Hash-map patterns: count, group, index

After the morning walk, members tip their baskets onto one long table and the club sorts the finds.

At the sorting table	In the code
one tally mark per find on the species sheet	count: key to a number (step 2)
the card box filed by basket number, so "whose basket was this?" is one pull	index: key to an item (step 4)
the old label printer writes every tray label as text, so basket 7 and basket "7" share a tray	object keys are strings (step 7)
the shelf puts numbered trays first, lowest number first, then named trays in arrival order	objects reorder integer-like keys (step 8)
the new label printer prints each label exactly as written on the basket, number or word	a Map (step 9)

In real terms: a hash map turns "find everything with this key" from a walk along the list into one lookup, if the key says exactly what belongs together.

Breaks down when: a tray is a physical place, while a key is a value the code computes: two equal-looking keys can still differ (a number and a string in a Map).

Contrasting case: five finds on a plate. Sorting by eye is fine; the pattern earns its keep as the list grows.

4.1 ExamHall EXH-4250: answer sheets by filtering per attempt

ExamHall built each candidate's answer sheet by filtering the whole event list once per attempt:

const sheets = attemptIds.map((id) => events.filter((e) => e.attemptId === id));

At exam size, 50,000 events from 2,500 attempts, the job crawled (EXH-4250). MR !4550 replaced it with one pass that files each event under its attempt in a Map (step 3). Both were counted at two smaller sizes, 20 events per attempt:

4.1 EXH-4250 filter per attempt, 4,000 events, 200 attempts: 800,000 event reads
4.1 MR !4550 one pass with a Map, same input: 12,200 counted operations (4,000 event reads, 4,200 Map calls, 4,000 list writes); the same sheets
4.1 EXH-4250 filter per attempt, 8,000 events, 400 attempts: 3,200,000 event reads
4.1 MR !4550 one pass with a Map, same input: 24,400 counted operations (8,000 event reads, 8,400 Map calls, 8,000 list writes); the same sheets
4.1 doubling events and attempts: filter ×4.00 · one pass ×2.00; reads = events × attempts, so 50,000 × 2,500 = 125,000,000 (arithmetic, not run)

Filtering reads every event once per attempt, so doubling both quadruples the reads (dsa-big-o-js-costs step 4). The one pass reads each event once and makes one or two Map calls per event, so doubling the events doubles the count.

4.2 Ticketing: seat sections as object keys and as Map keys

A box office counts sales per section; the sections arrive as numbers and as strings:

4.2 sales, in order: 101, 'VIP', '101', 7
4.2 plain object: 3 keys · '7':1 '101':2 'VIP':1
4.2 Map:          4 keys · 101:1 'VIP':1 '101':1 7:1

The object merged 101 and '101' (step 7) and moved the integer-like keys to the front, smallest first (step 8); the Map kept four keys, each with the type it was set with (step 9).