Expand description
Valence-wide read-through LRU for Model::get-shaped point reads.
Enable: default on; set VALENCE_READ_CACHE=0 to disable. Capacity: VALENCE_READ_CACHE_MAX (default 10_000).
Stores raw rows pre-privacy; callers still run check_read_privacy per request actor.
When ownership_unified_fetch_enabled is on, cache entries bundle the main row and ownership
gate status so warm reads avoid a second ownership trip.
Write paths invalidate via generated Model ops and OwnershipService.
Functions§
- get_
record_ via_ cache - Point-get with optional read-through cache (raw row, pre-privacy).
- get_
record_ with_ ownership_ bundle_ via_ cache - Unified
Model::getfetch: one backend round trip for row + ownership gate status (when cache cold). - invalidate
- Drop a cached row after a write or ownership status change.
- read_
cache_ enabled - Whether the read-through record cache is active (default on;
VALENCE_READ_CACHE=0disables).