Files
buzz/mobile/lib/shared/relay/relay_session.dart
Tom Brow 770ab7ff82 perf(mobile): decode relay query responses off the UI isolate (#8669)
### Context

Profiling on an iPhone 12 mini (Profile build, real community) showed
`RelaySessionNotifier._postForEvents` producing contiguous main-isolate
busy runs with p90 110 ms and max 124 ms (27 runs of 50 ms or longer).
Stacks were dominated by `jsonDecode` of `response.body`, with
`NostrEvent.fromJson` smaller. The whole HTTP response decode ran on the
UI isolate: gzip inflate (done by `dart:io`), the UTF-8 decode behind
`response.body`, JSON parsing, and event parsing.

### What changes

- Successful `/query` and `/profiles/search` responses (the only two
callers of `_postForEvents`) are now decoded in a single `Isolate.run`
from `response.bodyBytes`: inflate, UTF-8, JSON, validation,
`NostrEvent.fromJson`. The parsed events come back by ownership transfer
(`Isolate.exit`), not a copy.
- Gzip: the query transport now sends `Accept-Encoding: gzip` and sets
`autoUncompress = false`, so compressed bodies reach the isolate still
compressed. Whether bytes are gzip is decided by the gzip magic bytes,
so a transport that already inflated (or an injected test client) still
works. Error bodies are inflated inline (small).
- Unchanged: `FormatException` messages, the malformed-event wrapping,
`RelayException.body`, and the rate-limit gate reading the error body.
- Fencing (Review-Proven Rule 2): the `_disposed` / `_ageRestricted` /
`_contextGeneration` check still runs before the decode and now runs
again after it. If the session retires during the decode, the query
reports `StateError` even when the decode itself failed, matching the
old ordering where a retired session never decoded.

### Design notes

- `Isolate.run` per response instead of a long-lived worker: no
lifecycle or disposal to manage, no idle isolate heap on a small phone,
easy to fake in tests, and the same shape as `parseProfileEventBatch`.
The main-isolate side of a spawn measured at microseconds, so there is
no size threshold; small responses take the same path.
- Event id and signature verification does not happen on these events on
the main isolate today (nothing in the mobile app verifies queried
events; NIP-OA verification in profile parsing is already in an
isolate), so nothing was moved.
- Other HTTP decode sites found and left alone: invite create/join,
hosted join policy, BuilderLab account, media upload response, community
icon lookup, dev push lease NIP-11, archived-agents. These decode small,
rare responses (one small JSON object each), where an isolate adds
complexity without a measurable gain. Media and avatar byte downloads do
no JSON decoding. The mobile app has no `/count` caller.

### Validation

Added tests (`relay_session_test.dart`, through the production
`queryRelay` / `searchProfiles` path with the real isolate decoder):
parsing with non-ASCII content; profile search; gzip success body and
`Accept-Encoding`; gzip error body feeding the rate-limit gate and
`RelayException.body`; malformed JSON; non-list body; non-object
element; malformed event; invalid UTF-8; error status skips decoding;
session disposed, age-restricted, or retired-before-decode while a
decode is outstanding.

The fencing tests are falsifiable: removing the post-decode check fails
the disposed and age-restricted tests; removing the
retired-wins-over-decode-error branch fails its test; removing the
pre-decode check fails the never-decodes test.

`FakeRelay` (test helper) now injects an inline decoder, because
`fakeAsync` cannot deliver isolate results; it calls the same parse
function.

The default transport still honors `http.runWithClient`, which the
hosted journey tests use to fake the network
(`relay_http_query_client_test.dart`; fails without the fix).

`just mobile-check` and `just mobile-test` pass.

Host benchmark (AOT executable on an Apple silicon Mac, median of 25,
copy of the parse logic, main-isolate stall measured with a 0.5 ms
heartbeat timer):

| Response | Main-isolate blocked before | after |
| --- | --- | --- |
| 500 events, ~630 KB | 2.4 ms | 0 ms (max gap 0.05 ms) |
| 500 events, ~630 KB gzip | 3.2 ms | 0 ms (max gap 0.19 ms) |
| 500 events with 50 20 KB avatars, ~1.6 MB | 5.1 ms | 0 ms (max gap
0.28 ms) |
| 500 events, ~4.5 MB | 14.0 ms | 0 ms (max gap 0.27 ms) |
| 500 events, ~4.5 MB (827 KB gzip) | 17.4 ms | 0 ms (max gap 0.35 ms) |

Total latency is unchanged within noise (for example 14.5 ms to 14.9
ms). The host is much faster than the 12 mini, so absolute device
savings should be several times larger.

