IndexError in distributed.client.compute #9117
Description
Describe the issue:
I am getting IndexError: list index out of range when calling client.compute(collections). This is because expr.optimize is pruning my input collection from size 2 to a single key, which is causing and index out of bounds lookup over the names.
Minimal Complete Verifiable Example:
def test_mvp_index_out_of_bounds() -> None:
from dask.distributed import Client as DaskClient
from dask import delayed
dask_client = DaskClient(processes=False, n_workers=1, threads_per_worker=1)
pdf = pd.DataFrame({'a': [1, 2, 3]})
ddf = dd.from_pandas(pdf, npartitions=1)
def do_something(pdf: pd.DataFrame) -> None:
return None
dl = delayed(do_something)(pdf)
futures = dask_client.compute([ddf, dl])
results = dask_client.gather(futures)
print(len(results))
print(results)Anything else we need to know?:
Environment: Pytest
$ poetry show dask
name : dask
version : 2025.9.1
description : Parallel PyData with Task Scheduling
- Python version:
3.11.9 - Operating System: Debian Linux
- Install method (conda, pip, source): poetry