Struct ChartPatternDetector 

pub struct ChartPatternDetector {
    pub tolerance_pct: f64,
    pub flat_pct: f64,
}

Scans a swing-node sequence for pattern candidates and evicts overlapping lower-confidence ones, so the result is a ranked, non-redundant set rather than every geometrically-possible match.

Fields

tolerance_pct: f64

flat_pct: f64

How much a boundary may rise or fall across the window and still count as horizontal, as a percentage of the window’s opening gap between the two boundaries.

Its own threshold rather than a reuse of tolerance_pct: that one answers “are these two prices the same level”, this one answers “is this line flat”. Both are decisions, but they are not the same decision, and a caller that widens level matching to find more double tops should not thereby start seeing ascending triangles everywhere.

Implementations

impl ChartPatternDetector

pub fn new(tolerance_pct: f64) -> Self

pub fn with_flat_pct(self, flat_pct: f64) -> Self

Sets the flatness threshold for the ascending/descending triangle boundary.

pub fn scan(&self, nodes: &[ZigZagNode]) -> Vec<ChartPatternCandidate>

Detects all supported pattern kinds over nodes, then evicts overlapping candidates (sharing any swing node), keeping the highest-confidence one per overlapping cluster.