On the iPhone 12 mini (Profile builds of base and this branch, real
community over the production relay, two runs per scenario; main-isolate
stalls measured with a 4 ms heartbeat timer, since the wireless device
tunnel has no VM service):

| Scenario | Main-isolate `_postForEvents` time, base | this branch |
Max stall, base | this branch |
| --- | --- | --- | --- | --- |
| Cold start + 60 s (~16k events, 7.8 MB gzip) | 282 to 295 ms (max 116
ms) | 0.01 ms | 147 to 150 ms | 100 to 102 ms |
| Inbox + search (~7k events) | 122 to 123 ms (max 111 ms) | 0.01 ms |
138 to 142 ms | 97 to 99 ms |
| Open 6 channels (558 events) | 5 to 6 ms | 0.01 ms | 88 ms | 86 to 87
ms |
| Background 10 s + resume (~7k events) | 118 to 124 ms | 0.01 ms | n/a
(suspension) | n/a |

- The production relay serves gzip (about 4.4:1; the largest catch-up
response is 3.15 MB on the wire, 13.9 MB inflated, 7k events). Inside
the isolate that response takes 30 to 47 ms to inflate, 39 to 55 ms for
UTF-8, 65 to 105 ms for `jsonDecode` and 7 to 32 ms for
`NostrEvent.fromJson`.
- Many 16 to 90 ms stalls remain on both builds; they come from other
work (mostly markdown build and layout), not response decoding.
- Home, channels, inbox, search, profile and a thread render the same on
both builds.

buzz-review-completed

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Signed-off-by: Tom Brow <tomb@block.xyz>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-29 08:44:52 -07:00

