feat: Introduce a way to distinct documents

This commit is contained in:
Clément Renault 2018-10-17 13:35:34 +02:00
parent 3acac1458f
commit 37c709c9a9
8 changed files with 167 additions and 84 deletions

View File

@ -117,14 +117,14 @@ impl CsvIndexer {
}
}
fn insert_document_words<'a, I, A, B>(builder: &mut MetadataBuilder<A, B>, doc_index: u64, attr: u8, words: I)
fn insert_document_words<'a, I, A, B>(builder: &mut MetadataBuilder<A, B>, doc_id: u64, attr: u8, words: I)
where A: io::Write,
B: io::Write,
I: IntoIterator<Item=(usize, &'a str)>,
{
for (index, word) in words {
let doc_index = DocIndex {
document: doc_index,
document_id: doc_id,
attribute: attr,
attribute_index: index as u32,
};

View File

@ -122,14 +122,14 @@ impl JsonLinesIndexer {
}
}
fn insert_document_words<'a, I, A, B>(builder: &mut MetadataBuilder<A, B>, doc_index: u64, attr: u8, words: I)
fn insert_document_words<'a, I, A, B>(builder: &mut MetadataBuilder<A, B>, doc_id: u64, attr: u8, words: I)
where A: io::Write,
B: io::Write,
I: IntoIterator<Item=(usize, &'a str)>,
{
for (index, word) in words {
let doc_index = DocIndex {
document: doc_index,
document_id: doc_id,
attribute: attr,
attribute_index: index as u32,
};

View File

@ -5,8 +5,8 @@ use std::path::PathBuf;
use elapsed::measure_time;
use rocksdb::{DB, DBOptions, IngestExternalFileOptions};
use raptor::rank::{criterion, Config, RankedStream, Document};
use raptor::{automaton, Metadata, CommonWords};
use raptor::rank::{criterion, RankedStreamBuilder};
#[derive(Debug, StructOpt)]
pub struct CommandConsole {
@ -62,6 +62,13 @@ impl ConsoleSearch {
}
}
// "Sony" "PlayStation 4 500GB"
fn starts_with_playstation(doc: &Document, database: &DB) -> Vec<u8> {
let title_key = format!("{}-title", doc.id);
let title = database.get(title_key.as_bytes()).unwrap().unwrap();
title.get(0..4).map(|s| s.to_vec()).unwrap_or(Vec::new())
}
fn search(metadata: &Metadata, database: &DB, common_words: &CommonWords, query: &str) {
let mut automatons = Vec::new();
for query in query.split_whitespace().filter(|q| !common_words.contains(*q)) {
@ -69,10 +76,15 @@ fn search(metadata: &Metadata, database: &DB, common_words: &CommonWords, query:
automatons.push(lev);
}
let mut builder = RankedStreamBuilder::new(metadata, automatons);
builder.criteria(criterion::default());
let config = Config {
metadata: metadata,
automatons: automatons,
criteria: criterion::default(),
distinct: ((), 1),
};
let stream = RankedStream::new(config);
let mut stream = builder.build();
// let documents = stream.retrieve_distinct_documents(|doc| starts_with_playstation(doc, database), 0..20);
let documents = stream.retrieve_documents(0..20);
for document in documents {

View File

@ -7,11 +7,10 @@ use std::path::PathBuf;
use std::error::Error;
use std::sync::Arc;
use raptor::rank::{criterion, RankedStreamBuilder};
use raptor::rank::{criterion, Config, RankedStream};
use raptor::{automaton, Metadata, CommonWords};
use rocksdb::{DB, DBOptions, IngestExternalFileOptions};
use warp::Filter;
use structopt::StructOpt;
#[derive(Debug, StructOpt)]
@ -99,10 +98,14 @@ where M: AsRef<Metadata>,
automatons.push(lev);
}
let mut builder = RankedStreamBuilder::new(metadata.as_ref(), automatons);
builder.criteria(criterion::default());
let config = Config {
metadata: metadata.as_ref(),
automatons: automatons,
criteria: criterion::default(),
distinct: ((), 1),
};
let stream = RankedStream::new(config);
let mut stream = builder.build();
let documents = stream.retrieve_documents(0..20);
let mut body = Vec::new();

View File

@ -24,7 +24,7 @@ pub type DocumentId = u64;
#[repr(C)]
pub struct DocIndex {
/// The document identifier where the word was found.
pub document: DocumentId,
pub document_id: DocumentId,
/// The attribute identifier in the document
/// where the word was found.

