Review iso8601.py code comments (and code?)

[MetRonnie]

There are several cryptic comments such as

cylc-flow/cylc/flow/cycling/iso8601.py

Lines 669 to 677 in 5a19433

	# correct for year in 'now' if year only,
	# or year and time, specified in input
	if re.search(r"\(-\d{2}[);T]", value):
	now.year += 1
	# correct for month in 'now' if year and month only,
	# or year, month and time, specified in input
	elif re.search(r"\(-\d{4}[);T]", value):
	now.month_of_year += 1

cylc-flow/cylc/flow/cycling/iso8601.py

Lines 771 to 782 in 5a19433

	# ensure month and day from 'now' are not included
	# where they did not appear in the truncated datetime
	# NOTE: this may break when the order of tick over
	# for time point is reversed!!!
	# https://github.com/metomi/isodatetime/pull/101
	# case 1 - year only
	if re.search(r"\(-\d{2}[);T]", value):
	cycle_point.month_of_year = 1
	cycle_point.day_of_month = 1
	# case 2 - month only or year and month
	elif re.search(r"\(-(-\d{2}|\d{4})[;T)]", value):
	cycle_point.day_of_month = 1

In the latter, unfortunately the linked PR metomi/isodatetime#101 is not very illuminating to me! (And the discussion there appears as if it might be suffering from confusion over the "addition of truncated TimePoint" bug metomi/isodatetime#80, to boot)

Also, is it just me or are the regexes in the code alarmingly simple? We should see if we can find faults in them.

[hjoliver]

Maybe the authors of those comments will remember. The first were by @lhuggett ; the second by @oliver-sanders .

[lhuggett]

So I think what the first one was about was that if you include the year (and/or month) from the real "now" when you're not looking at a day as well, you end up with a false earliness, so you move it (them) forward one to make sure the next/previous works correctly. I do recall that I only found it by trial and error.

For the second one, I'm afraid I don't remember what the issue was about the potential breakage from the reversing of tickover order, only that it was something that we thought could be a problem but hadn't actually been demonstrated as one at that point.

[MetRonnie]

Thanks for the response, looks like the first one comes down to a possible bug in isodatetime:

>>> now = TimePointParser().parse("2010-08-08T15:40Z")
>>> point = TimePointParser(allow_truncated=True).parse("-10")
>>> now + point
<metomi.isodatetime.data.TimePoint: 2010-08-08T15:40:00Z>
>>> now == now + point
True

I think now + point ought to equal 2110-08-08T15:40:00Z. However this might be debatable

Possibly related to metomi/isodatetime#80

[MetRonnie]

For the second, I think I have figured it out: metomi/isodatetime#212