1169 lines
39 KiB
Dart

import 'dart:async';
import 'dart:convert';
import 'dart:math';
import 'package:http/http.dart' as http;
import 'package:nostr/nostr.dart' as nostr;
import 'package:pointycastle/digests/sha256.dart';
import 'package:uuid/uuid.dart';
import 'package:flutter/foundation.dart';
import 'package:hooks_riverpod/hooks_riverpod.dart';
import '../../features/age_gate/age_signal_provider.dart';
import '../auth/auth.dart';
import '../hosted/hosted_relay_auth.dart';
import '../hosted/builderlab_account.dart';
import 'nostr_filters.dart';
import 'nostr_models.dart';
import 'relay_client.dart';
import 'relay_closed_policy.dart';
import 'relay_http_query_client.dart';
import 'relay_provider.dart';
import 'relay_query_response.dart';
import 'relay_rate_limit_gate.dart';
import 'relay_reconnect_policy.dart';
import 'relay_session_types.dart';
import 'relay_socket.dart';
import 'package:buzz/shared/crypto/event_signing.dart';
export 'relay_session_types.dart';
part 'relay_session_auth.dart';
part 'relay_session_subscriptions.dart';
class RelaySessionNotifier extends Notifier<SessionState> {
RelaySessionNotifier({
http.Client? httpClient,
http.Client Function()? httpClientFactory,
RelaySocketFactory socketFactory = RelaySocket.new,
DateTime Function()? now,
RelayRateLimitGate? rateLimitGate,
RelayTimerFactory retryTimerFactory = Timer.new,
Future<void> Function(Duration) replayDelay = Future.delayed,
RelayTimerFactory reconnectTimerFactory = Timer.new,
Random? random,
RelayQueryDecoder queryDecoder = decodeRelayQueryResponse,
}) : _httpQueryClient = RelayHttpQueryClient(
client: httpClient,
clientFactory: httpClientFactory,
),
_socketFactory = socketFactory,
_now = now ?? DateTime.now,
_rateLimitGate = rateLimitGate ?? RelayRateLimitGate(),
_retryTimerFactory = retryTimerFactory,
_replayDelay = replayDelay,
_reconnectTimerFactory = reconnectTimerFactory,
_random = random ?? Random(),
_queryDecoder = queryDecoder;
final RelayHttpQueryClient _httpQueryClient;
final RelaySocketFactory _socketFactory;
final DateTime Function() _now;
final RelayRateLimitGate _rateLimitGate;
final RelayTimerFactory _retryTimerFactory;
final Future<void> Function(Duration) _replayDelay;
final RelayTimerFactory _reconnectTimerFactory;
final Random _random;
final RelayQueryDecoder _queryDecoder;
static const _baseReconnectDelayMs = relayReconnectBaseDelayMs;
static const _maxReconnectDelayMs = relayReconnectMaxDelayMs;
static const _eventBatchMs = 16;
static const _replayBatchSize = 8;
static const _replayInterBatchDelay = Duration(milliseconds: 50);
static const _maxRecentDeliveryKeys = 5000;
static const _backgroundGraceDuration = Duration(seconds: 5);
RelaySocket? _socket;
final Map<String, _HistorySubscription> _historySubscriptions = {};
final Map<String, _LiveSubscription> _liveSubscriptions = {};
final Map<String, _ClosedRetry> _pendingClosedRetries = {};
final Map<String, _PendingEvent> _pendingEvents = {};
final List<_BufferedEvent> _eventBuffer = [];
final Set<String> _recentDeliveryKeys = {};
Timer? _reconnectTimer;
Timer? _flushTimer;
Timer? _backgroundGraceTimer;
DateTime? _backgroundedAt;
DateTime? _connectedAt;
DateTime? _lastInboundAt;
int _reconnectDelayMs = _baseReconnectDelayMs;
int _subIdCounter = 0;
bool _disposed = false;
bool _ageRestricted = false;
bool _paused = false;
bool _hasConnectedOnce = false;
int _connectionGeneration = 0;
int _contextGeneration = 0;
final Map<Object, String> _visibleChannelsByOwner = {};
final Map<Object, Future<void> Function()> _beforePauseCallbacks = {};
bool _socketConnected = false;
bool _closedRetryReplayScheduled = false;
String? _relayUrl;
/// The relay origin this session serves. Switching community rebuilds the
/// session in place, so work begun for another community compares this
/// before reading from or publishing through it.
String? get relayUrl => _relayUrl;
@visibleForTesting
set debugRelayUrl(String? value) => _relayUrl = value;
@override
SessionState build() {
final config = ref.watch(relayConfigProvider);
_relayUrl = config.baseUrl;
final authState = ref.watch(authProvider);
_ageRestricted = ref.watch(ageSignalProvider) == AgeSignalState.restricted;
// Reset disposed flag — build() may re-run on the same Notifier instance
// after a provider dependency changes (e.g. auth completing).
_disposed = false;
ref.onDispose(_dispose);
// Auto-connect when authenticated and we have a signing key (NIP-42 AUTH).
final isAuthenticated = authState.value?.status == AuthStatus.authenticated;
if (!_ageRestricted && isAuthenticated && config.nsec != null) {
// Schedule connection after build completes.
Future.microtask(() => _connect(config));
}
return const SessionState(status: SessionStatus.disconnected);
}
/// Execute a one-shot query via the relay's HTTP bridge (`POST /query`).
Future<List<NostrEvent>> queryRelay(
List<NostrFilter> filters, {
Duration timeout = const Duration(seconds: 8),
}) => _postForEvents(
'/query',
filters.map((filter) => filter.toJson()).toList(),
timeout: timeout,
);
