Read a problem statement

Many wrong answers in interviews come from good code that solves the wrong problem. The statement is short, but almost every word in it is a rule.

This lesson gives you one way to read any statement, so you know what to build before you write a line of code.

Mark the words that matter

Here is a real problem from the A2Z sheet, Second Largest Element. Read it slowly, and underline every word that changes what you must build.

┌───────────────────────────────────────────────┐
│                                               │
│   Given nums, return the second largest       │
│                          ──────┬───────       │
│                                1              │
│   value. If it does not exist, return -1.     │
│   ──┬──        ──────┬───────  ────┬────      │
│     2                3             4          │
│   1 <= n <= 10^5   -10^4 <= nums[i] <= 10^4   │
│   ──────┬───────   ───────────┬────────────   │
│         5                     6               │
│   nums may contain duplicates.                │
│   ─────────────┬─────────────                 │
│                7                              │
│                                               │
└───────────────────────────────────────────────┘

1. Second largest. The maximum is not enough. You need the next value below it.
2. Value. Return a number from the array, not its position.
3. Does not exist. The answer can be missing. Find out when.
4. Return -1. The value that means "no answer".
5. n can be 1. One item has no second largest. And n up to 10⁵ means O(n log n) or faster.
6. Negative values. Any trick that assumes positive numbers will break.
7. Duplicates. In [8, 8, 7] the answer is 7, not 8.

Marks 3, 5 and 7 together answer "when does it not exist?": when the array has one item, or every item is the same.

The six-part read

Every statement, short or long, has the same six parts. Find each one before you think about a solution.

Then write the result down in your own words. In this course it becomes a problem card:

The "it asks" line is the test. If you cannot say the problem in one plain sentence, read it again. Do not start coding yet.

Make your own examples

The given examples show the normal case. Bugs hide in the cases they skip. Write 4 to 6 small inputs of your own, each one chosen for a reason.

Write the expected answer before you write code. Then the code has something to be checked against.

Dry run before you code

A common first idea: keep largest and second, and start second at -1, the "no answer" value. Walk the all-negative example through it by hand.

The start value -1 was bigger than every real value, so nothing could ever replace it. Mark 6 in the statement warned about this. Your own example caught it, before any code ran.

The fixed version:

JavaPythonTypeScript

int secondLargest(int[] nums) {
    int largest = Integer.MIN_VALUE; // the biggest value so far
    int second = Integer.MIN_VALUE; // the biggest value below largest
    for (int x : nums) {
        if (x > largest) {
            second = largest;
            largest = x;
        } else if (x < largest && x > second) {
            second = x;
        }
    }
    return second == Integer.MIN_VALUE ? -1 : second;
}

• A new maximum pushes the old one down to second.
• x < largest skips copies of the maximum, so [8, 8, 7] gives 7.
• If second never moved, no second value exists, so return -1.
• Both start below every real value: -Infinity in TypeScript, float("-inf") in Python. Java's int has no infinity, so it uses Integer.MIN_VALUE. That is safe here, because no value goes below -10⁴.

It returns the expected answer for all five inputs in the table. One pass, so O(n) time and O(1) extra space.

Edge cases and questions

The edge-case checklist

Run down this list for every problem. Most bugs in interviews come from one of these rows.

Ask before you code

In an interview, the statement is often vaguer than on a sheet. Ask, and say your assumption out loud.

• Size: can the input be empty? How large can n get?
• Values: can they be negative? Can they repeat?
• Output: a value or an index? In what order?
• No answer: what should I return? Here, -1 can also be a real answer, as in [5, -1]. Is that fine?
• Meaning: in [8, 8, 7], is the second largest 8 or 7?
• Input: may I change or sort the array in place? If not, work on a copy.

Each answer removes a guess. Interviewers notice when you ask.

Practice the read: Two Sum

Try the six-part read on the most asked interview problem. Fill in the table yourself first, then compare.

Two words in that table change the solution:

• Indices. Sorting the array loses the original positions. A sort-based idea must remember them.
• Not twice. With [3, 2, 4] and target 6, 3 + 3 is not allowed. The answer is [1, 2].

You meet Two Sum again in the Moderate part. You already know its traps.

Words that point to a pattern

Some words in a statement hint at the pattern before you think. Learn these pairs; every later lesson adds more.

These are first guesses, not rules. The limits are the other half of the signal, and they are the next lesson's topic. The pattern recognition chapter puts both together.

take U forwardYouTube

Read more

• Second Largest Element on takeUforward, the full statement to practise on
• Coding interview techniques from the Tech Interview Handbook, on finding an approach and then a faster one
• Coding interview cheatsheet, with what to clarify before you code and what to test after

Next: constraints decide the approach. The largest n tells you which Big O can pass, before you think of an idea.