Bounded least-recently-used cache: controlled algorithm specification.
This is a synthetic implementation fixture, not a published empirical paper. A cache stores key/value pairs with a positive integer capacity. Accessing an existing key marks it most recently used. Setting an existing key updates the value and refreshes its recency. Inserting a new key beyond capacity evicts the least recently used key. Missing keys raise KeyError. Nonpositive capacity raises ValueError. Keys follow Python dictionary hashability rules.
Implementation requires a cache class, a command-line demonstration, and tests. Use collections.OrderedDict or a equivalent deterministic implementation. Verify recency after get and update, eviction order, capacity one, missing key and invalid capacity. Use only Python standard library and pytest. No external dataset, model weight, numerical benchmark or networking is required. Complexity should be described as expected O(1) per operation for OrderedDict, not measured speedup. A small demonstration is synthetic and must not be described as a real experiment.
The verification covers functional examples and edge cases only; it does not reproduce a scientific benchmark. All experimental performance results remain unverified.