Files
Warp/server/internal/roomserver/query.go
T
tajniak81andClaude Opus 5 d187ad0d17
CI / Test (ubuntu-latest) (push) Canceled after 0s
CI / Test (windows-latest) (push) Canceled after 0s
CI / Storage tests on PostgreSQL (push) Canceled after 0s
CI / Complement (federation off) (push) Canceled after 0s
CI / golangci-lint (push) Canceled after 0s
CI / Smoke test the binary (push) Canceled after 0s
Phase 4: sync
Implements PLAN.md Phase 4: the read model. /sync, /messages and /context,
the filters that narrow them, and the streams a response carries that are
not room events — typing, receipts, presence and to-device messages. With
this a client can be used against Warp, which is milestone M1.

The one architectural choice is that the syncapi has no copy of the rooms:
it reads them through the roomserver. Dendrite gives its syncapi its own
tables and feeds them from the output stream, which is right for a design
where the two can be separate processes, and Dendrite's could be. Warp is
one process by decision, and there the copy buys nothing and costs the
hardest class of bug a homeserver has — a read model that disagrees with
the room, where a client is shown a message that was rejected and the only
way to find out is to compare two tables. So the roomserver grew a read
model (stream.go, roomsync.go, visibility.go) rather than exporting its
store: reading an event means parsing it under its room's version, and a
second copy of the version registry, the parser and the client projection
would be a second place for them to disagree.

syncapi: stream tokens naming a position in each of five independent
streams, so a client that has seen receipt 7 but only event 4 can say so in
one opaque string. A token with fewer positions than this build writes is
accepted and the rest read as zero, which is what lets a token stored
before a stream existed go on working. The PDU position comes from
MAX(stream_ordering) and not from the counter row the other four use: the
counter is advanced inside the writing transaction, so a token taken from
it would name a position whose event no reader can see yet, and the event
would be skipped for good once that transaction committed.

Long polling goes through a notifier that wakes by room and by user rather
than globally — an event wakes the listeners of its room, a membership
event also wakes the user it names, so an invite reaches a client that is
not in the room yet. A wake-up is only ever a hint: the listener re-reads
the streams and decides for itself what changed, which is what makes it
safe for the notifier to hold no state about what each client has seen. The
listener registers before the read, so an event landing in between leaves
the channel signalled and the sync reads again rather than blocking on a
change it just missed. EventPersisted must not block, because it is called
with the room's write lock still held.

"Something happened" is not "a token moved". A stream can move without this
user being concerned by it, and returning on every such move would turn a
long poll back into a busy loop. The same trap caught presence twice over:
a client says it is still online by syncing, so ?set_presence= makes every
sync a presence write, and a stream that moved each time would wake every
peer, which would sync, which would move it again. SetPresence therefore
reports whether anything a peer would notice actually changed, and a write
that only refreshes the last-active timestamp leaves the stream alone.

Timeline windows are named by their two ends and a direction, Low exclusive
and High inclusive in both directions, rather than by "from" and "to" as
the wire format spells them. Each of those would otherwise be the low bound
in one direction and the high bound in the other, which is exactly the
off-by-one every pagination bug is made of; the handlers translate once and
nothing below them thinks about it again. A sync's timeline is read
backwards and reported forwards, because a client wants the newest events
of the window it missed — a forward read with a limit hands it the oldest
and then calls the rest a gap.

visibility: history visibility is now a span rather than a yes or no, which
is what Phase 3 deferred here. A user who has left a room is served the
room as it was when they left — their state reads, their members and their
timeline — instead of being refused, and is bounded above at their own
membership event so a kicked client cannot go on streaming. Their span is
found from their own member events, so a user who joined, left and rejoined
sees their latest span. ?at= on /members is honoured for the same reason.
What is left is that the span is computed from the room's current
history_visibility where the spec judges each event by the one in force
when it was sent; that is a state read per event and belongs with open
question 1.

