Xarray resample().last() fails with Dask Cluster and flox for chunks containing only NaNs

[josephnowak]

I'm getting the following error when I have flox installed and I try to run a resample with the last method:

TypeError: float() argument must be a string or a real number, not 'ReprObject'

You should be able to reproduce the error using this script:

import pandas as pd
import xarray as xr
import numpy as np

from dask.distributed import Client

client = Client()

arr = xr.DataArray(
    np.nan,
    coords={
        "date": pd.to_datetime(["2025-03-24", "2025-06-23"]),
    },
    dims=["date"]
).chunk(date=(1, 1))

arr.resample(date="QE").last().compute()

Xarray show_versions:

INSTALLED VERSIONS
------------------
commit: None
python: 3.11.13 (main, Jul 11 2025, 22:36:59) [MSC v.1944 64 bit (AMD64)]
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 165 Stepping 2, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: ('es_ES', 'cp1252')
libhdf5: None
libnetcdf: None

xarray: 2025.7.1
pandas: 2.3.1
numpy: 2.1.3
scipy: 1.15.3
netCDF4: None
pydap: None
h5netcdf: None
h5py: None
zarr: 3.1.1
cftime: None
nc_time_axis: None
iris: None
bottleneck: 1.5.0
dask: 2025.7.0
distributed: 2025.7.0
matplotlib: None
cartopy: None
seaborn: None
numbagg: 0.9.0
fsspec: 2025.7.0
cupy: None
pint: None
sparse: 0.17.0
flox: 0.10.4
numpy_groupies: 0.11.3
setuptools: 80.3.1
pip: 25.1.1
conda: None
pytest: 8.4.1
mypy: None
IPython: 9.4.0
sphinx: None

[dcherian]

this doesn't error for me. Can you confirm in a new environment please?

[josephnowak]

It's strange, I created a new clean venv, and I could replicate the error. I tried on some other Windows machines, and the problem is also reproducible. I created a Dockerized version to try to isolate the script from any OS problem, and it raised the same error. Attached are the details of the Docker that should help to reproduce the error on your machine.

Dockerfile:

FROM python:3.11-slim

ENV PYTHONDONTWRITEBYTECODE=1 \
    PYTHONUNBUFFERED=1

WORKDIR /app

RUN python -m pip install --upgrade --no-cache-dir pip \
 && python -m pip install --no-cache-dir \
    "dask[complete]==2025.7.0" \
    "xarray==2025.8.0" \
    "flox==0.10.5"

# Copy only the files needed to run the script
COPY verify_error.py /app/

# Default command: run the script
CMD ["python", "verify_error.py"]

verify_error.py:

import pandas as pd
import xarray as xr
import numpy as np

from dask.distributed import Client


if __name__ == "__main__":
    client = Client()

    arr = xr.DataArray(
        np.nan,
        coords={
            "date": pd.to_datetime(["2025-03-24", "2025-06-23"]),
        },
        dims=["date"]
    ).chunk(date=(1, 1))

    arr.resample(date="QE").last().compute()

Complete traceback:

2025-08-18 10:37:12,316 - distributed.worker - ERROR - Compute Failed
Key:       ('groupby_nanlast-64f99473d79484d842dfd1c800caaf7c', 1)
State:     executing
Task:  <Task ('groupby_nanlast-64f99473d79484d842dfd1c800caaf7c', 1) _execute_subgraph(...)>
Exception: 'TypeError("float() argument must be a string or a real number, not \'ReprObject\'")'
Traceback: '  File "/usr/local/lib/python3.11/site-packages/flox/core.py", line 1705, in _reduce_blockwise\n    result = _finalize_results(results, agg, axis, expected_groups, reindex=reindex)\n       
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n  File "/usr/local/lib/python3.11/site-packages/flox/core.py", line 1410, in _finalize_results\n    finalized[agg.name] = finalized[agg.name].astype(agg.dtype["final"], copy=False)\n                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n'

Traceback (most recent call last):
  File "/app/verify_error.py", line 19, in <module>
    arr.resample(date="QE").last().compute()
  File "/usr/local/lib/python3.11/site-packages/xarray/core/dataarray.py", line 1241, in compute
    return new.load(**kwargs)
           ^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/xarray/core/dataarray.py", line 1167, in load
    ds = self._to_temp_dataset().load(**kwargs)
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/xarray/core/dataset.py", line 561, in load
    evaluated_data: tuple[np.ndarray[Any, Any], ...] = chunkmanager.compute(
                                                       ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/xarray/namedarray/daskmanager.py", line 85, in compute
    return compute(*data, **kwargs)  # type: ignore[no-untyped-call, no-any-return]
           ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/dask/base.py", line 681, in compute
    results = schedule(expr, keys, **kwargs)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/flox/core.py", line 1705, in _reduce_blockwise
    result = _finalize_results(results, agg, axis, expected_groups, reindex=reindex)
      ^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/site-packages/flox/core.py", line 1410, in _finalize_results
    finalized[agg.name] = finalized[agg.name].astype(agg.dtype["final"], copy=False)
      ~~~~~~~^^^^^^^^^^
TypeError: float() argument must be a string or a real number, not 'ReprObject'

Commands used to run Docker:

1. docker build -t flox-test .
2. docker run --rm -it flox-test

Notes:

• I tried with Python 3.13, and I got the error.
• I tried with 'bookworm', and I got the error

[dcherian]

Thanks @josephnowak this is now released to pypi and conda-forge