Add datetime and time pickers

[vidartf]

Resolves #2548

[vidartf]

Ready for review! :)

[pbugnion]

This looks great! I've played around with the interface. I've not looked at the code at all.

• In both the TimePicker and the DatetimePicker, the type of value can be None. It might be worth clarifying that in the docstring (right now, it suggests that these are always a valid datetime.time or datetime.datetime instance).
• In the TimePicker, the default value of step is 60 (at least on Chrome?). When you set "any", it defaults to 60? Is it worth specifying this? Also, you're allowed decimal values, which is not obvious. I think both these behaviours are sensible, but it may be worth clarifying this in the docstring description of the step parameter?
• In the TimePicker, the max value is inclusive, which is maybe counter-intuitive. Is it worth clarifying this in the docstring?
• In both docstrings, the examples still refer to ipydatetime.

In the DatetimePicker, there's some strange behaviour around zoned date times and offset-native datetimes. As far as I can tell (I haven't looked at the code), setting the value by playing around with the UI control sets a time-zone aware datetime (at least on Chrome). However, I can then set the value manually via an assignment to value, with an offset-native datetime.

The most straightforward way to exploit this discrepancy to cause an exception is:

1. Create and display a DatetimePicker

import ipywidgets as widgets
import datetime
dt = widgets.DatetimePicker()
dt

2. Set the control manually using the picker. dt.value is now offset-aware:

dt.value.tzname()  # => 'GMT' for me

3. Assign an offset-native datetime to min or max:

dt.max = datetime.datetime(2018, 1, 1)

gives:

~/oss/ipywidgets/ipywidgets/widgets/widget_datetime.py in _validate_max(self, proposal)
     82         if self.min and max < self.min:
     83             raise TraitError("setting max < min")
---> 84         if self.value and max < self.value:
     85             self.value = max
     86         return max

TypeError: can't compare offset-naive and offset-aware datetimes

It feels like the only sensible way to deal with this is to either only allow zoned datetimes or offset-native datetimes, but not a mix. From your entry in the documentation, it looks like you expect the user to select a zoned datetime, not an offset-native one. Maybe we could introduce a new traitlet, or allow customization of the current one?

[vidartf]

It feels like the only sensible way to deal with this is to either only allow zoned datetimes or offset-native datetimes, but not a mix.

Seems reasonable. I've also added a naive picker in ipydatetime, so I might just add validators to both of them to enforce it (should it coerce silently, or fail noisily?).

[vidartf]

@pbugnion I've added a separate picker for naive datetimes (NaiveDatetimePicker), and made DatetimePicker strictly timezone aware.

[vidartf]

Now green ✔️

[jasongrout]

From ipywidgets meeting today: we should have further discussion about serialization formats. CC @SylvainCorlay and @vidartf

[epifanio]

Hi, I was looking for a DatetimePicker in IPywidgets and I came across this :) what is the status? will it be implemented?

[vidartf]

It is currently available in the package ipydatetime. The hope/plan is to get this into core ipywidgets (since it is a "standard" browser control), but I need to update the PR with the latest changes from that package.

[davidbrochart]

@vidartf I rebased this PR on master and opened a PR in your repo: vidartf#1

[vidartf]

This should be ready now. @SylvainCorlay I didn't touch datetime serialization format. Let's discuss any changes in a followup issue/PR. For context: we were discussing using a different format for serialization, either Unix timestamp or Isoformat string (full version or limited to Python stdlib supported version).

[vidartf]

Don't merge until I've fixed the schemas becoming binary ⚠️

[jasongrout]

Don't merge until I've fixed the schemas becoming binary ⚠️

Marked as draft for now, then.

[vidartf]

Fixed 👍

[jasongrout]

@jtpio - is this good to merge?

[jtpio]

If anyone wants to try this branch on Binder:

https://mybinder.org/v2/gh/vidartf/ipywidgets/datetime?urlpath=lab/tree/docs/source/examples/

[jtpio]

Should we use a more descriptive name than update2 for this method?

[jtpio]

Looks like a couple of these headers here and below could be updated to match the existing ones.

[jtpio]

@vidartf mind rebasing on the latest master to pick up the update to TypeScript 4.3 to check everything compiles correctly?

[vidartf]

@jtpio Rebased, and all green ✔️

[jasongrout]

@jtpio, is this good to go in?

[jtpio]

I think so yes. In the dev meeting last week we had decided to wait until a couple of dependency update PRs were merged and a new alpha release cut. So it should be good to go now.

[jtpio]

Thanks!