Rewrite the synonym endpoint

This commit is contained in:
Quentin de Quelen 2019-12-11 16:49:09 +01:00
parent cc10804607
commit a4f26e8e48
7 changed files with 38 additions and 422 deletions

View File

@ -179,16 +179,8 @@ impl Index {
update::push_clear_all(writer, self.updates, self.updates_results)
}
pub fn synonyms_addition(&self) -> update::SynonymsAddition {
update::SynonymsAddition::new(
self.updates,
self.updates_results,
self.updates_notifier.clone(),
)
}
pub fn synonyms_deletion(&self) -> update::SynonymsDeletion {
update::SynonymsDeletion::new(
pub fn synonyms_update(&self) -> update::SynonymsUpdate {
update::SynonymsUpdate::new(
self.updates,
self.updates_results,
self.updates_notifier.clone(),

View File

@ -5,8 +5,7 @@ mod documents_deletion;
mod schema_update;
mod stop_words_addition;
mod stop_words_deletion;
mod synonyms_addition;
mod synonyms_deletion;
mod synonyms_update;
pub use self::clear_all::{apply_clear_all, push_clear_all};
pub use self::customs_update::{apply_customs_update, push_customs_update};
@ -17,8 +16,7 @@ pub use self::documents_deletion::{apply_documents_deletion, DocumentsDeletion};
pub use self::schema_update::{apply_schema_update, push_schema_update};
pub use self::stop_words_addition::{apply_stop_words_addition, StopWordsAddition};
pub use self::stop_words_deletion::{apply_stop_words_deletion, StopWordsDeletion};
pub use self::synonyms_addition::{apply_synonyms_addition, SynonymsAddition};
pub use self::synonyms_deletion::{apply_synonyms_deletion, SynonymsDeletion};
pub use self::synonyms_update::{apply_synonyms_update, SynonymsUpdate};
use std::cmp;
use std::collections::{BTreeMap, BTreeSet, HashMap};
@ -82,16 +80,9 @@ impl Update {
}
}
fn synonyms_addition(data: BTreeMap<String, Vec<String>>) -> Update {
fn synonyms_update(data: BTreeMap<String, Vec<String>>) -> Update {
Update {
data: UpdateData::SynonymsAddition(data),
enqueued_at: Utc::now(),
}
}
fn synonyms_deletion(data: BTreeMap<String, Option<Vec<String>>>) -> Update {
Update {
data: UpdateData::SynonymsDeletion(data),
data: UpdateData::SynonymsUpdate(data),
enqueued_at: Utc::now(),
}
}
@ -119,8 +110,7 @@ pub enum UpdateData {
DocumentsAddition(Vec<HashMap<String, serde_json::Value>>),
DocumentsPartial(Vec<HashMap<String, serde_json::Value>>),
DocumentsDeletion(Vec<DocumentId>),
SynonymsAddition(BTreeMap<String, Vec<String>>),
SynonymsDeletion(BTreeMap<String, Option<Vec<String>>>),
SynonymsUpdate(BTreeMap<String, Vec<String>>),
StopWordsAddition(BTreeSet<String>),
StopWordsDeletion(BTreeSet<String>),
}
@ -140,12 +130,9 @@ impl UpdateData {
UpdateData::DocumentsDeletion(deletion) => UpdateType::DocumentsDeletion {
number: deletion.len(),
},
UpdateData::SynonymsAddition(addition) => UpdateType::SynonymsAddition {
UpdateData::SynonymsUpdate(addition) => UpdateType::SynonymsUpdate {
number: addition.len(),
},
UpdateData::SynonymsDeletion(deletion) => UpdateType::SynonymsDeletion {
number: deletion.len(),
},
UpdateData::StopWordsAddition(addition) => UpdateType::StopWordsAddition {
number: addition.len(),
},
@ -165,8 +152,7 @@ pub enum UpdateType {
DocumentsAddition { number: usize },
DocumentsPartial { number: usize },
DocumentsDeletion { number: usize },
SynonymsAddition { number: usize },
SynonymsDeletion { number: usize },
SynonymsUpdate { number: usize },
StopWordsAddition { number: usize },
StopWordsDeletion { number: usize },
}
@ -361,25 +347,14 @@ pub fn update_task<'a, 'b>(
(update_type, result, start.elapsed())
}
UpdateData::SynonymsAddition(synonyms) => {
UpdateData::SynonymsUpdate(synonyms) => {
let start = Instant::now();
let update_type = UpdateType::SynonymsAddition {
let update_type = UpdateType::SynonymsUpdate {
number: synonyms.len(),
};
let result = apply_synonyms_addition(writer, index.main, index.synonyms, synonyms);
(update_type, result, start.elapsed())
}
UpdateData::SynonymsDeletion(synonyms) => {
let start = Instant::now();
let update_type = UpdateType::SynonymsDeletion {
number: synonyms.len(),
};
let result = apply_synonyms_deletion(writer, index.main, index.synonyms, synonyms);
let result = apply_synonyms_update(writer, index.main, index.synonyms, synonyms);
(update_type, result, start.elapsed())
}

