2021-06-16 18:33:33 +02:00
|
|
|
use std::{cmp, fmt, mem};
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-04-08 21:21:20 +02:00
|
|
|
use fst::Set;
|
2021-06-08 17:29:38 +02:00
|
|
|
use meilisearch_tokenizer::token::SeparatorKind;
|
|
|
|
use meilisearch_tokenizer::tokenizer::TokenStream;
|
|
|
|
use meilisearch_tokenizer::TokenKind;
|
2021-03-03 12:03:31 +01:00
|
|
|
use roaring::RoaringBitmap;
|
|
|
|
use slice_group_by::GroupBy;
|
|
|
|
|
2021-06-14 16:46:19 +02:00
|
|
|
use crate::{Index, Result};
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
type IsOptionalWord = bool;
|
|
|
|
type IsPrefix = bool;
|
|
|
|
|
|
|
|
#[derive(Clone, PartialEq, Eq, Hash)]
|
|
|
|
pub enum Operation {
|
|
|
|
And(Vec<Operation>),
|
2021-06-09 17:28:12 +02:00
|
|
|
// serie of consecutive non prefix and exact words
|
|
|
|
Phrase(Vec<String>),
|
2021-03-03 12:03:31 +01:00
|
|
|
Or(IsOptionalWord, Vec<Operation>),
|
|
|
|
Query(Query),
|
|
|
|
}
|
|
|
|
|
|
|
|
impl fmt::Debug for Operation {
|
|
|
|
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
|
|
|
|
fn pprint_tree(f: &mut fmt::Formatter<'_>, op: &Operation, depth: usize) -> fmt::Result {
|
|
|
|
match op {
|
|
|
|
Operation::And(children) => {
|
|
|
|
writeln!(f, "{:1$}AND", "", depth * 2)?;
|
|
|
|
children.iter().try_for_each(|c| pprint_tree(f, c, depth + 1))
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-06-09 17:28:12 +02:00
|
|
|
Operation::Phrase(children) => {
|
|
|
|
writeln!(f, "{:2$}PHRASE {:?}", "", children, depth * 2)
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
Operation::Or(true, children) => {
|
|
|
|
writeln!(f, "{:1$}OR(WORD)", "", depth * 2)?;
|
|
|
|
children.iter().try_for_each(|c| pprint_tree(f, c, depth + 1))
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
Operation::Or(false, children) => {
|
|
|
|
writeln!(f, "{:1$}OR", "", depth * 2)?;
|
|
|
|
children.iter().try_for_each(|c| pprint_tree(f, c, depth + 1))
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
Operation::Query(query) => writeln!(f, "{:2$}{:?}", "", query, depth * 2),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
pprint_tree(f, self, 0)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
impl Operation {
|
|
|
|
fn and(mut ops: Vec<Self>) -> Self {
|
|
|
|
if ops.len() == 1 {
|
|
|
|
ops.pop().unwrap()
|
|
|
|
} else {
|
|
|
|
Self::And(ops)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
pub fn or(word_branch: IsOptionalWord, mut ops: Vec<Self>) -> Self {
|
|
|
|
if ops.len() == 1 {
|
|
|
|
ops.pop().unwrap()
|
|
|
|
} else {
|
|
|
|
Self::Or(word_branch, ops)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-06-09 17:28:12 +02:00
|
|
|
fn phrase(mut words: Vec<String>) -> Self {
|
|
|
|
if words.len() == 1 {
|
2021-06-10 11:30:51 +02:00
|
|
|
Self::Query(Query { prefix: false, kind: QueryKind::exact(words.pop().unwrap()) })
|
2021-03-03 12:03:31 +01:00
|
|
|
} else {
|
2021-06-09 17:28:12 +02:00
|
|
|
Self::Phrase(words)
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
2021-02-22 17:17:01 +01:00
|
|
|
|
|
|
|
pub fn query(&self) -> Option<&Query> {
|
|
|
|
match self {
|
|
|
|
Operation::Query(query) => Some(query),
|
|
|
|
_ => None,
|
|
|
|
}
|
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
#[derive(Clone, Eq, PartialEq, Hash)]
|
|
|
|
pub struct Query {
|
|
|
|
pub prefix: IsPrefix,
|
|
|
|
pub kind: QueryKind,
|
|
|
|
}
|
|
|
|
|
|
|
|
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
|
|
|
|
pub enum QueryKind {
|
|
|
|
Tolerant { typo: u8, word: String },
|
|
|
|
Exact { original_typo: u8, word: String },
|
|
|
|
}
|
|
|
|
|
|
|
|
impl QueryKind {
|
2021-02-24 10:25:22 +01:00
|
|
|
pub fn exact(word: String) -> Self {
|
2021-03-03 12:03:31 +01:00
|
|
|
QueryKind::Exact { original_typo: 0, word }
|
|
|
|
}
|
|
|
|
|
2021-02-24 10:25:22 +01:00
|
|
|
pub fn exact_with_typo(original_typo: u8, word: String) -> Self {
|
|
|
|
QueryKind::Exact { original_typo, word }
|
|
|
|
}
|
|
|
|
|
|
|
|
pub fn tolerant(typo: u8, word: String) -> Self {
|
2021-03-03 12:03:31 +01:00
|
|
|
