snowykami/meilisearch
1
0
Fork 0
mirror of https://github.com/meilisearch/meilisearch.git synced 2024-11-23 10:37:41 +08:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Introduce a function to ignore useless paths

Browse Source
This commit is contained in:
Kerollmops 2020-06-13 00:17:43 +02:00
parent d02c5cb023
commit b9cc6c10af
No known key found for this signature in database
GPG Key ID: 92ADA4E935E71FA4
2 changed files with 30 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
src/best_proximity.rs
View File

@ -44,7 +44,14 @@ impl Node {
// TODO we must skip the successors that have already been seen // TODO we must skip the successors that have already been seen
// TODO we must skip the successors that doesn't return any documents // TODO we must skip the successors that doesn't return any documents
// this way we are able to skip entire paths // this way we are able to skip entire paths
fn successors(&self, positions: &[Vec<u32>], best_proximity: u32) -> Vec<(Node, u32)> { fn successors<F>(
&self,
positions: &[Vec<u32>],
best_proximity: u32,
mut contains_documents: F,
) -> Vec<(Node, u32)>
where F: FnMut((usize, u32), (usize, u32)) -> bool,
{
match self { match self {
Node::Uninit => { Node::Uninit => {
positions[0].iter().map(|p| { positions[0].iter().map(|p| {
@ -54,16 +61,19 @@ impl Node {
// We reached the highest layer // We reached the highest layer
n @ Node::Init { .. } if n.is_complete(positions) => vec![], n @ Node::Init { .. } if n.is_complete(positions) => vec![],
Node::Init { layer, position, acc_proximity } => { Node::Init { layer, position, acc_proximity } => {
let layer = layer + 1; positions[layer + 1].iter().filter_map(|p| {
positions[layer].iter().filter_map(|p| {
let proximity = positions_proximity(*position, *p); let proximity = positions_proximity(*position, *p);
let node = Node::Init { layer, position: *p, acc_proximity: acc_proximity + proximity }; let node = Node::Init { layer: layer + 1, position: *p, acc_proximity: acc_proximity + proximity };
if (contains_documents)((*layer, *position), (layer + 1, *p)) {
// We do not produce the nodes we have already seen in previous iterations loops. // We do not produce the nodes we have already seen in previous iterations loops.
if node.is_complete(positions) && acc_proximity + proximity < best_proximity { if node.is_complete(positions) && acc_proximity + proximity < best_proximity {
None None
} else { } else {
Some((node, proximity)) Some((node, proximity))
} }
} else {
None
}
}).collect() }).collect()
} }
} }
@ -84,18 +94,21 @@ impl Node {
} }
} }
pub struct BestProximity { pub struct BestProximity<F> {
positions: Vec<Vec<u32>>, positions: Vec<Vec<u32>>,
best_proximity: u32, best_proximity: u32,
contains_documents: F,
} }
impl BestProximity { impl<F> BestProximity<F> {
pub fn new(positions: Vec<Vec<u32>>) -> BestProximity { pub fn new(positions: Vec<Vec<u32>>, contains_documents: F) -> BestProximity<F> {
BestProximity { positions, best_proximity: 0 } BestProximity { positions, best_proximity: 0, contains_documents }
} }
} }
impl Iterator for BestProximity { impl<F> Iterator for BestProximity<F>
where F: FnMut((usize, u32), (usize, u32)) -> bool + Copy,
{
type Item = (u32, Vec<Vec<u32>>); type Item = (u32, Vec<Vec<u32>>);
fn next(&mut self) -> Option<Self::Item> { fn next(&mut self) -> Option<Self::Item> {
@ -107,7 +120,7 @@ impl Iterator for BestProximity {
let result = astar_bag( let result = astar_bag(
&Node::Uninit, // start &Node::Uninit, // start
|n| n.successors(&self.positions, self.best_proximity), |n| n.successors(&self.positions, self.best_proximity, self.contains_documents),
|_| 0, // heuristic |_| 0, // heuristic
|n| n.is_complete(&self.positions), // success |n| n.is_complete(&self.positions), // success
); );
@ -142,7 +155,7 @@ mod tests {
vec![ 1, ], vec![ 1, ],
vec![ 3, 6], vec![ 3, 6],
]; ];
let mut iter = BestProximity::new(positions); let mut iter = BestProximity::new(positions, |_, _| true);
assert_eq!(iter.next(), Some((1+2, vec![vec![0, 1, 3]]))); // 3 assert_eq!(iter.next(), Some((1+2, vec![vec![0, 1, 3]]))); // 3
assert_eq!(iter.next(), Some((2+2, vec![vec![2, 1, 3]]))); // 4 assert_eq!(iter.next(), Some((2+2, vec![vec![2, 1, 3]]))); // 4
@ -161,7 +174,7 @@ mod tests {
vec![ 1, 1000, 2001 ], vec![ 1, 1000, 2001 ],
vec![ 3, 6, 2002, 3000], vec![ 3, 6, 2002, 3000],
]; ];
let mut iter = BestProximity::new(positions); let mut iter = BestProximity::new(positions, |_, _| true);
assert_eq!(iter.next(), Some((1+1, vec![vec![2000, 2001, 2002]]))); // 2 assert_eq!(iter.next(), Some((1+1, vec![vec![2000, 2001, 2002]]))); // 2
assert_eq!(iter.next(), Some((1+2, vec![vec![0, 1, 3]]))); // 3 assert_eq!(iter.next(), Some((1+2, vec![vec![0, 1, 3]]))); // 3

2
src/lib.rs
View File

@ -119,7 +119,7 @@ impl Index {
let mut documents = Vec::new(); let mut documents = Vec::new();
for (proximity, mut positions) in BestProximity::new(positions) { for (proximity, mut positions) in BestProximity::new(positions, |_, _| true) {
// TODO we must ignore positions paths that gives nothing // TODO we must ignore positions paths that gives nothing
if (proximity as usize) < words.len() - 1 { if (proximity as usize) < words.len() - 1 {
eprintln!("Skipping too short proximities of {}.", proximity); eprintln!("Skipping too short proximities of {}.", proximity);