View File

@ -1,158 +0,0 @@
use std::collections::BTreeMap;
use std::iter::FromIterator;
use fst::{set::OpBuilder, SetBuilder};
use sdset::SetBuf;
use crate::database::{MainT, UpdateT};
use crate::automaton::normalize_str;
use crate::database::{UpdateEvent, UpdateEventsEmitter};
use crate::update::{next_update_id, Update};
use crate::{store, MResult};
pub struct SynonymsDeletion {
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
updates_notifier: UpdateEventsEmitter,
synonyms: BTreeMap<String, Option<Vec<String>>>,
}
impl SynonymsDeletion {
pub fn new(
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
updates_notifier: UpdateEventsEmitter,
) -> SynonymsDeletion {
SynonymsDeletion {
updates_store,
updates_results_store,
updates_notifier,
synonyms: BTreeMap::new(),
}
}
pub fn delete_all_alternatives_of<S: AsRef<str>>(&mut self, synonym: S) {
let synonym = normalize_str(synonym.as_ref());
self.synonyms.insert(synonym, None);
}
pub fn delete_specific_alternatives_of<S, T, I>(&mut self, synonym: S, alternatives: I)
where
S: AsRef<str>,
T: AsRef<str>,
I: Iterator<Item = T>,
{
let synonym = normalize_str(synonym.as_ref());
let value = self.synonyms.entry(synonym).or_insert(None);
let alternatives = alternatives.map(|s| s.as_ref().to_lowercase());
match value {
Some(v) => v.extend(alternatives),
None => *value = Some(Vec::from_iter(alternatives)),
}
}
pub fn finalize(self, writer: &mut heed::RwTxn<UpdateT>) -> MResult<u64> {
let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
let update_id = push_synonyms_deletion(
writer,
self.updates_store,
self.updates_results_store,
self.synonyms,
)?;
Ok(update_id)
}
}
pub fn push_synonyms_deletion(
writer: &mut heed::RwTxn<UpdateT>,
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
deletion: BTreeMap<String, Option<Vec<String>>>,
) -> MResult<u64> {
let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
let update = Update::synonyms_deletion(deletion);
updates_store.put_update(writer, last_update_id, &update)?;
Ok(last_update_id)
}
pub fn apply_synonyms_deletion(
writer: &mut heed::RwTxn<MainT>,
main_store: store::Main,
synonyms_store: store::Synonyms,
deletion: BTreeMap<String, Option<Vec<String>>>,
) -> MResult<()> {
let mut delete_whole_synonym_builder = SetBuilder::memory();
for (synonym, alternatives) in deletion {
match alternatives {
Some(alternatives) => {
let prev_alternatives = synonyms_store.synonyms(writer, synonym.as_bytes())?;
let prev_alternatives = match prev_alternatives {
Some(alternatives) => alternatives,
None => continue,
};
let delta_alternatives = {
let alternatives = SetBuf::from_dirty(alternatives);
let mut builder = SetBuilder::memory();
builder.extend_iter(alternatives).unwrap();
builder.into_inner().and_then(fst::Set::from_bytes).unwrap()
};
let op = OpBuilder::new()
.add(prev_alternatives.stream())
.add(delta_alternatives.stream())
.difference();
let (alternatives, empty_alternatives) = {
let mut builder = SetBuilder::memory();
let len = builder.get_ref().len();
builder.extend_stream(op).unwrap();
let is_empty = len == builder.get_ref().len();
let bytes = builder.into_inner().unwrap();
let alternatives = fst::Set::from_bytes(bytes).unwrap();
(alternatives, is_empty)
};
if empty_alternatives {
delete_whole_synonym_builder.insert(synonym.as_bytes())?;
} else {
synonyms_store.put_synonyms(writer, synonym.as_bytes(), &alternatives)?;
}
}
None => {
delete_whole_synonym_builder.insert(&synonym).unwrap();
synonyms_store.del_synonyms(writer, synonym.as_bytes())?;
}
}
}
let delta_synonyms = delete_whole_synonym_builder
.into_inner()
.and_then(fst::Set::from_bytes)
.unwrap();
let synonyms = match main_store.synonyms_fst(writer)? {
Some(synonyms) => {
let op = OpBuilder::new()
.add(synonyms.stream())
.add(delta_synonyms.stream())
.difference();
let mut synonyms_builder = SetBuilder::memory();
synonyms_builder.extend_stream(op).unwrap();
synonyms_builder
.into_inner()
.and_then(fst::Set::from_bytes)
.unwrap()
}
None => fst::Set::default(),
};
main_store.put_synonyms_fst(writer, &synonyms)?;
Ok(())
}