QueryKind::Tolerant { typo, word }
|
|
|
|
}
|
|
|
|
|
2021-02-24 17:44:35 +01:00
|
|
|
pub fn is_tolerant(&self) -> bool {
|
|
|
|
matches!(self, QueryKind::Tolerant { .. })
|
|
|
|
}
|
|
|
|
|
|
|
|
pub fn is_exact(&self) -> bool {
|
|
|
|
matches!(self, QueryKind::Exact { .. })
|
|
|
|
}
|
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
pub fn typo(&self) -> u8 {
|
|
|
|
match self {
|
|
|
|
QueryKind::Tolerant { typo, .. } => *typo,
|
|
|
|
QueryKind::Exact { original_typo, .. } => *original_typo,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
pub fn word(&self) -> &str {
|
|
|
|
match self {
|
|
|
|
QueryKind::Tolerant { word, .. } => word,
|
|
|
|
QueryKind::Exact { word, .. } => word,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
impl fmt::Debug for Query {
|
|
|
|
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
|
|
|
|
let Query { prefix, kind } = self;
|
|
|
|
let prefix = if *prefix { String::from("Prefix") } else { String::default() };
|
|
|
|
match kind {
|
|
|
|
QueryKind::Exact { word, .. } => {
|
|
|
|
f.debug_struct(&(prefix + "Exact")).field("word", &word).finish()
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
|
|
|
QueryKind::Tolerant { typo, word } => f
|
|
|
|
.debug_struct(&(prefix + "Tolerant"))
|
|
|
|
.field("word", &word)
|
|
|
|
.field("max typo", &typo)
|
|
|
|
.finish(),
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
trait Context {
|
|
|
|
fn word_docids(&self, word: &str) -> heed::Result<Option<RoaringBitmap>>;
|
2021-04-09 21:56:20 +02:00
|
|
|
fn synonyms<S: AsRef<str>>(&self, words: &[S]) -> heed::Result<Option<Vec<Vec<String>>>>;
|
2021-02-18 15:06:58 +01:00
|
|
|
fn word_documents_count(&self, word: &str) -> heed::Result<Option<u64>> {
|
|
|
|
match self.word_docids(word)? {
|
|
|
|
Some(rb) => Ok(Some(rb.len())),
|
|
|
|
None => Ok(None),
|
|
|
|
}
|
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/// The query tree builder is the interface to build a query tree.
|
|
|
|
pub struct QueryTreeBuilder<'a> {
|
|
|
|
rtxn: &'a heed::RoTxn<'a>,
|
|
|
|
index: &'a Index,
|
2021-02-25 10:49:25 +01:00
|
|
|
optional_words: bool,
|
|
|
|
authorize_typos: bool,
|
2021-04-13 19:10:58 +02:00
|
|
|
words_limit: Option<usize>,
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
impl<'a> Context for QueryTreeBuilder<'a> {
|
|
|
|
fn word_docids(&self, word: &str) -> heed::Result<Option<RoaringBitmap>> {
|
|
|
|
self.index.word_docids.get(self.rtxn, word)
|
|
|
|
}
|
|
|
|
|
2021-04-09 21:56:20 +02:00
|
|
|
fn synonyms<S: AsRef<str>>(&self, words: &[S]) -> heed::Result<Option<Vec<Vec<String>>>> {
|
2021-04-07 10:53:57 +02:00
|
|
|
self.index.words_synonyms(self.rtxn, words)
|
2021-02-18 15:06:58 +01:00
|
|
|
}
|
|
|
|
|
2021-04-07 10:53:57 +02:00
|
|
|
fn word_documents_count(&self, word: &str) -> heed::Result<Option<u64>> {
|
|
|
|
self.index.word_documents_count(self.rtxn, word)
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
impl<'a> QueryTreeBuilder<'a> {
|
|
|
|
/// Create a `QueryTreeBuilder` from a heed ReadOnly transaction `rtxn`
|
|
|
|
/// and an Index `index`.
|
|
|
|
pub fn new(rtxn: &'a heed::RoTxn<'a>, index: &'a Index) -> Self {
|
2021-04-13 19:10:58 +02:00
|
|
|
Self { rtxn, index, optional_words: true, authorize_typos: true, words_limit: None }
|
2021-02-25 10:49:25 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/// if `optional_words` is set to `false` the query tree will be
|
|
|
|
/// generated forcing all query words to be present in each matching documents
|
|
|
|
/// (the criterion `words` will be ignored).
|
|
|
|
/// default value if not called: `true`
|
2021-03-02 11:30:48 +01:00
|
|
|
#[allow(unused)]
|
2021-02-25 10:49:25 +01:00
|
|
|
pub fn optional_words(&mut self, optional_words: bool) -> &mut Self {
|
|
|
|
self.optional_words = optional_words;
|
|
|
|
self
|
|
|
|
}
|
|
|
|
|
|
|
|
/// if `authorize_typos` is set to `false` the query tree will be generated
|
|
|
|
/// forcing all query words to match documents without any typo
|
|
|
|
/// (the criterion `typo` will be ignored).
