feat(drive): prepare removal of user_id

this commit changes GET /api/<resources> to return resource caller has access to
    this is not anymmore resources users is owner of
This commit is contained in:
Edouard Vanbelle
2026-07-02 00:01:15 +02:00
parent 09339ea63f
commit 858139ef3b
18 changed files with 298 additions and 281 deletions
@@ -3,7 +3,7 @@
//! The repo deals only with the `storage.drives` table itself. Drive
//! membership lives in `storage.role_grants` (`resource_type='drive'`)
//! and is queried through the engine's existing grant paths;
//! `list_for_subjects` below resolves `role_grants` → `storage.drives`
//! `list_readable_by` below resolves `role_grants` → `storage.drives`
//! via a single join.
//!
//! See `migrations/20260802000000_drives_schema_additive.sql` for the
@@ -75,7 +75,7 @@ impl DrivePgRepository {
/// is declared owner→viewer (strongest→weakest), so `MIN` picks the
/// strongest of the caller's grants on the drive (direct +
/// group-mediated collapsed by GROUP BY). Used only by
/// `list_for_subjects`.
/// `list_readable_by`.
fn row_to_drive_with_name_and_role(
row: &sqlx::postgres::PgRow,
) -> Result<DriveWithRootName, DriveRepositoryError> {
@@ -463,13 +463,15 @@ impl DriveRepository for DrivePgRepository {
Self::row_to_drive_with_name(&row)
}
async fn list_for_subjects(
async fn list_readable_by(
&self,
subject_types: &[&str],
subject_ids: &[Uuid],
caller_id: Uuid,
) -> Result<Vec<DriveWithRootName>, DriveRepositoryError> {
// Joining role_grants → drives → folders returns every drive the
// expanded subject set can read, paired with its display name.
// 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 +
@@ -481,8 +483,6 @@ impl DriveRepository for DrivePgRepository {
// 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.
// Collapses direct + group-mediated grants on the same drive
// into one row alongside the existing GROUP BY.
let rows = sqlx::query(
r#"
SELECT d.id, d.kind, d.default_for_user, d.root_folder_id,
@@ -495,8 +495,11 @@ impl DriveRepository for DrivePgRepository {
JOIN storage.role_grants g
ON g.resource_type = 'drive'
AND g.resource_id = d.id
WHERE g.subject_type = ANY($1)
AND g.subject_id = ANY($2)
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,
@@ -505,16 +508,10 @@ impl DriveRepository for DrivePgRepository {
LOWER(f.name) ASC
"#,
)
.bind(
subject_types
.iter()
.map(|s| s.to_string())
.collect::<Vec<_>>(),
)
.bind(subject_ids)
.bind(caller_id)
.fetch_all(self.pool.as_ref())
.await
.map_err(|e| Self::map_sqlx_err("list_for_subjects", e))?;
.map_err(|e| Self::map_sqlx_err("list_readable_by", e))?;
rows.iter()
.map(Self::row_to_drive_with_name_and_role)
@@ -452,29 +452,28 @@ impl FileBlobReadRepository {
/// drive_id already ordered by capture date, so LIMIT stops the scan
/// early. Same O(LIMIT) shape as the pre-D7 `user_id`-keyed hot path.
///
/// Scope (`docs/plan/drive.md` §15): the caller's *effective subjects*
/// × drives with `policies.include_in_photo_index = true`. Default
/// personal drives always match because the flag is materialised to
/// `true` at drive creation (see
/// Scope (`docs/plan/drive.md` §15): drives with
/// `policies.include_in_photo_index = true` where the caller has a
/// direct grant (`subject_type = 'user'`) OR a grant on a group they
/// belong to transitively. Group membership is expanded inline by the
/// `storage.caller_group_ids(caller)` SQL function (migration
/// `20260901000002_caller_group_ids_function.sql`) — no ceremony at
/// the handler layer, no cross-space ambiguity from the earlier
/// parallel-arrays pattern.
///
/// Default personal drives always match because the flag is
/// materialised to `true` at drive creation (see
/// `DriveRepository::create_personal_drive_atomic` + the backfill
/// migration `20260901000000_default_personal_photo_music_flags.sql`)
/// — no per-kind carve-out needed. Non-default drives (secondary
/// personals, shared drives) surface here only after their owner
/// flips the flag on via the admin "Manage policies" modal.
