collective / icalendar / 31557281112

98%

push

github

Read a PERIOD written with dates as midnight date-times (#1645)

* Localize a date parsed with a TZID

Under pytz, a date with a TZID was read with the timezone's local mean
time offset instead of the offset in effect on that date.

AI model: Claude Opus 5 (claude-opus-5), used through Claude Code.
Prompt: implement the fix and tests for issue 1633, then investigate the
pytz-only test_equality.py failures that a new fixture exposed.
Output: (this commit)

* Read a date in a PERIOD as midnight

A PERIOD written with dates was accepted when a calendar was parsed and
then raised an AttributeError when it was written out again.

Fixes #1633

AI model: Claude Opus 5 (claude-opus-5), used through Claude Code.
Prompt: implement the fix and tests for issue 1633, converting date-only
periods rather than rejecting them, per the maintainers' decision on the
issue.
Output: (this commit)

* Take the timezone of the other half of a period

A period that mixed a date with a timezone aware date-time could not
compute its duration, so writing it out raised a TypeError.

AI model: Claude Opus 5 (claude-opus-5), used through Claude Code.
Prompt: keep the mixed date and date-time case in its own commit so that
it can be dropped on review.
Output: (this commit)

* Apply review wording and naming suggestions

Link the precise RFC section, spell datetime as one word, drop rst markup
from a test docstring, and reject date-time in the Vale vocabulary. Rename
the second news fragment so it links to the issue.

AI model: Claude Opus 5 (claude-opus-5), used through Claude Code.
Prompt: work through the review of PR 1645, starting with the mechanical
wording and naming suggestions.
Output: (this commit)

* Cover the gaps found in review

Add a converted start next to a named timezone, a comma separated period
list, and the end date that midnight puts before the start.

AI model: Claude Opus 5 (claude-opus-5), used through Claude Code.
Prompt: work through the review of PR 1645, t... (continued)

859 of 879 branches covered (97.72%)

Branch coverage included in aggregate %.

104 of 104 new or added lines in 4 files covered. (100.0%)

14096 of 14366 relevant lines covered (98.12%)

6.86 hits per line