[^1]: https://www.joelonsoftware.com/2007/01/21/the-big-picture/
currently, we store the "parent" recurring event as one document. then we store exception events as standalone one off calendar events that link back to the parent recurring event. unfortunately, we have to maintain two different types of exploded versions of recurring events (to handle notes on an instance).
eventually, we added a bunch of extra fields to the parent event like a pointer to the next upcoming instance of this event and denormalized full list of exceptions so that we can more easily handle application logic like sending out reminders, and letting you use arrows to navigate b/w instance s of recurring events.
so far with this approach of separating out the parents and children we've run into very few problems.
However, you would gonna need their sync-engine[0] on your own servers in order to avoid contacting Nylas servers. From what I understand, it's kind of a cloud API that makes some sense out of emailing protocols.
Calendar apps are deceptively hard. Very tricky UI problems alongside super complicated data transformations you need to do across a fragmented ecosystem of ICS/etc. Making something "just work" takes an incredible amount of focused time from an entire engineering team.
But here's a peek anyway if you're curious: https://www.dropbox.com/s/j1ry3qar45ozj7m/nylas-calendar.png...
PS: we are hiring ;) feel free to ping me directly