Module cache 

Scan-resistant caching with TinyLFU and byte-weighted eviction

Uses Moka’s TinyLFU eviction policy to handle scan workloads gracefully. Unlike timestamp-based eviction, TinyLFU tracks access frequency and rejects one-time “scan” entries that would pollute the cache.

Byte-weighted eviction: Caches track actual memory usage, not just entry counts. A 4MB emoji bitmap consumes 4000x more quota than a 1KB glyph. This prevents memory explosions from pathological fonts.

This prevents unbounded memory growth when processing many unique fonts (e.g., font matching across hundreds of candidates).

Structs

CacheManager

Centralized cache manager for shaping and glyph caches

CacheMetrics

Performance metrics for cache operations

CacheStats

Basic cache statistics

MultiLevelCache

Scan-resistant cache backed by Moka’s TinyLFU

RenderOutputCache

Byte-weighted cache for RenderOutput values

Constants

DEFAULT_CACHE_MAX_BYTES

Default cache byte limit: 512 MB

Functions

get_cache_manager

Returns a reference to the global CacheManager instance.

get_cache_max_bytes

Get the cache byte limit from environment or use default.