View File

@ -7,7 +7,6 @@ mod exact;
use std::vec;
use std::cmp::Ordering;
use std::ops::Deref;
use crate::rank::Document;
pub use self::{
@ -31,21 +30,21 @@ pub trait Criterion {
impl<'a, T: Criterion + ?Sized> Criterion for &'a T {
fn evaluate(&self, lhs: &Document, rhs: &Document) -> Ordering {
self.deref().evaluate(lhs, rhs)
(**self).evaluate(lhs, rhs)
}
fn eq(&self, lhs: &Document, rhs: &Document) -> bool {
self.deref().eq(lhs, rhs)
(**self).eq(lhs, rhs)
}
}
impl<T: Criterion + ?Sized> Criterion for Box<T> {
fn evaluate(&self, lhs: &Document, rhs: &Document) -> Ordering {
self.deref().evaluate(lhs, rhs)
(**self).evaluate(lhs, rhs)
}
fn eq(&self, lhs: &Document, rhs: &Document) -> bool {
self.deref().eq(lhs, rhs)
(**self).eq(lhs, rhs)
}
}

View File

@ -3,7 +3,7 @@ mod ranked_stream;
use crate::{Match, DocumentId};
pub use self::ranked_stream::{RankedStreamBuilder, RankedStream};
pub use self::ranked_stream::{Config, RankedStream};
#[inline]
fn match_query_index(a: &Match, b: &Match) -> bool {

