Struct PlannerStats 

pub struct PlannerStats {
    pub estimated_collection_sizes: DashMap<String, u64>,
    pub average_value_size: u64,
    pub fhe_op_latency_us: u64,
    pub fhe_comparison_cost: f64,
    pub fhe_boolean_cost: f64,
}

Statistics used for cost estimation

Maintains per-collection cardinality estimates and global latency hints so that the planner can make informed decisions.

Fields

estimated_collection_sizes: DashMap<String, u64>

Estimated row count per collection

average_value_size: u64

Average value size in bytes across all collections

fhe_op_latency_us: u64

Estimated microsecond latency of a single FHE gate operation

fhe_comparison_cost: f64

Cost of a single FHE comparison operation (Eq / Lt / Gt / Lte / Gte)

fhe_boolean_cost: f64

Cost of a single FHE boolean operation (And / Or)

Implementations

impl PlannerStats

pub fn set_collection_size(&self, collection: impl Into<String>, size: u64)

Update the estimated size for a collection

pub fn predicate_selectivity(&self, pred: &Predicate) -> f64

Estimate the fraction of rows a predicate will pass (0.0–1.0).

Heuristics (no histograms available):

• Eq → 0.001 (high selectivity, rare match)
• Lt/Gt/Lte/Gte → 0.3 (moderate selectivity)
• And(p1, p2) → s1 * s2
• Or(p1, p2) → 1 - (1-s1)*(1-s2)
• Not(p) → 1 - s

pub fn predicate_fhe_cost(&self, pred: &Predicate) -> f64

Estimate the relative FHE cost of evaluating a predicate (in abstract units).

• Leaf comparisons (Eq/Lt/Gt/Lte/Gte) each cost 1.0 comparison unit.
• And/Or cost 1.0 boolean-op + sum of children costs.
• Not costs 0.5 boolean-op + child cost.

Trait Implementations

impl Default for PlannerStats

fn default() -> Self

Returns the “default value” for a type.