Struct PartitionTree 

pub struct PartitionTree {
    pub cells: Vec<Cell>,
    pub split_order: Vec<usize>,
    pub k: usize,
}

Attribute Partition Tree (Algorithm 1 from the paper).

Recursive balanced partition on attributes. Each leaf is a cell of points sharing the same attribute combination.

Fields

cells: Vec<Cell>

All leaf cells.

split_order: Vec<usize>

Attribute split order (permutation of [0..k]).

k: usize

Number of attribute dimensions.

Implementations

impl PartitionTree

pub fn build(store: &PointStore) -> Self

Build the partition tree from a PointStore. Split order: most-distinct-values first (information gain heuristic).

pub fn filter_cells(&self, constraints: &[(usize, Vec<u32>)]) -> Vec<usize>

Find all cells compatible with a filter. A cell is compatible if for every constrained attribute j, the cell’s value on j is in the allowed set.

pub fn count_points(&self, cell_indices: &[usize]) -> usize

Total number of points across given cell indices.

pub fn collect_points(&self, cell_indices: &[usize]) -> Vec<u32>

Get all point ids in the given cell indices.

pub fn cell_of(&self, store: &PointStore, point_id: u32) -> Option<usize>

Find which cell a point belongs to. Returns cell index.