/// Finds people by profile name, for typeahead surfaces.
///
/// NIP-50 search over kind:0 matches the whole profile JSON, so text in
/// avatar data URLs, `about` or `nip05` can fill every result page and hide
/// a real name match. The relay's `POST /profiles/search` applies the same
/// prefix rule to just `name` and `display_name` and puts exact and prefix
/// name matches first.
///
/// A relay without that endpoint answers 404 or 405. The search then uses
/// NIP-50 ([NostrFilters.searchUsers]), and so do later searches on that
/// relay for [profileSearchRetry]; after that the endpoint is tried again.
/// A query or limit the endpoint does not accept also uses NIP-50.
Future<List<NostrEvent>> searchProfiles(
String query, {
int limit = 50,
Duration timeout = const Duration(seconds: 8),
}) async {
final text = query.trim();
final relay = ref.read(relayConfigProvider).baseUrl;
final fallback = _profileSearchFallback;
if (text.isNotEmpty &&
text.runes.length <= _profileSearchQueryMax &&
limit >= 1 &&
limit <= _profileSearchLimitMax &&
(fallback == null ||
fallback.relay != relay ||
!_now().isBefore(fallback.until))) {
try {
final events = await _postForEvents('/profiles/search', {
'query': text,
'limit': limit,
}, timeout: timeout);
return [
for (final event in events)
if (event.kind == 0) event,
];
} on RelayException catch (error) {
if (error.statusCode != 404 && error.statusCode != 405) rethrow;
_profileSearchFallback = (
relay: relay,
until: _now().add(profileSearchRetry),
);
}
}
return queryRelay([
NostrFilters.searchUsers(text, limit: limit),
], timeout: timeout);
}
/// How long a relay's 404 or 405 keeps [searchProfiles] on NIP-50.
static const profileSearchRetry = Duration(minutes: 5);
static const _profileSearchQueryMax = 100;
static const _profileSearchLimitMax = 100;
({String relay, DateTime until})? _profileSearchFallback;
/// POSTs [payload] as JSON to relay [path] with NIP-98 auth and decodes
/// the event array it answers.
Future<List<NostrEvent>> _postForEvents(
String path,
Object payload, {
required Duration timeout,
}) async {
if (_disposed || _ageRestricted) {
throw StateError('Relay session is unavailable');
}
final generation = _contextGeneration;
final config = ref.read(relayConfigProvider);
final url = Uri.parse(config.baseUrl).resolve(path).toString();
final bodyBytes = utf8.encode(jsonEncode(payload));
// Reuse the session transport on success. A timeout rotates immediately
// for new queries, then closes the retired client after its peers finish.
final response = await _httpQueryClient.post(
Uri.parse(url),
headers: {
'Authorization': buildNip98AuthHeader(
method: 'POST',
url: url,
bodyBytes: bodyBytes,
nsec: config.nsec,
pubkey: ref.read(myPubkeyProvider),
),
'Content-Type': 'application/json',
},
body: bodyBytes,
timeout: timeout,
decorate: (client, active) => HostedRelayHttpClient(
ref.read(hostedRelayAuthProvider),
config,
inner: client,
canSend: () =>
active() && !_disposed && generation == _contextGeneration,
),
);
bool retired() =>
_disposed || _ageRestricted || generation != _contextGeneration;
if (retired()) {
throw StateError('Relay query belongs to a retired session');
}
if (response.statusCode < 200 || response.statusCode >= 300) {
final body = relayResponseText(
response.bodyBytes,
response.headers['content-type'],
);
_activateRateLimitGateFromHttpError(body);
throw RelayException(response.statusCode, body);
}
// Decoding runs off the UI isolate, so the session can retire meanwhile.
// A retired session reports that, not whatever its bytes decoded to.
final List<NostrEvent> events;
try {
events = await _queryDecoder(response);
} catch (_) {
if (retired()) {
throw StateError('Relay query belongs to a retired session');
}
rethrow;
}
if (retired()) {
throw StateError('Relay query belongs to a retired session');
}
return events;
}
void _activateRateLimitGateFromHttpError(String body) {
final dynamic decoded;
try {
decoded = jsonDecode(body);
} on FormatException {
return;
}
if (decoded is! Map<String, dynamic>) return;
final message = decoded['error'];
if (message is! String ||
classifyRelayClosed(message) != RelayClosedClass.rateLimited) {
return;
}
_rateLimitGate.activate(parseRateLimitRetrySeconds(message));
}
/// Fetch historical events matching [filter]. Sends REQ, collects events
/// until EOSE, then resolves. One-shot subscription.
Future<List<NostrEvent>> fetchHistory(
NostrFilter filter, {
Duration timeout = const Duration(seconds: 8),
}) async {
if (_rateLimitGate.isActive) await _rateLimitGate.wait();
if (_disposed) throw StateError('Relay session is disposed');
final subId = _nextSubId('h');
final completer = Completer<List<NostrEvent>>();
final timer = Timer(timeout, () {
final sub = _historySubscriptions.remove(subId);
if (sub != null && !sub.completer.isCompleted) {
sub.completer.completeError(
TimeoutException('Relay history request timed out after $timeout'),
);
}
if (sub?.heldFilter == null) _sendClose(subId);
});
final subscription = _HistorySubscription(
completer: completer,
