FastPrepMessage Event Aggregation
Problems/Message Event Aggregation

Message Event Aggregation

OpenAI logoOpenAI● MediumFULLTIMENEW GRADPHONE SCREENONSITE INTERVIEW

Updated Nov 2026

QuestionFunctionExamplesConstraints

Problem statement

Source note: 🐣 Backed by a real interview report. This existing practice problem was cleaned up against the source on 07-20-2026. Parts 1–3 are source-confirmed: five-minute message aggregation using user_id, chat_id, and timestamp with bounded memory; react/end_chat active-chat tracking; and out-of-order timestamps. Part 4 was not available. Because the source did not specify a callable function signature, the event and query input encoding, sample cases, numeric bounds, the exact five-minute boundary, equal-timestamp ordering, or the returned user-summary format, these implementation details were added so the task can run and be judged on FastPrep. Match rate: approximately 80%.

This exercise combines the first three parts of a message-event aggregation task.

You are given a list of chat events and a list of query timestamps. Each event is encoded as a comma-separated string:

timestamp,userId,chatId,eventType

eventType is one of:

  • message: a message sent in the chat.
  • react: the user reacted in the chat, making the chat active for that user unless a later end_chat event ends it.
  • end_chat: the user ended the chat.

For each query time q, consider events in the inclusive five-minute window [q - 300, q]. Timestamps are integers measured in seconds.

For every user with at least one event in the window, compute:

  • messageCount: the number of message events for that user in the window.
  • activeChatCount: the number of distinct chats whose latest react or end_chat event in the window is react. A message event does not change a chat's active state.

Events may be given out of timestamp order. Process them by timestamp; when timestamps are equal, preserve their original input order.

Return one string for each query, in the original query order. Format a user's result as userId:messageCount:activeChatCount, sort users lexicographically by userId, and join entries with semicolons. If the window contains no events, return an empty string for that query.

Memory requirement

As the five-minute window advances, discard expired aggregation state. Apart from storage used to order the input and hold the output, retained counting and chat-state data must be proportional to the events and user-chat pairs in the current window, not to every chat seen earlier.

Function

aggregateMessageEvents(events: String[], queryTimes: int[]) → String[]

Examples

Example 1

events = ["10,u1,c1,message", "20,u1,c1,react", "80,u2,c5,message", "90,u2,c5,react", "200,u1,c1,end_chat", "260,u1,c2,react", "400,u1,c2,message"]queryTimes = [260, 500]return = ["u1:1:1;u2:1:1", "u1:1:1"]

At time 260, each user has one message in the window. Chat c2 is active for u1, while c5 is active for u2. At time 500, the window starts at 200; only u1 has events in the window, with one message and one active chat.

Example 2

events = ["300,u2,c8,end_chat", "40,u1,c1,react", "100,u1,c1,end_chat", "70,u1,c2,react", "20,u2,c8,react", "60,u1,c2,message", "110,u2,c8,message"]queryTimes = [120, 320]return = ["u1:1:1;u2:1:1", "u1:1:1;u2:1:0"]

The events are not given in timestamp order. At time 120, chat c8 is active for u2. At time 320, the later end_chat event makes it inactive. Chat c2 remains active for u1 at both query times.

Constraints

  • 0 <= events.length <= 10^5
  • 0 <= queryTimes.length <= 10^5
  • 0 <= timestamp, q <= 10^9
  • Every event has exactly four comma-separated fields and a valid eventType.
  • userId and chatId contain no commas, colons, or semicolons.

More OpenAI Onsite Interview questions

  • Sparse Plant Infection SimulationPHONE SCREEN · ONSITE INTERVIEW · Seen Nov 2026
  • Rectangle Point Coverage CountsPHONE SCREEN · ONSITE INTERVIEW · Seen Nov 2026
  • 15-Minute Chat Session TrackerPHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Segmented Durable Key-Value StorePHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Durable Work Queue OperationsPHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Versioned Social NetworkPHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Budgeted API Inference with Out-of-Order FeedbackONSITE INTERVIEW · Seen Oct 2026
  • Plant Infection Simulation, Part 4: Death CountdownPHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Count Friend CirclesPHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026
  • Incremental Hard Attention with a KV CachePHONE SCREEN · ONSITE INTERVIEW · Seen Oct 2026

Similar questions at other companies

  • Merchant Payout Scheduler, Part 2: Per-Merchant SchedulesStripe · PHONE SCREENStripe interview questions
  • Count Zero Request ServersSalesforce · ONSITE INTERVIEWSalesforce interview questions
  • In-Memory Filesystem CommandsPerplexity AI · PHONE SCREEN · ONSITE INTERVIEWPerplexity AI interview questions
  • Top K Best-Selling Products in the Last N DaysAmazon · ONSITE INTERVIEW
  • Pins Violation Log QueriesPinterest · PHONE SCREENPinterest interview questions
  • Camel CardsRippling · PHONE SCREEN · ONSITE INTERVIEWRippling interview questions
  • Shopping Cart ReceiptSony Interactive Entertainment PlayStation · PHONE SCREEN
  • Banking System with Top ActivityCapital One · ONSITE INTERVIEWCapital One interview questions
  • Linked Perception HistoryWaymo · ONSITE INTERVIEWWaymo interview questions
  • Banking System with Pending Transfer AcceptanceCircle · PHONE SCREENCircle interview questions
All OpenAI interview questionsSee OpenAI hiring insights
Saved
class Solution {
    public String[] aggregateMessageEvents(String[] events, int[] queryTimes) {
      // write your code here
    }
}
Checking account…