Failing test: test_merge_and_average fails with a TypeError with pandas 2.0.3

[olp-cs]

The test is failing on platforms:

• Linux-6.1.85+-x86_64-with-glibc2.35
  • Python 3.10.12
  • pandas 2.0.3
  • numpy 1.25.2
• Windows-11-10.0.22631-SP0
  • Python 3.12
  • pandas 2.2.2
  • numpy 2.0.0

Update: The test passes with Python 3.8 and pandas 1.3.0; it seems to be a backward compatibility issue with pandas.

Failed test:

test_GEOparse.py:626 (TestGSE.test_merge_and_average)

TypeError: agg function failed [how->mean,dtype->object]

TypeError: Could not convert string 'DNA segment, Chr 8, ERATO Doi 594, expressed' to numeric

The test fails on this line:

..\src\GEOparse\GEOTypes.py:445: in annotate_and_average

tmp_data = tmp_data.groupby(group_by_column).mean()[[expression_column]]

where

• tmp_data is a pandas dataframe that contains both numeric and string columns (attached: tmp_data.csv);
• expression_column = 'VALUE'
• group_by_column = 'GB_ACC'

Jupyter notebook reproducing the issue:
https://gist.github.com/olp-cs/9902b5cdc554afbf3faa7127ee602f20

Would it make sense to filter the columns first, to keep the numerical ones only?