TTL Key/Value Store

Implement an in-memory Store with set, get, delete, and scan. A time-to-live (TTL) deadline makes a value disappear at an exact injected-clock boundary.

Requirements

• Store(now) receives a zero-argument clock function.
• set(key, value, ttl=None) stores a value. ttl is measured in seconds.
• Keys, prefixes, and values are strings. Empty strings are valid; stored None / Java null values are outside scope.
• Each operation samples the clock once. A TTL deadline is that sample plus ttl; a scan checks every key against its one sample.
• A zero or negative TTL makes the new value immediately expired, including when it replaces a live value.
• Overwriting a key replaces both its value and expiry; ttl=None clears any prior expiry.
• get(key) returns the value or None when missing or expired.
• delete(key) removes a key and returns True if it existed and was not expired.
• scan(prefix) returns sorted (key, value) pairs for non-expired keys whose key starts with prefix.
• Expiry is inclusive: a key is expired when now() >= expires_at.
• An empty prefix matches all live keys. Scan results are sorted lexicographically by key in the chosen language.

Example

A value written at time 0 with ttl=10 stays visible at 9.999 and expires at exactly 10. Overwriting it without a TTL clears the old deadline instead of letting the stale expiration delete the replacement.

The following assertions test that inclusive expiry boundary:

clock = {"t": 0}
store = Store(lambda: clock["t"])
store.set("user:1", "Ada", ttl=10)
assert store.get("user:1") == "Ada"
clock["t"] = 10
assert store.get("user:1") is None

Constraints

• Don't use sleep in tests or implementation.
• Use standard-library Python only.
• Calls are sequential; the injected clock is finite and nondecreasing, and TTL deadlines are finite.
• Apply the same expiry rule on reads, deletes, and prefix scans.