BUG: rank is not supported for double[pyarrow] dtype

[dylan-lee94]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
import numpy as np
pd.options.mode.dtype_backend = 'pyarrow'

df = pd.DataFrame({
    'tags': pd.Series([1,1,1,2,2,2,3,3,3,4,4,4,5,5,5],dtype='int64[pyarrow]'),
    'value': pd.Series(np.random.rand(15),dtype='double[pyarrow]')
    })

result = df['value'].rank()
result = df.groupby('tags')['value'].rank()

Issue Description

When grouping a dataframe and applying the rank function on a column with data type double[pyarrow] I get the following error:
TypeError: rank is not supported for double[pyarrow] dtype

However, applying the rank function without groupby works. This leads me to believe that the error message is misleading and that in fact the rank function does support data type double[pyarrow].

Expected Behavior

The rank function works in combination with groupby for data type double[pyarrow].

Installed Versions

Replace this line with the output of pd.show_versions()

[mroeschke]

Thanks for the report.

Looks like for rank specifically we prevent casting to the original data type, probably because rank should be numeric and we don't want to cast the type back to a non-numeric type.

pandas/pandas/core/groupby/ops.py

Lines 376 to 379 in 2a983bb

	if self.how in self.cast_blocklist:
	# i.e. how in ["rank"], since other cast_blocklist methods don't go
	# through cython_operation
	return res_values

In the short term, maybe an exception should be make here if the original dtype isn't numeric? In the longer term, I think we'll integrate a way to dispatch to pyarrrows groupby aggregation methods.

[topper-123]

This looks fixed:

import pandas as pd
import numpy as np
# pd.options.mode.dtype_backend = 'pyarrow'

df = pd.DataFrame({
    'tags': pd.Series([1,1,1,2,2,2,3,3,3,4,4,4,5,5,5],dtype='int64[pyarrow]'),
    'value': pd.Series(np.random.rand(15),dtype='double[pyarrow]')
    })

result = df['value'].rank()
result = df.groupby('tags')['value'].rank()
print(result)

with the result being:

0     3.0
1     2.0
2     1.0
3     2.0
4     1.0
5     3.0
6     1.0
7     2.0
8     3.0
9     3.0
10    1.0
11    2.0
12    1.0
13    3.0
14    2.0
Name: value, dtype: double[pyarrow]

Before I close it, do you agree this works as intended now?

[topper-123]

Ok, I'm closing, you can open again if you see something that needs to be discussed.