Duplicate Records Within a Time Window

Senior phone-screen stream-processing problem: given log records of `id, text, title, timestamp`, emit every record whose content repeats within a 60-second window, with an explicit follow-up to evict expired entries from the tracking structure so memory stays bounded by the window. Candidates often shorthand it as "the LRU question", but the real pattern is sliding-window deduplication. Requirements Given a stream of log records, each a line of the form id, text, title, timestamp, output every record whose content duplicates an earlier record within a 60-second window. Records arrive in timestamp order. A record is reported when a matching record exists no more than 60 seconds before it. Space optimization is an explicit follow-up: once a tracked record is more than 60 seconds older than the current record's timestamp it can never match again — remove it from the tracking structure so memory stays bounded by the window, not by the stream. Clarify before coding: What defines a duplicate — the text alone, or the (text, title) pair? The exact output format — full original lines vs record ids. Examples Input: id1, t1, t1, 2 id2, t2, t2, 4 id3, t1, t1, 50 id4, t1, t1, 80 Output: ['id3, t1, t1, 50', 'id4, t1, t1, 80'] id3 is reported because it repeats id1's content 48 seconds later; id4 is reported because it repeats id3's content 30 seconds later — inside the window relative to id3, even though it is 78 seconds after id1. Notes Easy to misread: candidates often shorthand this as "the LRU question", but the actual ask is sliding-window duplicate detection — the LRU-adjacent part is only the eviction of expired entries. Recognizing the pattern late costs significant time in the round. The space discussion is graded explicitly: a hashmap of content → most recent in-window timestamp plus eviction of expired entries keeps memory proportional to the 60-second window. Preparation Implement sliding-window duplicate detection with a hashmap keyed on content mapping to the most recent in-window timestamp, plus an eviction pass dropping entries older than the window; test duplicates that straddle the 60-second boundary. Rehearse the clarification script for stream-processing prompts — duplicate-key definition, output format, timestamp-ordering guarantees — before writing any code.