SNOW-710684: Unable to use Cursor method fetch_arrow_all() without pandas installed

[ir3456]

Please answer these questions before submitting your issue. Thanks!

1. What version of Python are you using?

   3.10.8

2. What operating system and processor architecture are you using?

   macOS-12.4-arm64-arm-64bit

3. What are the component versions in the environment (pip freeze)?

   pyarrow==8.0.0
   snowflake-connector-python==2.8.3

4. What did you do?

I tried to use the Cursor method fetch_arrow_all() without pandas as a dependency which threw the following error:
NameError: name 'concat_tables' is not defined in the snowflake/connector/result_set.py module.

concat_tables and pyarrow are only imported if the installed_pandas variable is True. Tracing installed_pandas back, it is True when from pandas import DataFrame runs without an error.

Essentially, pyarrow and concat_tables are only imported if pandas is installed in the environment, thus using the fetch_arrow_all() method unnecessarily requires pandas as a dependency.

For use cases not requiring pandas (such as using polars) it would be nice not to have pandas as an unnecessary dependency but still be able to use the pyarrow functionality.

[sfc-gh-sfan]

Today pandas related operations depend on pyarrow, but not vice versa, so disentangling them seems to make sense so that we could provide arrow support without requiring pandas.

@sfc-gh-mkeller Let me know what you think.

[sfc-gh-achandrasekaran]

This is a feature request. We will look into it in the near future.

[axelwas]

Any updates on this ? It would be great to have native polars support !

[MelleVessies]

Any updates? The fix seems rather trivial. In options.py the function _import_or_missing_pandas_option tries to import both pandas and arrow in one go. Adding a extra function _import_or_missing_arrow_option seems like a logical solution. Ofcourse you still need to fix the arrow and pandas imports where-ever they are used.