Skip to content

fix(sync): reconcile the cache instead of rebuilding it #236

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
romaintb merged 1 commit into from
Sep 5, 2026
Merged

fix(sync): reconcile the cache instead of rebuilding it #236

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
273 changes: 201 additions & 72 deletions src/sync/storage.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,19 +25,25 @@ impl SyncService {
.await
.context("Failed to start cache refresh transaction")?;

// A full remote snapshot is authoritative. Clear only this backend's cached
// objects inside the same transaction so removed remote objects do not linger,
// while any later write failure still restores the previous cache on rollback.
task::Entity::delete_many()
// A full remote snapshot is authoritative about *which* objects exist, so drop the
// rows it stopped returning. It says nothing about their local UUIDs, which the UI
// keys its selection and cursor on, so the rows it still returns are updated in
// place by the batches below rather than deleted and re-inserted.
//
// Parent links are detached first. A parent the remote stopped returning is about to
// be deleted, and Todoist never returns completed tasks, so completing a parent makes
// it vanish while its still-open subtasks keep coming. The task hierarchy FK cascades
// and the project one restricts, so leaving the links in place would either delete a
// live subtask or fail the delete outright. The batches relink from the snapshot.
task::Entity::update_many()
.col_expr(
task::Column::ParentUuid,
sea_orm::sea_query::Expr::value(Option::<Uuid>::None),
)
.filter(task::Column::BackendUuid.eq(self.backend_uuid))
.exec(&transaction)
.await
.context("Failed to clear cached tasks")?;
section::Entity::delete_many()
.filter(section::Column::BackendUuid.eq(self.backend_uuid))
.exec(&transaction)
.await
.context("Failed to clear cached sections")?;
.context("Failed to detach cached task hierarchy")?;
project::Entity::update_many()
.col_expr(
project::Column::ParentUuid,
Expand All @@ -47,16 +53,36 @@ impl SyncService {
.exec(&transaction)
.await
.context("Failed to detach cached project hierarchy")?;

task::Entity::delete_many()
.filter(task::Column::BackendUuid.eq(self.backend_uuid))
.filter(task::Column::RemoteId.is_not_in(tasks.iter().map(|task| task.remote_id.as_str())))
.exec(&transaction)
.await
.context("Failed to drop tasks missing from the snapshot")?;
// An empty section list is ambiguous: `sync` turns a failed section fetch into an
// empty slice and carries on, so treat it as "nothing to reconcile". Every other
// fetch aborts the sync on failure, where empty honestly means empty.
if !sections.is_empty() {
section::Entity::delete_many()
.filter(section::Column::BackendUuid.eq(self.backend_uuid))
.filter(section::Column::RemoteId.is_not_in(sections.iter().map(|s| s.remote_id.as_str())))
.exec(&transaction)
.await
.context("Failed to drop sections missing from the snapshot")?;
}
project::Entity::delete_many()
.filter(project::Column::BackendUuid.eq(self.backend_uuid))
.filter(project::Column::RemoteId.is_not_in(projects.iter().map(|p| p.remote_id.as_str())))
.exec(&transaction)
.await
.context("Failed to clear cached projects")?;
.context("Failed to drop projects missing from the snapshot")?;
label::Entity::delete_many()
.filter(label::Column::BackendUuid.eq(self.backend_uuid))
.filter(label::Column::RemoteId.is_not_in(labels.iter().map(|l| l.remote_id.as_str())))
.exec(&transaction)
.await
.context("Failed to clear cached labels")?;
.context("Failed to drop labels missing from the snapshot")?;

self.store_projects_batch(&transaction, projects)
.await
Expand Down Expand Up @@ -464,16 +490,16 @@ impl SyncService {
#[cfg(test)]
mod tests {
use super::*;
use crate::backend::{BackendProject, BackendSection};
use crate::backend_registry::BackendRegistry;
use crate::backend::{BackendProject, BackendSection, BackendTask};
use crate::entities::backend;
use sea_orm::{EntityTrait, Set};
use std::path::PathBuf;
use std::sync::Arc;
use tokio::sync::Mutex;