///
/// `subject_types` / `subject_ids` are the caller expanded through
/// their group memberships (`AuthorizationEngine::
/// expand_subject_for_listing`); the arrays reach into the ANY()
/// predicates so a group-mediated grant on a drive counts too.
pub async fn list_media_files(
&self,
subject_types: &[&str],
subject_ids: &[Uuid],
caller_id: Uuid,
before: Option<i64>,
limit: i64,
) -> Result<(Vec<File>, Vec<i64>, Vec<(Option<i32>, Option<i32>)>), DomainError> {
let subject_types_owned: Vec<String> =
subject_types.iter().map(|s| s.to_string()).collect();
let rows: Vec<MediaFileRow> = sqlx::query_as(
r#"
SELECT fi.id::text, fi.name, fi.folder_id::text, fo.path,
@@ -495,21 +494,23 @@ impl FileBlobReadRepository {
JOIN storage.role_grants g
ON g.resource_type = 'drive'
AND g.resource_id = d.id
WHERE g.subject_type = ANY($1)
AND g.subject_id = ANY($2)
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())
AND (d.policies->>'include_in_photo_index')::boolean = true
)
AND NOT fi.is_trashed
AND (fi.mime_type LIKE 'image/%' OR fi.mime_type LIKE 'video/%')
AND ($3::bigint IS NULL
OR EXTRACT(EPOCH FROM fi.media_sort_date)::bigint < $3::bigint)
AND ($2::bigint IS NULL
OR EXTRACT(EPOCH FROM fi.media_sort_date)::bigint < $2::bigint)
ORDER BY fi.media_sort_date DESC
LIMIT $4
LIMIT $3
"#,
)
.bind(&subject_types_owned)
.bind(subject_ids)
.bind(caller_id)
.bind(before)
.bind(limit)
.fetch_all(self.pool.as_ref())
@@ -537,21 +538,18 @@ impl FileBlobReadRepository {
/// Scope: same `include_in_photo_index` predicate as
/// `list_media_files` (§15). Places is the map view over the same
/// content set the Photos timeline shows, so the two surfaces MUST
/// agree on drive scope. If a drive is opt-out for Photos its
/// geotagged files never appear on the map either.
/// agree on drive scope. Group membership is expanded inline by
/// `storage.caller_group_ids(caller)`.
///
/// This query is a per-cell aggregate (group by rounded lat/lng
/// bucket) rather than an ORDER BY / LIMIT hot path — the plain
/// `idx_files_drive_id` is sufficient to seek by drive.
pub async fn list_geo_clusters(
&self,
subject_types: &[&str],
subject_ids: &[Uuid],
caller_id: Uuid,
bounds: GeoBounds,
cell: f64,
) -> Result<Vec<GeoCluster>, DomainError> {
let subject_types_owned: Vec<String> =
subject_types.iter().map(|s| s.to_string()).collect();
let rows: Vec<(i64, f64, f64, String)> = sqlx::query_as(
r#"
SELECT count(*) AS n,
@@ -566,21 +564,23 @@ impl FileBlobReadRepository {
JOIN storage.role_grants g
ON g.resource_type = 'drive'
AND g.resource_id = d.id
WHERE g.subject_type = ANY($1)
AND g.subject_id = ANY($2)
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())
AND (d.policies->>'include_in_photo_index')::boolean = true
)
AND NOT fi.is_trashed
AND fm.latitude IS NOT NULL
AND fm.longitude IS NOT NULL
AND fm.longitude BETWEEN $3 AND $4
AND fm.latitude BETWEEN $5 AND $6
GROUP BY round(fm.longitude / $7), round(fm.latitude / $7)
AND fm.longitude BETWEEN $2 AND $3
AND fm.latitude BETWEEN $4 AND $5
GROUP BY round(fm.longitude / $6), round(fm.latitude / $6)
"#,
)
.bind(&subject_types_owned)
.bind(subject_ids)
