feat(backend): policy rules show what they inherit, files of a chat or folder go to trash in one batch, trash list is paged

backend/src/utils/policy/models.py

@@ -79,6 +79,24 @@ class FolderSpec:
     is_chatlist: bool = False
 
 
+class PolicyParent(BaseModel):
+    scope_type: ScopeType
+    scope_id: int | None
+    title: str | None = None
+    toggles: CaptureToggles
+
+
+class PolicyRuleView(PolicyRecord):
+    title: str | None = None
+    parent: PolicyParent | None = None
+
+
+class PolicyScopeView(BaseModel):
+    title: str | None
+    rule: PolicyRecord | None
+    parent: PolicyParent | None
+
+
 @dataclass(slots=True)
 class PolicySet:
     chat: dict[int, CaptureToggles] = field(default_factory=dict)

backend/src/utils/policy/overview.py

@@ -0,0 +1,103 @@
+import asyncpg
+
+from utils.policy import repository
+from utils.policy.models import (
+    ChatKind,
+    ChatMeta,
+    FolderSpec,
+    PolicyParent,
+    PolicyRecord,
+    PolicyRuleView,
+    PolicyScopeView,
+    PolicySet,
+    ScopeType,
+)
+from utils.policy.resolver import resolve_parent
+from utils.read import chats
+from utils.read.models import ChatListItem
+
+
+def _meta(chat_id: int, chat: ChatListItem | None) -> ChatMeta:
+    if chat_id > 0:
+        kind = ChatKind.DM
+    elif chat is not None and chat.is_broadcast:
+        kind = ChatKind.CHANNEL
+    else:
+        kind = ChatKind.GROUP
+    return ChatMeta(
+        chat_id=chat_id,
+        kind=kind,
+        is_bot=bool(chat and chat.is_bot),
+        is_contact=chat.is_contact if chat else None,
+    )
+
+
+def _without_chats(policies: PolicySet) -> PolicySet:
+    return PolicySet(folder=policies.folder, defaults=policies.defaults)
+
+
+def _chat_parent(
+    chat_id: int,
+    chat: ChatListItem | None,
+    folders: list[FolderSpec],
+    policies: PolicySet,
+) -> PolicyParent:
+    return resolve_parent(_meta(chat_id, chat), folders, _without_chats(policies))
+
+
+async def list_rules(pool: asyncpg.Pool, account_id: int) -> list[PolicyRuleView]:
+    records = await repository.list_policies(pool, account_id)
+    folders = await repository.list_folders(pool, account_id)
+    policies = await repository.load_policy_set(pool, account_id)
+    chat_ids = [
+        record.scope_id
+        for record in records
+        if record.scope_type is ScopeType.CHAT and record.scope_id is not None
+    ]
+    known = {
+        chat.chat_id: chat for chat in await chats.get_chats(pool, account_id, chat_ids)
+    }
+    titles = {folder.folder_id: folder.title for folder in folders}
+    rules: list[PolicyRuleView] = []
+    for record in records:
+        rule = PolicyRuleView(**record.model_dump())
+        if record.scope_type is ScopeType.FOLDER and record.scope_id is not None:
+            rule.title = titles.get(record.scope_id)
+        if record.scope_type is ScopeType.CHAT and record.scope_id is not None:
+            chat = known.get(record.scope_id)
+            rule.title = chat.title if chat else None
+            rule.parent = _chat_parent(record.scope_id, chat, folders, policies)
+        rules.append(rule)
+    return rules
+
+
+async def _rule(
+    pool: asyncpg.Pool, account_id: int, scope_type: ScopeType, scope_id: int
+) -> PolicyRecord | None:
+    return await repository.find_policy(pool, account_id, scope_type, scope_id)
+
+
+async def chat_scope(
+    pool: asyncpg.Pool, account_id: int, chat_id: int
+) -> PolicyScopeView:
+    folders = await repository.list_folders(pool, account_id)
+    policies = await repository.load_policy_set(pool, account_id)
+    chat = await chats.get_chat(pool, account_id, chat_id)
+    return PolicyScopeView(
+        title=chat.title if chat else None,
+        rule=await _rule(pool, account_id, ScopeType.CHAT, chat_id),
+        parent=_chat_parent(chat_id, chat, folders, policies),
+    )
+
+
+async def folder_scope(
+    pool: asyncpg.Pool, account_id: int, folder_id: int
+) -> PolicyScopeView | None:
+    folder = await repository.get_folder(pool, account_id, folder_id)
+    if folder is None:
+        return None
+    return PolicyScopeView(
+        title=folder.title,
+        rule=await _rule(pool, account_id, ScopeType.FOLDER, folder_id),
+        parent=None,
+    )

