Polyfit fails with few non-NaN values

[sfinkens]

What happened:
A linear DataArray.polyfit seems to fail if there are less than 3 non-NaN elements along the fitting dimension.

Traceback

TypeError: only size-1 arrays can be converted to Python scalars

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "polyfit.py", line 6, in <module>
    out = arr.polyfit(dim='x', deg=1)
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/xarray/core/dataarray.py", line 3455, in polyfit
    return self._to_temp_dataset().polyfit(
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/xarray/core/dataset.py", line 5962, in polyfit
    coeffs, residuals = duck_array_ops.least_squares(
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/xarray/core/duck_array_ops.py", line 625, in least_squares
    return nputils.least_squares(lhs, rhs, rcond=rcond, skipna=skipna)
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/xarray/core/nputils.py", line 239, in least_squares
    out[:, nan_cols] = np.apply_along_axis(
  File "<__array_function__ internals>", line 5, in apply_along_axis
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/numpy/lib/shape_base.py", line 379, in apply_along_axis
    res = asanyarray(func1d(inarr_view[ind0], *args, **kwargs))
  File "/home/stephan/venv/variogram/lib/python3.8/site-packages/xarray/core/nputils.py", line 227, in _nanpolyfit_1d
    out[:-1], out[-1], _, _ = np.linalg.lstsq(x[~mask, :], arr[~mask], rcond=rcond)
ValueError: setting an array element with a sequence.

I've played around with the degree a little bit and the error seems to occur as soon as (# of non-NaN values - degree) < 2

What you expected to happen:

The fit to succeed - I think two non-NaN values should be enough for a linear fit. I also noticed that there is no RankWarning: Polyfit may be poorly conditioned if the degree is larger than the number of non-NaN values.

Minimal Complete Verifiable Example:

import xarray as xr
import numpy as np

arr = xr.DataArray([np.nan, 1, 2], dims='x', coords={'x': [0, 1, 2]})
arr.polyfit(dim='x', deg=1)

Anything else we need to know?:

Environment:

Output of xr.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.8.2 (default, Apr 27 2020, 15:53:34)
[GCC 9.3.0]
python-bits: 64
OS: Linux
OS-release: 5.4.0-39-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8
libhdf5: 1.10.4
libnetcdf: 4.6.3

xarray: 0.15.2.dev112+g54b9450b
pandas: 1.0.5
numpy: 1.19.0
scipy: 1.5.0
netCDF4: 1.5.3
pydap: None
h5netcdf: None
h5py: None
Nio: None
zarr: 2.4.0
cftime: 1.1.3
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.19.0
distributed: 2.19.0
matplotlib: 3.2.2
cartopy: None
seaborn: None
numbagg: None
pint: None
setuptools: 44.0.0
pip: 20.0.2
conda: None
pytest: None
IPython: 7.16.1
sphinx: None

[dcherian]

Thanks for the report @sfinkens

cc @aulemahal