Typing is held in memory and nowhere else: notifications expire in seconds,
are worthless across a restart, and arrive one per keystroke-burst per
room, so persisting them would be the busiest table in the database holding
the least valuable data in it. Receipts only ever advance, compared on the
position of the event acknowledged rather than on arrival, because two of a
user's clients read at different speeds and the slower one must not drag
the marker backwards. To-device queues are per device and drained by the
next sync rather than by the one that delivered them, so a client that
crashes in between gets the batch again — the right way round, since a lost
key share leaves a message permanently undecryptable and a duplicate is
ignored. A wildcard addressee is expanded when sent, not stored, because
"every device" means the ones that existed then.

Presence is off by default, which answers PLAN.md open question 3. It is
the noisiest thing in Matrix and the one genuinely fan-out notification
here; a private homeserver gains little from it. With it off the endpoints
refuse rather than disappearing, because a client that gets a 404 from
/presence concludes the server is too old to have it and stops asking.

storage: one migration per driver adding the receipt, to-device, presence
and filter tables with a counter row each, and a stream position on the
existing account data table rather than a parallel one — there is exactly
one current value per (user, room, type), and a second table would be a
second thing to keep in step. Account data is reported in an event list but
is not an event, so it gets its own response type: a room event with an
empty sender and no timestamp would make a client decide what that meant.

Deliberately missing, each named in docs/sync.md: notification counts (they
need push rules, Phase 8), device lists (the keyserver, Phase 7), anything
remote (Phase 5), threads (Phase 10). device_lists is sent as two empty
lists because a client reads a missing one as malformed; the notification
counts are left out entirely, because a zero there is a claim and an absent
field is not.

Tested through HTTP rather than against the builder, because what is worth
pinning down is the response a client sees — which section a room lands in,
whether a token handed back can be handed in again, whether paging
terminates and loses nothing. The storage tests run against both SQL
drivers; the notifier has its own tests for the concurrent case. A live
server was driven end to end, which is what caught the two bugs the unit
tests had missed: a backwards /messages page must be in reverse
chronological order, and `start` must name where the page began rather than
position zero when ?from= was omitted.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-10-16 11:55:27 +02:00

