Schedule Generation Performance Optimization

[jack-of-some-trades]

Small, but notable, performance increase to the _tryholidays() function.

In testing I found that generating market calendars and date ranges comprised a significant portion of what my code was spending time doing. While the majority of the time spent is in panda's date_range function I did find that the _tryholidays function was abnormally slow.

It turns out that some calendars, such as NYSE, use Holiday Calendars for Special opens/closes that behave more like Ad-hoc Days as than Holidays. This is a format that Exchange Calendars follows as well. Currently when making a schedule, these dates are retrieved through Pandas' Holiday Calendar System which is designed around reoccurring holidays making the date retrieval far slower than it needs to be.

I added a check that manually retrieves all the single occurrence holiday dates if it can, and defers to Pandas' Holiday Calendar system if it needs to. This is a pretty niche optimization since it is most noticeable when generating a lot of schedules, but since I found it I figured I'd share it.

Before the Optimization:

After the Optimization:

Code Executed during profiling:
import pandas_market_calendars as mcal
from timeit import timeit
NYSE = mcal.get_calendar("NYSE")
special_dates = lambda: NYSE.special_dates("market_open", "2025-01-01", "2025-02-01")
print(f"{timeit(special_dates, number=6000)/6000 = }")

[rsheftel]

Great improvement. thank you