timeout: timer,
);
_historySubscriptions[subId] = subscription;
// The relay rejects a REQ sent before NIP-42 AUTH completes. Hold it for
// the authenticated connection instead of paying a rejection and retry.
if (_socketConnected) {
_sendReq(subId, filter);
} else {
subscription.heldFilter = filter;
}
return completer.future;
}
Future<void Function()> subscribe(
NostrFilter filter,
void Function(NostrEvent) onEvent, {
void Function(String message)? onClosed,
}) => _subscribe(filter, onEvent, onClosed: onClosed);
/// Subscribe to a live stream and observe its recovery lifecycle.
///
/// The returned future completes after initial EOSE (or the existing
/// fallback timeout) and yields a cleanup callback. [onStatusChanged] emits
/// [RelaySubscriptionStatus.ready] at each EOSE after buffered replay events
/// have been delivered, and [RelaySubscriptionStatus.retrying] immediately
/// when a retryable or rate-limited CLOSED begins backoff. Terminal CLOSED
/// invokes [onClosed] and removes the subscription instead of retrying it.
/// Calling the cleanup callback cancels pending retries and sends CLOSE.
Future<void Function()> subscribeWithStatus(
NostrFilter filter,
void Function(NostrEvent) onEvent, {
void Function(String message)? onClosed,
required void Function(RelaySubscriptionStatus status) onStatusChanged,
}) => _subscribe(
filter,
onEvent,
onClosed: onClosed,
onStatusChanged: onStatusChanged,
);
Future<void Function()> _subscribe(
NostrFilter filter,
void Function(NostrEvent) onEvent, {
void Function(String message)? onClosed,
void Function(RelaySubscriptionStatus status)? onStatusChanged,
}) async {
if (_disposed) throw StateError('Relay session is disposed');
final subId = _nextSubId('l');
final readyCompleter = Completer<void>();
final subscription = _LiveSubscription(
filter: filter,
onEvent: onEvent,
onClosed: onClosed,
onStatusChanged: onStatusChanged,
readyCompleter: readyCompleter,
);
_liveSubscriptions[subId] = subscription;
// Before AUTH the relay closes it as auth-required, which is terminal.
// The replay on connect sends it instead.
if (_socketConnected) _sendLiveReq(subId, subscription);
try {
await readyCompleter.future.timeout(
const Duration(milliseconds: 500),
onTimeout: () {},
);
} catch (_) {
_liveSubscriptions.remove(subId);
_recentDeliveryKeys.removeWhere((key) => key.startsWith('$subId:'));
rethrow;
}
final liveSub = _liveSubscriptions[subId];
if (liveSub != null && liveSub.readyCompleter == readyCompleter) {
liveSub.readyCompleter = null;
}
return () => _unsubscribe(subId);
}
Future<NostrEvent> publish(
NostrEvent event, {
Duration timeout = const Duration(seconds: 8),
}) async {
final generation = _connectionGeneration;
if (_rateLimitGate.isActive) await _rateLimitGate.wait();
if (!_isActiveConnection(generation) || !_socketConnected) {
throw StateError('Relay session is not connected');
}
final completer = Completer<NostrEvent>();
final timer = Timer(timeout, () {
final pending = _pendingEvents.remove(event.id);
if (pending != null && !pending.completer.isCompleted) {
pending.completer.completeError(
TimeoutException(
'Event ${event.id} not acknowledged within $timeout',
),
);
}
});
_pendingEvents[event.id] = _PendingEvent(
completer: completer,
timeout: timer,
);
_socket?.send(['EVENT', event.toJson()]);
return completer.future;
}
void sendRaw(List<dynamic> payload) {
_socket?.send(payload);
}
@visibleForTesting
void debugHandleMessage(List<dynamic> data) => _handleMessage(data);
@visibleForTesting
void debugFlushEventBuffer() => _flushEventBuffer();
@visibleForTesting
Future<void> debugHandleConnected() =>
_handleConnected(_connectionGeneration);
@visibleForTesting
Future<void> debugReplayLiveSubscriptions() =>
_replayLiveSubscriptions(_connectionGeneration);
@visibleForTesting
void debugDispose() => _dispose();
@visibleForTesting
void debugSupersedeConnection() => _connectionGeneration++;
@visibleForTesting
void debugHandleDisconnected([Object? error]) {
_socketConnected = false;
_handleDisconnected(_connectionGeneration, error);
}
@visibleForTesting
void debugResetClosedRetriesForDisconnect() {
_socketConnected = false;
_resetAllClosedRetries();
}
@visibleForTesting
void debugSetSessionStatus(SessionStatus status) {
_socketConnected = status == SessionStatus.connected;
}
@visibleForTesting
void debugPauseNow() => _pauseNow();
@visibleForTesting
void debugHandleSocketMessageForTest(List<dynamic> data) =>
_handleMessage(data);
@visibleForTesting
void debugAttachSocketForTest(RelaySocket socket) {
_socket?.dispose();
_socket = socket;
_socketConnected = true;
}
/// Registers a visible channel and returns an owner-scoped release callback.
/// The most recently registered owner is prioritized during reconnect replay.
void Function() registerVisibleChannel(String channelId) {
final owner = Object();
_visibleChannelsByOwner[owner] = channelId;
return () => _visibleChannelsByOwner.remove(owner);
}
/// Force a reconnect (e.g., returning from background).
Future<void> reconnect() async {
_socketConnected = false;
