BUG: Grouper.cov emits warning unlike DataFrame.cov

[ghost]

Like #27470, another case where Grouper and DataFrame provide essentially the same method but diverge in behavior.

import pandas as pd

df=pd.DataFrame(dict(a=[0,4,8],b=[2,7,12],c=[3,9,11]))
g=df.groupby([0,1,1])

# single row frame silently emits NaNs
df.iloc[[0]].cov()

# single row group emits multiple warnings
g.cov()
RuntimeWarning: Degrees of freedom <= 0 for slice 

[jbrockmendel]

I'm not seeing warnings produced for the first call but now the second. Is the behavior you expect?

[mroeschke]

I see warnings in both cases now and since they are both coming from numpy I don't necessarily thing we should be testing this. So closing

In [2]: df.iloc[[0]].cov()
pandas/core/frame.py:9516: RuntimeWarning: Degrees of freedom <= 0 for slice
  base_cov = np.cov(mat.T, ddof=ddof)
python3.8/site-packages/numpy/lib/function_base.py:2542: RuntimeWarning: divide by zero encountered in true_divide
  c *= np.true_divide(1, fact)
python3.8/site-packages/numpy/lib/function_base.py:2542: RuntimeWarning: invalid value encountered in multiply
  c *= np.true_divide(1, fact)
Out[2]:
    a   b   c
a NaN NaN NaN
b NaN NaN NaN
c NaN NaN NaN

In [3]: g.cov()
pandas/core/frame.py:9516: RuntimeWarning: Degrees of freedom <= 0 for slice
  base_cov = np.cov(mat.T, ddof=ddof)
python3.8/site-packages/numpy/lib/function_base.py:2542: RuntimeWarning: divide by zero encountered in true_divide
  c *= np.true_divide(1, fact)
python3.8/site-packages/numpy/lib/function_base.py:2542: RuntimeWarning: invalid value encountered in multiply
  c *= np.true_divide(1, fact)
Out[3]:
        a     b    c
0 a   NaN   NaN  NaN
  b   NaN   NaN  NaN
  c   NaN   NaN  NaN
1 a   8.0  10.0  4.0
  b  10.0  12.5  5.0
  c   4.0   5.0  2.0