#[tokio::test]
async fn failed_snapshot_write_preserves_the_previous_cache() {
let db_path = std::env::temp_dir().join(format!("terminalist-snapshot-{}.db", Uuid::new_v4()));
/// A sync service over a throwaway database file, with its backend row already present.
async fn test_service(label: &str) -> (PathBuf, Arc<Mutex<LocalStorage>>, SyncService) {
let db_path = std::env::temp_dir().join(format!("terminalist-{label}-{}.db", Uuid::new_v4()));
let storage = LocalStorage::new_at(db_path.clone()).await.unwrap();
let backend_uuid = Uuid::new_v4();

Expand All @@ -490,22 +516,52 @@ mod tests {
.unwrap();

let storage = Arc::new(Mutex::new(storage));
let service = SyncService {
backend_registry: Arc::new(BackendRegistry::new(storage.clone())),
backend_uuid,
storage: storage.clone(),
sync_in_progress: Arc::new(Mutex::new(false)),
debug_mode: false,
};
let service = SyncService::new_for_test(storage.clone(), backend_uuid);
(db_path, storage, service)
}

/// Closing the pool before removing the file keeps Windows and NFS happy.
async fn teardown(db_path: PathBuf, storage: Arc<Mutex<LocalStorage>>) {
storage.lock().await.conn.clone().close().await.unwrap();
let _ = std::fs::remove_file(db_path);
}

let cached_project = BackendProject {
remote_id: "cached-project".to_string(),
name: "Cached project".to_string(),
fn project(remote_id: &str, order_index: i32) -> BackendProject {
BackendProject {
remote_id: remote_id.to_string(),
name: format!("Project {remote_id}"),
is_favorite: false,
is_inbox: false,
order_index: 1,
order_index,
parent_remote_id: None,
};
}
}

fn task(remote_id: &str, project_remote_id: &str, parent_remote_id: Option<&str>) -> BackendTask {
BackendTask {
remote_id: remote_id.to_string(),
content: format!("Task {remote_id}"),
description: None,
project_remote_id: project_remote_id.to_string(),
section_remote_id: None,
parent_remote_id: parent_remote_id.map(str::to_string),
priority: 1,
order_index: 1,
due_date: None,
due_datetime: None,
is_recurring: false,
deadline: None,
duration: None,
is_completed: false,
labels: Vec::new(),
}
}

#[tokio::test]
async fn failed_snapshot_write_preserves_the_previous_cache() {
let (db_path, storage, service) = test_service("snapshot").await;

let cached_project = project("cached-project", 1);
{
let storage = storage.lock().await;
service
Expand All @@ -514,12 +570,8 @@ mod tests {
.unwrap();
}

let replacement_project = BackendProject {
remote_id: "replacement-project".to_string(),
name: "Replacement project".to_string(),
order_index: 2,
..cached_project
};
// The section names a project the snapshot does not carry, so storing it fails and
// must take the whole transaction down with it.
let invalid_section = BackendSection {
remote_id: "invalid-section".to_string(),
name: "Invalid section".to_string(),
Expand All @@ -529,7 +581,13 @@ mod tests {
{
let storage = storage.lock().await;
let result = service
.store_snapshot(&storage, &[replacement_project], &[], &[invalid_section], &[])
.store_snapshot(
&storage,
&[project("replacement-project", 2)],
&[],
&[invalid_section],
&[],
)
.await;
assert!(result.is_err());

Expand All @@ -538,56 +596,127 @@ mod tests {
assert_eq!(projects[0].remote_id, "cached-project");
}

storage.lock().await.conn.clone().close().await.unwrap();
let _ = std::fs::remove_file(db_path);
teardown(db_path, storage).await;
}

#[tokio::test]
async fn successful_snapshot_removes_objects_missing_from_backend() {
let db_path = std::env::temp_dir().join(format!("terminalist-replace-snapshot-{}.db", Uuid::new_v4()));
let storage = LocalStorage::new_at(db_path.clone()).await.unwrap();
let backend_uuid = Uuid::new_v4();
let (db_path, storage, service) = test_service("replace-snapshot").await;

backend::Entity::insert(backend::ActiveModel {
uuid: Set(backend_uuid),
backend_type: Set("test".to_string()),
name: Set("Test".to_string()),
is_enabled: Set(true),
credentials: Set("{}".to_string()),
settings: Set("{}".to_string()),
})
.exec(&storage.conn)
.await
.unwrap();
{
let storage = storage.lock().await;
service
.store_snapshot(&storage, &[project("old-project", 1)], &[], &[], &[])
.await
.unwrap();
service
.store_snapshot(&storage, &[project("new-project", 2)], &[], &[], &[])
.await
.unwrap();

