[FEATURE] add log_dtypes to pandas_utils

[koaning]

I might also want to log the dtypes between each step in a pandas pipeline. It's probably best to add a separate logger for both the dtype as well as the shape.

Should the column names be seperate too? Maybe worth a discussion.

It would also be nice if these features were documented on the documentation page together with some other pandas_utils functions.

[koaning]

After thinking about it, we probably want to have these three:

• log_shape
• log_names
• log_dtypes

Note too sure about log_names and log_dtypes.

[david26694]

Is this still open? I'd be happy to take it

[koaning]

It is still open, but let's first make sure we're clear on what will get implemented. Do you want to implement log_dtypes or also log_names?

Relevant: there's a related issue too #371.

[david26694]

I was thinking about both. I don't fully understand what is happening in #371, why is matplotlib generating warnings and what would the print step do?

[koaning]

The print step would do a very similar thing to what log_step would do but it would use print instead of using the python logging framework. The reason why matplotlib is generating warnings is because we've configured a logger to work in the notebook. This allows you to receive logs from log_step (this is awesome for production) but you might also get the logs from other libraries (which is not awesome in a notebook).

[koaning]

This got adressed.