alberttwong opened a new issue, #40919:
URL: https://github.com/apache/beam/issues/40919

   ### What would you like to happen?
   
   Please validate Apache Beam's Python SDK with the latest published PyArrow, 
**26.0.0**, and raise the supported upper bound to include it once the relevant 
tests pass.
   
   Apache Beam 2.77.0 currently declares `pyarrow>=14.0.1,<26.0.0`. The same 
cap is present in [sdks/python/setup.py on current 
master](https://github.com/apache/beam/blob/b88840e0d3624dc2bec79a5b6f0292b640761ef2/sdks/python/setup.py#L152).
 The nearby comment asks that the PyArrow test matrix in `sdks/python/tox.ini` 
be updated together with this bound.
   
   ### Use case / motivation
   
   We are testing native Arrow Database Connectivity (ADBC) data workflows 
alongside Beam. Our other current analytics integrations use PyArrow 26, but 
Beam's cap forces a separate environment or a downgrade to PyArrow 25. This 
request concerns dependency interoperability; it does not assume PyArrow 26 
runtime compatibility is already proven.
   
   As a related upstream example, MLflow has already accepted this upgrade: 
[mlflow/mlflow#27201](https://github.com/mlflow/mlflow/issues/27201), fixed on 
its main branch by 
[mlflow/mlflow#27216](https://github.com/mlflow/mlflow/pull/27216). Beam still 
needs its own validation and dependency update.
   
   ### Reproduction
   
   On macOS arm64, Python 3.12.15:
   
   ```sh
   uv venv --python 3.12 /tmp/beam-pyarrow26
   uv pip install --dry-run --python /tmp/beam-pyarrow26/bin/python \
     'apache-beam==2.77.0' 'pyarrow==26.0.0'
   ```
   
   Observed resolver result:
   
   ```text
   No solution found when resolving dependencies
   apache-beam>=2.77.0 depends on pyarrow>=14.0.1,<26.0.0
   you require pyarrow==26.0.0
   your requirements are unsatisfiable
   ```
   
   Verified October 30, 2026. PyPI lists PyArrow 26.0.0 as the latest release 
and requires Python >=3.11, so that Python-version boundary should be 
considered when updating supported environments.
   
   ### Suggested scope
   
   - Exercise the Arrow conversion/schema, Parquet I/O, and applicable 
batch/dataframe paths on PyArrow 26.
   - Update the SDK dependency range and corresponding PyArrow CI/tox 
environments after validation.
   - Retain compatible resolution for supported older Python versions and 
document any version-specific limits.
   
   I have reproduced the package-resolution restriction; I have not bypassed 
the constraint or run Beam's upstream test suite against PyArrow 26.
   
   ### Issue Priority
   
   Priority: 3 (dependency interoperability improvement).
   
   ### Issue Component
   
   Component: sdk-py-core.
   
   Prepared with AI assistance (OpenAI Codex).
   


-- 
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]

Reply via email to