حل آموزشی تمرین‌های فصل 3Chapter 3 exercise solutions, explained

11 تمرین ۱ — مدل‌سازی کاتالوگ فروشگاهExercise 1 — modelling a shop catalog

بازگشت به صورت تمرینBack to the exercise · ارزیابی و سطحAssessment and level

پاسخ ۶۰ثانیه‌ایThe 60-second answer

صورت مسئله و فرض‌هاProblem and assumptions

چرا این روش و این عددها؟Why this method and these numbers?

مدل ذهنیMental model

حل گام‌به‌گامStep by step

طراحی مشخصConcrete design

مثال محاسبه‌شدهWorked example

دانلود نمودار مستقلDownload standalone diagram

دانلود SVG مستقلDownload standalone SVG

گزینه‌های جایگزینAlternatives

سناریوهای خرابیFailure scenarios

کد قابل اجراRunnable code

اجراشده و تأییدشده در همین محیطExecuted and verified in this environment

این کد آموزشی و ساده‌سازی‌شده است؛ فقط برای آشنایی با ایده. کد واقعی پروداکشن ملاحظات بسیار بیشتری دارد و هیچ‌کس این‌شکلی کد پروداکشن نمی‌زند.Simplified teaching code for building intuition only — not production code; nobody ships production code like this.

import java.util.*;

// Teaching demo: catalog over in-memory rows — keyset pagination + UNIQUE guard.
// Mirrors the relational schema: brands, products, product_categories, reviews
// with UNIQUE(product_id, user_id). DDL lives in prose; this is runnable logic.
public class CatalogDemo {
    record Brand(int id, String name, String logoUrl, int logoV) {}
    record Product(int id, int brandId, String name, long price) {}
    record Review(int id, int productId, int userId, int stars) {}

    public static void main(String[] a) {
        List<Brand> brands = List.of(
            new Brand(1, "Acme", "acme-v3.png", 3),
            new Brand(2, "Globex", "globex-v1.png", 1));
        List<Product> products = List.of(
            new Product(1, 1, "Desk lamp A", 2900),
            new Product(2, 1, "Desk lamp B", 3400),
            new Product(3, 2, "Chair C", 9900),
            new Product(4, 2, "Floor lamp D", 4500));
        // product -> categories (join table rows)
        Map<Integer, List<String>> prodCats = Map.of(
            1, List.of("lamps", "desk"), 2, List.of("lamps"), 3, List.of("chairs"), 4, List.of("lamps"));
        List<Review> reviews = new ArrayList<>(List.of(
            new Review(1, 1, 101, 5), new Review(2, 1, 102, 4),
            new Review(3, 1, 103, 5), new Review(4, 2, 101, 3)));

        // 1) product page: one JOIN-equivalent (product + brand)
        Product p = products.get(0);
        Brand b = brands.stream().filter(x -> x.id() == p.brandId()).findFirst().orElseThrow();
        System.out.println("product page = " + p.name() + " / " + b.name()
            + " / cats=" + prodCats.get(p.id()));

        // 2) category page via keyset: WHERE (id > last) ORDER BY id LIMIT 2
        Map<String, List<Integer>> catIndex = new HashMap<>();
        for (Product x : products)
            for (String c : prodCats.getOrDefault(x.id(), List.of()))
                catIndex.computeIfAbsent(c, k -> new ArrayList<>()).add(x.id());
        catIndex.values().forEach(Collections::sort);
        List<Integer> page1 = keyset(catIndex.get("lamps"), 0, 2);
        List<Integer> page2 = keyset(catIndex.get("lamps"), page1.get(page1.size() - 1), 2);
        System.out.println("category lamps page1 = " + page1);
        System.out.println("category lamps page2 = " + page2);

        // 3) reviews of product 1, first two via keyset on (product_id, id)
        List<Review> r1 = reviews.stream()
            .filter(r -> r.productId() == 1).sorted(Comparator.comparingInt(Review::id))
            .limit(2).toList();
        System.out.println("reviews p1 first2 = " + r1.stream()
            .map(r -> "u" + r.userId() + ":" + r.stars()).toList());