|
|
|
|
/// default value if not called: `true`
|
2021-03-02 11:30:48 +01:00
|
|
|
#[allow(unused)]
|
2021-02-25 10:49:25 +01:00
|
|
|
pub fn authorize_typos(&mut self, authorize_typos: bool) -> &mut Self {
|
|
|
|
self.authorize_typos = authorize_typos;
|
|
|
|
self
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
|
2021-04-13 19:10:58 +02:00
|
|
|
/// Limit words and phrases that will be taken for query building.
|
|
|
|
/// Any beyond `words_limit` will be ignored.
|
|
|
|
pub fn words_limit(&mut self, words_limit: usize) -> &mut Self {
|
|
|
|
self.words_limit = Some(words_limit);
|
|
|
|
self
|
|
|
|
}
|
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
/// Build the query tree:
|
|
|
|
/// - if `optional_words` is set to `false` the query tree will be
|
|
|
|
/// generated forcing all query words to be present in each matching documents
|
|
|
|
/// (the criterion `words` will be ignored)
|
|
|
|
/// - if `authorize_typos` is set to `false` the query tree will be generated
|
|
|
|
/// forcing all query words to match documents without any typo
|
|
|
|
/// (the criterion `typo` will be ignored)
|
2021-06-14 16:46:19 +02:00
|
|
|
pub fn build(&self, query: TokenStream) -> Result<Option<(Operation, PrimitiveQuery)>> {
|
2021-04-08 21:21:20 +02:00
|
|
|
let stop_words = self.index.stop_words(self.rtxn)?;
|
2021-04-13 19:10:58 +02:00
|
|
|
let primitive_query = create_primitive_query(query, stop_words, self.words_limit);
|
2021-03-03 12:03:31 +01:00
|
|
|
if !primitive_query.is_empty() {
|
2021-06-16 18:33:33 +02:00
|
|
|
let qt = create_query_tree(
|
|
|
|
self,
|
|
|
|
self.optional_words,
|
|
|
|
self.authorize_typos,
|
|
|
|
&primitive_query,
|
|
|
|
)?;
|
2021-05-04 13:44:55 +02:00
|
|
|
Ok(Some((qt, primitive_query)))
|
2021-03-03 12:03:31 +01:00
|
|
|
} else {
|
|
|
|
Ok(None)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Split the word depending on the frequency of subwords in the database documents.
|
2021-05-10 10:27:18 +02:00
|
|
|
fn split_best_frequency(ctx: &impl Context, word: &str) -> heed::Result<Option<Operation>> {
|
2021-03-03 12:03:31 +01:00
|
|
|
let chars = word.char_indices().skip(1);
|
|
|
|
let mut best = None;
|
|
|
|
|
|
|
|
for (i, _) in chars {
|
|
|
|
let (left, right) = word.split_at(i);
|
|
|
|
|
2021-02-18 15:06:58 +01:00
|
|
|
let left_freq = ctx.word_documents_count(left)?.unwrap_or(0);
|
|
|
|
let right_freq = ctx.word_documents_count(right)?.unwrap_or(0);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
let min_freq = cmp::min(left_freq, right_freq);
|
|
|
|
if min_freq != 0 && best.map_or(true, |(old, _, _)| min_freq > old) {
|
|
|
|
best = Some((min_freq, left, right));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
Ok(best.map(|(_, left, right)| Operation::Phrase(vec![left.to_string(), right.to_string()])))
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/// Return the `QueryKind` of a word depending on `authorize_typos`
|
|
|
|
/// and the provided word length.
|
|
|
|
fn typos(word: String, authorize_typos: bool) -> QueryKind {
|
|
|
|
if authorize_typos {
|
2021-09-28 12:10:43 +02:00
|
|
|
match word.chars().count() {
|
2021-03-03 12:03:31 +01:00
|
|
|
0..=4 => QueryKind::exact(word),
|
|
|
|
5..=8 => QueryKind::tolerant(1, word),
|
2021-06-16 18:33:33 +02:00
|
|
|
_ => QueryKind::tolerant(2, word),
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
QueryKind::exact(word)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-04-09 21:56:20 +02:00
|
|
|
/// Fetch synonyms from the `Context` for the provided word
|
2021-03-03 12:03:31 +01:00
|
|
|
/// and create the list of operations for the query tree
|
2021-04-09 21:56:20 +02:00
|
|
|
fn synonyms(ctx: &impl Context, word: &[&str]) -> heed::Result<Option<Vec<Operation>>> {
|
|
|
|
let synonyms = ctx.synonyms(word)?;
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
Ok(synonyms.map(|synonyms| {
|
2021-06-16 18:33:33 +02:00
|
|
|
synonyms
|
|
|
|
.into_iter()
|
|
|
|
.map(|synonym| {
|
|
|
|
let words = synonym
|
|
|
|
.into_iter()
|
|
|
|
.map(|word| {
|
|
|
|
Operation::Query(Query { prefix: false, kind: QueryKind::exact(word) })
|
|
|
|
})
|
|
|
|
.collect();
|
|
|
|
Operation::and(words)
|
|
|
|
})
|
|
|
|
.collect()
|
2021-03-03 12:03:31 +01:00
|
|
|
}))
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Main function that creates the final query tree from the primitive query.
