Files
tantivy/src/query/phrase_query/phrase_scorer.rs
2022-11-09 11:20:41 +01:00

487 lines
15 KiB
Rust

use std::cmp::Ordering;
use crate::docset::{DocSet, TERMINATED};
use crate::fieldnorm::FieldNormReader;
use crate::postings::Postings;
use crate::query::bm25::Bm25Weight;
use crate::query::{Intersection, Scorer};
use crate::{DocId, Score};
struct PostingsWithOffset<TPostings> {
offset: u32,
postings: TPostings,
}
impl<TPostings: Postings> PostingsWithOffset<TPostings> {
pub fn new(segment_postings: TPostings, offset: u32) -> PostingsWithOffset<TPostings> {
PostingsWithOffset {
offset,
postings: segment_postings,
}
}
pub fn positions(&mut self, output: &mut Vec<u32>) {
self.postings.positions_with_offset(self.offset, output)
}
}
impl<TPostings: Postings> DocSet for PostingsWithOffset<TPostings> {
fn advance(&mut self) -> DocId {
self.postings.advance()
}
fn seek(&mut self, target: DocId) -> DocId {
self.postings.seek(target)
}
fn doc(&self) -> DocId {
self.postings.doc()
}
fn size_hint(&self) -> u32 {
self.postings.size_hint()
}
}
pub struct PhraseScorer<TPostings: Postings> {
intersection_docset: Intersection<PostingsWithOffset<TPostings>, PostingsWithOffset<TPostings>>,
num_terms: usize,
left: Vec<u32>,
right: Vec<u32>,
phrase_count: u32,
fieldnorm_reader: FieldNormReader,
similarity_weight_opt: Option<Bm25Weight>,
slop: u32,
}
/// Returns true if and only if the two sorted arrays contain a common element
fn intersection_exists(left: &[u32], right: &[u32]) -> bool {
let mut left_index = 0;
let mut right_index = 0;
while left_index < left.len() && right_index < right.len() {
let left_val = left[left_index];
let right_val = right[right_index];
match left_val.cmp(&right_val) {
Ordering::Less => {
left_index += 1;
}
Ordering::Equal => {
return true;
}
Ordering::Greater => {
right_index += 1;
}
}
}
false
}
fn intersection_count(left: &[u32], right: &[u32]) -> usize {
let mut left_index = 0;
let mut right_index = 0;
let mut count = 0;
while left_index < left.len() && right_index < right.len() {
let left_val = left[left_index];
let right_val = right[right_index];
match left_val.cmp(&right_val) {
Ordering::Less => {
left_index += 1;
}
Ordering::Equal => {
count += 1;
left_index += 1;
right_index += 1;
}
Ordering::Greater => {
right_index += 1;
}
}
}
count
}
/// Intersect twos sorted arrays `left` and `right` and outputs the
/// resulting array in left.
///
/// Returns the length of the intersection
fn intersection(left: &mut [u32], right: &[u32]) -> usize {
let mut left_index = 0;
let mut right_index = 0;
let mut count = 0;
let left_len = left.len();
let right_len = right.len();
while left_index < left_len && right_index < right_len {
let left_val = left[left_index];
let right_val = right[right_index];
match left_val.cmp(&right_val) {
Ordering::Less => {
left_index += 1;
}
Ordering::Equal => {
left[count] = left_val;
count += 1;
left_index += 1;
right_index += 1;
}
Ordering::Greater => {
right_index += 1;
}
}
}
count
}
/// Intersect twos sorted arrays `left` and `right` and outputs the
/// resulting array in left.
///
/// Condition for match is that the value stored in left is less than or equal to
/// the value in right and that the distance to the previous token is lte to the slop.
///
/// Returns the length of the intersection
fn intersection_with_slop(left: &mut [u32], right: &[u32], slop: u32) -> usize {
let mut left_index = 0;
let mut right_index = 0;
let mut count = 0;
let left_len = left.len();
let right_len = right.len();
while left_index < left_len && right_index < right_len {
let left_val = left[left_index];
let right_val = right[right_index];
// The three conditions are:
// left_val < right_slop -> left index increment.
// right_slop <= left_val <= right -> find the best match.
