Enum CandidateCount 

#[non_exhaustive]
pub enum CandidateCount {
    Exact(usize),
    AtLeast(usize),
}

What the counting probe learned about the candidate set.

The distinction is the point: SQLite has no histograms and sqlite_stat1 carries average rows-per-key, which estimates an equality predicate and not multi-hop reachability. So the count is measured, by running the traversal under a cap — and a probe that hits its cap has not measured anything except that the set is too big to care about the exact size.

The cap became a real one at 0.15.10 (D-252). Until then the traversal ran to completion and the tail was dropped afterwards, so AtLeast recorded that the answer had been trimmed and never that any work had been saved — C-8’s finding that probe_cap “bounds memory, not work”. It is now TraversalBuilder::limit, which stops the recursion, and the variant means what its name always claimed.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Exact(usize)

The traversal ran to the end of the graph and returned this many ids.

AtLeast(usize)

The walk stopped at the cap. The true count is at least this.

This is the id count, not the cap, and the two can differ now that the ceiling is on the walk rather than on the list it produced: the walk dedupes on (node, depth) and the projection drops retired concepts, so a walk cut at 10 rows can yield 8 ids. Reporting 8 is the true lower bound; reporting 10 would be a number nothing counted.

Implementations

impl CandidateCount

pub fn lower_bound(self) -> usize

The number to compute with. For a capped probe this understates the true count, which is the safe direction: it makes PreFilterCTE look cheaper than it is, and PreFilterCTE is the exact strategy.

It also now reflects what the traversal was paid for rather than what survived a truncation, which is the half of C-8 the cost model cared about: the estimator was pricing strategies against a candidate count the walk had already exceeded.

pub fn is_capped(self) -> bool

Trait Implementations

impl Clone for CandidateCount

fn clone(&self) -> CandidateCount

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for CandidateCount

impl Debug for CandidateCount

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for CandidateCount

impl PartialEq for CandidateCount

fn eq(&self, other: &CandidateCount) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for CandidateCount