|
|
|
|
fn create_query_tree(
|
|
|
|
ctx: &impl Context,
|
|
|
|
optional_words: bool,
|
|
|
|
authorize_typos: bool,
|
2021-05-04 13:44:55 +02:00
|
|
|
query: &[PrimitiveQueryPart],
|
2021-06-16 18:33:33 +02:00
|
|
|
) -> Result<Operation> {
|
2021-03-03 12:03:31 +01:00
|
|
|
/// Matches on the `PrimitiveQueryPart` and create an operation from it.
|
|
|
|
fn resolve_primitive_part(
|
|
|
|
ctx: &impl Context,
|
|
|
|
authorize_typos: bool,
|
|
|
|
part: PrimitiveQueryPart,
|
2021-06-16 18:33:33 +02:00
|
|
|
) -> Result<Operation> {
|
2021-03-03 12:03:31 +01:00
|
|
|
match part {
|
|
|
|
// 1. try to split word in 2
|
|
|
|
// 2. try to fetch synonyms
|
|
|
|
// 3. create an operation containing the word
|
|
|
|
// 4. wrap all in an OR operation
|
|
|
|
PrimitiveQueryPart::Word(word, prefix) => {
|
|
|
|
let mut children = synonyms(ctx, &[&word])?.unwrap_or_default();
|
|
|
|
if let Some(child) = split_best_frequency(ctx, &word)? {
|
|
|
|
children.push(child);
|
|
|
|
}
|
2021-06-16 18:33:33 +02:00
|
|
|
children
|
|
|
|
.push(Operation::Query(Query { prefix, kind: typos(word, authorize_typos) }));
|
2021-03-03 12:03:31 +01:00
|
|
|
Ok(Operation::or(false, children))
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
// create a CONSECUTIVE operation wrapping all word in the phrase
|
2021-06-16 18:33:33 +02:00
|
|
|
PrimitiveQueryPart::Phrase(words) => Ok(Operation::phrase(words)),
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Create all ngrams 1..=3 generating query tree branches.
|
|
|
|
fn ngrams(
|
|
|
|
ctx: &impl Context,
|
|
|
|
authorize_typos: bool,
|
|
|
|
query: &[PrimitiveQueryPart],
|
2021-06-16 18:33:33 +02:00
|
|
|
) -> Result<Operation> {
|
2021-03-03 12:03:31 +01:00
|
|
|
const MAX_NGRAM: usize = 3;
|
|
|
|
let mut op_children = Vec::new();
|
|
|
|
|
2021-04-08 21:21:20 +02:00
|
|
|
for sub_query in query.linear_group_by(|a, b| !(a.is_phrase() || b.is_phrase())) {
|
2021-03-03 12:03:31 +01:00
|
|
|
let mut or_op_children = Vec::new();
|
|
|
|
|
|
|
|
for ngram in 1..=MAX_NGRAM.min(sub_query.len()) {
|
|
|
|
if let Some(group) = sub_query.get(..ngram) {
|
|
|
|
let mut and_op_children = Vec::new();
|
|
|
|
let tail = &sub_query[ngram..];
|
|
|
|
let is_last = tail.is_empty();
|
|
|
|
|
|
|
|
match group {
|
|
|
|
[part] => {
|
2021-06-16 18:33:33 +02:00
|
|
|
let operation =
|
|
|
|
resolve_primitive_part(ctx, authorize_typos, part.clone())?;
|
2021-03-03 12:03:31 +01:00
|
|
|
and_op_children.push(operation);
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
words => {
|
2021-04-08 21:21:20 +02:00
|
|
|
let is_prefix = words.last().map_or(false, |part| part.is_prefix());
|
2021-06-16 18:33:33 +02:00
|
|
|
let words: Vec<_> = words
|
|
|
|
.iter()
|
|
|
|
.filter_map(|part| {
|
|
|
|
if let PrimitiveQueryPart::Word(word, _) = part {
|
|
|
|
Some(word.as_str())
|
|
|
|
} else {
|
|
|
|
None
|
|
|
|
}
|
|
|
|
})
|
|
|
|
.collect();
|
2021-03-03 12:03:31 +01:00
|
|
|
let mut operations = synonyms(ctx, &words)?.unwrap_or_default();
|
|
|
|
let concat = words.concat();
|
2021-06-16 18:33:33 +02:00
|
|
|
let query =
|
|
|
|
Query { prefix: is_prefix, kind: typos(concat, authorize_typos) };
|
2021-04-08 15:12:37 +02:00
|
|
|
operations.push(Operation::Query(query));
|
|
|
|
and_op_children.push(Operation::or(false, operations));
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if !is_last {
|
|
|
|
let ngrams = ngrams(ctx, authorize_typos, tail)?;
|
|
|
|
and_op_children.push(ngrams);
|
|
|
|
}
|
|
|
|
or_op_children.push(Operation::and(and_op_children));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
op_children.push(Operation::or(false, or_op_children));
|
|
|
|
}
|
|
|
|
|
|
|
|
Ok(Operation::and(op_children))
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Create a new branch removing the last non-phrase query parts.