// left_val > right -> right index increment.
let right_slop = if right_val >= slop {
right_val - slop
} else {
0
};
if left_val < right_slop {
left_index += 1;
} else if right_slop <= left_val && left_val <= right_val {
while left_index + 1 < left_len {
// there could be a better match
let next_left_val = left[left_index + 1];
if next_left_val > right_val {
// the next value is outside the range, so current one is the best.
break;
}
// the next value is better.
left_index += 1;
}
// store the match in left.
left[count] = right_val;
count += 1;
left_index += 1;
right_index += 1;
} else if left_val > right_val {
right_index += 1;
}
}
count
}
fn intersection_count_with_slop(left: &[u32], right: &[u32], slop: u32) -> usize {
let mut left_index = 0;
let mut right_index = 0;
let mut count = 0;
let left_len = left.len();
let right_len = right.len();
while left_index < left_len && right_index < right_len {
let left_val = left[left_index];
let right_val = right[right_index];
let right_slop = if right_val >= slop {
right_val - slop
} else {
0
};
if left_val < right_slop {
left_index += 1;
} else if right_slop <= left_val && left_val <= right_val {
while left_index + 1 < left_len {
let next_left_val = left[left_index + 1];
if next_left_val > right_val {
break;
}
left_index += 1;
}
count += 1;
left_index += 1;
right_index += 1;
} else if left_val > right_val {
right_index += 1;
}
}
count
}
fn intersection_exists_with_slop(left: &[u32], right: &[u32], slop: u32) -> bool {
let mut left_index = 0;
let mut right_index = 0;
let left_len = left.len();
let right_len = right.len();
while left_index < left_len && right_index < right_len {
let left_val = left[left_index];
let right_val = right[right_index];
let right_slop = if right_val >= slop {
right_val - slop
} else {
0
};
if left_val < right_slop {
left_index += 1;
} else if right_slop <= left_val && left_val <= right_val {
return true;
} else if left_val > right_val {
right_index += 1;
}
}
false
}
impl<TPostings: Postings> PhraseScorer<TPostings> {
// If similarity_weight is None, then scoring is disabled.
pub fn new(
term_postings: Vec<(usize, TPostings)>,
similarity_weight_opt: Option<Bm25Weight>,
fieldnorm_reader: FieldNormReader,
slop: u32,
) -> PhraseScorer<TPostings> {
let max_offset = term_postings
.iter()
.map(|&(offset, _)| offset)
.max()
.unwrap_or(0);
let num_docsets = term_postings.len();
let postings_with_offsets = term_postings
.into_iter()
.map(|(offset, postings)| {
PostingsWithOffset::new(postings, (max_offset - offset) as u32)
})
.collect::<Vec<_>>();
let mut scorer = PhraseScorer {
intersection_docset: Intersection::new(postings_with_offsets),
num_terms: num_docsets,
left: Vec::with_capacity(100),
right: Vec::with_capacity(100),
phrase_count: 0u32,
similarity_weight_opt,
fieldnorm_reader,
slop,
};
if scorer.doc() != TERMINATED && !scorer.phrase_match() {
scorer.advance();
}
scorer
}
pub fn phrase_count(&self) -> u32 {
self.phrase_count
}
fn phrase_match(&mut self) -> bool {
if self.similarity_weight_opt.is_some() {
let count = self.compute_phrase_count();
self.phrase_count = count;
count > 0u32
} else {
self.phrase_exists()
}
}
fn phrase_exists(&mut self) -> bool {
let intersection_len = self.compute_phrase_match();
if self.has_slop() {
return intersection_exists_with_slop(
&self.left[..intersection_len],
&self.right[..],
self.slop,
);
}
intersection_exists(&self.left[..intersection_len], &self.right[..])
}
fn compute_phrase_count(&mut self) -> u32 {
let intersection_len = self.compute_phrase_match();
if self.has_slop() {
return intersection_count_with_slop(
&self.left[..intersection_len],
&self.right[..],
self.slop,
) as u32;
}
intersection_count(&self.left[..intersection_len], &self.right[..]) as u32
}
fn compute_phrase_match(&mut self) -> usize {
{
self.intersection_docset
.docset_mut_specialized(0)
.positions(&mut self.left);
}
let mut intersection_len = self.left.len();
for i in 1..self.num_terms - 1 {
{
self.intersection_docset
.docset_mut_specialized(i)
.positions(&mut self.right);
}
intersection_len = if self.has_slop() {
intersection_with_slop(
&mut self.left[..intersection_len],
&self.right[..],
self.slop,
)
} else {
intersection(&mut self.left[..intersection_len], &self.right[..])
