When planning out my solution for this code test I made a few assumptions based upon how I have tackled a problem like this previously.
I've assumed that this rest application represents a service that would be run in a highly available manner, so for that reason I've designed the billing logic to be an endpoint specific to an individual invoice that would in a real scenario be triggered by an external scheduling tool (either a cron, a job that's been put on a queue, etc).
This is because if the server itself is controlling when these are all billed, it could open itself up to race conditions, and wouldn't be as fault tolerant. Due to this application handling payments, it needs the atomicity that could be provided from an external scheduler triggering it.
- Scheduling
- Do we consider timezones? Just run on the first of every month UTC?
- Would the scheduler take into account failures or just try to re-run that next time? etc?
- Billing Logic
- Do we create invoices for next month after successful billing?