backend/src/utils/policy/resolver.py

@@ -1,11 +1,14 @@
 from collections.abc import Iterable
 
 from utils.policy.models import (
+    DEFAULT_SCOPE_BY_KIND,
     CaptureToggles,
     ChatKind,
     ChatMeta,
     FolderSpec,
+    PolicyParent,
     PolicySet,
+    ScopeType,
 )
 
 
@@ -33,14 +36,29 @@ def folder_contains(folder: FolderSpec, chat: ChatMeta) -> bool:
     return _category_match(folder, chat)
 
 
+def resolve_parent(
+    chat: ChatMeta, folders: Iterable[FolderSpec], policies: PolicySet
+) -> PolicyParent:
+    for folder in sorted(folders, key=lambda f: f.order_index):
+        toggles = policies.folder.get(folder.folder_id)
+        if toggles is not None and folder_contains(folder, chat):
+            return PolicyParent(
+                scope_type=ScopeType.FOLDER,
+                scope_id=folder.folder_id,
+                title=folder.title,
+                toggles=toggles,
+            )
+    return PolicyParent(
+        scope_type=DEFAULT_SCOPE_BY_KIND[chat.kind],
+        scope_id=None,
+        toggles=policies.defaults.get(chat.kind, CaptureToggles()),
+    )
+
+
 def resolve(
     chat: ChatMeta, folders: Iterable[FolderSpec], policies: PolicySet
 ) -> CaptureToggles:
     chat_override = policies.chat.get(chat.chat_id)
     if chat_override is not None:
         return chat_override
-    for folder in sorted(folders, key=lambda f: f.order_index):
-        toggles = policies.folder.get(folder.folder_id)
-        if toggles is not None and folder_contains(folder, chat):
-            return toggles
-    return policies.defaults.get(chat.kind, CaptureToggles())
+    return resolve_parent(chat, folders, policies).toggles

backend/src/utils/read/chats.py

@@ -176,6 +176,18 @@ async def list_chats(
     return [_chat_item(row) for row in rows]
 
 
+async def folder_chat_ids(
+    pool: asyncpg.Pool, account_id: int, folder: FolderSpec
+) -> list[int]:
+    rows_sql = _CHAT_ROWS.format(ids=_ALL_IDS)
+    rows = await pool.fetch(
+        f"SELECT chat.chat_id FROM ({rows_sql}) chat WHERE {_folder_filter(1)}",  # noqa: S608
+        account_id,
+        *_folder_params(folder),
+    )
+    return [row["chat_id"] for row in rows]
+
+
 async def get_chat(
     pool: asyncpg.Pool, account_id: int, chat_id: int
 ) -> ChatListItem | None:

backend/src/utils/read/models.py

@@ -287,6 +287,14 @@ class TrashOutcome(BaseModel):
     blocked: list[int] = []
 
 
+class StorageScopeView(BaseModel):
+    chats: int
+    files: int
+    bytes: int
+    kept: int
+    gone: int
+
+
 class AvatarRef(BaseModel):
     unique_id: str
     storage_key: str | None