.bind(caller_id)
.bind(bounds.west)
.bind(bounds.east)
.bind(bounds.south)
@@ -392,47 +392,55 @@ impl FolderRepository for FolderDbRepository {
.collect()
}
#[allow(clippy::type_complexity)]
async fn list_folders_by_owner(
async fn list_root_folders_for_caller(
&self,
parent_id: Option<&str>,
owner_id: Uuid,
caller_id: Uuid,
) -> Result<Vec<Folder>, DomainError> {
let rows: Vec<FolderRow> = if let Some(pid) = parent_id {
sqlx::query_as(
r#"
SELECT id::text, name, path, parent_id::text, user_id, drive_id,
EXTRACT(EPOCH FROM created_at)::bigint,
EXTRACT(EPOCH FROM updated_at)::bigint,
EXTRACT(EPOCH FROM tree_modified_at)::bigint,
created_by, updated_by
FROM storage.folders
WHERE parent_id = $1::uuid AND user_id = $2 AND NOT is_trashed
ORDER BY name
"#,
)
.bind(pid)
.bind(owner_id)
.fetch_all(self.pool())
.await
} else {
sqlx::query_as(
r#"
SELECT id::text, name, path, parent_id::text, user_id, drive_id,
EXTRACT(EPOCH FROM created_at)::bigint,
EXTRACT(EPOCH FROM updated_at)::bigint,
EXTRACT(EPOCH FROM tree_modified_at)::bigint,
created_by, updated_by
FROM storage.folders
WHERE parent_id IS NULL AND user_id = $1 AND NOT is_trashed
ORDER BY name
"#,
)
.bind(owner_id)
.fetch_all(self.pool())
.await
}
.map_err(|e| DomainError::internal_error("FolderDb", format!("list_by_owner: {e}")))?;
// Drive-scoped root-folder listing: return every root folder
// whose drive the caller has any role_grant on. Group
// memberships (direct + transitive) resolve inline via
// `storage.caller_group_ids($1)`.
//
// Closes `bug_root_folder_listing_legacy_user_id`: pre-D7 this
// query filtered on `folders.user_id = $caller`, which returned
// rows admin had created for other users' drives without ever
// getting a role on them. The drive-membership predicate below
// makes the "admin's own listing" correct without a separate
// filter.
//
// `caller_role` is NOT surfaced here — see the memory
// `project_caller_role_on_file_folder_dto` and the note at the
// top of `folder_repository.rs`. Frontend cross-references
// `/api/drives::caller_role` via `folder.drive_id`.
let rows: Vec<FolderRow> = sqlx::query_as(
r#"
SELECT f.id::text, f.name, f.path, f.parent_id::text, f.user_id, f.drive_id,
EXTRACT(EPOCH FROM f.created_at)::bigint,
EXTRACT(EPOCH FROM f.updated_at)::bigint,
EXTRACT(EPOCH FROM f.tree_modified_at)::bigint,
f.created_by, f.updated_by
FROM storage.folders f
WHERE f.parent_id IS NULL
AND NOT f.is_trashed
AND EXISTS (
SELECT 1
FROM storage.role_grants g
WHERE g.resource_type = 'drive'
AND g.resource_id = f.drive_id
AND (g.expires_at IS NULL OR g.expires_at > NOW())
AND (
(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)))
)
)
ORDER BY f.name
"#,
)
.bind(caller_id)
.fetch_all(self.pool())
.await
.map_err(|e| DomainError::internal_error("FolderDb", format!("list_root_folders: {e}")))?;
rows.into_iter()
.map(|(id, name, path, pid, uid, did, ca, ma, tma, cb, ub)| {
@@ -512,60 +520,52 @@ impl FolderRepository for FolderDbRepository {
Ok((folders?, total))
}
/// Paginated folder listing filtered by owner — single query with
/// `COUNT(*) OVER()` to avoid a separate COUNT round-trip.
