fix the whole batchKind thingy

This commit is contained in:
Tamo 2022-09-09 01:09:50 +02:00 committed by Clément Renault
parent af0f5d6c0c
commit 6b9689a1c0
No known key found for this signature in database
GPG Key ID: 92ADA4E935E71FA4
4 changed files with 292 additions and 5 deletions

View File

@ -2,7 +2,8 @@ use crate::{
task::{KindWithContent, Status}, task::{KindWithContent, Status},
Error, IndexScheduler, Result, TaskId, Error, IndexScheduler, Result, TaskId,
}; };
use milli::heed::RoTxn; use milli::{heed::RoTxn, update::IndexDocumentsMethod};
use uuid::Uuid;
use crate::{task::Kind, Task}; use crate::{task::Kind, Task};
@ -10,9 +11,53 @@ pub(crate) enum Batch {
Cancel(Task), Cancel(Task),
Snapshot(Vec<Task>), Snapshot(Vec<Task>),
Dump(Vec<Task>), Dump(Vec<Task>),
Contiguous { tasks: Vec<Task>, kind: Kind }, DocumentAddition {
One(Task), index_uid: String,
Empty, tasks: Vec<Task>,
primary_key: Option<String>,
content_files: Vec<Uuid>,
},
}
impl IndexScheduler {
/*
pub(crate) fn process_batch(&self, wtxn: &mut RwTxn, batch: Batch) -> Result<Vec<Task>> {
match batch {
Batch::DocumentAddition {
tasks,
primary_key,
content_files,
index_uid,
} => {
let index = self.create_index(wtxn, &index_uid)?;
let ret = index.update_documents(
IndexDocumentsMethod::UpdateDocuments,
primary_key,
self.file_store,
content_files,
)?;
assert_eq!(ret.len(), tasks.len(), "Update documents must return the same number of `Result` than the number of tasks.");
Ok(tasks
.into_iter()
.zip(ret)
.map(|(mut task, res)| match res {
Ok(info) => {
task.status = Status::Succeeded;
task.info = Some(info.to_string());
}
Err(error) => {
task.status = Status::Failed;
task.error = Some(error.to_string());
}
})
.collect())
}
_ => unreachable!(),
}
}
*/
} }
impl IndexScheduler { impl IndexScheduler {
@ -21,7 +66,7 @@ impl IndexScheduler {
/// 2. We get the *next* snapshot to process. /// 2. We get the *next* snapshot to process.
/// 3. We get the *next* dump to process. /// 3. We get the *next* dump to process.
/// 4. We get the *next* tasks to process for a specific index. /// 4. We get the *next* tasks to process for a specific index.
pub(crate) fn get_next_batch(&self, rtxn: &RoTxn) -> Result<Batch> { pub(crate) fn create_next_batch(&self, rtxn: &RoTxn) -> Result<Batch> {
let enqueued = &self.get_status(rtxn, Status::Enqueued)?; let enqueued = &self.get_status(rtxn, Status::Enqueued)?;
let to_cancel = self.get_kind(rtxn, Kind::CancelTask)? & enqueued; let to_cancel = self.get_kind(rtxn, Kind::CancelTask)? & enqueued;
@ -122,3 +167,233 @@ impl Batch {
} }
} }
} }
pub(crate) enum BatchKind {
ClearAll {
ids: Vec<TaskId>,
},
DocumentAddition {
addition_ids: Vec<TaskId>,
},
DocumentDeletion {
deletion_ids: Vec<TaskId>,
},
ClearAllAndSettings {
other: Vec<TaskId>,
settings_ids: Vec<TaskId>,
},
SettingsAndDocumentAddition {
settings_ids: Vec<TaskId>,
addition_ids: Vec<TaskId>,
},
Settings {
settings_ids: Vec<TaskId>,
},
DeleteIndex {
ids: Vec<TaskId>,
},
CreateIndex {
id: TaskId,
},
SwapIndex {
id: TaskId,
},
RenameIndex {
id: TaskId,
},
}
impl BatchKind {
/// return true if you must stop right there.
pub fn new(task_id: TaskId, kind: Kind) -> (Self, bool) {
match kind {
Kind::CreateIndex => (BatchKind::CreateIndex { id: task_id }, true),
Kind::DeleteIndex => (BatchKind::DeleteIndex { ids: vec![task_id] }, true),
Kind::RenameIndex => (BatchKind::RenameIndex { id: task_id }, true),
Kind::SwapIndex => (BatchKind::SwapIndex { id: task_id }, true),
Kind::ClearAllDocuments => (BatchKind::ClearAll { ids: vec![task_id] }, false),
Kind::DocumentAddition => (
BatchKind::DocumentAddition {
addition_ids: vec![task_id],
},
false,
),
Kind::DocumentDeletion => (
BatchKind::DocumentDeletion {
deletion_ids: vec![task_id],
},
false,
),
Kind::Settings => (
BatchKind::Settings {
settings_ids: vec![task_id],
},
false,
),
Kind::DumpExport | Kind::Snapshot | Kind::CancelTask => unreachable!(),
}
}
/// Return true if you must stop.
fn accumulate(&mut self, id: TaskId, kind: Kind) -> bool {
match (self, kind) {
// must handle the deleteIndex
(_, Kind::CreateIndex | Kind::RenameIndex | Kind::SwapIndex) => true,
(BatchKind::ClearAll { ids }, Kind::ClearAllDocuments | Kind::DocumentDeletion) => {
ids.push(id);
false
}
(BatchKind::ClearAll { .. }, Kind::DocumentAddition | Kind::Settings) => true,
(BatchKind::DocumentAddition { addition_ids }, Kind::ClearAllDocuments) => {
addition_ids.push(id);
*self = BatchKind::ClearAll {
ids: addition_ids.clone(),
};
false
}
(BatchKind::DocumentAddition { addition_ids }, Kind::DocumentAddition) => {
addition_ids.push(id);
false
}
(BatchKind::DocumentAddition { .. }, Kind::DocumentDeletion) => true,
(BatchKind::DocumentAddition { addition_ids }, Kind::Settings) => {
*self = BatchKind::SettingsAndDocumentAddition {
settings_ids: vec![id],
addition_ids: addition_ids.clone(),
};
false
}
(BatchKind::DocumentDeletion { deletion_ids }, Kind::ClearAllDocuments) => {
deletion_ids.push(id);
*self = BatchKind::ClearAll {
ids: deletion_ids.clone(),
};
false
}
(BatchKind::DocumentDeletion { .. }, Kind::DocumentAddition) => true,
(BatchKind::DocumentDeletion { deletion_ids }, Kind::DocumentDeletion) => {
deletion_ids.push(id);
false
}
(BatchKind::DocumentDeletion { .. }, Kind::Settings) => true,
(BatchKind::Settings { settings_ids }, Kind::ClearAllDocuments) => {
*self = BatchKind::ClearAllAndSettings {
settings_ids: settings_ids.clone(),
other: vec![id],
};
false
}
(BatchKind::Settings { .. }, Kind::DocumentAddition) => true,
(BatchKind::Settings { .. }, Kind::DocumentDeletion) => true,
(BatchKind::Settings { settings_ids }, Kind::Settings) => {
settings_ids.push(id);
false
}
(
BatchKind::ClearAllAndSettings {
other,
settings_ids,
},
Kind::ClearAllDocuments,
) => {
other.push(id);
false
}
(BatchKind::ClearAllAndSettings { .. }, Kind::DocumentAddition) => true,
(
BatchKind::ClearAllAndSettings {
other,
settings_ids,
},
Kind::DocumentDeletion,
) => {
other.push(id);
false
}
(
BatchKind::ClearAllAndSettings {
settings_ids,
other,
},
Kind::Settings,
) => {
settings_ids.push(id);
false
}
(
BatchKind::SettingsAndDocumentAddition {
settings_ids,
addition_ids,
},
Kind::ClearAllDocuments,
) => {
addition_ids.push(id);
*self = BatchKind::ClearAllAndSettings {
settings_ids: settings_ids.clone(),
other: addition_ids.clone(),
};
false
}
(
BatchKind::SettingsAndDocumentAddition {
settings_ids,
addition_ids,
},
Kind::DocumentAddition,
) => {
addition_ids.push(id);
false
}
(
BatchKind::SettingsAndDocumentAddition {
settings_ids,
addition_ids,
},
Kind::DocumentDeletion,
) => true,
(
BatchKind::SettingsAndDocumentAddition {
settings_ids,
addition_ids,
},
Kind::Settings,
) => {
settings_ids.push(id);
false
}
(_, Kind::CancelTask | Kind::DumpExport | Kind::Snapshot) => unreachable!(),
(
BatchKind::CreateIndex { .. }
| BatchKind::DeleteIndex { .. }
| BatchKind::SwapIndex { .. }
| BatchKind::RenameIndex { .. },
_,
) => {
unreachable!()
}
}
}
}
pub fn autobatcher(enqueued: Vec<(TaskId, Kind)>) -> Option<BatchKind> {
let mut enqueued = enqueued.into_iter();
let (id, kind) = enqueued.next()?;
let (mut acc, is_finished) = BatchKind::new(id, kind);
if is_finished {
return Some(acc);
}
for (id, kind) in enqueued {
if acc.accumulate(id, kind) {
break;
}
}
Some(acc)
}