        // 4) UNIQUE(product_id, user_id) guard: dup (1,101) rejected
        Set<String> uniq = new HashSet<>();
        for (Review r : reviews) uniq.add(r.productId() + ":" + r.userId());
        Review dup = new Review(9, 1, 101, 5);
        String key = dup.productId() + ":" + dup.userId();
        if (!uniq.add(key)) {
            System.out.println("UNIQUE holds: dup (1,101) rejected");
        } else {
            System.out.println("UNIQUE BROKEN: dup accepted");
        }

        // 5) brand change cost: 1 row vs rewriting every product doc
        System.out.println("brand rename cost = 1 row (relational) vs "
            + products.stream().filter(x -> x.brandId() == 1).count()
            + " doc rewrites (embedded)");
        System.out.println("OK");
    }

    static List<Integer> keyset(List<Integer> sorted, int last, int n) {
        List<Integer> out = new ArrayList<>();
        for (int id : sorted) if (id > last && out.size() < n) out.add(id);
        return out;
    }
}

آزمون صحت و معیار عملیاتیCorrectness checks and operational criteria

پاسخ سؤال‌های پیگیریFollow-up answers

اشتباهات رایجCommon mistakes

منابع:Sources: Amazon · SDE III interview preparation · Atlassian · engineering interview handbook

22 تمرین ۲ — تصمیم نرمال‌سازیExercise 2 — the normalization call

بازگشت به صورت تمرینBack to the exercise · ارزیابی و سطحAssessment and level

پاسخ ۶۰ثانیه‌ایThe 60-second answer

صورت مسئله و فرض‌هاProblem and assumptions

چرا این روش و این عددها؟Why this method and these numbers?

مدل ذهنیMental model

حل گام‌به‌گامStep by step

طراحی مشخصConcrete design

مثال محاسبه‌شدهWorked example

دانلود نمودار مستقلDownload standalone diagram

دانلود SVG مستقلDownload standalone SVG

گزینه‌های جایگزینAlternatives

سناریوهای خرابیFailure scenarios

کد قابل اجراRunnable code

اجراشده و تأییدشده در همین محیطExecuted and verified in this environment

این کد آموزشی و ساده‌سازی‌شده است؛ فقط برای آشنایی با ایده. کد واقعی پروداکشن ملاحظات بسیار بیشتری دارد و هیچ‌کس این‌شکلی کد پروداکشن نمی‌زند.Simplified teaching code for building intuition only — not production code; nobody ships production code like this.

import java.util.*;

// Teaching math: reference vs embedding vs derived read model for one brand
// logo shown on N products. Teaching numbers only (amozeshi), not benchmarks.
public class LogoMath {
    public static void main(String[] a) {
        double n = 100_000;      // products of one brand
        double rps = 800;        // product-page reads/s
        double lookupMs = 1.5;   // brand lookup cost
        double backfill = 2_000; // derived-view apply rate/s
        double budgetS = 5;      // acceptable freshness bound (teaching)

        double readTax = rps * lookupMs / 1000; // core-s per s
        double fanout = n;                       // sync-embed writes per change
        double mixedWindow = n / backfill;       // seconds of old/new mix
        long coverage = (long) (backfill * budgetS);

        System.out.println("read tax core-s/s (reference) = " + readTax);
        System.out.println("embed fan-out writes           = " + (long) fanout);
        System.out.println("mixed window s (embed)         = " + mixedWindow);
        System.out.println("lag-budget coverage (5 s)      = " + coverage);
        System.out.println("reference change cost          = 1 row");
        System.out.println("derived view                   = "
            + (long) fanout + " async versioned writes, resumable from cursor");

        String verdict;
        if (mixedWindow > budgetS * 4) {
            verdict = "VERDICT: sync embedding ruled out (50 s mix >> 5 s budget); "
                + "reference wins when freshness must be immediate, "
                + "versioned read model wins when 5 s lag is fine.";
        } else {
            verdict = "VERDICT: window fits the budget; embedding acceptable.";
        }
        System.out.println(verdict);
        System.out.println("OK");
    }
}

آزمون صحت و معیار عملیاتیCorrectness checks and operational criteria

پاسخ سؤال‌های پیگیریFollow-up answers

اشتباهات رایجCommon mistakes

منابع:Sources: Atlassian · engineering interview handbook · Debezium · outbox event router

33 تمرین ۳ — event sourcing برای رزرو سینماExercise 3 — event sourcing for cinema booking

بازگشت به صورت تمرینBack to the exercise · ارزیابی و سطحAssessment and level

پاسخ ۶۰ثانیه‌ایThe 60-second answer

صورت مسئله و فرض‌هاProblem and assumptions

چرا این روش و این عددها؟Why this method and these numbers?