|
|
|
|
fn optional_word(
|
|
|
|
ctx: &impl Context,
|
|
|
|
authorize_typos: bool,
|
|
|
|
query: PrimitiveQuery,
|
2021-06-16 18:33:33 +02:00
|
|
|
) -> Result<Operation> {
|
2021-02-18 22:18:36 +01:00
|
|
|
let number_phrases = query.iter().filter(|p| p.is_phrase()).count();
|
2021-03-03 12:03:31 +01:00
|
|
|
let mut operation_children = Vec::new();
|
|
|
|
|
2021-02-18 22:18:36 +01:00
|
|
|
let start = number_phrases + (number_phrases == 0) as usize;
|
|
|
|
for len in start..=query.len() {
|
|
|
|
let mut word_count = len - number_phrases;
|
2021-06-16 18:33:33 +02:00
|
|
|
let query: Vec<_> = query
|
|
|
|
.iter()
|
|
|
|
.filter(|p| {
|
|
|
|
if p.is_phrase() {
|
|
|
|
true
|
|
|
|
} else if word_count != 0 {
|
|
|
|
word_count -= 1;
|
|
|
|
true
|
|
|
|
} else {
|
|
|
|
false
|
|
|
|
}
|
|
|
|
})
|
|
|
|
.cloned()
|
|
|
|
.collect();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-02-18 22:18:36 +01:00
|
|
|
let ngrams = ngrams(ctx, authorize_typos, &query)?;
|
2021-03-03 12:03:31 +01:00
|
|
|
operation_children.push(ngrams);
|
|
|
|
}
|
|
|
|
|
|
|
|
Ok(Operation::or(true, operation_children))
|
|
|
|
}
|
|
|
|
|
|
|
|
if optional_words {
|
2021-05-04 13:44:55 +02:00
|
|
|
optional_word(ctx, authorize_typos, query.to_vec())
|
2021-03-03 12:03:31 +01:00
|
|
|
} else {
|
2021-05-04 13:44:55 +02:00
|
|
|
ngrams(ctx, authorize_typos, query)
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-05-04 13:44:55 +02:00
|
|
|
pub type PrimitiveQuery = Vec<PrimitiveQueryPart>;
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
#[derive(Debug, Clone)]
|
2021-05-04 13:44:55 +02:00
|
|
|
pub enum PrimitiveQueryPart {
|
2021-03-03 12:03:31 +01:00
|
|
|
Phrase(Vec<String>),
|
|
|
|
Word(String, IsPrefix),
|
|
|
|
}
|
|
|
|
|
|
|
|
impl PrimitiveQueryPart {
|
|
|
|
fn is_phrase(&self) -> bool {
|
|
|
|
matches!(self, Self::Phrase(_))
|
|
|
|
}
|
|
|
|
|
|
|
|
fn is_prefix(&self) -> bool {
|
|
|
|
matches!(self, Self::Word(_, is_prefix) if *is_prefix)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Create primitive query from tokenized query string,
|
|
|
|
/// the primitive query is an intermediate state to build the query tree.
|
2021-06-16 18:33:33 +02:00
|
|
|
fn create_primitive_query(
|
|
|
|
query: TokenStream,
|
|
|
|
stop_words: Option<Set<&[u8]>>,
|
|
|
|
words_limit: Option<usize>,
|
|
|
|
) -> PrimitiveQuery {
|
2021-03-03 12:03:31 +01:00
|
|
|
let mut primitive_query = Vec::new();
|
|
|
|
let mut phrase = Vec::new();
|
|
|
|
let mut quoted = false;
|
|
|
|
|
2021-04-13 19:10:58 +02:00
|
|
|
let parts_limit = words_limit.unwrap_or(usize::MAX);
|
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
let mut peekable = query.peekable();
|
|
|
|
while let Some(token) = peekable.next() {
|
2021-04-13 19:10:58 +02:00
|
|
|
// early return if word limit is exceeded
|
2021-06-16 18:33:33 +02:00
|
|
|
if primitive_query.len() >= parts_limit {
|
|
|
|
return primitive_query;
|
|
|
|
}
|
2021-04-13 19:10:58 +02:00
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
match token.kind {
|
2021-06-16 18:33:33 +02:00
|
|
|
TokenKind::Word | TokenKind::StopWord => {
|
2021-03-03 12:03:31 +01:00
|
|
|
// 1. if the word is quoted we push it in a phrase-buffer waiting for the ending quote,
|
2021-04-08 21:21:20 +02:00
|
|
|
// 2. if the word is not the last token of the query and is not a stop_word we push it as a non-prefix word,
|
2021-03-03 12:03:31 +01:00
|
|
|
// 3. if the word is the last token of the query we push it as a prefix word.
