fix phrase search

This commit is contained in:
ad hoc 2022-02-01 20:10:16 +01:00
parent 38d23546a5
commit d852dc0d2b
No known key found for this signature in database
GPG Key ID: 4F00A782990CC643
2 changed files with 46 additions and 15 deletions

View File

@ -448,8 +448,10 @@ impl Search {
#[derive(Debug, StructOpt)]
struct SettingsUpdate {
#[structopt(short, long)]
#[structopt(long)]
filterable_attributes: Option<Vec<String>>,
#[structopt(long)]
criteria: Option<Vec<String>>,
}
impl Performer for SettingsUpdate {
@ -468,6 +470,14 @@ impl Performer for SettingsUpdate {
}
}
if let Some(criteria) = self.criteria {
if !criteria.is_empty() {
update.set_criteria(criteria);
} else {
update.reset_criteria();
}
}
let mut bars = Vec::new();
let progesses = MultiProgress::new();
for _ in 0..4 {

View File

@ -1,6 +1,7 @@
use std::borrow::Cow;
use std::collections::HashMap;
use itertools::Itertools;
use roaring::RoaringBitmap;
use self::asc_desc::AscDesc;
@ -318,23 +319,43 @@ pub fn resolve_query_tree<'t>(
}
Phrase(words) => {
let mut candidates = RoaringBitmap::new();
let mut first_loop = true;
for slice in words.windows(2) {
let (left, right) = (&slice[0], &slice[1]);
match ctx.word_pair_proximity_docids(left, right, 1)? {
Some(pair_docids) => {
if pair_docids.is_empty() {
return Ok(RoaringBitmap::new());
} else if first_loop {
candidates = pair_docids;
first_loop = false;
} else {
candidates &= pair_docids;
}
}
let mut first_iter = true;
let winsize = words.len().min(7);
for win in words.windows(winsize) {
// Get all the word pairs and their compute their relative distance
let dists = win
.iter()
.enumerate()
.cartesian_product(win.iter().enumerate())
.filter(|(x, y)| y > x)
.map(|((pos1, s1), (pos2, s2))| (s1, s2, pos2 - pos1));
let mut bitmaps = Vec::with_capacity(winsize.pow(2));
for (s1, s2, d) in dists {
match ctx.word_pair_proximity_docids(s1, s2, d as u8)? {
Some(m) => bitmaps.push(m),
None => return Ok(RoaringBitmap::new()),
}
}
// We sort the bitmaps so that we perform the small intersections first, which is faster.
bitmaps.sort_unstable_by(|a, b| a.len().cmp(&b.len()));
for bitmap in bitmaps {
if first_iter {
candidates = bitmap;
first_iter = false;
} else {
candidates &= bitmap;
}
// There will be no match, return early
if candidates.is_empty() {
break;
}
}
}
Ok(candidates)
}
Or(_, ops) => {