مدل ذهنیMental model

حل گام‌به‌گامStep by step

طراحی مشخصConcrete design

مثال محاسبه‌شدهWorked example

دانلود نمودار مستقلDownload standalone diagram

دانلود نمودار مستقلDownload standalone diagram

دانلود SVG مستقلDownload standalone SVG

گزینه‌های جایگزینAlternatives

سناریوهای خرابیFailure scenarios

کد قابل اجراRunnable code

اجراشده و تأییدشده در همین محیطExecuted and verified in this environment

این کد آموزشی و ساده‌سازی‌شده است؛ فقط برای آشنایی با ایده. کد واقعی پروداکشن ملاحظات بسیار بیشتری دارد و هیچ‌کس این‌شکلی کد پروداکشن نمی‌زند.Simplified teaching code for building intuition only — not production code; nobody ships production code like this.

import java.util.*;
import java.util.concurrent.*;
import java.util.concurrent.atomic.*;

// Teaching demo: one seat, two concurrent conditional-append requests —
// exactly one wins — plus an idempotent resend guard for the ticket send.
// Conditional append = append iff seat_version == read version.
public class SeatRace {
    // seat -> versioned holder; version -1 = free
    static class SeatStore {
        final ConcurrentHashMap<String, AtomicLong> versions = new ConcurrentHashMap<>();
        final ConcurrentHashMap<String, String> holders = new ConcurrentHashMap<>();
        final Set<String> seenReq = ConcurrentHashMap.newKeySet();

        String hold(String seat, String req, String holder, long readV) {
            if (!seenReq.add(req)) return "dup-ignored";
            AtomicLong v = versions.computeIfAbsent(seat, k -> new AtomicLong(-1));
            if (v.compareAndSet(readV, readV + 1)) {
                holders.put(seat, holder);
                return "held";
            }
            return "rejected";
        }
    }

    public static void main(String[] a) throws Exception {
        SeatStore store = new SeatStore();
        CountDownLatch go = new CountDownLatch(1);
        ExecutorService pool = Executors.newFixedThreadPool(2);
        List<Future<String>> fs = new ArrayList<>();
        // Both racers read version -1 (free) before either appends.
        fs.add(pool.submit(() -> { go.await(); return store.hold("A5", "req-1", "u1", -1); }));
        fs.add(pool.submit(() -> { go.await(); return store.hold("A5", "req-2", "u2", -1); }));
        go.countDown();
        List<String> res = List.of(fs.get(0).get(), fs.get(1).get());
        pool.shutdown();
        long held = res.stream().filter("held"::equals).count();
        long rejected = res.stream().filter("rejected"::equals).count();
        System.out.println("race results      = " + res);
        System.out.println("exactly one held  = " + (held == 1 && rejected == 1));

        // Idempotent resend guard: retry with the same key changes nothing.
        System.out.println("resend req-1      = " + store.hold("A5", "req-1", "u1", -1));

        // Ticket send behind the event-id key: first sends, repeat blocked.
        Set<String> sent = ConcurrentHashMap.newKeySet();
        System.out.println("send e-77         = " + send(sent, "e-77", "A5"));
        System.out.println("send e-77 again   = " + send(sent, "e-77", "A5"));
        System.out.println(held == 1 && rejected == 1 ? "OK" : "RACE BROKEN");
    }

    static String send(Set<String> sent, String ev, String seat) {
        return sent.add(ev) ? "ticket:" + seat : "resend-blocked";
    }
}

آزمون صحت و معیار عملیاتیCorrectness checks and operational criteria

پاسخ سؤال‌های پیگیریFollow-up answers

اشتباهات رایجCommon mistakes

منابع:Sources: AWS · safe retries with idempotent APIs · Amazon · SDE III interview preparation