refactor: CreateVersion and run_threaded

This commit is contained in:
Elijah McMorris 2024-08-27 18:59:25 -07:00
parent 17a115f473
commit b885315cd7
Signed by: NexVeridian
SSH key fingerprint: SHA256:bsA1SKZxuEcEVHAy3gY1HUeM5ykRJl0U0kQHQn0hMg8
5 changed files with 210 additions and 235 deletions

View file

@ -51,7 +51,7 @@ impl File_Format {
}
}
pub async fn create_db_entity(db: &Surreal<impl Connection>, line: &str) -> Result<(), Error> {
pub async fn create_entity(db: &Surreal<impl Connection>, line: &str) -> Result<(), Error> {
let line = line.trim().trim_end_matches(',').to_string();
if line == "[" || line == "]" {
return Ok(());
@ -76,105 +76,6 @@ pub async fn create_db_entity(db: &Surreal<impl Connection>, line: &str) -> Resu
Ok(())
}
pub async fn create_db_entities(
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
) -> Result<(), Error> {
let mut counter = 0;
for line in lines {
create_db_entity(db, line).await?;
counter += 1;
if counter % 100 == 0 {
if let Some(ref p) = pb {
p.inc(100)
}
}
}
Ok(())
}
pub async fn create_db_entities_bulk(
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
batch_size: usize,
) -> Result<(), Error> {
let lines = lines
.iter()
.map(|line| line.trim().trim_end_matches(',').to_string())
.filter(|line| line != "[" && line != "]")
.collect::<Vec<String>>();
let mut data_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
let mut claims_vec: Vec<Claims> = Vec::with_capacity(batch_size);
let mut property_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
let mut lexeme_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
for line in lines {
let json: Value = from_str(&line).expect("Failed to parse JSON");
let data = Entity::from_json(json).expect("Failed to parse JSON");
let (claims, data) = EntityMini::from_entity(data);
match data.id.clone().expect("No ID").tb.as_str() {
"Property" => property_vec.push(data),
"Lexeme" => lexeme_vec.push(data),
"Entity" => data_vec.push(data),
_ => panic!("Unknown table"),
}
claims_vec.push(claims);
}
db.insert::<Vec<EntityMini>>("Entity")
.content(data_vec)
.await?;
db.insert::<Vec<Claims>>("Claims")
.content(claims_vec)
.await?;
db.insert::<Vec<EntityMini>>("Property")
.content(property_vec)
.await?;
db.insert::<Vec<EntityMini>>("Lexeme")
.content(lexeme_vec)
.await?;
if let Some(ref p) = pb {
p.inc(batch_size as u64)
}
Ok(())
}
pub async fn create_db_entities_bulk_filter(
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
batch_size: usize,
) -> Result<(), Error> {
let db_mem = init_db::create_db_mem().await?;
create_db_entities_bulk(&db_mem, lines, &None, batch_size).await?;
let filter = tokio::fs::read_to_string(&*FILTER_PATH).await?;
db_mem.query(filter).await?;
let file_name: String = rand::thread_rng()
.sample_iter(&Alphanumeric)
.take(30)
.map(char::from)
.collect();
let file_path = format!("data/temp/{}.surql", file_name);
tokio::fs::create_dir_all("data/temp").await?;
db_mem.export(&file_path).await?;
db.import(&file_path).await?;
tokio::fs::remove_file(&file_path).await?;
if let Some(ref p) = pb {
p.inc(batch_size as u64)
}
Ok(())
}
#[derive(Clone, Copy)]
pub enum CreateVersion {
Single,
@ -191,95 +92,182 @@ impl CreateVersion {
batch_size: usize,
) -> bool {
match self {
CreateVersion::Single => create_db_entities(db, chunk, pb).await.is_ok(),
CreateVersion::Bulk => create_db_entities_bulk(db, chunk, pb, batch_size)
.await
.is_ok(),
CreateVersion::BulkFilter => create_db_entities_bulk_filter(db, chunk, pb, batch_size)
CreateVersion::Single => self.create_single(db, chunk, pb).await.is_ok(),
CreateVersion::Bulk => self.create_bulk(db, chunk, pb, batch_size).await.is_ok(),
CreateVersion::BulkFilter => self
.create_bulk_filter(db, chunk, pb, batch_size)
.await
.is_ok(),
}
}
}
pub async fn create_db_entities_threaded(
dbo: Option<Surreal<impl Connection>>, // None::<Surreal<Client>>
reader: Box<dyn BufRead>,
pb: Option<ProgressBar>,
batch_size: usize,
batch_num: usize,
create_version: CreateVersion,
) -> Result<(), Error> {
let mut futures = Vec::new();
let mut chunk = Vec::with_capacity(batch_size);
let mut chunk_counter = 0;
let mut lines = reader.lines();
let mut last_loop = false;
pub async fn run_threaded(
self,
dbo: Option<Surreal<impl Connection>>,
reader: Box<dyn BufRead>, // None::<Surreal<Client>>
pb: Option<ProgressBar>,
batch_size: usize,
batch_num: usize,
) -> Result<(), Error> {
let mut lines = reader.lines().peekable();
let mut futures = Vec::new();
