Observations

When running a few notebooks on pandas 2.x, errors like those can be observed:

TypeError: Could not convert string 'BerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlin' to numeric

-- https://github.com/crate/cratedb-examples/actions/runs/8975962618/job/24651748395?pr=430#step:6:1839
-- https://github.com/crate/cratedb-examples/actions/runs/8975962618/job/24651748121?pr=430#step:6:825

References

• Time Series: Modernize notebooks to use recent versions of pandas and SQLAlchemy #387
• Update to pandas 2.2 in /topic/timeseries (blocked by PyCaret) #477

Evaluations

• It looks like it is a data shape error.
• Apparently, Google Colab now strictly uses and requires pandas 2.x since 2024-05-13?

Thoughts

It looks like it is a data shape error. Maybe the way the notebooks are working with pandas needs an update when using more recent pandas 2.x? The string repetition flaw reminds me of the famous »Wat« talk by Gary Bernhardt. ;]

Workaround

As a temporary measure, tests stopped including the corresponding notebook. It will get skipped per cfd1a6c, on behalf of the relevant modernization patch.

• Time Series: Update to pandas 2.0 and SQLAlchemy 2.0 #430

Time Series: Skip testing notebooks not compatible with pandas 2.x

• exploratory_data_analysis.ipynb
• time-series-decomposition.ipynb

They are not ready for pandas 2.x yet, and block others from being
upgraded.

Originally posted by @amotl in #430 (comment)