detect Path objects in argument iterable and auto-use ns.P about natsort HOT 4 CLOSED

I am very reluctant to have natsort do any automatic modification of the algorithm settings based on the input data because I don't want to remove control from the user. Also, scanning the input data and doing type checking like this for a less common use case would undo most of the optimizations I have added in the most recent releases.

I am wondering, is there a reason that you aren't sure the type of the incoming data? If you know that you want to sort paths, you can just always add the ns.P modifier since non-path strings will still get sorted correctly, it will just eat up some runtime.

Would the following be acceptable alternatives?

>>> # Convert to string if the input is a Path object.
>>> natsorted(mylist, key=lambda x: str(x) if isinstance(x, pathlib.Path) else x, alg=ns.P)

or

>>> # Just convert everything to strings.
>>> natsorted(mylist, key=lambda x: str(x), alg=ns.P)

I might be open to adding an option that will do the checking and turning on of the path algorithm, but the user has to request this explicitly.

from natsort.

Oh, of course I know the type of the data and key argument works. At the end it's all about practicality vs. speed. What about using the first key argument (convert only Path instances to str) when ns.P is on?

from natsort.

I will look into this.

from natsort.

I have fixed this in a recent commit. I will release as an incremental update (version 3.5.2) in a few days.

from natsort.