Module render_cache 

Content-addressed cache over backend rendering. Content-addressed render cache.

Caches backend rendering output keyed on the source IR’s content hash and the backend’s platform identifier. A cache hit skips the entire generate_resource call — rendering becomes a hashmap lookup when nothing in the IR changed.

Invariants

• Purity is required. Cache hits short-circuit Backend::apply; if a backend has observable side effects (touching the filesystem, writing a log, reading env vars), the cache hides them. All existing backends are pure renderers — this is safe for them.
• Schema version. The cache key includes a schema version so Backend trait upgrades that change output invalidate old entries automatically. Bump SCHEMA_VERSION whenever the Backend contract changes in a user-visible way.
• Deterministic inputs. Because content_hash is over canonical emission and canonical emission is deterministic (proven by proptest), two semantically-equal IR values ALWAYS produce the same cache key.

Usage

use iac_forge::render_cache::RenderCache;
let mut cache = RenderCache::new();
let artifacts = cache.render(backend, &input);
// Second call with the same (backend, input) is a lookup.
let artifacts2 = cache.render(backend, &input);

Structs

CacheKey

Cache key: (schema_version, backend_platform, ir_content_hash).

CacheStats

A stat block for cache observability — useful in tests and telemetry.

RenderCache

In-memory render cache.

Constants

SCHEMA_VERSION

Schema version embedded in every cache key. Bump when the Backend trait’s output contract changes in a way that would invalidate cached artifacts (new mandatory field, changed path scheme, etc.).