View File

@ -1,6 +1,6 @@
use std::collections::BTreeMap;
use fst::{set::OpBuilder, SetBuilder};
use fst::SetBuilder;
use sdset::SetBuf;
use crate::database::{MainT, UpdateT};
@ -9,20 +9,20 @@ use crate::database::{UpdateEvent, UpdateEventsEmitter};
use crate::update::{next_update_id, Update};
use crate::{store, MResult};
pub struct SynonymsAddition {
pub struct SynonymsUpdate {
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
updates_notifier: UpdateEventsEmitter,
synonyms: BTreeMap<String, Vec<String>>,
}
impl SynonymsAddition {
impl SynonymsUpdate {
pub fn new(
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
updates_notifier: UpdateEventsEmitter,
) -> SynonymsAddition {
SynonymsAddition {
) -> SynonymsUpdate {
SynonymsUpdate {
updates_store,
updates_results_store,
updates_notifier,
@ -46,7 +46,7 @@ impl SynonymsAddition {
pub fn finalize(self, writer: &mut heed::RwTxn<UpdateT>) -> MResult<u64> {
let _ = self.updates_notifier.send(UpdateEvent::NewUpdate);
let update_id = push_synonyms_addition(
let update_id = push_synonyms_update(
writer,
self.updates_store,
self.updates_results_store,
@ -56,7 +56,7 @@ impl SynonymsAddition {
}
}
pub fn push_synonyms_addition(
pub fn push_synonyms_update(
writer: &mut heed::RwTxn<UpdateT>,
updates_store: store::Updates,
updates_results_store: store::UpdatesResults,
@ -64,20 +64,20 @@ pub fn push_synonyms_addition(
) -> MResult<u64> {
let last_update_id = next_update_id(writer, updates_store, updates_results_store)?;
let update = Update::synonyms_addition(addition);
let update = Update::synonyms_update(addition);
updates_store.put_update(writer, last_update_id, &update)?;
Ok(last_update_id)
}
pub fn apply_synonyms_addition(
pub fn apply_synonyms_update(
writer: &mut heed::RwTxn<MainT>,
main_store: store::Main,
synonyms_store: store::Synonyms,
addition: BTreeMap<String, Vec<String>>,
) -> MResult<()> {
let mut synonyms_builder = SetBuilder::memory();
synonyms_store.clear(writer)?;
for (word, alternatives) in addition {
synonyms_builder.insert(&word).unwrap();
@ -92,28 +92,11 @@ pub fn apply_synonyms_addition(
synonyms_store.put_synonyms(writer, word.as_bytes(), &alternatives)?;
}
let delta_synonyms = synonyms_builder
let synonyms = synonyms_builder
.into_inner()
.and_then(fst::Set::from_bytes)
.unwrap();
let synonyms = match main_store.synonyms_fst(writer)? {
Some(synonyms) => {
let op = OpBuilder::new()
.add(synonyms.stream())
.add(delta_synonyms.stream())
.r#union();
let mut synonyms_builder = SetBuilder::memory();
synonyms_builder.extend_stream(op).unwrap();
synonyms_builder
.into_inner()
.and_then(fst::Set::from_bytes)
.unwrap()
}
None => delta_synonyms,
};
main_store.put_synonyms_fst(writer, &synonyms)?;
Ok(())

