fix(python): Selectors by_name and by_dtype should allow empty list as input

[29antonioac]

Closes #10855

Currently the selectors are being created with an unpacked list F.col(*all_names), but since F.col can manage iterables, by not unpacking we allow selecting nothing by inputing an empty list.

import polars as pl
import polars.selectors as cs

cols_to_process = [] # This could come from an input file

# This works now
df = pl.DataFrame({"a": [1,2,3]}) 
df_processed = df.with_columns(cs.by_name(cols_to_process) * 2)

Fixing this in my opinion gives more consistency on how some functions work: with_columns also works with empty lists for example.

[alexander-beedie]

Hmm... not quite sure about this; I can see why it would work for this case, but may need some other/related tweaks if we want this behaviour. I'll be able to review it properly tomorrow night though 👍

[29antonioac]

Thanks! Happy to try to improve this if you think this could be useful!

[29antonioac]

Hi, I just had another look at the code of the selectors and F.col and I can't see a reason why this could be a problem. Probably since I'm not used to this codebase I'm missing a lot of stuff so feel free to correct me 😄

Looks like the empty list is fully covered in F.col so I think it shouldn't be an issue. It looks every case is controlled in that function.

This could increase the consistency across other selectors too like datetime() (it uses a list to create the selector instead of an unpacked iterable), and consistency with the use of different ones as numeric() (if there's no numeric columns in the dataframe, it returns nothing). Any thoughts?

Thanks a lot!

[alexander-beedie]

Apologies, got swamped by a few things; will get this sorted out today tomorrow 👌

[stinodego]

I looked into this, and it's a good change 👍 there's no need to unpack the lists here. Thanks!