await _socket?.disconnect();
_reconnectDelayMs = _baseReconnectDelayMs;
final config = ref.read(relayConfigProvider);
await _connect(config);
}
/// Registers work that must settle before the background grace disconnect.
void Function() registerBeforePause(Future<void> Function() callback) {
final owner = Object();
_beforePauseCallbacks[owner] = callback;
return () => _beforePauseCallbacks.remove(owner);
}
/// Called by the app lifecycle provider when the app goes to background.
void onAppPaused() {
_backgroundedAt = _now();
_backgroundGraceTimer?.cancel();
_backgroundGraceTimer = Timer(_backgroundGraceDuration, () {
unawaited(_pauseAfterCallbacks());
});
}
Future<void> _pauseAfterCallbacks() async {
final callbacks = _beforePauseCallbacks.values.toList();
try {
await Future.wait(callbacks.map((callback) => callback()));
} catch (error) {
debugPrint('Background cleanup failed: $error');
}
if (_backgroundedAt != null) _pauseNow();
}
void _pauseNow() {
_paused = true;
_socketConnected = false;
_reconnectTimer?.cancel();
_cancelAllHistory(Exception('App moved to background'));
_rejectAllPending(Exception('App moved to background'));
_socket?.disconnect();
state = const SessionState(status: SessionStatus.disconnected);
}
/// Called by the app lifecycle provider when the app returns to foreground.
void onAppResumed() {
_paused = false;
final backgroundedAt = _backgroundedAt;
_backgroundedAt = null;
_backgroundGraceTimer?.cancel();
_backgroundGraceTimer = null;
final backgroundedLongEnoughToRequireReconnect =
backgroundedAt != null &&
_now().difference(backgroundedAt) >= _backgroundGraceDuration;
if (!backgroundedLongEnoughToRequireReconnect &&
state.status == SessionStatus.connected) {
return;
}
// Cancel any in-flight reconnect backoff timer so we reconnect immediately
// instead of waiting for the (possibly large) exponential delay.
_reconnectTimer?.cancel();
_reconnectDelayMs = _baseReconnectDelayMs;
final config = ref.read(relayConfigProvider);
_connect(config);
}
Future<void> _connect(RelayConfig config, {bool renewed = false}) async {
if (_disposed || _ageRestricted) return;
// Whatever connection came before ends here, capped at its last frame.
_markLiveSubscriptionsClosed();
_connectedAt = null;
final generation = ++_connectionGeneration;
// A replacement socket is unauthenticated until its AUTH completes.
_socketConnected = false;
state = SessionState(
status: _hasConnectedOnce
? SessionStatus.reconnecting
: SessionStatus.connecting,
reconnectAttempt: state.reconnectAttempt,
);
_socket?.dispose();
late final RelaySocket socket;
socket = _socketFactory(
wsUrl: config.wsUrl,
nsec: config.nsec,
onMessage: (message) {
if (generation == _connectionGeneration) _handleMessage(message);
},
onConnected: () => _handleConnected(generation),
onDisconnected: (error) {
if (_disposed || generation != _connectionGeneration) return;
final auth = ref.read(hostedRelayAuthProvider);
if (auth.protects(config) && relayUpgradeRefused(error)) {
auth.rejected(
config,
socket.headers['Nostr-Federated-Identity'],
terminal: renewed,
);
if (!renewed) {
unawaited(_connect(config, renewed: true));
return;
}
error = AccountFailure.denied;
}
_handleDisconnected(generation, error, closeCode: socket.closeCode);
},
);
_socket = socket;
try {
final auth = ref.read(hostedRelayAuthProvider);
if (auth.protects(config)) socket.headers = await auth.headers(config);
if (_disposed || generation != _connectionGeneration) return;
await socket.connect();
} on AccountFailure catch (error) {
_handleDisconnected(generation, error);
}
}
Future<void> _handleConnected(int generation) async {
if (_disposed || generation != _connectionGeneration) return;
_socketConnected = true;
ref.read(hostedRelayAuthProvider).connected(ref.read(relayConfigProvider));
_hasConnectedOnce = true;
// The backoff is not reset here: a client bounced repeatedly must keep
// backing off. [_handleDisconnected] resets it after a stable connection.
_connectedAt = _now();
_lastInboundAt = _connectedAt;
state = const SessionState(status: SessionStatus.connected);
await _replayLiveSubscriptions(generation);
}
void _handleDisconnected(int generation, Object? error, {int? closeCode}) {
if (_disposed || generation != _connectionGeneration) return;
_socketConnected = false;
final connectedAt = _connectedAt;
_connectedAt = null;
if (connectedAt != null &&
isRelayConnectionStable(_now().difference(connectedAt))) {
_reconnectDelayMs = _baseReconnectDelayMs;
}
_cancelAllHistory(error);
_rejectAllPending(error);
_resetAllClosedRetries();
_eventBuffer.clear();
_flushTimer?.cancel();
_flushTimer = null;
if (error is RelayAuthRejectedException ||
(error is AccountFailure && error != AccountFailure.unavailable)) {
_reconnectTimer?.cancel();
state = const SessionState(status: SessionStatus.disconnected);
return;
}
_scheduleReconnect(closeCode);
}
void _scheduleReconnect(int? closeCode) {
if (_disposed || _paused) return;
final attempt = state.reconnectAttempt + 1;
state = SessionState(
status: SessionStatus.reconnecting,