View File

@ -80,6 +80,10 @@ pub struct IndexScheduler {
} }
impl IndexScheduler { impl IndexScheduler {
pub fn create_index(&self, rwtxn: &mut RwTxn, name: &str) -> Result<Index> {
todo!()
}
/// Return the index corresponding to the name. If it wasn't opened before /// Return the index corresponding to the name. If it wasn't opened before
/// it'll be opened. But if it doesn't exist on disk it'll throw an /// it'll be opened. But if it doesn't exist on disk it'll throw an
/// `IndexNotFound` error. /// `IndexNotFound` error.
@ -226,6 +230,9 @@ impl IndexScheduler {
let res = self.process_batch(&mut wtxn, &mut batch); let res = self.process_batch(&mut wtxn, &mut batch);
// TODO: TAMO: do this later // TODO: TAMO: do this later
// must delete the file on disk
// in case of error, must update the tasks with the error
// in case of « success » we must update all the task on disk
// self.handle_batch_result(res); // self.handle_batch_result(res);
match wtxn.commit() { match wtxn.commit() {

View File

@ -27,6 +27,9 @@ pub struct Task {
#[serde(with = "time::serde::rfc3339::option")] #[serde(with = "time::serde::rfc3339::option")]
pub finished_at: Option<OffsetDateTime>, pub finished_at: Option<OffsetDateTime>,
pub error: Option<String>,
pub info: Option<String>,
pub status: Status, pub status: Status,
pub kind: KindWithContent, pub kind: KindWithContent,
} }

View File

@ -67,6 +67,8 @@ impl IndexScheduler {
})?; })?;
} }
// TODO: TAMO: update the task in `all_tasks`
Ok(()) Ok(())
} }