Marco Gorelli - Polars and time zones: everything you need to know | PyData Global 2023

When dealing with time zones, use specialized libraries like Polars rather than handling them manually due to complexities with DST, time zone changes, and cross-country differences

Datetime values have two key components: time unit (smallest representable time increment) and time zone

The difference between calendar duration (1d) and fixed duration (24h) is important - a calendar day can be 23, 24, or 25 hours due to DST changes

Store datetimes in UTC for consistency, but perform logic in local time zones when working with DST-affected data

Convert_timezone vs replace_timezone:

• Convert answers “what time is it now in another location?”
• Replace changes the zone while keeping the same local time

Polars handles time zones by deferring to the Rust ChronoTZ library, offering better post-2038 support compared to pandas’ PYTZ

When reading timezone-aware data from CSVs in Polars, it converts everything to UTC when strings contain UTC offsets

Time zone boundaries and day lengths vary by location due to DST transitions happening at different dates globally

Using PyArrow strings instead of regular strings improves performance when working with datetime data

For datetime interpolation and manipulation, Polars offers expressions and the .dt namespace with timezone-aware operations