432 lines
14 KiB
Go

package roomserver
import (
"context"
"encoding/json"
"fmt"
"github.com/dariuszjaniak/warp/server/internal/matrix/auth"
"github.com/dariuszjaniak/warp/server/internal/matrix/event"
"github.com/dariuszjaniak/warp/server/internal/matrix/roomversion"
"github.com/dariuszjaniak/warp/server/internal/storage"
)
// currentState reads a room's current state as the authorisation rules see
// it.
func (r *API) currentState(ctx context.Context, room storage.Room, rules roomversion.Rules) (auth.State, error) {
stored, err := r.store.CurrentStateEvents(ctx, room.ID)
if err != nil {
return nil, err
}
events, err := parseEvents(stored, rules)
if err != nil {
return nil, err
}
return auth.NewState(events), nil
}
// ClientEvent is an event in the shape the Client-Server API returns it.
//
// It is not the event as it was signed. A client is given the event ID
// explicitly — in every room version since 3 it is a hash, not a property —
// and is not given the hashes, signatures, auth_events or depth, which are
// a server's business. What it does get that the signed event never carried
// is `unsigned`, which is per-server annotation: the age of the event, the
// state it replaced, and the redaction that stripped it.
type ClientEvent struct {
EventID string `json:"event_id"`
// RoomID is omitted when empty, because a /sync response carries its
// events without one: they are already nested under the room they
// belong to, and repeating it on every event of a client with a
// thousand rooms is a measurable part of every response. Every other
// endpoint sets it, and the spec has a separate name for the shape
// without it — ClientEventWithoutRoomID — rather than two event types.
RoomID string `json:"room_id,omitempty"`
Type string `json:"type"`
StateKey *string `json:"state_key,omitempty"`
Sender string `json:"sender"`
Content json.RawMessage `json:"content"`
OriginServerTS int64 `json:"origin_server_ts"`
Unsigned json.RawMessage `json:"unsigned,omitempty"`
// Redacts is only on an m.room.redaction event, and only in the room
// versions that keep it at the top level. Clients still read it there.
Redacts string `json:"redacts,omitempty"`
}
// clientEvent projects a stored event into the client shape.
func (r *API) clientEvent(stored storage.StoredEvent, e *event.Event, roomID string) ClientEvent {
out := ClientEvent{
EventID: e.EventID(),
RoomID: roomID,
Type: e.Type(),
Sender: e.Sender(),
Content: e.Content(),
OriginServerTS: e.OriginServerTS(),
Redacts: e.Redacts(),
}
if stateKey, ok := e.StateKey(); ok {
out.StateKey = &stateKey
}
out.Unsigned = r.unsigned(stored, e)
return out
}
// unsigned builds the `unsigned` block a client is given.
//
// `age` is computed on the way out rather than stored: it is how long ago
// the event happened as far as this server can tell, and a value written
// down at persist time would be wrong by however long the event sat in the
// database.
func (r *API) unsigned(stored storage.StoredEvent, e *event.Event) json.RawMessage {
block := map[string]json.RawMessage{}
if existing := e.Unsigned(); len(existing) > 0 {
if err := json.Unmarshal(existing, &block); err != nil {
// The unsigned block is this server's own annotation and is
// written as an object everywhere; anything else is a bug here
// rather than something a client sent, and dropping it is
// better than failing the read.
r.log.Warn("stored event has an unreadable unsigned block", "event_id", e.EventID())
block = map[string]json.RawMessage{}
}
}
if age := r.nowMS() - e.OriginServerTS(); age >= 0 {
block["age"] = mustMarshal(age)
}
if stored.RedactedBy != "" {
block["redacted_because"] = mustMarshal(map[string]any{"event_id": stored.RedactedBy})
}
if len(block) == 0 {
return nil
}
return mustMarshal(block)
}
// Event returns one event of a room, if the user may see it.
func (r *API) Event(ctx context.Context, roomID, userID, eventID string) (ClientEvent, error) {
_, rules, err := r.roomAndRules(ctx, roomID)
if err != nil {
return ClientEvent{}, err
}
visible, err := r.requireReadable(ctx, roomID, userID)
if err != nil {
return ClientEvent{}, err
}
stored, err := r.store.Event(ctx, roomID, eventID)
if err != nil {
return ClientEvent{}, err
}
// An event outside the user's window is reported as unknown rather
// than as forbidden: whether a particular event ID exists in a room is
// itself something they are not entitled to know.
if !visible.Contains(stored.StreamOrdering) {
return ClientEvent{}, fmt.Errorf("event %q: %w", eventID, ErrNoSuchEvent)
}
e, err := parseEvent(stored, rules)
if err != nil {
return ClientEvent{}, err
}
