URL Shortener Service: machine coding practice

URL Shortener Service

Scenario. Marketing wants short links for campaigns: some permanent, some that expire after a promo ends, and they want click counts. Build the core service (no HTTP layer here, just the domain logic).

What to build

Package shortener.

Base62 (static utility)

• Alphabet: 0-9, then A-Z, then a-z (so 0 → "0", 61 → "z", 62 → "10").
• String encode(long n) for n >= 0 (negative → IllegalArgumentException).
• long decode(String s) (empty, null, or invalid char → IllegalArgumentException).

UrlShortener(TimeSource time)

Method · Behaviour

String shorten(String longUrl, long ttlMillis) · Generate a code. ttlMillis <= 0 means never expires.

String shortenWithAlias(String longUrl, String alias, long ttlMillis) · Use a custom code.

Optional<String> resolve(String code) · Long URL if code exists and is not expired. Counts a hit.

LinkStats stats(String code) · Hits and last access time. Unknown code → NoSuchElementException.

Rules

• URL validation: must start with http:// or https://, have something after the scheme, no whitespace,

max 2048 chars → otherwise IllegalArgumentException.

• Generated codes are Base62, between 6 and 10 characters, and must never collide with each other or with

custom aliases. A counter-based scheme is fine. (Pad/offset it so codes are at least 6 chars.)

• Dedup: calling shorten with the same URL again returns the same code, as long as that earlier

generated link has not expired. (Aliases are never reused by dedup.)

• Aliases: must match [A-Za-z0-9_-]{3,32}, else IllegalArgumentException. Already taken (by an alias or a

generated code, even if expired) → IllegalStateException.

• Expiry: a link created at t with ttl d resolves while now < t + d and is expired at now >= t + d.

Expired links return Optional.empty() and do not count hits.

• Stats: hits counts successful resolves; lastAccessMillis is the time of the latest successful resolve,

or -1 if never accessed.

• Thread-safe: concurrent shorten calls with *different* URLs must get distinct codes; concurrent resolves must

not lose hits.

Interview talking points

• Counter + Base62 vs random codes vs hashing the URL: collision handling, predictability, sharding the counter.
• Storage schema: links(code PK, long_url, created_at, expires_at, is_alias), clicks as a separate counter table.
• 301 vs 302 redirects and what that does to your click analytics.