View File

@ -6,7 +6,7 @@ pub enum UpdateOperation {
ClearAllDocuments,
DocumentsAddition,
DocumentsDeletion,
SynonymsAddition,
SynonymsUpdate,
SynonymsDeletion,
StopWordsAddition,
StopWordsDeletion,
@ -22,7 +22,7 @@ impl fmt::Display for UpdateOperation {
ClearAllDocuments => write!(f, "ClearAllDocuments"),
DocumentsAddition => write!(f, "DocumentsAddition"),
DocumentsDeletion => write!(f, "DocumentsDeletion"),
SynonymsAddition => write!(f, "SynonymsAddition"),
SynonymsUpdate => write!(f, "SynonymsUpdate"),
SynonymsDeletion => write!(f, "SynonymsDelettion"),
StopWordsAddition => write!(f, "StopWordsAddition"),
StopWordsDeletion => write!(f, "StopWordsDeletion"),

View File

@ -60,21 +60,9 @@ pub fn load_routes(app: &mut tide::App<Data>) {
.post(document::delete_multiple_documents);
});
router.at("/synonyms").nest(|router| {
router
.at("/")
.get(synonym::list)
.post(synonym::create)
.delete(synonym::clear);
router
.at("/:synonym")
router.at("/synonyms")
.get(synonym::get)
.put(synonym::update)
.delete(synonym::delete);
router.at("/batch").post(synonym::batch_write);
});
.post(synonym::update);
router.at("/stop-words").nest(|router| {
router

