Practice · Code Quality and Working with AI Assistants

Q1

Which variable name is best for the number of days left before an assignment is due?

A daysUntilDeadline B theNumberOfDaysThatRemainBeforeTheAssignmentIsDueForThisStudent C d D days2

Q2

Which comment adds the most value?

A // TODO fix later B // TAs see real names so they can handle misconduct cases (policy MOD-4) C i++; // add one to i D // this is a method

Q3

if (secondsSinceLogin > 86400) logout(); — which smell is this, and what is the fix?

A Dead code; delete it B Magic number; replace with a named constant such as SESSION_LIMIT_SECONDS = 24 * 60 * 60 C Feature envy; move method D Long method; extract method

Q4

The same 15-line validation of a post's text appears in three controllers. When the maximum length changes, a developer updates two of them and forgets the third. Which smell caused the bug?

A Speculative generality B Primitive obsession C Long parameter list D Duplicated code

Q5

A method createMatch(String name1, String email1, String day1, String start1, String end1, String name2, …) has 10 parameters. Which refactoring fits best?

A Introduce parameter objects (e.g., Student, TimeSlot) B Add more comments C Make all parameters global variables D Inline method

Q6

Adding a new notification channel requires editing the same switch (channel) statement in Notifier, Settings, Stats and AdminPage. This smell is called…

A Shotgun surgery / switch on type — fix by replacing the conditional with polymorphism (one class per channel) B Good design C Dead code D Large class

Q7

Which statement correctly defines refactoring?

A Adding new features quickly B Changing the internal structure of code to make it easier to understand and modify, without changing its observable behaviour C Rewriting a module from scratch with a new design D Fixing bugs reported by users

Q8

What should you have before refactoring a method?

A Permission to skip code review B Tests that cover the behaviour you are going to touch C A longer deadline D A new framework

Q9

Why should refactoring and new features not be mixed in the same commit?

A Refactoring is only allowed on Fridays. B Git does not allow it. C If a test fails, you cannot tell whether the restructuring or the new behaviour caused it; small separate steps are easy to verify and undo. D New features never need tests.

Q10

What is the cyclomatic complexity of this method?

int level(int score, boolean late) {
    if (score >= 90 && !late) {
        return 3;
    } else if (score >= 60) {
        return 2;
    }
    return late ? 0 : 1;
}

A 5 B 4 C 6 D 3

Q11

Why is a method with cyclomatic complexity 25 a problem?

A It has many independent paths, so it is hard to understand and needs many test cases to cover; bugs tend to cluster there. B Java forbids complexity above 20. C It compiles more slowly. D It uses too much memory.

Q12

Which defect can static analysis not detect?

A An unclosed database connection B A possible null dereference C A method that computes the late penalty as 5% per day when the syllabus says 10% per day D Unused private methods

Q13

What is wrong with this code?

try {
    repository.save(post);
} catch (Exception e) {
}

A Nothing; it prevents crashes. B It should catch Throwable instead. C It should print "error" to the console and continue. D The exception is swallowed: the save silently fails, nobody is told, and the bug is very hard to find.

Q14

Which Java construct guarantees that a database connection is closed even when an exception is thrown?

A Calling System.gc() B A comment saying "close this later" C A while (true) loop D try-with-resources: try (Connection c = dataSource.getConnection()) { ... }

Q15

Which prompt works most like a specification?

A "Make it fast and clean." B "Write a matching function." C "Write the best possible matching algorithm using modern techniques." D "Java 8, no new libraries. Write matchPartners(Student me, List<Student> others): return names of students sharing ≥ 1 course whose free slots overlap mine by ≥ 30 min; exclude me; sort by overlap desc, then name; touching slots do not count; list your assumptions; do not change Student."

Q16

An AI assistant adds the dependency com.studyhelp:json-fastparse:2.1.0 to the build file. Nobody on the team has heard of it. What is the best action?