reconnectAttempt: attempt,
);
_reconnectTimer?.cancel();
final delay = relayReconnectDelay(
backoffMs: _reconnectDelayMs,
closeCode: closeCode,
jitterUnit: _random.nextDouble(),
spreadUnit: _random.nextDouble(),
);
_reconnectTimer = _reconnectTimerFactory(delay, () {
_reconnectDelayMs = min(_reconnectDelayMs * 2, _maxReconnectDelayMs);
final config = ref.read(relayConfigProvider);
_connect(config);
});
}
/// Records when each subscription that was live on the closing connection
/// stopped being live, so the next replay can resume from there.
///
/// Liveness ends at the last frame received, not at the close: after a
/// device sleep the loss is only noticed on wake, and nothing arrived in
/// between.
void _markLiveSubscriptionsClosed() {
final now = _now();
final inbound = _lastInboundAt;
final end = inbound == null || inbound.isAfter(now) ? now : inbound;
final closedAt = end.millisecondsSinceEpoch ~/ 1000;
for (final subscription in _liveSubscriptions.values) {
if (!subscription.liveOnConnection) continue;
subscription.liveOnConnection = false;
subscription.liveUntil = closedAt;
}
}
/// Replay all live subscriptions after a reconnect, with a time skew to
/// catch events that occurred during the disconnect.
Future<void> _replayLiveSubscriptions(int generation) async {
if (_rateLimitGate.isActive) await _rateLimitGate.wait();
if (!_isActiveConnection(generation)) return;
_sendHeldHistory();
final entries = _liveSubscriptions.entries.toList();
final visibleChannelId = _visibleChannelsByOwner.isEmpty
? null
: _visibleChannelsByOwner.values.last;
if (visibleChannelId != null) {
entries.sort((left, right) {
final leftVisible =
left.value.filter.tags['#h']?.contains(visibleChannelId) ?? false;
final rightVisible =
right.value.filter.tags['#h']?.contains(visibleChannelId) ?? false;
if (leftVisible == rightVisible) return 0;
return leftVisible ? -1 : 1;
});
}
await _sendReplayBatches(entries, generation);
}
Future<void> _replayPendingClosedRetries(int generation) async {
if (!_isActiveConnection(generation)) return;
final entries = _pendingClosedRetries.entries
.where((entry) => entry.value.generation == generation)
.map(
(entry) => MapEntry<String, _LiveSubscription>(
entry.key,
entry.value.subscription,
),
)
.toList();
await _sendReplayBatches(entries, generation, pendingClosedRetries: true);
}
Future<void> _sendReplayBatches(
List<MapEntry<String, _LiveSubscription>> entries,
int generation, {
bool pendingClosedRetries = false,
}) async {
for (var i = 0; i < entries.length; i += _replayBatchSize) {
if (_rateLimitGate.isActive) await _rateLimitGate.wait();
if (!_isActiveConnection(generation)) return;
final batch = entries.sublist(
i,
min(i + _replayBatchSize, entries.length),
);
for (final entry in batch) {
if (_liveSubscriptions[entry.key] != entry.value) continue;
if (pendingClosedRetries) {
final pendingRetry = _pendingClosedRetries[entry.key];
if (pendingRetry?.subscription != entry.value ||
pendingRetry?.generation != generation) {
continue;
}
_pendingClosedRetries.remove(entry.key);
}
_sendLiveReq(entry.key, entry.value);
}
if (i + _replayBatchSize < entries.length) {
await _replayDelay(_replayInterBatchDelay);
}
}
}
bool _isActiveConnection(int generation) =>
!_disposed && generation == _connectionGeneration;
NostrFilter _replayFilter(_LiveSubscription subscription) {
final filter = subscription.filter;
final since = relayResumeSince(
originalSince: filter.since,
until: filter.until,
lastSeenCreatedAt: subscription.lastSeenCreatedAt,
liveUntil: subscription.liveUntil,
);
return since == null || since == filter.since
? filter
: filter.copyWithSince(since);
}
void _handleMessage(List<dynamic> data) {
_lastInboundAt = _now();
if (data.isEmpty) return;
final type = data[0] as String;
switch (type) {
case 'EVENT':
_handleEvent(data);
case 'EOSE':
_handleEose(data);
case 'CLOSED':
_handleClosed(data);
case 'OK':
_handleOk(data);
}
}
void _handleEvent(List<dynamic> data) {
if (data.length < 3) return;
final subId = data[1] as String;
final eventJson = data[2] as Map<String, dynamic>;
final event = NostrEvent.fromJson(eventJson);
// History subscriptions accumulate immediately.
final historySub = _historySubscriptions[subId];
if (historySub != null) {
historySub.events.add(event);
return;
}
// Live subscriptions get batched. An EVENT proves the stream is active,
// but not that a retry replay is complete; only EOSE is that boundary.
final liveSub = _liveSubscriptions[subId];
if (liveSub != null) {
_resetClosedRetry(liveSub);
// Track last seen timestamp for reconnect replay.
if (liveSub.lastSeenCreatedAt == null ||
event.createdAt > liveSub.lastSeenCreatedAt!) {
liveSub.lastSeenCreatedAt = event.createdAt;
}
_eventBuffer.add(_BufferedEvent(subId, event));
_scheduleFlush();
}
}
void _handleEose(List<dynamic> data) {
if (data.length < 2) return;
final subId = data[1] as String;
// History subscription: resolve with collected events.
final historySub = _historySubscriptions.remove(subId);
if (historySub != null) {
historySub.timeout.cancel();