|
|
|
|
if quoted {
|
|
|
|
phrase.push(token.word.to_string());
|
|
|
|
} else if peekable.peek().is_some() {
|
2021-06-16 18:33:33 +02:00
|
|
|
if !stop_words
|
|
|
|
.as_ref()
|
|
|
|
.map_or(false, |swords| swords.contains(token.word.as_ref()))
|
|
|
|
{
|
|
|
|
primitive_query
|
|
|
|
.push(PrimitiveQueryPart::Word(token.word.to_string(), false));
|
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
} else {
|
|
|
|
primitive_query.push(PrimitiveQueryPart::Word(token.word.to_string(), true));
|
|
|
|
}
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-06-08 17:29:38 +02:00
|
|
|
TokenKind::Separator(separator_kind) => {
|
2021-03-03 12:03:31 +01:00
|
|
|
let quote_count = token.word.chars().filter(|&s| s == '"').count();
|
|
|
|
// swap quoted state if we encounter a double quote
|
|
|
|
if quote_count % 2 != 0 {
|
|
|
|
quoted = !quoted;
|
|
|
|
}
|
2021-06-08 17:52:37 +02:00
|
|
|
// if there is a quote or a hard separator we close the phrase.
|
2021-06-16 18:33:33 +02:00
|
|
|
if !phrase.is_empty() && (quote_count > 0 || separator_kind == SeparatorKind::Hard)
|
|
|
|
{
|
2021-03-03 12:03:31 +01:00
|
|
|
primitive_query.push(PrimitiveQueryPart::Phrase(mem::take(&mut phrase)));
|
|
|
|
}
|
2021-06-16 18:33:33 +02:00
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
_ => (),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If a quote is never closed, we consider all of the end of the query as a phrase.
|
|
|
|
if !phrase.is_empty() {
|
|
|
|
primitive_query.push(PrimitiveQueryPart::Phrase(mem::take(&mut phrase)));
|
|
|
|
}
|
|
|
|
|
|
|
|
primitive_query
|
|
|
|
}
|
|
|
|
|
2021-02-23 15:50:33 +01:00
|
|
|
/// Returns the maximum number of typos that this Operation allows.
|
|
|
|
pub fn maximum_typo(operation: &Operation) -> usize {
|
2021-06-16 18:33:33 +02:00
|
|
|
use Operation::{And, Or, Phrase, Query};
|
2021-02-23 15:50:33 +01:00
|
|
|
match operation {
|
|
|
|
Or(_, ops) => ops.iter().map(maximum_typo).max().unwrap_or(0),
|
2021-06-09 17:28:12 +02:00
|
|
|
And(ops) => ops.iter().map(maximum_typo).sum::<usize>(),
|
2021-02-23 15:50:33 +01:00
|
|
|
Query(q) => q.kind.typo() as usize,
|
2021-06-09 17:28:12 +02:00
|
|
|
// no typo allowed in phrases
|
|
|
|
Phrase(_) => 0,
|
2021-02-23 15:50:33 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-02-23 15:53:24 +01:00
|
|
|
/// Returns the maximum proximity that this Operation allows.
|
|
|
|
pub fn maximum_proximity(operation: &Operation) -> usize {
|
2021-06-16 18:33:33 +02:00
|
|
|
use Operation::{And, Or, Phrase, Query};
|
2021-02-23 15:53:24 +01:00
|
|
|
match operation {
|
|
|
|
Or(_, ops) => ops.iter().map(maximum_proximity).max().unwrap_or(0),
|
2021-02-24 15:36:57 +01:00
|
|
|
And(ops) => {
|
2021-06-16 18:33:33 +02:00
|
|
|
ops.iter().map(maximum_proximity).sum::<usize>() + ops.len().saturating_sub(1) * 7
|
|
|
|
}
|
2021-06-09 17:28:12 +02:00
|
|
|
Query(_) | Phrase(_) => 0,
|
2021-02-23 15:53:24 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
#[cfg(test)]
|
|
|
|
mod test {
|
2021-03-02 11:30:48 +01:00
|
|
|
use std::collections::HashMap;
|
|
|
|
|
2021-06-01 11:48:56 +02:00
|
|
|
use maplit::hashmap;
|
2021-03-03 12:03:31 +01:00
|
|
|
use meilisearch_tokenizer::{Analyzer, AnalyzerConfig};
|
2021-06-16 18:33:33 +02:00
|
|
|
use rand::rngs::StdRng;
|
|
|
|
use rand::{Rng, SeedableRng};
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
use super::*;
|
2021-03-02 11:30:48 +01:00
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
#[derive(Debug)]
|
|
|
|
struct TestContext {
|
|
|
|
synonyms: HashMap<Vec<String>, Vec<Vec<String>>>,
|
|
|
|
postings: HashMap<String, RoaringBitmap>,
|
|
|
|
}
|
|
|
|
|
|
|
|
impl TestContext {
|
|
|
|
fn build(
|
|
|
|
&self,
|
|
|
|
optional_words: bool,
|
|
|
|
authorize_typos: bool,
|
2021-04-13 19:10:58 +02:00
|
|
|
words_limit: Option<usize>,
|
2021-03-03 12:03:31 +01:00
|
|
|
query: TokenStream,
|
2021-06-16 18:33:33 +02:00
|
|
|
) -> Result<Option<(Operation, PrimitiveQuery)>> {
|
2021-04-13 19:10:58 +02:00
|
|
|
let primitive_query = create_primitive_query(query, None, words_limit);
|
2021-03-03 12:03:31 +01:00
|
|
|
if !primitive_query.is_empty() {
|
2021-06-16 18:33:33 +02:00
|
|
|
let qt =
|
|
|
|
create_query_tree(self, optional_words, authorize_typos, &primitive_query)?;
|
2021-05-04 13:44:55 +02:00
|
|
|
Ok(Some((qt, primitive_query)))
|
2021-03-03 12:03:31 +01:00
|
|
|
} else {
|
|
|
|
Ok(None)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
impl Context for TestContext {
|
|
|
|
fn word_docids(&self, word: &str) -> heed::Result<Option<RoaringBitmap>> {
|
|
|
|
Ok(self.postings.get(word).cloned())
|
|
|
|
}
|
|
|
|
|
2021-04-09 21:56:20 +02:00
|
|
|
fn synonyms<S: AsRef<str>>(&self, words: &[S]) -> heed::Result<Option<Vec<Vec<String>>>> {
|
|
|
|
let words: Vec<_> = words.iter().map(|s| s.as_ref().to_owned()).collect();
|
2021-03-03 12:03:31 +01:00
|
|
|
Ok(self.synonyms.get(&words).cloned())
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
impl Default for TestContext {
|
|
|
|
fn default() -> TestContext {
|
|
|
|
let mut rng = StdRng::seed_from_u64(102);
|
|
|
|
let rng = &mut rng;
|
|
|
|
|
|
|
|
fn random_postings<R: Rng>(rng: &mut R, len: usize) -> RoaringBitmap {