#[allow(clippy::type_complexity)]
async fn list_folders_by_owner_paginated(
/// Paginated companion to `list_root_folders_for_caller` — same
/// drive-membership predicate, adds LIMIT/OFFSET and an optional
/// window-function COUNT so total pages can be surfaced without a
/// second round-trip.
async fn list_root_folders_for_caller_paginated(
&self,
parent_id: Option<&str>,
owner_id: Uuid,
caller_id: Uuid,
offset: usize,
limit: usize,
include_total: bool,
) -> Result<(Vec<Folder>, Option<usize>), DomainError> {
let rows: Vec<FolderRowPaginated> = if let Some(pid) = parent_id {
sqlx::query_as(
r#"
SELECT id::text, name, path, parent_id::text, user_id, drive_id,
EXTRACT(EPOCH FROM created_at)::bigint,
EXTRACT(EPOCH FROM updated_at)::bigint,
EXTRACT(EPOCH FROM tree_modified_at)::bigint,
created_by, updated_by,
COUNT(*) OVER() AS total_count
FROM storage.folders
WHERE parent_id = $1::uuid AND user_id = $2 AND NOT is_trashed
ORDER BY name
LIMIT $3 OFFSET $4
"#,
)
.bind(pid)
.bind(owner_id)
.bind(limit as i64)
.bind(offset as i64)
.fetch_all(self.pool())
.await
} else {
sqlx::query_as(
r#"
SELECT id::text, name, path, parent_id::text, user_id, drive_id,
EXTRACT(EPOCH FROM created_at)::bigint,
EXTRACT(EPOCH FROM updated_at)::bigint,
EXTRACT(EPOCH FROM tree_modified_at)::bigint,
created_by, updated_by,
COUNT(*) OVER() AS total_count
FROM storage.folders
WHERE parent_id IS NULL AND user_id = $1 AND NOT is_trashed
ORDER BY name
LIMIT $2 OFFSET $3
"#,
)
.bind(owner_id)
.bind(limit as i64)
.bind(offset as i64)
.fetch_all(self.pool())
.await
}
.map_err(|e| DomainError::internal_error("FolderDb", format!("paginate_by_owner: {e}")))?;
let rows: Vec<FolderRowPaginated> = sqlx::query_as(
r#"
SELECT f.id::text, f.name, f.path, f.parent_id::text, f.user_id, f.drive_id,
EXTRACT(EPOCH FROM f.created_at)::bigint,
EXTRACT(EPOCH FROM f.updated_at)::bigint,
EXTRACT(EPOCH FROM f.tree_modified_at)::bigint,
f.created_by, f.updated_by,
COUNT(*) OVER() AS total_count
FROM storage.folders f
WHERE f.parent_id IS NULL
AND NOT f.is_trashed
AND EXISTS (
SELECT 1
FROM storage.role_grants g
WHERE g.resource_type = 'drive'
AND g.resource_id = f.drive_id
AND (g.expires_at IS NULL OR g.expires_at > NOW())
AND (
(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)))
)
)
ORDER BY f.name
LIMIT $2 OFFSET $3
"#,
)
.bind(caller_id)
.bind(limit as i64)
.bind(offset as i64)
.fetch_all(self.pool())
.await
.map_err(|e| {
DomainError::internal_error("FolderDb", format!("list_root_folders_paginated: {e}"))
})?;
let total = if include_total {
Some(rows.first().map_or(0, |r| r.11) as usize)
+9 -4
View File
@@ -377,10 +377,15 @@ impl PgAclEngine {
/// Public wrapper around `subject_match_set` for callers that need
/// the expanded `(subject_types, subject_ids)` pair without invoking
/// the engine's full `check`/`require` pipeline. Used by
/// `GET /api/drives` (and future drive-aware listing surfaces) to
/// ask the `DriveRepository` for every drive the caller can read,
/// reusing the engine's cached group-expansion logic.
/// the engine's full `check`/`require` pipeline.
///
/// **Retained for legacy callers only** — new listing queries embed
/// the `storage.caller_group_ids` PostgreSQL function inline (see
/// migration `20260901000002_caller_group_ids_function.sql`) and
/// take a bare `caller_id: Uuid` instead of the pre-expanded arrays.
/// The engine's Moka cache still backs the fast path for per-request
/// AuthZ decisions (`check_inner`, `drive_role_cache`) where the
/// same subject is looked up repeatedly.
pub async fn expand_subject_for_listing(
&self,
subject: Subject,