Preparing your workspaceloading the brief, studio & editor…

Loadingone moment…

Assign events to event-time windows · Python scenario · D8LooP

Reading is open to everyone. Running code and saving drafts need an account so your work is yours and comes back on your next visit.

CODE WORKSPACE

Assign events to event-time windows

The streaming aggregate counts events in five-minute windows. It currently buckets them by when the consumer read them, which means a mobile client that was offline for an hour has its events counted in the wrong window — and the daily totals still reconcile, so nobody has noticed. Write assign_windows(events, window_seconds, allowed_lateness). Each event carries an event_time and a processing_time.

Function to write

assign_windows(events: list[dict], window_seconds: int, allowed_lateness: int) -> dict

A dict with windows — each with start, end and sorted event ids — and late, the events that missed theirs.

How to approach it

Convert both timestamps to epoch seconds; the window start is then one modulo away.

Sample cases

+ 2 held back until you submit

a five-minute tumble

An event on the window boundary, one that is late but inside the grace period, one hours late, and one unparseable.

Input

Argument 1

id	event_time	processing_time
e1	2026-03-02T09:00:10Z	2026-03-02T09:00:12Z
e2	2026-03-02T09:04:59Z	2026-03-02T09:05:01Z
e3	2026-03-02T09:05:00Z	2026-03-02T09:05:02Z
e4	2026-03-02T09:01:00Z	2026-03-02T09:07:30Z
e5	2026-03-02T09:02:00Z	2026-03-02T11:00:00Z
e6	not a timestamp	2026-03-02T09:06:00Z

Argument 2

Argument 3

Returns

{
  'windows': [
    {
      'window_start': '2026-03-02T09:00:00Z',
      'window_end': '2026-03-02T09:05:00Z',
      'events': [
        'e1',
        'e2'
      ]
    },
    {
      'window_start': '2026-03-02T09:05:00Z',
      'window_end': '2026-03-02T09:10:00Z',
      'events': [
        'e3'
      ]
    }
  ],
  'late': [
    {
      'id': 'e4',
      'reason': 'too_late',
      'window_start': '2026-03-02T09:00:00Z'
    },
    {
      'id': 'e5',
      'reason': 'too_late',
      'window_start': '2026-03-02T09:00:00Z'
    },
    {
      'id': 'e6',
      'reason': 'unusable_timestamp',
      'window_start': None
    }
  ]
}

no events

An idle period produces no windows and no late events rather than an empty window.

Input

Argument 1

[] (empty list)

Argument 2

Argument 3

Returns

{
  'windows': [],
  'late': []
}

no grace at all

With zero lateness allowed, anything arriving after its window closed is diverted.

Input

Argument 1

id	event_time	processing_time
e1	2026-03-02T09:00:10Z	2026-03-02T09:00:12Z
e2	2026-03-02T09:04:59Z	2026-03-02T09:05:01Z
e3	2026-03-02T09:05:00Z	2026-03-02T09:05:02Z
e4	2026-03-02T09:01:00Z	2026-03-02T09:07:30Z
e5	2026-03-02T09:02:00Z	2026-03-02T11:00:00Z
e6	not a timestamp	2026-03-02T09:06:00Z

Argument 2

Argument 3

Returns

{
  'windows': [
    {
      'window_start': '2026-03-02T09:00:00Z',
      'window_end': '2026-03-02T09:05:00Z',
      'events': [
        'e1'
      ]
    },
    {
      'window_start': '2026-03-02T09:05:00Z',
      'window_end': '2026-03-02T09:10:00Z',
      'events': [
        'e3'
      ]
    }
  ],
  'late': [
    {
      'id': 'e2',
      'reason': 'too_late',
      'window_start': '2026-03-02T09:00:00Z'
    },
    {
      'id': 'e4',
      'reason': 'too_late',
      'window_start': '2026-03-02T09:00:00Z'
    },
    {
      'id': 'e5',
      'reason': 'too_late',
      'window_start': '2026-03-02T09:00:00Z'
    },
    {
      'id': 'e6',
      'reason': 'unusable_timestamp',
      'window_start': None
    }
  ]
}

Constraints

1Windows are tumbling and aligned to the epoch: an event belongs to the window starting at the largest multiple of window_seconds at or before its event time.
2Windows are half-open. An event exactly on a boundary belongs to the window that starts there, not the one that ends there.
3An event is too late when its processing time is more than allowed_lateness seconds after its window's end. Put it in late with the window it should have been in, rather than in that window.
4An event whose event time or processing time cannot be parsed goes in late with the reason unusable_timestamp and no window.
5Timestamps arrive with a Z, with an offset, or with no zone. Treat a zoneless timestamp as UTC and compare everything in UTC.
6Report the windows sorted by start, each with its start, its end, and its event ids sorted.
7Emit only windows that have events. An empty window is not a fact you have.
8A window of zero seconds or less describes nothing: return empty results.

Worked example

e2 happened at 09:04:59 and e3 at 09:05:00, one second apart. They land in different windows — the first in the one starting at 09:00, the second in the one starting at 09:05 — because the boundary belongs to the window that starts on it. Getting that backwards moves one event per boundary, forever, in a way that never shows up in a total. e4 happened at 09:01 and arrived at 09:07:30. Its window closed at 09:05, so it is two and a half minutes late — inside a two-minute grace? No: 09:07:30 is 150 seconds after 09:05, and the allowance is 120, so it is diverted. e5 happened at 09:02 and arrived at 11:00, nearly two hours late. Counting either of them in the window they arrived in is how the mobile clients' traffic ended up attributed to the moment their signal came back.

What this tests

The distinction between event time and processing time, which is the central idea in stream processing, and floor division on an aligned grid. The boundary rule and the lateness rule are both places where a defensible-looking choice is wrong.

PYTHONassign_windows(events: list[dict], window_seconds: int, allowed_lateness: int) -> dict

AI evaluation

The reviewer reads your query.

Submit for review to find out what your query gets right, what it gets wrong, and how it compares with the best working query for this exercise.

Desktop only

This scenario runs a full workspace — editor, canvas and results side by side. It needs a laptop or desktop to be usable. Open this page on a bigger screen to start building.

Read the library instead Back to modules