How do I practice coding and algorithm questions?

Use PracHub's coding console to write, test, and debug your solutions in Python or JavaScript. View hints, test against sample inputs, and compare with official solutions.

What difficulty level is this coding question?

This is a medium difficulty Coding & Algorithms question, commonly asked during Onsite rounds at Google.

What role is this question designed for?

This question is commonly asked for Software Engineer candidates at Google during technical interviews.

Design autocomplete with Trie | Google Coding Question

Design autocomplete with Trie

Company: Google

Role: Software Engineer

Category: Coding & Algorithms

Difficulty: medium

Interview Round: Onsite

Design and implement an autocomplete service using a prefix tree (Trie). Support operations insert(word, weight), update(word, newWeight), delete(word), and topK(prefix, k) that returns the k highest-weight words sharing the prefix, breaking ties lexicographically. Assume lowercase English letters. Explain the data structures you maintain at each node to make topK efficient, analyze time and space complexity for each operation, and discuss trade-offs (e.g., storing heaps/lists at nodes vs. computing on demand). Dry-run your solution on: insert('cat', 5), insert('car', 3), insert('cart', 8), insert('dog', 2), topK('ca', 2), update('car', 9), topK('ca', 2), delete('cart'), topK('ca', 2).

Quick Answer: This question evaluates understanding of Trie-based string indexing, augmented per-node metadata for efficient prefix queries, and algorithmic analysis of insert, update, delete, and topK operations including their time and space complexity.

Design and implement an **autocomplete service** backed by a prefix tree (Trie). The service stores a dynamic dictionary of words, each with an integer **weight** (a popularity/frequency score), and supports four operations: - `insert(word, weight)` — add a word with the given weight (overwrite if it already exists). - `update(word, newWeight)` — change the weight of an existing word (no-op if absent). - `delete(word)` — remove a word from the dictionary (no-op if absent). - `topK(prefix, k)` — return the `k` highest-weight words sharing `prefix`, breaking ties **lexicographically** (smaller string first). If fewer than `k` words match, return all of them. An empty prefix matches every word. `topK` is the hot path — called far more often than mutations — so it is acceptable to do extra work on mutation to keep queries cheap. The classic design caches a bounded, sorted top list of `(weight, word)` at each node (ordered by `(-weight, word)`); a mutation re-derives only the caches on the root path of the affected word. **Executable form for this console.** Implement `solution(operations)`, a driver that replays a list of operations against your autocomplete service and returns the outputs of the `topK` calls in order. Each operation is one of: - `['insert', word, weight]` - `['update', word, weight]` - `['delete', word]` - `['topK', prefix, k]` → append the resulting list of words to the output. Return the list of `topK` results, one entry per `topK` operation. **Example.** For the sequence `insert('cat',5), insert('car',3), insert('cart',8), insert('dog',2), topK('ca',2), update('car',9), topK('ca',2), delete('cart'), topK('ca',2)` the output is `[['cart','cat'], ['car','cart'], ['car','cat']]`.

Constraints

Words consist of lowercase English letters a-z only (branching factor <= 26).
Weights are integers fitting in a machine word; ties broken by lexicographic order of the word.
topK is the hot path; mutations are comparatively rare.
Dictionary size: thousands to low millions of words, single-machine in-memory.
insert on an existing word overwrites its weight; update/delete on a missing word is a no-op.
topK with an empty prefix returns the global top-k; a prefix with no matches returns [].

Examples

Input: ([['insert', 'cat', 5], ['insert', 'car', 3], ['insert', 'cart', 8], ['insert', 'dog', 2], ['topK', 'ca', 2], ['update', 'car', 9], ['topK', 'ca', 2], ['delete', 'cart'], ['topK', 'ca', 2]],)

Expected Output: [['cart', 'cat'], ['car', 'cart'], ['car', 'cat']]

Explanation: The question's exact dry run. After the inserts, 'ca' subtree holds cart:8, cat:5, car:3, so topK('ca',2)=[cart,cat]. update('car',9) makes car the top, so topK=[car,cart]. delete('cart') leaves car:9, cat:5, so topK=[car,cat]. 'dog' lives in a different subtree and is never touched.

Input: ([['topK', 'ca', 3]],)

Expected Output: [[]]

Explanation: Querying a prefix on an empty dictionary: the descent hits a missing child immediately, so topK returns [].

Hints

A Trie answers 'which words share this prefix?' for free: after walking the prefix, the node you land on roots the subtree of all matching words. The open question is getting top-k by weight out of that subtree without scanning it on every query.
Propagate the best candidates up the spine: at each node cache a small sorted list of its subtree's top words keyed by (-weight, word). Then topK is just walk-to-prefix-node and read the cached list — O(P + k).
A weight change or deletion can only affect the cached lists of ancestors on the path from the word's terminal node to the root. Re-derive each node's cache bottom-up along that path from its own word plus its children's caches; every node off the path stays valid.
Include the node's OWN terminal word when rebuilding, not just children — an internal node can also be a complete word (e.g. 'car' is a prefix of 'cart').

Quick Overview

This question evaluates understanding of Trie-based string indexing, augmented per-node metadata for efficient prefix queries, and algorithmic analysis of insert, update, delete, and topK operations including their time and space complexity.