Making icalendar.from_ical independent of pytz and usable with all timezone implementations

[niccokunzmann]

At the present moment v5.0.0a2, I understand that icalendar components can be constructed as objects with any timezone implementation (pytz, zoneinfo, tz).

As pytz is more and more outdated,
I would like to be able to choose the timezone implementation that we use when we parse an ICS file.
At the moment, I believe that this is the step forward to make the icalendar step in up to the advances Python developers have made with the latest timezone implementations.

Proposal 1

Proposal for the interface:

Calendar/Event/Component
    .from_ical(..., timezone=pytz.tzinfo)
    .from_ical(..., timezone=tz.gettz)
    .from_ical(..., timezone=zoneinfo.ZoneInfo)

Any object returned from these calls should not use pytz but the timezone specified.

Proposal 2

As icalendar has a global state to parse timezones set to pytz, we could set an other global default:

icalendar.set_timezone_implementation(pytz.tzinfo)
icalendar.set_timezone_implementation(tz.gettz)
icalendar.set_timezone_implementation(zoneinfo.ZoneInfo)

The global default will be something like a lazy pytz (as proposed by #294) that imports pytz only if no other implementation is set.
Calling icalendar.set_timezone_implementation will set the default parameter from Proposal 1.

Implementation

As we use pytest to run the tests and @jacadzaca is already heavily refactoring the tests, see #400, we can use a parameter for parsing to pass the desired timezone implementation.
Parametrizing timezone implementations shows that pytz will stand out and always look a bit different.
The recurring-ical-events library shows that test parametrization can be used to run all tests for pytz and zoneinfo.

Datetime/Timezone implementations

These timezone implementations should be supported:

• pytz
• zoneinfo/backports.zoneinfo
• arrow Use arrow library #181
• dateutil
• custom

Related

• refactor test_issue_237 #418 (comment)
• test_fixed_issues restructure #400 (comment)
• Don't Require pytz #294
• Parametrizing a timezone implementation
• recurring-ical-events timezone implementation test parametrization
• Test Restructuring #374

---

Comments and ideas are very much appreciated on this as this is a deep change - feels like it to me. Also, I wrote this now without reading through all the related issues. Please comment.

[jacadzaca]

I like the first option better - if people would like to use pytz in one section (e.g. legacy code) and ZoneInfo in other (e.g. new code) they easily can.

[jacadzaca]

running grep 'import pytz' on the tests folder shows these tests still use it explicitly (in order of # of pytz usages):
test_unit_cal.py
test_recurrence.py
test_timezoned.py
test_unit_prop.py

[jacadzaca]

While refactoring test_timezone.py, I noticed that inside cal.py icalendar uses the pytz specific field _utc_transition_times, which seem to have an equivalent in ZoneInfo , but it's unaccesible e.g:

from zoneinfo import ZoneInfo

ZoneInfo('Europe/Warsaw')._trans_utc

throws an AttributeError: 'zoneinfo.ZoneInfo' object has no attribute '_trans_utc'

[niccokunzmann]

This is now an issue and in progress: #609 and #623.