return r.clientEvent(stored, e, roomID), nil
}
// RoomState returns a room's state in the client shape: the current state
// for a member, and the state as it was when they left for a user who has.
func (r *API) RoomState(ctx context.Context, roomID, userID string) ([]ClientEvent, error) {
_, rules, err := r.roomAndRules(ctx, roomID)
if err != nil {
return nil, err
}
visible, err := r.requireReadable(ctx, roomID, userID)
if err != nil {
return nil, err
}
return r.stateAt(ctx, roomID, rules, visible.To)
}
// StateEventContent returns the content of one piece of a room's current
// state, which is what GET /rooms/{roomID}/state/{type}/{key} answers.
//
// The endpoint returns the content alone, not the event: a quirk of the
// Client-Server API that predates the rest of it, and one clients depend on.
func (r *API) StateEventContent(ctx context.Context, roomID, userID, eventType, stateKey string) (json.RawMessage, error) {
e, err := r.StateEvent(ctx, roomID, userID, eventType, stateKey)
if err != nil {
return nil, err
}
return e.Content, nil
}
// StateEvent returns one piece of a room's current state as a whole event.
//
// It is what ?format=event asks for on the state endpoint, and the only way
// a client can see who set a piece of state and when.
func (r *API) StateEvent(ctx context.Context, roomID, userID, eventType, stateKey string) (ClientEvent, error) {
_, rules, err := r.roomAndRules(ctx, roomID)
if err != nil {
return ClientEvent{}, err
}
visible, err := r.requireReadable(ctx, roomID, userID)
if err != nil {
return ClientEvent{}, err
}
// A member reads the one row they asked for. A user who has left has to
// be served from the state as of their departure, which means reading
// the snapshot and picking the tuple out of it — the room has moved on
// since, and the current row is not theirs to see.
if visible.To != 0 {
state, err := r.stateAt(ctx, roomID, rules, visible.To)
if err != nil {
return ClientEvent{}, err
}
for _, e := range state {
if e.Type == eventType && e.StateKey != nil && *e.StateKey == stateKey {
return e, nil
}
}
return ClientEvent{}, fmt.Errorf("state %s/%s of %q: %w",
eventType, stateKey, roomID, ErrNoSuchEvent)
}
stored, err := r.store.CurrentStateEvent(ctx, roomID, eventType, stateKey)
if err != nil {
return ClientEvent{}, err
}
e, err := parseEvent(stored, rules)
if err != nil {
return ClientEvent{}, err
}
return r.clientEvent(stored, e, roomID), nil
}
// MembersRequest filters GET /rooms/{roomID}/members.
type MembersRequest struct {
RoomID string
// User is the full user ID asking.
User string
// Membership and NotMembership filter the result; either may be empty.
Membership string
NotMembership string
// At is the stream position to report the membership as of, or 0 for
// the current one.
//
// It is what ?at= asks for, and what a client that is rendering an
// older part of a room's timeline needs: the names and avatars to show
// against those messages are the ones their senders had then, not now.
At int64
}
// Members returns a room's m.room.member events.
func (r *API) Members(ctx context.Context, req MembersRequest) ([]ClientEvent, error) {
_, rules, err := r.roomAndRules(ctx, req.RoomID)
if err != nil {
return nil, err
}
visible, err := r.requireReadable(ctx, req.RoomID, req.User)
if err != nil {
return nil, err
}
// The position to report as of: the one asked for, held inside what
// the user may see. A user who has left is pinned to their departure
// whatever they asked for — the room has moved on since, and the
// current membership is not theirs to read.
at := req.At
if visible.To != 0 && (at == 0 || at > visible.To) {
at = visible.To
}
if at != 0 && at <= visible.From {
return nil, fmt.Errorf("%w: %s may not see %s as it was at %d",
ErrForbidden, req.User, req.RoomID, req.At)
}
var members []ClientEvent
if at != 0 {
state, err := r.stateAt(ctx, req.RoomID, rules, at)
if err != nil {
return nil, err
}
members = filterClientEvents(state, storage.TimelineFilter{Types: []string{event.TypeMember}})
} else {
var filter []string
if req.Membership != "" {
filter = append(filter, req.Membership)
}
stored, err := r.store.Members(ctx, req.RoomID, filter)
if err != nil {
return nil, err
}
if members, err = r.clientEvents(stored, rules, req.RoomID); err != nil {
return nil, err
}
}
out := make([]ClientEvent, 0, len(members))
for _, e := range members {
membership := membershipOf(e)
if req.Membership != "" && membership != req.Membership {
continue
}
if req.NotMembership != "" && membership == req.NotMembership {
continue
}
out = append(out, e)
}
return out, nil
}
// membershipOf reads content.membership out of a projected member event.
