# Schedule Weekly Deployment Windows The source reports a weekly deployment-window scheduler. The absolute-time horizon, cyclic `start == end` rule, and cross-week merge behavior below are deterministic practice clarifications where the report does not supply an exact contract. Implement a scheduler over a repeating UTC week. Let `W = 10080` minutes. All intervals are half-open. ```python def schedule_deployments(part: str, input_csv: list[str]) -> list[list[int]]: ... ``` `part` is exactly `"1"` or `"2"`. Input rows are syntactically valid CSV with base-10 integers. ## Part 1: Allowed Time Minus Freeze Time Every row is: ```text start,end,type ``` `type` is `allowed` or `freeze`, and `0 <= start < end <= W`. Compute the set covered by at least one allowed interval and no freeze interval. Return its maximal intervals as `[[start, end], ...]`, sorted by start. Merge touching output intervals. Freeze always wins. Example: ```text 540,600,allowed 570,585,freeze ``` returns `[[540, 570], [585, 600]]`. No allowed rows returns `[]`. `[0, W)` represents a full-week source interval. ## Part 2: Absolute Time and a Repeating Local Schedule The first row is: ```text utc_now,lead_time_minutes,min_continuous_minutes,k ``` `utc_now` is a nonnegative absolute UTC minute on an unbounded timeline, not a minute-of-week value. The other three values are nonnegative integers. Define: ```text earliest = utc_now + lead_time_minutes horizon = [earliest, earliest + W) ``` Search exactly that one-week horizon. Every remaining row is: ```text start,end,type,timezone_offset_minutes ``` Local endpoints lie in `[0, W)`. Each row denotes a cyclic weekly interval: - If `start == end`, it denotes the full local week. - Otherwise its duration is `(end - start) mod W`, in `1..W-1`, walking forward from `start` to `end`. The offset means `UTC = local - timezone_offset_minutes`. For a non-full interval, compute: ```text utc_start = (start - timezone_offset_minutes) mod W utc_end_unwrapped = utc_start + duration ``` Split at multiples of `W` to obtain canonical UTC-week pieces. A full-week local interval remains a full-week UTC interval. Union allowed pieces, union freeze pieces, and subtract freeze from allowed. Treat the resulting UTC schedule as repeating every `W` minutes. Expand enough neighboring occurrences to cover `horizon`, intersect with the horizon, and merge touching absolute intervals. In particular, canonical pieces ending at `W` and beginning at `0` are one continuous interval across a week boundary when their repeated occurrences touch. Discard empty intervals and any interval whose clipped length is less than `min_continuous_minutes`. Return the first at most `k` maximal qualifying intervals in chronological order, using absolute unwrapped endpoints. If `k == 0`, return `[]`. If the deployable schedule covers the full week continuously, the only horizon result is `[earliest, earliest + W]`, provided `k > 0` and `W >= min_continuous_minutes`. A full-week freeze removes all output. ## Constraints and Edge Cases - Part 2 offsets may be any integer; normalization is modulo `W`. - When `min_continuous_minutes == 0`, every nonempty clipped interval qualifies. - A lead time may cross any number of week boundaries because `utc_now` is absolute. - Output is maximal only within the requested horizon; clipping at either horizon boundary is expected. - Do not iterate minute by minute. Base running time on sorting, union, subtraction, and the number of interval endpoints. Test local and UTC wrap, `start == end`, lead times beyond one week, a cut through a window, cross-week adjacency, touching allowed and freeze endpoints, full-week allowed and freeze rows, `k == 0`, and an empty allowed set.