|
|
|
|
let mut values = Vec::<u32>::with_capacity(len);
|
|
|
|
while values.len() != len {
|
|
|
|
values.push(rng.gen());
|
|
|
|
}
|
|
|
|
values.sort_unstable();
|
|
|
|
|
|
|
|
RoaringBitmap::from_sorted_iter(values.into_iter())
|
|
|
|
}
|
|
|
|
|
|
|
|
TestContext {
|
2021-06-16 18:33:33 +02:00
|
|
|
synonyms: hashmap! {
|
2021-03-03 12:03:31 +01:00
|
|
|
vec![String::from("hello")] => vec![
|
|
|
|
vec![String::from("hi")],
|
|
|
|
vec![String::from("good"), String::from("morning")],
|
|
|
|
],
|
|
|
|
vec![String::from("world")] => vec![
|
|
|
|
vec![String::from("earth")],
|
|
|
|
vec![String::from("nature")],
|
|
|
|
],
|
|
|
|
// new york city
|
|
|
|
vec![String::from("nyc")] => vec![
|
|
|
|
vec![String::from("new"), String::from("york")],
|
|
|
|
vec![String::from("new"), String::from("york"), String::from("city")],
|
|
|
|
],
|
|
|
|
vec![String::from("new"), String::from("york")] => vec![
|
|
|
|
vec![String::from("nyc")],
|
|
|
|
vec![String::from("new"), String::from("york"), String::from("city")],
|
|
|
|
],
|
|
|
|
vec![String::from("new"), String::from("york"), String::from("city")] => vec![
|
|
|
|
vec![String::from("nyc")],
|
|
|
|
vec![String::from("new"), String::from("york")],
|
|
|
|
],
|
|
|
|
},
|
2021-06-16 18:33:33 +02:00
|
|
|
postings: hashmap! {
|
2021-03-03 12:03:31 +01:00
|
|
|
String::from("hello") => random_postings(rng, 1500),
|
|
|
|
String::from("hi") => random_postings(rng, 4000),
|
|
|
|
String::from("word") => random_postings(rng, 2500),
|
|
|
|
String::from("split") => random_postings(rng, 400),
|
|
|
|
String::from("ngrams") => random_postings(rng, 1400),
|
|
|
|
String::from("world") => random_postings(rng, 15_000),
|
|
|
|
String::from("earth") => random_postings(rng, 8000),
|
|
|
|
String::from("2021") => random_postings(rng, 100),
|
|
|
|
String::from("2020") => random_postings(rng, 500),
|
|
|
|
String::from("is") => random_postings(rng, 50_000),
|
|
|
|
String::from("this") => random_postings(rng, 50_000),
|
|
|
|
String::from("good") => random_postings(rng, 1250),
|
|
|
|
String::from("morning") => random_postings(rng, 125),
|
|
|
|
},
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn prefix() {
|
|
|
|
let query = "hey friends";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: true,
|
|
|
|
kind: QueryKind::tolerant(1, "friends".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: true,
|
|
|
|
kind: QueryKind::tolerant(2, "heyfriends".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn no_prefix() {
|
|
|
|
let query = "hey friends ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "friends".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "heyfriends".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn synonyms() {
|
|
|
|
let query = "hello world ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hi".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("good".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("morning".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "hello".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("earth".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("nature".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "world".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "helloworld".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn complex_synonyms() {
|
|
|
|
let query = "new york city ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("new".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("york".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("city".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "yorkcity".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
|
|
|
Operation::And(vec![
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("nyc".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("new".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("york".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("city".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "newyork".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("city".to_string()),
|
|
|
|
}),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("nyc".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("new".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("york".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "newyorkcity".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn ngrams() {
|
|
|
|
let query = "n grams ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("n".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "grams".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "ngrams".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn word_split() {
|
|
|
|
let query = "wordsplit fish ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::Phrase(vec!["word".to_string(), "split".to_string()]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "wordsplit".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("fish".to_string()),
|
|
|
|
}),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "wordsplitfish".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
2021-03-03 12:03:31 +01:00
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn phrase() {
