Time Slot Generator from Weekly Working Hours

Phone-screen. Given a `start_date`, `end_date`, and a weekly-repeating `working_hours` config, generate every 30-minute slot inside the range, with strict boundary handling for the first and last day. Requirements Inputs: a start_date/end_date range and a working_hours table that repeats weekly (e.g. Mon 09:00-12:00, 13:00-17:00, Tue 09:00-12:00, ...). Output: every 30-minute slot that falls inside the range AND inside that day's working hours, sorted by start time. Boundary handling: compare the full datetime, not just the date, on the first and last day. A slot must never start before start_date and never end after end_date. Checking only the date would wrongly include slots before the start time or after the end time on the boundary days. The round requires the candidate to write their own test cases. The canonical OA shape: def generate_time_slots( start_date: str, # "YYYY-MM-DD HH:MM" end_date: str, # "YYYY-MM-DD HH:MM" working_hours: list[dict], ) -> list[tuple[str, str]]: ... # Each working_hours row: {"day_of_week": int, "start_time": "HH:MM", "end_time": "HH:MM"} # day_of_week: 0 = Monday, 1 = Tuesday, ..., 6 = Sunday # start_time inclusive; end_time exclusive, but a slot ending exactly at end_time is valid. # Returns 30-minute slots as ("YYYY-MM-DD HH:MM", "YYYY-MM-DD HH:MM") tuples, sorted by start. # All datetimes share one timezone (no tz/DST in the base ask). Slot rules Each slot is exactly 30 minutes long, generated in 30-minute increments counted from the working interval's start_time. A slot must be fully contained inside a working-hours interval for that weekday: slot_start >= work_start and slot_end <= work_end. A slot start must be >= start_date; a slot end must be <= end_date. A slot ending exactly at end_date is included. If start_date >= end_date, return an empty list. Overlapping/touching working intervals on the same weekday should be merged so duplicate slots are not emitted. Examples start_date = "2026-03-02 10:15" (Mon), end_date = "2026-03-04 11:45" (Wed), with working hours Mon 09:00-12:00, Tue 13:00-14:00, Wed 10:00-12:00: [ ("2026-03-02 10:30", "2026-03-02 11:00"), ("2026-03-02 11:00", "2026-03-02 11:30"), ("2026-03-02 11:30", "2026-03-02 12:00"), ("2026-03-03 13:00", "2026-03-03 13:30"), ("2026-03-03 13:30", "2026-03-03 14:00"), ("2026-03-04 10:00", "2026-03-04 10:30"), ("2026-03-04 10:30", "2026-03-04 11:00"), ("2026-03-04 11:00", "2026-03-04 11:30"), ] The Monday 10:00->10:30 slot is dropped (starts before start_date); the Wednesday 11:30->12:00 slot is dropped (ends after end_date). Notes The source report is light on Parts 2/3 detail — assume the interviewer will follow up on time-zone handling, DST, or recurring exceptions. Off-by-one on the last slot (slot ends exactly at end-of-day) is the most likely source of bugs. Boundary edge cases to test Start mid-slot: start_date = "...09:01", interval 09:00-10:00 → the 09:00->09:30 slot is excluded (starts before start_date); first emitted slot is 09:30->10:00. End mid-slot: end_date = "...09:59", interval 09:00-10:00 → the 09:30->10:00 slot is excluded (ends after end_date); only 09:00->09:30 is emitted. No matching weekday in range: range falls only on a weekday with no working_hours row → return []. Input invariants to validate When normalizing the working_hours table, enforce these per-row before generating slots: day_of_week must be in 0..6; otherwise raise ValueError. Within a single row, start_time must be strictly before end_time; otherwise raise ValueError. Interval merge for overlapping/touching same-weekday rows: sort the day's intervals by start, then fold left — open a new merged interval when start_time > merged[-1].end, else extend merged[-1].end = max(merged[-1].end, end_time). Using > (not >=) treats touching intervals (...12:00 then 12:00...) as one, so the boundary slot is not emitted twice. Suggested approach Parse both bounds to datetime; group working_hours by day_of_week; iterate calendar days from start_date.date() through end_date.date(). For each matching interval, build concrete datetimes for that date, advance a cursor in 30-minute steps to the first boundary >= max(work_start, start_date), then emit while cursor + 30min <= min(work_end, end_date). Jump the cursor in O(1) instead of stepping: boundaries are counted from work_start, so the first valid start is work_start + ceil((lower_bound - work_start) / 30min) * 30min where lower_bound = max(work_start, start_date) (return work_start directly when lower_bound <= work_start). This avoids a per-slot loop just to skip past start_date on the first day. Complexity O(D * H + S) time, O(S) space — D calendar days, H intervals/day, S emitted slots. Preparation Drill timezone-aware datetime arithmetic in your language of choice (zoneinfo in Python, java.time in Java). Pre-write a generator that walks a date range day by day and yields slots inside per-day working-hour intervals. Have a test-case template ready: same-day range, multi-day range, range that starts mid-slot, range that ends mid-slot.