let storage = Arc::new(Mutex::new(storage));
let service = SyncService::new_for_test(storage.clone(), backend_uuid);
let old_project = BackendProject {
remote_id: "old-project".to_string(),
name: "Old project".to_string(),
is_favorite: false,
is_inbox: false,
order_index: 1,
parent_remote_id: None,
};
let new_project = BackendProject {
remote_id: "new-project".to_string(),
name: "New project".to_string(),
order_index: 2,
..old_project.clone()
let projects = ProjectRepository::get_all(&storage.conn).await.unwrap();
assert_eq!(projects.len(), 1);
assert_eq!(projects[0].remote_id, "new-project");
}

teardown(db_path, storage).await;
}

/// The UI keys its selection on local UUIDs, so a row the remote still returns has to
/// keep the UUID it had, and a subtask has to survive its parent being completed away.
#[tokio::test]
async fn snapshot_keeps_uuids_and_orphans_subtasks_of_vanished_parents() {
let (db_path, storage, service) = test_service("stable-uuids").await;
let backend_uuid = service.backend_uuid;

let (project_uuid, subtask_uuid) = {
let storage = storage.lock().await;
service
.store_snapshot(
&storage,
&[project("p1", 1)],
&[],
&[],
&[task("parent", "p1", None), task("subtask", "p1", Some("parent"))],
)
.await
.unwrap();

let subtask = TaskRepository::get_by_remote_id(&storage.conn, &backend_uuid, "subtask")
.await
.unwrap()
.expect("subtask cached");
let parent = TaskRepository::get_by_remote_id(&storage.conn, &backend_uuid, "parent")
.await
.unwrap()
.expect("parent cached");
assert_eq!(subtask.parent_uuid, Some(parent.uuid));

(
ProjectRepository::get_all(&storage.conn).await.unwrap()[0].uuid,
subtask.uuid,
)
};

// Completing the parent makes Todoist stop returning it, while the subtask keeps
// coming. Renaming the project proves the row was updated rather than replaced.
{
let storage = storage.lock().await;
service.store_snapshot(&storage, &[old_project], &[], &[], &[]).await.unwrap();
service.store_snapshot(&storage, &[new_project], &[], &[], &[]).await.unwrap();
let mut renamed = project("p1", 1);
renamed.name = "Renamed".to_string();
service
.store_snapshot(&storage, &[renamed], &[], &[], &[task("subtask", "p1", Some("parent"))])
.await
.unwrap();

let projects = ProjectRepository::get_all(&storage.conn).await.unwrap();
assert_eq!(projects.len(), 1);
assert_eq!(projects[0].remote_id, "new-project");
assert_eq!(projects[0].uuid, project_uuid, "project uuid churned across a sync");
assert_eq!(projects[0].name, "Renamed");

let subtask = TaskRepository::get_by_remote_id(&storage.conn, &backend_uuid, "subtask")
.await
.unwrap()
.expect("subtask survives its parent being completed");
assert_eq!(subtask.uuid, subtask_uuid, "task uuid churned across a sync");
assert_eq!(subtask.parent_uuid, None);

assert!(TaskRepository::get_by_remote_id(&storage.conn, &backend_uuid, "parent")
.await
.unwrap()
.is_none());
}

storage.lock().await.conn.clone().close().await.unwrap();
let _ = std::fs::remove_file(db_path);
teardown(db_path, storage).await;
}

/// `sync` turns a failed section fetch into an empty slice, so an empty section list
/// cannot be read as "the remote has no sections".
#[tokio::test]
async fn empty_section_list_leaves_cached_sections_alone() {
let (db_path, storage, service) = test_service("empty-sections").await;

let section = BackendSection {
remote_id: "s1".to_string(),
name: "Section".to_string(),
project_remote_id: "p1".to_string(),
order_index: 1,
};
{
let storage = storage.lock().await;
service
.store_snapshot(&storage, &[project("p1", 1)], &[], &[section], &[])
.await
.unwrap();
service
.store_snapshot(&storage, &[project("p1", 1)], &[], &[], &[])
.await
.unwrap();

let sections = SectionRepository::get_all(&storage.conn).await.unwrap();
assert_eq!(sections.len(), 1);
}

teardown(db_path, storage).await;
}
}