|
|
|
|
let query = "\"hey friends\" \" \" \"wooop";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
|
|
|
let expected = Operation::And(vec![
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Phrase(vec!["hey".to_string(), "friends".to_string()]),
|
2021-03-03 12:03:31 +01:00
|
|
|
Operation::Query(Query { prefix: false, kind: QueryKind::exact("wooop".to_string()) }),
|
|
|
|
]);
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
2021-06-08 17:29:38 +02:00
|
|
|
#[test]
|
|
|
|
fn phrase_with_hard_separator() {
|
|
|
|
let query = "\"hey friends. wooop wooop\"";
|
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
|
|
|
let expected = Operation::And(vec![
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Phrase(vec!["hey".to_string(), "friends".to_string()]),
|
|
|
|
Operation::Phrase(vec!["wooop".to_string(), "wooop".to_string()]),
|
2021-06-08 17:29:38 +02:00
|
|
|
]);
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, true, None, tokens).unwrap().unwrap();
|
2021-06-08 17:29:38 +02:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
2021-03-03 12:03:31 +01:00
|
|
|
#[test]
|
|
|
|
fn optional_word() {
|
|
|
|
let query = "hey my friend ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
true,
|
|
|
|
vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
2021-03-03 12:03:31 +01:00
|
|
|
Operation::And(vec![
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("my".to_string()),
|
|
|
|
}),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "heymy".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("my".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "friend".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "myfriend".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
]),
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "heymy".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "friend".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(2, "heymyfriend".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
],
|
|
|
|
);
|
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(true, true, None, tokens).unwrap().unwrap();
|
2021-02-18 22:18:36 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn optional_word_phrase() {
|
|
|
|
let query = "\"hey my\"";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-02-18 22:18:36 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Phrase(vec!["hey".to_string(), "my".to_string()]);
|
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(true, true, None, tokens).unwrap().unwrap();
|
2021-02-18 22:18:36 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn optional_word_multiple_phrases() {
|
|
|
|
let query = r#""hey" my good "friend""#;
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-02-18 22:18:36 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
true,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("friend".to_string()),
|
|
|
|
}),
|
2021-03-03 12:03:31 +01:00
|
|
|
]),
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("my".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("friend".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("my".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("good".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::tolerant(1, "mygood".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("friend".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
],
|
|
|
|
);
|
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(true, true, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
|
|
|
#[test]
|
|
|
|
fn no_typo() {
|
|
|
|
let query = "hey friends ";
|
2021-03-29 19:15:47 +02:00
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
2021-03-03 12:03:31 +01:00
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let expected = Operation::Or(
|
|
|
|
false,
|
|
|
|
vec![
|
|
|
|
Operation::And(vec![
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("hey".to_string()),
|
|
|
|
}),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("friends".to_string()),
|
|
|
|
}),
|
|
|
|
]),
|
|
|
|
Operation::Query(Query {
|
|
|
|
prefix: false,
|
|
|
|
kind: QueryKind::exact("heyfriends".to_string()),
|
|
|
|
}),
|
|
|
|
],
|
|
|
|
);
|
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, false, None, tokens).unwrap().unwrap();
|
2021-03-03 12:03:31 +01:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
|
|
|
|
2021-04-13 19:10:58 +02:00
|
|
|
#[test]
|
|
|
|
fn words_limit() {
|
|
|
|
let query = "\"hey my\" good friend";
|
|
|
|
let analyzer = Analyzer::new(AnalyzerConfig::<Vec<u8>>::default());
|
|
|
|
let result = analyzer.analyze(query);
|
|
|
|
let tokens = result.tokens();
|
|
|
|
|
|
|
|
let expected = Operation::And(vec![
|
2021-06-16 18:33:33 +02:00
|
|
|
Operation::Phrase(vec!["hey".to_string(), "my".to_string()]),
|
2021-04-13 19:10:58 +02:00
|
|
|
Operation::Query(Query { prefix: false, kind: QueryKind::exact("good".to_string()) }),
|
|
|
|
]);
|
|
|
|
|
2021-06-16 18:33:33 +02:00
|
|
|
let (query_tree, _) =
|
|
|
|
TestContext::default().build(false, false, Some(2), tokens).unwrap().unwrap();
|
2021-04-13 19:10:58 +02:00
|
|
|
|
|
|
|
assert_eq!(expected, query_tree);
|
|
|
|
}
|
2021-03-03 12:03:31 +01:00
|
|
|
}
|