if (!historySub.completer.isCompleted) {
historySub.completer.complete(historySub.events);
}
_sendClose(subId);
return;
}
// Live subscription: flush replay callbacks before signaling ready. This
// ordering matters for retry replays, whose original ready completer has
// already been released.
final liveSub = _liveSubscriptions[subId];
if (liveSub != null) {
// A second REQ on this socket makes its EOSE ambiguous: it may belong
// to the REQ it replaced, so it cannot vouch for the replacement.
liveSub.liveOnConnection =
liveSub.ambiguousGeneration != _connectionGeneration;
_resetClosedRetry(liveSub);
_flushBufferedEventsNow();
liveSub.onStatusChanged?.call(RelaySubscriptionStatus.ready);
}
if (liveSub != null &&
liveSub.readyCompleter != null &&
!liveSub.readyCompleter!.isCompleted) {
liveSub.readyCompleter!.complete();
liveSub.readyCompleter = null;
}
}
void _handleClosed(List<dynamic> data) {
if (data.length < 2) return;
final subId = data[1] as String;
final message = data.length >= 3 && data[2] is String
? data[2] as String
: 'subscription closed by relay';
final closedClass = classifyRelayClosed(message);
final historySub = _historySubscriptions.remove(subId);
if (historySub != null) {
if (closedClass == RelayClosedClass.rateLimited) {
_rateLimitGate.activate(parseRateLimitRetrySeconds(message));
}
historySub.timeout.cancel();
if (!historySub.completer.isCompleted) {
historySub.completer.completeError(Exception(message));
}
return;
}
final liveSub = _liveSubscriptions[subId];
if (liveSub == null) return;
final readyCompleter = liveSub.readyCompleter;
// The stream is interrupted until the retry's EOSE.
liveSub.liveOnConnection = false;
// A deadline is terminal for one-shot history, but a live subscription
// retries it: same filter under backoff, which can keep retrying at the
// cap. `since` is not advanced, so events missed while closed aren't
// skipped.
if (closedClass == RelayClosedClass.terminal &&
!isRelayDeadlineError(message)) {
if (readyCompleter != null && !readyCompleter.isCompleted) {
readyCompleter.completeError(Exception(message));
}
liveSub.onClosed?.call(message);
_removeLiveSubscription(subId, liveSub);
return;
}
if (readyCompleter != null && !readyCompleter.isCompleted) {
readyCompleter.complete();
liveSub.readyCompleter = null;
}
liveSub.onStatusChanged?.call(RelaySubscriptionStatus.retrying);
if (liveSub.closedRetryTimer != null) return;
final attempt = liveSub.closedRetryAttempt;
final backoffMs = attempt >= 5
? _maxReconnectDelayMs
: _baseReconnectDelayMs * (1 << attempt);
var delayMs = backoffMs;
if (closedClass == RelayClosedClass.rateLimited) {
final retrySeconds = parseRateLimitRetrySeconds(message);
_rateLimitGate.activate(retrySeconds);
final fallbackMs =
(retrySeconds != null
? min(retrySeconds, RelayRateLimitGate.maxRetrySeconds)
: RelayRateLimitGate.defaultRetrySeconds) *
1000;
delayMs = max(
backoffMs,
_rateLimitGate.remainingMs() == 0
? fallbackMs
: _rateLimitGate.remainingMs(),
);
}
liveSub.closedRetryAttempt = attempt + 1;
final retryGeneration = _connectionGeneration;
liveSub.closedRetryTimer = _retryTimerFactory(
Duration(milliseconds: delayMs),
() async {
liveSub.closedRetryTimer = null;
if (!_isActiveConnection(retryGeneration) ||
_liveSubscriptions[subId] != liveSub) {
return;
}
if (_rateLimitGate.isActive) await _rateLimitGate.wait();
if (!_isActiveConnection(retryGeneration) ||
_liveSubscriptions[subId] != liveSub ||
!_socketConnected) {
return;
}
_pendingClosedRetries[subId] = _ClosedRetry(
subscription: liveSub,
generation: retryGeneration,
);
_scheduleClosedRetryReplay(retryGeneration);
},
);
}
void _scheduleClosedRetryReplay(int generation) {
if (_closedRetryReplayScheduled) return;
_closedRetryReplayScheduled = true;
scheduleMicrotask(() async {
try {
await _replayPendingClosedRetries(generation);
} finally {
_closedRetryReplayScheduled = false;
_pendingClosedRetries.removeWhere(
(_, retry) => retry.generation != _connectionGeneration,
);
if (_pendingClosedRetries.values.any(
(retry) => retry.generation == _connectionGeneration,
)) {
_scheduleClosedRetryReplay(_connectionGeneration);
}
}
});
}
void _handleOk(List<dynamic> data) {
if (data.length < 3) return;
final eventId = data[1] as String;
final accepted = data[2] as bool;
final message = data.length > 3 && data[3] is String
? data[3] as String
: '';
final pending = _pendingEvents.remove(eventId);
if (pending == null) return;
pending.timeout.cancel();
if (accepted) {
// We don't have the full event here; create a minimal placeholder.
// Command kinds (e.g. 41010, 30620, 46020) return "response:{...}" in
// the OK message — preserve it in `content` so callers can parse it.
if (!pending.completer.isCompleted) {
pending.completer.complete(
NostrEvent(
id: eventId,
pubkey: '',
createdAt: 0,
kind: 0,
tags: [],
content: message,
sig: '',
),
);
}
} else {
// Back-pressure now arrives here rather than as a NOTICE: the relay
// rejects an over-quota EVENT on the OK channel so this pending publish
// can be settled at all. