cjw/Oxicloud
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

perf: round 4 — one-pass row paths, drive-selector cache, CalDAV single-parse, streamed Azure, batched hydration

Browse Source 
Nine benchmark-gated changes (benches/ROUND4.md; every one ships with a
BEFORE/AFTER bench + equivalence gate, rollback rule as ROUND2/3):

- Row→entity path build: one-pass StoragePath::from_folder_and_name /
  from_joined + normalize_storage_name_owned + alloc-free Display —
  743→417 ns/file-row (1.78x), −5 allocs/row on every listing surface.
- WebDAV drive-selector: per-user readable_cache (single-flight, 30 s
  TTL, explicit invalidation incl. membership + group changes) replaces
  the grants join per request — 441 µs → 0.8 µs (~550x), 0 queries warm.
- CalDAV from_ical/update_ical_data: 8 full IcalParser runs per VEVENT
  → 1 (7.1x per PUT, 4.4x on 50-event imports); alloc-free split_vevents,
  chunk scan without the whole-body uppercase copy (1.4x), borrowed-key
  UID grouping (1.3x), REPORT props no longer cloned.
- PROPFIND emit: partition Vecs dropped (single-pass 404 list) + stack
  rendered RFC 3339/2822 dates, sizes, quoted etags (common::fmt,
  chrono-byte-identical, sweep-tested) on both DAV surfaces — 1.22x
  per page, 17.9→12.0 allocs/row.
- Grant-listing hydration: calendars/address books/playlists batch
  hydrate via = ANY($1) — 15 serial queries → 1 (~13x per sync poll).
- user-flags cache: get→insert → try_get_with single-flight (32→1
  queries per cold herd).
- Azure downloads: whole-blob Vec buffering → streamed SDK pages —
  TTFB 349→4 ms (87x), peak heap 480→1.9 MiB (254x) on 256 MiB blobs;
  new OXICLOUD_AZURE_ENDPOINT_URL override (Azurite/bench hook).
- Face indexing: unbounded per-image tokio::spawn → core-count
  semaphore, permit before blob read — peak heap 1175→176 MiB (6.7x).

Checks: cargo fmt, clippy --all-features --all-targets -D warnings,
cargo test --workspace (523 passed) + --features test_utils. hurl API
suite and dockerized integration DB not runnable in this environment —
left to CI.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017aJu9ghvuT8WqC31ZEGTBA
This commit is contained in:
Claude
2026-07-17 13:48:37 +00:00
parent 8a73607229
commit 12dc648cff
44 changed files with 5092 additions and 402 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/infrastructure/repositories/pg/drive_pg_repository.rs
+129 -53
View File
@@ -41,6 +41,27 @@ pub struct DrivePgRepository {
/// provisioning idempotency check (`NotFound` → create) always sees
/// the live table.
default_drive_cache: Cache<Uuid, DriveWithRootName>,
/// caller_id → every drive the caller can read (the full
/// role_grants ⋈ drives ⋈ folders join of [`list_readable_by`],
/// including the transitive-group expansion).
///
/// Re-resolved before this cache existed on EVERY native `/webdav`
/// request that names an explicit drive selector (all verbs; MOVE
/// and COPY twice), plus per-request in search, trash listing and
/// the `GET /api/drives` picker — the heaviest per-request query
/// left on the DAV path after CHROOT-CACHE. Concurrent misses are
/// coalesced (`try_get_with`), errors are never cached.
///
/// Freshness: every membership/lifecycle mutation that flows
/// through this repository or `DriveManagementService` invalidates
/// explicitly (per-user when the subject is a User, whole cache for
/// Group subjects, whose transitive membership is not resolvable
/// here). Residual staleness — a root-folder rename or a grant
/// written by a path that can't reach this cache — is bounded by
/// the same 30 s TTL the sibling caches accept; actual permission
/// enforcement is unaffected (the ACL engine re-checks per
/// operation with its own invalidation).
readable_cache: Cache<Uuid, Arc<Vec<DriveWithRootName>>>,
}
impl DrivePgRepository {
@@ -51,9 +72,27 @@ impl DrivePgRepository {
.max_capacity(DEFAULT_DRIVE_CACHE_CAPACITY)
.time_to_live(DEFAULT_DRIVE_CACHE_TTL)
.build(),
readable_cache: Cache::builder()
.max_capacity(DEFAULT_DRIVE_CACHE_CAPACITY)
.time_to_live(DEFAULT_DRIVE_CACHE_TTL)
.build(),
}
}
/// Drop the cached readable-drive list for one user (their grant set
/// changed: membership write, personal-drive provisioning, …).
pub async fn invalidate_readable_for_user(&self, user_id: Uuid) {
self.readable_cache.invalidate(&user_id).await;
}
/// Drop every cached readable-drive list. Used when the affected
/// user set is unknown at this layer: group-subject grants, drive
/// deletion, policy edits. All are admin-rare; repopulation costs
/// one join per active caller.
pub fn invalidate_readable_all(&self) {
self.readable_cache.invalidate_all();
}
fn map_sqlx_err(context: &'static str, e: sqlx::Error) -> DriveRepositoryError {
if let sqlx::Error::Database(ref dberr) = e
&& let Some(code) = dberr.code()
@@ -112,6 +151,63 @@ impl DrivePgRepository {
dwr.caller_role = role_str.as_deref().and_then(Role::parse);
Ok(dwr)
}
/// The uncached grants join behind [`DriveRepository::list_readable_by`].
///
/// Joining role_grants → drives → folders returns every drive the
/// caller can read, paired with its display name. Group
/// memberships (direct + transitive) are expanded inline by
/// `storage.caller_group_ids($caller)` — no Rust-side ceremony.
///
/// ORDER BY puts default drives first (so the picker UI doesn't
/// need a follow-up sort), then alphabetical by name. GROUP BY
/// collapses duplicate role_grants on the same drive (direct +
/// group-mediated) and sidesteps PostgreSQL's "ORDER BY
/// expression must appear in select list" rule that SELECT
/// DISTINCT imposes.
/// `MIN(g.role)` picks the caller's strongest role on each drive:
/// `storage.grant_role` is declared `owner → viewer` (strongest →
/// weakest), so MIN returns the strongest. Cast `::text` matches
/// the codebase convention for reading enum columns into Rust
/// (see `pg_acl_engine.rs`); `Role::parse` handles the trip back.
async fn query_readable_by(
&self,
caller_id: Uuid,
) -> Result<Vec<DriveWithRootName>, DriveRepositoryError> {
let rows = sqlx::query(
r#"
SELECT d.id, d.kind, d.default_for_user, d.root_folder_id,
d.quota_bytes, d.used_bytes, d.policies,
d.created_at, d.updated_at,
f.name AS root_folder_name,
MIN(g.role)::text AS caller_role
FROM storage.drives d
JOIN storage.folders f ON f.id = d.root_folder_id
JOIN storage.role_grants g
ON g.resource_type = 'drive'
AND g.resource_id = d.id
WHERE (
(g.subject_type = 'user' AND g.subject_id = $1)
OR (g.subject_type = 'group' AND g.subject_id IN
(SELECT storage.caller_group_ids($1)))
)
AND (g.expires_at IS NULL OR g.expires_at > NOW())
GROUP BY d.id, d.kind, d.default_for_user, d.root_folder_id,
d.quota_bytes, d.used_bytes, d.policies,
d.created_at, d.updated_at, f.name
ORDER BY (d.default_for_user IS NULL) ASC,
LOWER(f.name) ASC
"#,
)
.bind(caller_id)
.fetch_all(self.pool.as_ref())
.await
.map_err(|e| Self::map_sqlx_err("list_readable_by", e))?;
rows.iter()
.map(Self::row_to_drive_with_name_and_role)
.collect()
}
}
#[async_trait::async_trait]
@@ -239,6 +335,8 @@ impl DriveRepository for DrivePgRepository {
// Drop any cached default-drive resolution for this user (a stale
// NotFound is never cached, but be explicit about the write path).
self.default_drive_cache.invalidate(&owner_id).await;
// The owner gained a drive — their readable list changed too.
self.invalidate_readable_for_user(owner_id).await;
Self::row_to_drive_with_name(&row)
}
@@ -347,6 +445,16 @@ impl DriveRepository for DrivePgRepository {
.await
.map_err(|e| Self::map_sqlx_err("create_shared_drive_atomic.commit", e))?;
// The owner grant written above changes the grantee's readable
// list. User subjects invalidate precisely; Group subjects fall
// back to a full clear (transitive members unknown here).
match owner_subject {
crate::domain::services::authorization::Subject::User(uid) => {
self.invalidate_readable_for_user(uid).await;
}
_ => self.invalidate_readable_all(),
}
Self::row_to_drive_with_name(&row)
}
@@ -425,10 +533,11 @@ impl DriveRepository for DrivePgRepository {
tx.commit()
.await
.map_err(|e| Self::map_sqlx_err("delete_atomic.commit", e))?;
// We only have the drive id here; the cache is keyed by user.
// Deletion is rare — clearing the whole cache is the simple,
// We only have the drive id here; the caches are keyed by user.
// Deletion is rare — clearing them whole is the simple,
// always-correct move (repopulates at one query per active user).
self.default_drive_cache.invalidate_all();
self.invalidate_readable_all();
Ok(())
}
@@ -513,55 +622,21 @@ impl DriveRepository for DrivePgRepository {
&self,
caller_id: Uuid,
) -> Result<Vec<DriveWithRootName>, DriveRepositoryError> {
// Joining role_grants → drives → folders returns every drive the
// caller can read, paired with its display name. Group
// memberships (direct + transitive) are expanded inline by
// `storage.caller_group_ids($caller)` — no Rust-side ceremony.
//
// ORDER BY puts default drives first (so the picker UI doesn't
// need a follow-up sort), then alphabetical by name. GROUP BY
// collapses duplicate role_grants on the same drive (direct +
// group-mediated) and sidesteps PostgreSQL's "ORDER BY
// expression must appear in select list" rule that SELECT
// DISTINCT imposes.
// `MIN(g.role)` picks the caller's strongest role on each drive:
// `storage.grant_role` is declared `owner → viewer` (strongest →
// weakest), so MIN returns the strongest. Cast `::text` matches
// the codebase convention for reading enum columns into Rust
// (see `pg_acl_engine.rs`); `Role::parse` handles the trip back.
let rows = sqlx::query(
r#"
SELECT d.id, d.kind, d.default_for_user, d.root_folder_id,
d.quota_bytes, d.used_bytes, d.policies,
d.created_at, d.updated_at,
f.name AS root_folder_name,
MIN(g.role)::text AS caller_role
FROM storage.drives d
JOIN storage.folders f ON f.id = d.root_folder_id
JOIN storage.role_grants g
ON g.resource_type = 'drive'
AND g.resource_id = d.id
WHERE (
(g.subject_type = 'user' AND g.subject_id = $1)
OR (g.subject_type = 'group' AND g.subject_id IN
(SELECT storage.caller_group_ids($1)))
)
AND (g.expires_at IS NULL OR g.expires_at > NOW())
GROUP BY d.id, d.kind, d.default_for_user, d.root_folder_id,
d.quota_bytes, d.used_bytes, d.policies,
d.created_at, d.updated_at, f.name
ORDER BY (d.default_for_user IS NULL) ASC,
LOWER(f.name) ASC
"#,
)
.bind(caller_id)
.fetch_all(self.pool.as_ref())
.await
.map_err(|e| Self::map_sqlx_err("list_readable_by", e))?;
rows.iter()
.map(Self::row_to_drive_with_name_and_role)
.collect()
// Serve from the per-user cache; concurrent misses for the same
// caller are coalesced into one join (`try_get_with`), and errors
// are never cached. See the `readable_cache` field docs for the
// freshness/invalidation contract.
let cached = self
.readable_cache
.try_get_with(caller_id, async move {
self.query_readable_by(caller_id).await.map(Arc::new)
})
.await
.map_err(|e: Arc<DriveRepositoryError>| {
Arc::try_unwrap(e)
.unwrap_or_else(|shared| DriveRepositoryError::StorageError(shared.to_string()))
})?;
Ok((*cached).clone())
}
async fn list_all(&self) -> Result<Vec<DriveWithRootName>, DriveRepositoryError> {
@@ -717,9 +792,10 @@ impl DriveRepository for DrivePgRepository {
.ok_or_else(|| DriveRepositoryError::NotFound(drive_id.to_string()))?
.0;
// Policy edits must not serve a stale `policies` bag from the
// default-drive cache (keyed by user, and we only have the drive
// id) — clear it; policy edits are admin-rare.
// user-keyed caches (we only have the drive id) — clear both;
// policy edits are admin-rare.
self.default_drive_cache.invalidate_all();
self.invalidate_readable_all();
Ok(crate::domain::entities::drive::DrivePolicies::from_value(
&raw,
))