loop {
let line = lines.next();
match line {
Some(line) => chunk.push(line?),
None => last_loop = true,
};
while lines.peek().is_some() {
let chunk: Vec<String> = lines
.by_ref()
.take(batch_size)
.filter_map(Result::ok)
.collect();
if chunk.len() >= batch_size || last_loop {
let dbo = dbo.clone();
let pb = pb.clone();
futures.push(self.spawn_chunk(dbo.clone(), chunk, pb.clone(), batch_size));
futures.push(tokio::spawn(async move {
let mut retries = 0;
loop {
match dbo {
Some(ref db) => {
if create_version.run(db, &chunk, &pb, batch_size).await {
break;
}
if db.use_ns("wikidata").use_db("wikidata").await.is_err() {
continue;
};
}
None => {
let db = if let Ok(db) = init_db::create_db_ws().await {
db
} else {
continue;
};
if create_version.run(&db, &chunk, &pb, batch_size).await {
break;
}
if futures.len() >= batch_num {
join_all(futures).await;
futures = Vec::new();
}
}
join_all(futures).await;
Ok(())
}
fn spawn_chunk(
&self,
dbo: Option<Surreal<impl Connection>>,
chunk: Vec<String>,
pb: Option<ProgressBar>,
batch_size: usize,
) -> tokio::task::JoinHandle<()> {
let create_version = *self;
tokio::spawn(async move {
let mut retries = 0;
loop {
match dbo {
Some(ref db) => {
if create_version.run(db, &chunk, &pb, batch_size).await {
break;
}
}
if retries >= 60 * 10 {
panic!("Failed to create entities, too many retries");
None => {
let db = match init_db::create_db_ws().await {
Ok(db) => db,
Err(_) => continue,
};
if create_version.run(&db, &chunk, &pb, batch_size).await {
break;
}
}
retries += 1;
sleep(Duration::from_millis(250)).await;
}
}));
chunk_counter += 1;
chunk = Vec::with_capacity(batch_size);
}
if chunk_counter >= batch_num || last_loop {
join_all(futures).await;
futures = Vec::new();
chunk_counter = 0;
}
if last_loop {
break;
}
if retries >= 60 * 10 {
panic!("Failed to create entities, too many retries");
}
retries += 1;
sleep(Duration::from_millis(250)).await;
}
})
}
match dbo {
Some(db) => {
create_db_entities(&db, &chunk, &pb).await?;
}
None => {
create_db_entities(&init_db::create_db_ws().await?, &chunk, &pb).await?;
async fn create_single(
self,
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
) -> Result<(), Error> {
let mut counter = 0;
for line in lines {
create_entity(db, line).await?;
counter += 1;
if counter % 100 == 0 {
if let Some(ref p) = pb {
p.inc(100)
}
}
}
Ok(())
}
async fn create_bulk(
self,
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
batch_size: usize,
) -> Result<(), Error> {
let lines = lines
.iter()
.map(|line| line.trim().trim_end_matches(',').to_string())
.filter(|line| line != "[" && line != "]")
.collect::<Vec<String>>();
let mut data_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
let mut claims_vec: Vec<Claims> = Vec::with_capacity(batch_size);
let mut property_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
let mut lexeme_vec: Vec<EntityMini> = Vec::with_capacity(batch_size);
for line in lines {
let json: Value = from_str(&line).expect("Failed to parse JSON");
let data = Entity::from_json(json).expect("Failed to parse JSON");
let (claims, data) = EntityMini::from_entity(data);
match data.id.clone().expect("No ID").tb.as_str() {
"Property" => property_vec.push(data),
"Lexeme" => lexeme_vec.push(data),
"Entity" => data_vec.push(data),
_ => panic!("Unknown table"),
}
claims_vec.push(claims);
}
db.insert::<Vec<EntityMini>>("Entity")
.content(data_vec)
.await?;
db.insert::<Vec<Claims>>("Claims")
.content(claims_vec)
.await?;
db.insert::<Vec<EntityMini>>("Property")
.content(property_vec)
.await?;
db.insert::<Vec<EntityMini>>("Lexeme")
.content(lexeme_vec)
.await?;
if let Some(ref p) = pb {
p.inc(batch_size as u64)
}
Ok(())
}
async fn create_bulk_filter(
self,
db: &Surreal<impl Connection>,
lines: &[String],
pb: &Option<ProgressBar>,
batch_size: usize,
) -> Result<(), Error> {
let db_mem = init_db::create_db_mem().await?;
self.create_bulk(&db_mem, lines, &None, batch_size).await?;
let filter = tokio::fs::read_to_string(&*FILTER_PATH).await?;
db_mem.query(filter).await?;
let file_name: String = rand::thread_rng()
.sample_iter(&Alphanumeric)
.take(30)
.map(char::from)
.collect();
let file_path = format!("data/temp/{}.surql", file_name);
tokio::fs::create_dir_all("data/temp").await?;
db_mem.export(&file_path).await?;
db.import(&file_path).await?;
tokio::fs::remove_file(&file_path).await?;
if let Some(ref p) = pb {
p.inc(batch_size as u64)
}
Ok(())
}
join_all(futures).await;
Ok(())
}