Add missing test cases for next()/previous()

[MetRonnie]

We were not testing the case of next() and previous() where the truncated properties were the same as the current datetime. E.g.

>>> now = 2020-10-19T13:53Z

>>> next(--10-19)
2020-10-19T00:00Z

>>> previous(--10-19)
2019-10-19T00:00Z

This behaviour does not seem to have been rigorously defined before? Adding these test cases ensures we do not break compatibility

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• No dependency changes
• Tests are included
• No changelog entry needed
• No docs needed
• If this is a bug fix, PR should be raised against the relevant ?.?.x branch.

[wxtim]

It strikes me that we also ought to test for

>>> now = 2020-10-19T13:53Z

>>> next(T-53)
2020-10-19T13:53Z

>>> previous(T-53)
2019-10-19T12:53Z

Although in the latter case I'm really not clear what I'd expect the user to expect the correct answer to be. I think it probably should be the previous hour.

[MetRonnie]

Hang on a minute, it strikes me that

>>> now = 2020-10-19T13:53Z
>>> next(--10-19)
2020-10-19T00:00Z

is a bug Something is wrong :( because it returns a datetime that is in the past!

It should either be

2021-10-19T00:00Z  # next year

or

2020-10-19T13:53Z  # same datetime

[wxtim]

is a bug because it returns a datetime that is in the past!

That's definitely a bug. 🐞 ← (a well spotted bug)

[oliver-sanders]

Yes, I would expect next to return a time in the future too.

[MetRonnie]

So the question Flag this as a question for the next Cylc project meeting. is, which of these should be the behaviour?

2021-10-19T00:00Z  # next year

or

2020-10-19T13:53Z  # same datetime

[oliver-sanders]

The one that's in the future ("next year") because just as T00 is short for T0000 and 01 is short for 01T0000, --10-19 is short for --10-19T0000.

I.E. matching date-time components should not be a special case.

[wxtim]

Nice use of xfail. Worth putting a comment with a link to #5795 in the code nearby?

[MetRonnie]

Worth putting a comment with a link to #5795 in the code nearby?

A curious dev can use git blame