Bump pandas and sqlalchemy

[surister]

Summary of the changes / Why this is an improvement

Update pandas and sqlalchemy to latest (or fairly modern versions)

Checklist

• Link to issue this PR refers to (if applicable): Time Series: Modernize notebooks to use recent versions of pandas and SQLAlchemy #387
• CLA is signed

[surister]

Found some issues with Pandas while running all the notebooks, will try to fix later

[amotl]

Thanks a stack for working on this. There is an error on CI wrt. to PyCaret. I hope it will turn out well. Maybe PyCaret exclusively needs pandas<2 or SQLAlchemy<2?

File /opt/hostedtoolcache/Python/3.11.8/x64/lib/python3.11/site-packages/pycaret/internal/memory.py:25
ImportError: cannot import name '_format_load_msg' from 'joblib.memory' (/opt/hostedtoolcache/Python/3.11.8/x64/lib/python3.11/site-packages/joblib/memory.py)

-- https://github.com/crate/cratedb-examples/actions/runs/8604266577/job/23577980009?pr=400#step:6:859

If this is the case, we may need to split up the ingredients to be separated by pandas/SQLAlchemy major versions in one way or another, either by providing dedicated requirements.txt files, or by separating notebooks into different folders.

[amotl]

There is an error on CI wrt. to PyCaret: ImportError: cannot import name '_format_load_msg' from 'joblib.memory'.

I discovered the same error here:

• Update MLflow from 2.10.2 to 2.11.3 mlflow-cratedb#124

This adjustment aims to mitigate it. Effectively, just downgrade to joblib 1.3.

• crate/mlflow-cratedb@49cb2b6

See also:

• [BUG]: Can't use RegressionExperiment with the latest version of joblib (1.4.0) pycaret/pycaret#3959
• TSML/PyCaret woes: Downgrade to Joblib 1.3 and Python 3.10 #401

[amotl]

There is an error with PyCaret.

There is also a compatibility flaw with Dask and most recent Python 3.11.9, orthogonal to the Joblib 1.4 compatibility issue.

• [BUG]: TypeError: descriptor '__call__' for 'type' objects doesn't apply to a 'property' object pycaret/pycaret#3960 (comment)

[amotl]

The TSML stack modulo LLM/LangChain needed two fixes, as outlined above. After rebasing, relevant obstacles should be remedied for you.

• TSML/PyCaret woes: Downgrade to Joblib 1.3 and Python 3.10 #401
• TSML: Update to dask 2024.4.1 and mlflow-cratedb 2.11.3 #402

[amotl]

More havocs caused by recent updates had to be remedied, to make relevant artefacts work on Google Colab again. Thanks for your support, @marijaselakovic.

• TSML dependencies: Explicitly add dask-expr and distributed #403
• CI/TSML: Fix Time Series builds by downgrading to Python 3.11.8 #404

@surister: You may now want to rebase again, to be in a better position than before.

[amotl]

Hi there. I just rebased your patch, also resolving slight conflicts, and would like to verify it on CI. However, I can't push easily, because you forked the repository into a private one on your account. May I ask you to add corresponding permissions, so I can push, if you are fine with that?

$ git push --force
error: Authentication error: Authentication required: You must have push access to verify locks
error: failed to push some refs to 'github.com:surister/cratedb-examples.git'

[amotl]

This patch is currently blocked on errors like this, evaluated on behalf of GH-430 ¹:

TypeError: Could not convert string 'BerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlinBerlin' to numeric

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

Footnotes

1. @surister: Feel free to close your PR, and continue working on collab/update-pandas-sqlalchemy, directly from within this repository. It will save you a few cycles on rebasing. ↩