DEPR: _timelex is deprecated in dateutil

[gfyoung]

See: #21137

The question is: how exactly do we replicate the functionality?

[jreback]

cc @pganssle

[pganssle]

To be clear, _timelex is not deprecated in dateutil, it's private. If you need that functionality I recommend copy-pasting it or using @jbrockmendel's sleek new implementation (dateutil/dateutil#732). dateutil is under a permissive license so the only requirement is that you retain the copyright notice (which I can't waive, unfortunately, because I'm not the original author of _timelex). When some form of _timelex is eventually made public, you can switch to that.

[gfyoung]

@pganssle :

To be clear, _timelex is not deprecated in dateutil, it's private.

Ah, thank you for clarifying that. Admittedly, the message associated with the DeprecationWarning had a little bit of a cognitive dissonance upon first read.

When some form of _timelex is eventually made public, you can switch to that.

Do you have a timeline by any chance for making _timelex public? If it's not on the horizon, we could copy-paste as an immediate fix (@jreback , we do that with clipboard IINM).

[jreback]

@gfyoung I would just in-line the code from above

[jbrockmendel]

Do you have a timeline by any chance for making _timelex public?

There's some discussion in dateutil#729. Short answer: "no".

If it's not on the horizon, we could copy-paste as an immediate fix

Definitely should use the version in dateutil#732 as it is much simpler and more performant. I think even that can be pared down to a single call to re.findall.

[gfyoung]

@jbrockmendel : Thanks for clarification! Since you're working directly on this, you probably would know best as to when we should port over a copy of the implementation.