`
field
definitions, so no field-id constants are duplicated.
- `write()` validates that each batch is exactly the two required,
non-nullable,
correctly-typed columns carrying the reserved field ids, and rejects a
closed
writer before validating — a clear error instead of a confusing failure
deeper
in the Parquet writer.
- `close()` stamps `DataContentType::PositionDeletes`, propagates the
partition
key, and leaves `sort_order_id` null per the spec.
- Modeled on the existing `equality_delete_writer.rs`.
This is the first of a short series building out the delete-file write path.
Deliberately out of scope here, planned as follow-ups: enforcing the
`(file_path, pos)` sort order (the `write()` docs state the caller's
responsibility until then), setting `referenced_data_file`, and a
higher-level
`DeltaWriter` (#2218) that routes inserts/deletes to the data / position /
equality writers.
## Are these changes tested?
Yes — 12 unit tests in the new module:
- schema shape (reserved field ids, names, types, required flag, non-null
Arrow
projection);
- a Parquet round-trip for single and multiple `write()` calls (write → read
back → compare), asserting content type, record count, partition, and file
size;
- partition + `partition_spec_id` propagation on `close()`;
- the validation matrix — wrong column count, missing/wrong field ids (both
columns), wrong Arrow types (including `LargeUtf8` `file_path`, the common
DataFusion / DuckDB / Polars shape), nullable columns, and write/close
after
close.
`cargo test -p iceberg`, `cargo clippy`, and `cargo fmt --check` are clean.
## AI assistance disclosure
Per the project's guidelines for AI-assisted contributions: this change was
drafted with AI assistance (Claude Code), then reviewed, tested, and
validated
by the author — including a multi-agent code-review pass before opening.
`cargo test`, `clippy`, and `fmt` were run locally.
--
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]