View File

@ -1,9 +1,9 @@
use std::collections::HashMap;
use http::StatusCode;
use serde::{Deserialize, Serialize};
use tide::response::IntoResponse;
use tide::{Context, Response};
use indexmap::IndexMap;
use crate::error::{ResponseError, SResult};
use crate::helpers::tide::ContextExt;
@ -11,23 +11,7 @@ use crate::models::token::ACL::*;
use crate::routes::document::IndexUpdateResponse;
use crate::Data;
#[derive(Clone, Serialize, Deserialize)]
#[serde(untagged)]
pub enum Synonym {
OneWay(SynonymOneWay),
MultiWay { synonyms: Vec<String> },
}
#[derive(Clone, Serialize, Deserialize)]
#[serde(rename_all = "camelCase", deny_unknown_fields)]
pub struct SynonymOneWay {
pub input: String,
pub synonyms: Vec<String>,
}
pub type Synonyms = Vec<Synonym>;
pub async fn list(ctx: Context<Data>) -> SResult<Response> {
pub async fn get(ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsRead)?;
let index = ctx.index()?;
@ -42,7 +26,7 @@ pub async fn list(ctx: Context<Data>) -> SResult<Response> {
let synonyms_fst = synonyms_fst.unwrap_or_default();
let synonyms_list = synonyms_fst.stream().into_strs().map_err(ResponseError::internal)?;
let mut response = HashMap::new();
let mut response = IndexMap::new();
let index_synonyms = &index.synonyms;
@ -60,171 +44,23 @@ pub async fn list(ctx: Context<Data>) -> SResult<Response> {
Ok(tide::response::json(response))
}
pub async fn get(ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsRead)?;
let synonym = ctx.url_param("synonym")?;
let index = ctx.index()?;
let db = &ctx.state().db;
let reader = db.main_read_txn().map_err(ResponseError::internal)?;
let synonym_list = index
.synonyms
.synonyms(&reader, synonym.as_bytes())
.map_err(ResponseError::internal)?;
let list = match synonym_list {
Some(list) => list.stream().into_strs().map_err(ResponseError::internal)?,
None => Vec::new(),
};
Ok(tide::response::json(list))
}
pub async fn create(mut ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsWrite)?;
let data: Synonym = ctx.body_json().await.map_err(ResponseError::bad_request)?;
let index = ctx.index()?;
let db = &ctx.state().db;
let mut writer = db.update_write_txn().map_err(ResponseError::internal)?;
let mut synonyms_addition = index.synonyms_addition();
match data.clone() {
Synonym::OneWay(content) => {
synonyms_addition.add_synonym(content.input, content.synonyms.into_iter())
}
Synonym::MultiWay { mut synonyms } => {
if synonyms.len() > 1 {
for _ in 0..synonyms.len() {
let (first, elems) = synonyms.split_first().unwrap();
synonyms_addition.add_synonym(first, elems.iter());
synonyms.rotate_left(1);
}
}
}
}
let update_id = synonyms_addition
.finalize(&mut writer)
.map_err(ResponseError::internal)?;
writer.commit().map_err(ResponseError::internal)?;
let response_body = IndexUpdateResponse { update_id };
Ok(tide::response::json(response_body)
.with_status(StatusCode::ACCEPTED)
.into_response())
}
pub async fn update(mut ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsWrite)?;
let synonym = ctx.url_param("synonym")?;
let data: HashMap<String, Vec<String>> = ctx.body_json().await.map_err(ResponseError::bad_request)?;
let index = ctx.index()?;
let data: Vec<String> = ctx.body_json().await.map_err(ResponseError::bad_request)?;
let db = &ctx.state().db;
let mut writer = db.update_write_txn().map_err(ResponseError::internal)?;
let mut synonyms_addition = index.synonyms_addition();
synonyms_addition.add_synonym(synonym.clone(), data.clone().into_iter());
let update_id = synonyms_addition
.finalize(&mut writer)
.map_err(ResponseError::internal)?;
writer.commit().map_err(ResponseError::internal)?;
let response_body = IndexUpdateResponse { update_id };
Ok(tide::response::json(response_body)
.with_status(StatusCode::ACCEPTED)
.into_response())
}
pub async fn delete(ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsWrite)?;
let synonym = ctx.url_param("synonym")?;
let index = ctx.index()?;
let db = &ctx.state().db;
let mut writer = db.update_write_txn().map_err(ResponseError::internal)?;
let mut synonyms_deletion = index.synonyms_deletion();
synonyms_deletion.delete_all_alternatives_of(synonym);
let update_id = synonyms_deletion
.finalize(&mut writer)
.map_err(ResponseError::internal)?;
writer.commit().map_err(ResponseError::internal)?;
let response_body = IndexUpdateResponse { update_id };
Ok(tide::response::json(response_body)
.with_status(StatusCode::ACCEPTED)
.into_response())
}
pub async fn batch_write(mut ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsWrite)?;
let data: Synonyms = ctx.body_json().await.map_err(ResponseError::bad_request)?;
let index = ctx.index()?;
let db = &ctx.state().db;
let mut writer = db.update_write_txn().map_err(ResponseError::internal)?;
let mut synonyms_addition = index.synonyms_addition();
for raw in data {
match raw {
Synonym::OneWay(content) => {
synonyms_addition.add_synonym(content.input, content.synonyms.into_iter())
}
Synonym::MultiWay { mut synonyms } => {
if synonyms.len() > 1 {
for _ in 0..synonyms.len() {
let (first, elems) = synonyms.split_first().unwrap();
synonyms_addition.add_synonym(first, elems.iter());
synonyms.rotate_left(1);
}
}
}
}
}
let update_id = synonyms_addition
.finalize(&mut writer)
.map_err(ResponseError::internal)?;
writer.commit().map_err(ResponseError::internal)?;
let response_body = IndexUpdateResponse { update_id };
Ok(tide::response::json(response_body)
.with_status(StatusCode::ACCEPTED)
.into_response())
}
pub async fn clear(ctx: Context<Data>) -> SResult<Response> {
ctx.is_allowed(SettingsWrite)?;
let index = ctx.index()?;
let db = &ctx.state().db;
let reader = db.main_read_txn().map_err(ResponseError::internal)?;
let mut writer = db.update_write_txn().map_err(ResponseError::internal)?;
let synonyms_fst = index
.main
.synonyms_fst(&reader)
.map_err(ResponseError::internal)?;
let synonyms_fst = synonyms_fst.unwrap_or_default();
let synonyms_list = synonyms_fst.stream().into_strs().map_err(ResponseError::internal)?;
let mut synonyms_deletion = index.synonyms_deletion();
for synonym in synonyms_list {
synonyms_deletion.delete_all_alternatives_of(synonym);
}
let update_id = synonyms_deletion
let mut synonyms_update = index.synonyms_update();
for (input, synonyms) in data {
synonyms_update.add_synonym(input, synonyms.into_iter());
}
let update_id = synonyms_update
.finalize(&mut writer)
.map_err(ResponseError::internal)?;