};
if intersection_len == 0 {
return 0;
}
}
self.intersection_docset
.docset_mut_specialized(self.num_terms - 1)
.positions(&mut self.right);
intersection_len
}
fn has_slop(&self) -> bool {
self.slop > 0
}
}
impl<TPostings: Postings> DocSet for PhraseScorer<TPostings> {
fn advance(&mut self) -> DocId {
loop {
let doc = self.intersection_docset.advance();
if doc == TERMINATED || self.phrase_match() {
return doc;
}
}
}
fn seek(&mut self, target: DocId) -> DocId {
debug_assert!(target >= self.doc());
let doc = self.intersection_docset.seek(target);
if doc == TERMINATED || self.phrase_match() {
return doc;
}
self.advance()
}
fn doc(&self) -> DocId {
self.intersection_docset.doc()
}
fn size_hint(&self) -> u32 {
self.intersection_docset.size_hint()
}
}
impl<TPostings: Postings> Scorer for PhraseScorer<TPostings> {
fn score(&mut self) -> Score {
let doc = self.doc();
let fieldnorm_id = self.fieldnorm_reader.fieldnorm_id(doc);
if let Some(similarity_weight) = self.similarity_weight_opt.as_ref() {
similarity_weight.score(fieldnorm_id, self.phrase_count)
} else {
1.0f32
}
}
}
#[cfg(test)]
mod tests {
use super::{intersection, intersection_count, intersection_with_slop};
fn test_intersection_sym(left: &[u32], right: &[u32], expected: &[u32]) {
test_intersection_aux(left, right, expected, 0);
test_intersection_aux(right, left, expected, 0);
}
fn test_intersection_aux(left: &[u32], right: &[u32], expected: &[u32], slop: u32) {
let mut left_vec = Vec::from(left);
let left_mut = &mut left_vec[..];
if slop == 0 {
let left_mut = &mut left_vec[..];
assert_eq!(intersection_count(left_mut, right), expected.len());
let count = intersection(left_mut, right);
assert_eq!(&left_mut[..count], expected);
return;
}
let mut right_vec = Vec::from(right);
let right_mut = &mut right_vec[..];
let count = intersection_with_slop(left_mut, right_mut, slop);
assert_eq!(&left_mut[..count], expected);
}
#[test]
fn test_intersection() {
test_intersection_sym(&[1], &[1], &[1]);
test_intersection_sym(&[1], &[2], &[]);
test_intersection_sym(&[], &[2], &[]);
test_intersection_sym(&[5, 7], &[1, 5, 10, 12], &[5]);
test_intersection_sym(&[1, 5, 6, 9, 10, 12], &[6, 8, 9, 12], &[6, 9, 12]);
}
#[test]
fn test_slop() {
// The slop is not symmetric. It does not allow for the phrase to be out of order.
test_intersection_aux(&[1], &[2], &[2], 1);
test_intersection_aux(&[1], &[3], &[], 1);
test_intersection_aux(&[1], &[3], &[3], 2);
test_intersection_aux(&[], &[2], &[], 100000);
test_intersection_aux(&[5, 7, 11], &[1, 5, 10, 12], &[5, 12], 1);
test_intersection_aux(&[1, 5, 6, 9, 10, 12], &[6, 8, 9, 12], &[6, 9, 12], 1);
test_intersection_aux(&[1, 5, 6, 9, 10, 12], &[6, 8, 9, 12], &[6, 9, 12], 10);
test_intersection_aux(&[1, 3, 5], &[2, 4, 6], &[2, 4, 6], 1);
test_intersection_aux(&[1, 3, 5], &[2, 4, 6], &[], 0);
}
fn test_merge(left: &[u32], right: &[u32], expected_left: &[u32], slop: u32) {
let mut left_vec = Vec::from(left);
let left_mut = &mut left_vec[..];
let mut right_vec = Vec::from(right);
let right_mut = &mut right_vec[..];
let count = intersection_with_slop(left_mut, right_mut, slop);
assert_eq!(&left_mut[..count], expected_left);
}
#[test]
fn test_merge_slop() {
test_merge(&[1, 2], &[1], &[1], 1);
test_merge(&[3], &[4], &[4], 2);
test_merge(&[3], &[4], &[4], 2);
test_merge(&[1, 5, 6, 9, 10, 12], &[6, 8, 9, 12], &[6, 9, 12], 10);
}
}
#[cfg(all(test, feature = "unstable"))]
mod bench {
use test::Bencher;
use super::{intersection, intersection_count};
#[bench]
fn bench_intersection_short(b: &mut Bencher) {
b.iter(|| {
let mut left = [1, 5, 10, 12];
let right = [5, 7];
intersection(&mut left, &right);
});
}
#[bench]
fn bench_intersection_count_short(b: &mut Bencher) {
b.iter(|| {
let left = [1, 5, 10, 12];
let right = [5, 7];
intersection_count(&left, &right);
});
}
}