Found an issue with running on MacOS for `ballista==53.0.0` wheels on
TestPyPI. Tested on linux successfully, passed all the examples from
python/README.
MacOS Issue:
Constructing the in-process BallistaScheduler or BallistaExecutor
segfaults. Reproduced on Python 3.10 and 3.12, with both uv and pip/venv.
Minimal repro on MacOS:
```
uv run --python 3.10 --with "ballista==53.0.0" \
--index-url https://test.pypi.org/simple/ \
--extra-index-url https://pypi.org/simple/ \
--index-strategy unsafe-best-match \
python -X faulthandler -u -c \
"from ballista import BallistaScheduler; BallistaScheduler()"
```
Output:
"""
Fatal Python error: Segmentation fault
Current thread 0x00000001effd1e80 (most recent call first):
File "<string>", line 1 in <module>
Extension modules: pyarrow.lib (total: 1)
"""
Linux via docker:
```
docker run --rm --platform linux/amd64 -v "$PWD/python/testdata:/data:ro"
python:3.10-slim bash -c '
pip install --quiet --index-url https://test.pypi.org/simple/ \
--extra-index-url https://pypi.org/simple/ "ballista==53.0.0"
datafusion &&
python -u <<PY
from ballista import BallistaSessionContext, BallistaScheduler,
BallistaExecutor
from datafusion import col, lit
import time
sched = BallistaScheduler(bind_port=50050); sched.start()
execu = BallistaExecutor(bind_port=50051, scheduler_port=50050);
execu.start()
time.sleep(3)
ctx = BallistaSessionContext("df://localhost:50050")
ctx.sql("create external table t stored as parquet location
\"/data/test.parquet\"")
ctx.sql("select * from t limit 5").show()
ctx.sql("select count(*) as n from t").show()
df = ctx.read_parquet("/data/test.parquet").filter(col("id") >
lit(4)).limit(5)
batches = df.collect()
print("rows:", sum(b.num_rows for b in batches))
execu.close(); sched.close()
PY'
```
Outputs:
"""
DataFrame()
+----+----------+-------------+--------------+---------+------------+-----------+------------+------------------+------------+---------------------+
| id | bool_col | tinyint_col | smallint_col | int_col | bigint_col |
float_col | double_col | date_string_col | string_col | timestamp_col
|
+----+----------+-------------+--------------+---------+------------+-----------+------------+------------------+------------+---------------------+
| 4 | true | 0 | 0 | 0 | 0 | 0.0
| 0.0 | 30332f30312f3039 | 30 | 2009-03-01T00:00:00 |
| 5 | false | 1 | 1 | 1 | 10 | 1.1
| 10.1 | 30332f30312f3039 | 31 | 2009-03-01T00:01:00 |
| 6 | true | 0 | 0 | 0 | 0 | 0.0
| 0.0 | 30342f30312f3039 | 30 | 2009-04-01T00:00:00 |
| 7 | false | 1 | 1 | 1 | 10 | 1.1
| 10.1 | 30342f30312f3039 | 31 | 2009-04-01T00:01:00 |
| 2 | true | 0 | 0 | 0 | 0 | 0.0
| 0.0 | 30322f30312f3039 | 30 | 2009-02-01T00:00:00 |
+----+----------+-------------+--------------+---------+------------+-----------+------------+------------------+------------+---------------------+
DataFrame()
+---+
| n |
+---+
| 8 |
+---+
rows: 3
"""
Best,
Kevin Liu
On Sun, May 24, 2026 at 7:49 AM Shekhar Rajak <[email protected]>
wrote:
> +1 (non-binding) — installed from test.pypi.org and ran the smoke import.
> $ pip show ballista
> python -c "import ballista; print(ballista.__version__)"
> Name: ballista
> Version: 53.0.0
> Summary: Python client for Apache Arrow Ballista Distributed SQL Query
> Engine
> Home-page: https://datafusion.apache.org/ballista/
> Author:
> Author-email:
> License:
> Location: /private/tmp/ballista-rc-verify/lib/python3.13/site-packages
> Requires: datafusion, pyarrow
> Required-by: 53.0.0
> Result: `from ballista import BallistaSessionContext; print('ok')` -> ok
>
>
>
>
> Regards,
> Shekharrajak
>
>
> On Sunday 24 May 2026 at 06:53:18 am GMT+5:30, Andy Grove <
> [email protected]> wrote:
>
> I have published a test version of Ballista to test.pypi.org [1] and I
> am looking for help testing this.
>
> Instructions for installing Ballista from test.pypi.org can be found
> in the release verification documentation [2].
>
> Please note that this is NOT an official Apache release. This is a
> test of the new PyPi publishing process.
>
> This release was built from GitHub tag 53.0.0-rc1-pypitest-3.
>
> I plan on creating an official 53.x.x release to PyPi pretty soon,
> once I have feedback from this test.
>
> Thanks,
>
> Andy.
>
> [1] https://test.pypi.org/project/ballista/
> [2]
> https://github.com/apache/datafusion-ballista/blob/main/dev/release/README.md#optional-verify-the-python-wheels-from-testpypi
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>
