[BUG] Test Failures caused because of pandas 3

[satvshr]

Metadata

• Reference Issue: fixes [BUG] Test failures #1627

• What does this PR implement/fix? Explain your changes.
  This PR fixes the 7 recurring bugs across all current PRs because of pandas 3:

1. test_get_data_pandas bug: Solved type error for dataframe columns having str datatype for pandas==3 and object for older versions.
2. test_get_sparse_dataset_dataframe, test_get_sparse_dataset_rowid_and_ignore_and_target, and test_get_sparse_dataset_dataframe_with_target bug: typecasting type_ to a np array.
3. bugs in test_flow_functions.py: ext_version can now be nan because of pandas 3.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.75%. Comparing base (cf8e9db) to head (8010a20).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1628      +/-   ##
==========================================
+ Coverage   52.36%   52.75%   +0.39%     
==========================================
  Files          36       36              
  Lines        4333     4333              
==========================================
+ Hits         2269     2286      +17     
+ Misses       2064     2047      -17     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[geetu040]

Really nice, this fixes the bugs. Just left a comment for discussion, otherwise LGTM!

[geetu040]

Isn't that cheating? why would it ever fail for "name" column if we are dynamically extracting the dtype. Is it to make sure it works with pandas<3.0.0?

would make sense to replace "object" simply with "str" in col_dtype if we are setting pandas>=3.0.0 in dependencies

CC: @fkiraly, are we bounding to this pandas version in pyproject.toml?

[satvshr]

Isn't that cheating?

It is definitely cheating! Thought it was better to get a quick fix out and then fine-tune the fixes later on. It was between this or bounding pandas, and for me this was the better/less intense option between the 2.