//
// A member event that reached a room always has one — the authorisation
// rules refuse one that does not — so an empty answer means the event is
// not a member event at all.
func membershipOf(e ClientEvent) string {
var content struct {
Membership string `json:"membership"`
}
_ = json.Unmarshal(e.Content, &content)
return content.Membership
}
// RoomMember is one joined member, as /joined_members reports them.
type RoomMember struct {
DisplayName string `json:"display_name,omitempty"`
AvatarURL string `json:"avatar_url,omitempty"`
}
// JoinedMembers returns the joined members of a room with their profiles.
//
// Unlike /members this one is for members only: the spec says so, and the
// reason is that it exists to let a client that is in a room render it, not
// to let anyone enumerate who is in it.
func (r *API) JoinedMembers(ctx context.Context, roomID, userID string) (map[string]RoomMember, error) {
if err := r.requireJoined(ctx, roomID, userID); err != nil {
return nil, err
}
_, rules, err := r.roomAndRules(ctx, roomID)
if err != nil {
return nil, err
}
stored, err := r.store.Members(ctx, roomID, []string{event.MembershipJoin})
if err != nil {
return nil, err
}
out := make(map[string]RoomMember, len(stored))
for _, s := range stored {
e, err := parseEvent(s, rules)
if err != nil {
return nil, err
}
stateKey, ok := e.StateKey()
if !ok {
continue
}
var content struct {
DisplayName string `json:"displayname"`
AvatarURL string `json:"avatar_url"`
}
// A member event with a malformed profile still counts as a member;
// the profile is simply not shown.
_ = json.Unmarshal(e.Content(), &content)
out[stateKey] = RoomMember{DisplayName: content.DisplayName, AvatarURL: content.AvatarURL}
}
return out, nil
}
// JoinedRooms lists the rooms a user is joined to.
func (r *API) JoinedRooms(ctx context.Context, userID string) ([]string, error) {
rooms, err := r.store.RoomsForUser(ctx, userID, []string{event.MembershipJoin})
if err != nil {
return nil, err
}
if rooms == nil {
rooms = []string{}
}
return rooms, nil
}
// Membership returns a user's membership in a room, which is empty for a
// user who was never in it.
func (r *API) Membership(ctx context.Context, roomID, userID string) (string, error) {
entry, err := r.store.Membership(ctx, roomID, userID)
if err != nil {
return "", err
}
return entry.Membership, nil
}
// roomAndRules reads a room and the rules its events are under.
func (r *API) roomAndRules(ctx context.Context, roomID string) (storage.Room, roomversion.Rules, error) {
room, err := r.store.Room(ctx, roomID)
if err != nil {
return storage.Room{}, roomversion.Rules{}, err
}
rules, err := r.rules(room)
if err != nil {
return storage.Room{}, roomversion.Rules{}, err
}
return room, rules, nil
}
// requireJoined refuses a caller who is not currently in the room.
func (r *API) requireJoined(ctx context.Context, roomID, userID string) error {
entry, err := r.store.Membership(ctx, roomID, userID)
if err != nil {
return err
}
if entry.Membership != event.MembershipJoin {
return fmt.Errorf("%w: %s is not joined to %s", ErrForbidden, userID, roomID)
}
return nil
}
// CanSendStateEvent reports whether a user may send a state event of a type
// into a room.
//
// It exists for the two endpoints that are *about* a room rather than in it
// — publishing the room in the directory, and deleting one of its aliases —
// where the spec's permission is "could set the canonical alias". Nothing
// that sends an event uses it: an event's permission is decided by the
// authorisation rules when it is sent, and asking twice would be two
// answers to maintain.
func (r *API) CanSendStateEvent(ctx context.Context, roomID, userID, eventType string) (bool, error) {
room, rules, err := r.roomAndRules(ctx, roomID)
if err != nil {
return false, err
}
st, err := r.currentState(ctx, room, rules)
if err != nil {
return false, err
}
if st.Membership(userID) != event.MembershipJoin {
return false, nil
}
create := st.Create()
if create == nil {
return false, fmt.Errorf("%w: %s has no create event in its state", ErrRejected, roomID)
}
powerLevels, err := auth.ResolvePowerLevels(st.PowerLevels(), create, rules)
if err != nil {
return false, fmt.Errorf("%w: %v", ErrRejected, err)
}
return powerLevels.UserLevel(userID) >= powerLevels.SendLevel(eventType, true), nil
}
// RoomVersion returns a room's version, for the handful of callers that
// need to report it.
func (r *API) RoomVersion(ctx context.Context, roomID string) (string, error) {
room, err := r.store.Room(ctx, roomID)
if err != nil {
return "", err
}
return room.Version, nil
}