View File

@ -1,6 +1,8 @@
use std::collections::HashMap;
use std::hash::Hash;
use std::ops::Range;
use std::rc::Rc;
use std::{mem, vec, cmp};
use std::{mem, vec};
use fnv::FnvHashMap;
use fst::Streamer;
@ -9,52 +11,41 @@ use group_by::GroupByMut;
use crate::automaton::{DfaExt, AutomatonExt};
use crate::metadata::Metadata;
use crate::metadata::ops::OpBuilder;
use crate::rank::criterion::Criterion;
use crate::rank::criterion::{self, Criterion};
use crate::rank::Document;
use crate::Match;
use crate::{Match, DocumentId};
#[derive(Clone)]
pub struct RankedStreamBuilder<'m, C> {
metadata: &'m Metadata,
automatons: Vec<Rc<DfaExt>>,
criteria: Vec<C>,
pub struct Config<'m, C, F> {
pub metadata: &'m Metadata,
pub automatons: Vec<DfaExt>,
pub criteria: Vec<C>,
pub distinct: (F, usize),
}
impl<'m, C> RankedStreamBuilder<'m, C> {
pub fn new(metadata: &'m Metadata, automatons: Vec<DfaExt>) -> Self {
RankedStreamBuilder {
metadata: metadata,
automatons: automatons.into_iter().map(Rc::new).collect(),
criteria: Vec::new(), // hummm... prefer the criterion::default() ones !
}
}
pub struct RankedStream<'m, C, F> {
stream: crate::metadata::ops::Union<'m>,
automatons: Vec<Rc<DfaExt>>,
criteria: Vec<C>,
distinct: (F, usize),
}
pub fn criteria(&mut self, criteria: Vec<C>) {
self.criteria = criteria;
}
pub fn build(&self) -> RankedStream<C> {
let mut builder = OpBuilder::with_automatons(self.automatons.clone());
builder.push(self.metadata);
impl<'m, C, F> RankedStream<'m, C, F> {
pub fn new(config: Config<'m, C, F>) -> Self {
let automatons: Vec<_> = config.automatons.into_iter().map(Rc::new).collect();
let mut builder = OpBuilder::with_automatons(automatons.clone());
builder.push(config.metadata);
RankedStream {
stream: builder.union(),
automatons: &self.automatons,
criteria: &self.criteria,
automatons: automatons,
criteria: config.criteria,
distinct: config.distinct,
}
}
}
pub struct RankedStream<'a, 'm, C> {
stream: crate::metadata::ops::Union<'m>,
automatons: &'a [Rc<DfaExt>],
criteria: &'a [C],
}
impl<'a, 'm, C> RankedStream<'a, 'm, C> {
pub fn retrieve_documents(&mut self, range: Range<usize>) -> Vec<Document>
where C: Criterion
{
impl<'m, C, F> RankedStream<'m, C, F> {
fn retrieve_all_documents(&mut self) -> Vec<Document> {
let mut matches = FnvHashMap::default();
while let Some((string, indexed_values)) = self.stream.next() {
@ -63,55 +54,133 @@ impl<'a, 'm, C> RankedStream<'a, 'm, C> {
let distance = automaton.eval(string).to_u8();
let is_exact = distance == 0 && string.len() == automaton.query_len();
for di in iv.doc_indexes.as_slice() {
for doc_index in iv.doc_indexes.as_slice() {
let match_ = Match {
query_index: iv.index as u32,
distance: distance,
attribute: di.attribute,
attribute_index: di.attribute_index,
attribute: doc_index.attribute,
attribute_index: doc_index.attribute_index,
is_exact: is_exact,
};
matches.entry(di.document).or_insert_with(Vec::new).push(match_);
matches.entry(doc_index.document_id).or_insert_with(Vec::new).push(match_);
}
}
}
// collect matches from an HashMap into a Vec
let mut documents: Vec<_> = matches.into_iter().map(|(id, mut matches)| {
matches.into_iter().map(|(id, mut matches)| {
matches.sort_unstable();
unsafe { Document::from_sorted_matches(id, matches) }
}).collect();
}).collect()
}
}
impl<'a, C, F> RankedStream<'a, C, F>
where C: Criterion
{
pub fn retrieve_documents(mut self, range: Range<usize>) -> Vec<Document> {
let mut documents = self.retrieve_all_documents();
let mut groups = vec![documents.as_mut_slice()];
for criterion in self.criteria {
let tmp_groups = mem::replace(&mut groups, Vec::new());
let mut current_range = Range { start: 0, end: 0 };
'grp: for group in tmp_groups {
current_range.end += group.len();
// if a part of the current group is in the range returned
// we must sort it and emit the sub-groups
if current_range.contains(&range.start) {
group.sort_unstable_by(|a, b| criterion.evaluate(a, b));
for group in GroupByMut::new(group, |a, b| criterion.eq(a, b)) {
groups.push(group);
if current_range.end >= range.end { break 'grp }
}
} else {
groups.push(group)
for group in tmp_groups {
group.sort_unstable_by(|a, b| criterion.evaluate(a, b));
for group in GroupByMut::new(group, |a, b| criterion.eq(a, b)) {
groups.push(group);
}
current_range.start = current_range.end;
}
}
// TODO find a better algorithm, here we allocate for too many documents
// and we do a useless allocation, we should reuse the documents Vec
let start = cmp::min(range.start, documents.len());
let mut documents = documents.split_off(start);
documents.truncate(range.len());
documents
documents[range].to_vec()
}
pub fn retrieve_distinct_documents<K>(mut self, range: Range<usize>) -> Vec<Document>
where F: Fn(&DocumentId) -> K,
K: Hash + Eq,
{
let mut documents = self.retrieve_all_documents();
let mut groups = vec![documents.as_mut_slice()];
for criterion in self.criteria {
let tmp_groups = mem::replace(&mut groups, Vec::new());
for group in tmp_groups {
group.sort_unstable_by(|a, b| criterion.evaluate(a, b));
for group in GroupByMut::new(group, |a, b| criterion.eq(a, b)) {
groups.push(group);
}
}
}
let mut out_documents = Vec::with_capacity(range.len());
let (distinct, limit) = self.distinct;
let mut seen = DistinctMap::new(limit);
for document in documents {
let key = distinct(&document.id);
let accepted = seen.digest(key);
if accepted {
if seen.len() == range.end { break }
if seen.len() >= range.start {
out_documents.push(document);
}
}
}
out_documents
}
}
pub struct DistinctMap<K> {
inner: HashMap<K, usize>,
limit: usize,
len: usize,
}
impl<K: Hash + Eq> DistinctMap<K> {
pub fn new(limit: usize) -> Self {
DistinctMap {
inner: HashMap::new(),
limit: limit,
len: 0,
}
}
pub fn digest(&mut self, key: K) -> bool {
let seen = self.inner.entry(key).or_insert(0);
if *seen < self.limit { *seen += 1; self.len += 1; true } else { false }
}
pub fn len(&self) -> usize {
self.len
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn easy_distinct_map() {
let mut map = DistinctMap::new(2);
for x in &[1, 1, 1, 2, 3, 4, 5, 6, 6, 6, 6, 6] {
map.digest(x);
}
assert_eq!(map.len(), 8);
let mut map = DistinctMap::new(2);
assert_eq!(map.digest(1), true);
assert_eq!(map.digest(1), true);
assert_eq!(map.digest(1), false);
assert_eq!(map.digest(1), false);
assert_eq!(map.digest(2), true);
assert_eq!(map.digest(3), true);
assert_eq!(map.digest(2), true);
assert_eq!(map.digest(2), false);
assert_eq!(map.len(), 5);
}
}