Autocomplete & Typeahead

Also Known As

TL;DR

Explanation

Autocomplete strategies: prefix matching (startsWith), edge n-gram tokenisation (index every prefix), and completion suggesters (Elasticsearch's specialised structure). Performance: debounce input (wait 200-300ms after last keystroke before querying), cache recent queries, limit result count (5-10), and use a search engine not LIKE 'term%' SQL for large datasets. UX considerations: keyboard navigation, screen reader announcements (aria-live), and not autocompleting too aggressively.

Diagram

flowchart LR
    USER[User types ph] --> DEBOUNCE[Debounce 150ms]
    DEBOUNCE --> CACHE{Browser cache<br/>for this prefix?}
    CACHE -->|hit| SHOW[Show suggestions instantly]
    CACHE -->|miss| API[GET /autocomplete?q=ph]
    API --> TRIE2[Server-side Trie<br/>or search engine prefix]
    TRIE2 --> RANK[Rank by frequency<br/>personalisation]
    RANK --> RESP[Top 5-10 suggestions<br/>highlight matched prefix]
    RESP --> CACHE
    RESP --> SHOW
style SHOW fill:#238636,color:#fff
style RANK fill:#1f6feb,color:#fff
style DEBOUNCE fill:#d29922,color:#fff

Common Misconception

Why It Matters

Common Mistakes

• No debounce — a query fires on every keystroke, overwhelming the server for fast typists.
• Not caching autocomplete results — the same prefix is often typed repeatedly; cache aggressively.
• Showing too many suggestions — 10+ results require scrolling; 5-7 is the usability sweet spot.
• Not handling 'no results' gracefully — showing an empty dropdown is confusing; hide it or show a fallback.

Code Examples

// No debounce — fires on every keystroke:
document.getElementById('search').addEventListener('input', async (e) => {
    const results = await fetch('/api/suggest?q=' + e.target.value);
    // 'P' → 'PH' → 'PHP' → 3 requests in 100ms, all potentially in-flight
});

// Debounced + cached:
const cache = new Map();
let timer;
document.getElementById('search').addEventListener('input', (e) => {
    clearTimeout(timer);
    const q = e.target.value.trim();
    if (q.length < 2) return hideSuggestions();
    timer = setTimeout(async () => {
        if (cache.has(q)) return showSuggestions(cache.get(q));
        const results = await fetch('/api/suggest?q=' + encodeURIComponent(q)).then(r => r.json());
        cache.set(q, results); // Cache for session
        showSuggestions(results);
    }, 250); // 250ms debounce
});

Tags

🧠 Logic Engine

References

• https://www.elastic.co/guide/en/elasticsearch/reference/current/search-suggesters.html