backend/src/utils/read/storage.py

@@ -4,10 +4,13 @@ from collections import defaultdict
 import asyncpg
 
 from utils.files import media_file_name
+from utils.policy import repository as policy_repository
+from utils.read import chats
 from utils.read.models import (
     Page,
     StorageFileView,
     StorageGroupView,
+    StorageScopeView,
     StorageSummaryView,
     TrashItemView,
 )
@@ -71,6 +74,46 @@ SELECT storage_key, chat_id, message_id, kind, mime, file_name, file_size,
     trashed_at
 FROM media_trash WHERE account_id = $1
 ORDER BY file_size DESC NULLS LAST, storage_key
+LIMIT $2 OFFSET $3
+"""
+
+SCOPE_BLOBS = """
+WITH inside AS (
+    SELECT DISTINCT storage_key FROM media
+    WHERE account_id = $1 AND chat_id = ANY($2::bigint[])
+      AND downloaded AND storage_key IS NOT NULL
+), blobs AS (
+    SELECT m.storage_key, max(m.file_size) AS file_size,
+        min(m.id) FILTER (
+            WHERE m.account_id = $1 AND m.chat_id = ANY($2::bigint[])
+        ) AS media_id,
+        bool_and(m.account_id = $1 AND m.chat_id = ANY($2::bigint[])) AS own
+    FROM media m JOIN inside i ON i.storage_key = m.storage_key
+    GROUP BY m.storage_key
+)
+SELECT b.storage_key, b.file_size, r.chat_id, r.message_id, r.kind, r.mime,
+    b.own
+    AND NOT EXISTS (SELECT 1 FROM stories s WHERE s.storage_key = b.storage_key)
+    AND NOT EXISTS (SELECT 1 FROM avatars a WHERE a.storage_key = b.storage_key)
+    AND NOT EXISTS (
+        SELECT 1 FROM custom_emoji e WHERE e.storage_key = b.storage_key
+    ) AS free
+FROM blobs b JOIN media r ON r.id = b.media_id
+"""
+
+_SCOPE_GONE = """
+SELECT (
+    SELECT count(DISTINCT m.storage_key) FROM media m
+    JOIN messages g ON g.account_id = m.account_id AND g.chat_id = m.chat_id
+        AND g.message_id = m.message_id
+    WHERE m.account_id = $1 AND m.chat_id = ANY($2::bigint[])
+      AND m.downloaded AND m.storage_key IS NOT NULL
+      AND g.deleted_at IS NOT NULL
+) + (
+    SELECT count(DISTINCT storage_key) FROM media
+    WHERE account_id = $1 AND chat_id = ANY($2::bigint[])
+      AND downloaded AND storage_key IS NOT NULL AND ttl_seconds IS NOT NULL
+)
 """
 
 _TRASH_REFS = """
@@ -179,20 +222,26 @@ async def list_files(
     return list(files.values())
 
 
-async def list_trash(pool: asyncpg.Pool, account_id: int) -> list[TrashItemView]:
-    rows = await pool.fetch(_TRASH, account_id)
+async def list_trash(
+    pool: asyncpg.Pool, account_id: int, page: Page
+) -> list[TrashItemView]:
+    rows = await pool.fetch(_TRASH, account_id, page.capped_limit, page.offset)
     refs = await pool.fetch(
         _TRASH_REFS, account_id, [row["storage_key"] for row in rows], REFS_CHECKED
     )
     messages = await _messages(pool, account_id, refs)
     alive: dict[str, bool] = defaultdict(bool)
     counts: dict[str, int] = {}
+    names: dict[str, str] = {}
     for ref, message in zip(refs, messages, strict=True):
-        alive[ref["storage_key"]] |= _alive(ref, message)
-        counts[ref["storage_key"]] = ref["refs"]
+        key = ref["storage_key"]
+        alive[key] |= _alive(ref, message)
+        counts[key] = ref["refs"]
+        if message is not None and message["original_name"]:
+            names.setdefault(key, message["original_name"])
     return [
         TrashItemView(
-            **dict(row),
+            **{**dict(row), "file_name": _trash_name(row, names)},
             purge_at=row["trashed_at"] + RETENTION,
             refs=counts.get(row["storage_key"], 0),
             refetchable=alive.get(row["storage_key"], True),
@@ -201,6 +250,40 @@ async def list_trash(pool: asyncpg.Pool, account_id: int) -> list[TrashItemView]
     ]
 