A Rename it to something shorter. B Verify that the package really exists, who publishes it, its popularity, licence and security record — or use a well-known library already in the project. Invented package names can be registered by attackers ("slopsquatting"). C Add it to all projects. D Keep it; the AI must know it.

Q17

StudyBuddy targets Java 8. An assistant generates code using var and List.of(...). What kind of failure is this?

A Wrong-version code: features from later Java versions that will not compile for the project's target B A performance problem C A security vulnerability D A hallucinated package

Q18

An AI generated a method and its unit tests. All tests pass. Why is this weak evidence of correctness?

A Unit tests are never useful. B AI cannot write Java tests. C Tests generated from the code often assert what the code currently does — including its bugs — instead of what the requirement says. D Passing tests always prove correctness.

Q19

An assistant "refactors" the late-penalty method and also changes Math.round to Math.floor. All other lines are just renamed. What is the risk?

A A silent behaviour change hidden inside a refactoring — scores now round differently, which tests should catch but a quick visual review may miss. B The code becomes more secure. C None; renaming is safe. D The method will not compile.

Q20

An AI-generated DAO method builds SQL as "SELECT * FROM posts WHERE author = '" + name + "'". What is the main problem?

A It uses too many quotes. B It is too slow. C It is vulnerable to SQL injection; use a PreparedStatement with parameters. D Nothing; it works for normal names.

Q21

A coding agent was asked to make the failing test testLatePenalty pass. Its pull request changes the test's expected value from 70 to 80. What should the reviewer do?

A Ask the agent to add more tests with value 80. B Approve; the test passes now. C Delete the test. D Reject: the agent weakened the test instead of fixing the code; check the requirement to decide which value is correct.

Q22

Which is the safest way to run a coding agent on StudyBuddy?

A On a teammate's laptop while they are away B On the production server with the database password, so it can test with real data C In a sandbox copy of the repository with limited permissions, test data only, and its changes going through normal PR review and CI D With automatic merge to main if it says the task is done

Q23

Industry analyses of large code bases (such as GitClear's 2024 study) reported what trend as AI coding assistance spread?

A The disappearance of code review B More copy-pasted code and more code rewritten within a short time after being written C No measurable change in any metric D Less duplicated code and less churn

Q24

Which is the best practice when a method is hard to understand because of deep nesting?

A Add a comment at every level explaining the nesting B Increase the indentation width to 8 spaces C Put the whole method on one line D Use guard clauses (handle special cases first and return) and extract nested blocks into well-named methods

Q25

Which item belongs on a review checklist specifically because code is AI-generated?

A Count the number of lines. B Check that the code was written quickly. C Check that every called method and every new dependency really exists in the version we use. D Check that the author used their favourite editor.

Q26 Short answer

Refactor the following method: remove the smells, keep the behaviour, and name each refactoring you applied.

double calc(int d, double s) {
    if (d > 0) {
        if (d <= 7) {
            return s - s * 0.1 * d;
        } else {
            return 0;
        }
    } else {
        return s;
    }
}

(d = days late, s = score; 10% penalty per day, zero after 7 days.)

Q27 Short answer

List six failure modes of AI-generated code and, for each, one concrete check that would catch it during review.

Q28 Short answer

Compute the cyclomatic complexity of this method and propose how to reduce it.

String badge(Student s) {
    if (s == null) return "none";
    int n = 0;
    for (Post p : s.posts()) {
        if (p.isEndorsed() || p.votes() > 10) n++;
    }
    if (n >= 20 && s.isActive()) return "gold";
    else if (n >= 10) return "silver";
    else if (n >= 3) return "bronze";
    return "none";
}

Q29 Short answer

Your teammate says: "The AI wrote the tests too, and they all pass, so the pull request is ready." Write a short, constructive review comment (4–6 sentences) explaining what else is needed.

Q30 Short answer

Explain what "slopsquatting" is and describe three practices that protect StudyBuddy against it.