 
+def _trash_name(row: asyncpg.Record, names: dict[str, str]) -> str:
+    original = names.get(row["storage_key"])
+    if original is None:
+        return row["file_name"]
+    return media_file_name(row["kind"], row["mime"], row["message_id"], original)
+
+
+async def scope_chat_ids(
+    pool: asyncpg.Pool, account_id: int, chat_id: int | None, folder_id: int | None
+) -> list[int] | None:
+    if chat_id is not None:
+        return [chat_id]
+    if folder_id is None:
+        return None
+    folder = await policy_repository.get_folder(pool, account_id, folder_id)
+    if folder is None:
+        return None
+    return await chats.folder_chat_ids(pool, account_id, folder)
+
+
+async def get_scope(
+    pool: asyncpg.Pool, account_id: int, chat_ids: list[int]
+) -> StorageScopeView:
+    blobs = await pool.fetch(SCOPE_BLOBS, account_id, chat_ids)
+    free = [blob for blob in blobs if blob["free"]]
+    return StorageScopeView(
+        chats=len(chat_ids),
+        files=len(free),
+        bytes=sum(blob["file_size"] or 0 for blob in free),
+        kept=len(blobs) - len(free),
+        gone=await pool.fetchval(_SCOPE_GONE, account_id, chat_ids) if free else 0,
+    )
+
+
 async def get_trash_item(pool: asyncpg.Pool, storage_key: str) -> asyncpg.Record | None:
     return await pool.fetchrow(
         "SELECT kind, mime, file_name FROM media_trash WHERE storage_key = $1",

backend/src/utils/trash.py

@@ -4,6 +4,7 @@ from datetime import timedelta
 
 import asyncpg
 
+from utils.files import media_file_name
 from utils.read.media import get_media
 from utils.read.models import TrashOutcome
 from utils.storage import ContentAddressedStorage
@@ -12,6 +13,7 @@ logger = logging.getLogger(__name__)
 
 RETENTION = timedelta(days=7)
 PURGE_INTERVAL_SECONDS = 3600
+BATCH = 200
 
 _SHARED_ELSEWHERE = """
 SELECT EXISTS (SELECT 1 FROM stories WHERE storage_key = $1)
@@ -47,6 +49,12 @@ _FORGET_KEYS = """
 UPDATE media SET evicted_key = NULL WHERE evicted_key = ANY($1::text[])
 """
 
+_EVICT_MANY = """
+UPDATE media SET evicted_key = storage_key, storage_key = NULL,
+    downloaded = false, evicted_at = now()
+WHERE storage_key = ANY($1::text[])
+"""
+
 
 async def move_to_trash(
     pool: asyncpg.Pool,
@@ -86,12 +94,65 @@ async def move_to_trash(
     return outcome
 
 
+def _move_batch(storage: ContentAddressedStorage, keys: list[str]) -> None:
+    moved: list[str] = []
+    try:
+        for key in keys:
+            if storage.exists(key):
+                storage.trash(key)
+                moved.append(key)
+    except OSError:
+        for key in moved:
+            storage.restore(key)
+        raise
+
+
+async def trash_blobs(
+    pool: asyncpg.Pool,
+    storage: ContentAddressedStorage,
+    account_id: int,
+    blobs: list[asyncpg.Record],
+) -> TrashOutcome:
+    outcome = TrashOutcome()
+    for start in range(0, len(blobs), BATCH):
+        batch = blobs[start : start + BATCH]
+        keys = [blob["storage_key"] for blob in batch]
+        async with pool.acquire() as conn, conn.transaction():
+            await conn.execute(_EVICT_MANY, keys)
+            await conn.executemany(
+                _INSERT,
+                [
+                    (
+                        blob["storage_key"],
+                        account_id,
+                        blob["chat_id"],
+                        blob["message_id"],
+                        blob["kind"],
+                        blob["mime"],
+                        media_file_name(blob["kind"], blob["mime"], blob["message_id"]),
+                        blob["file_size"],
+                    )
+                    for blob in batch
+                ],
+            )
+            await asyncio.to_thread(_move_batch, storage, keys)
+        outcome.keys.extend(keys)
+        outcome.files += len(batch)
+        outcome.bytes += sum(blob["file_size"] or 0 for blob in batch)
+    return outcome
+
+
 async def restore(
     pool: asyncpg.Pool,
     storage: ContentAddressedStorage,
     account_id: int,
-    keys: list[str],
+    keys: list[str] | None,
 ) -> TrashOutcome:
+    if keys is None:
+        rows = await pool.fetch(
+            "SELECT storage_key FROM media_trash WHERE account_id = $1", account_id
+        )
+        keys = [row["storage_key"] for row in rows]
     outcome = TrashOutcome()
     for key in keys:
         async with pool.acquire() as conn, conn.transaction():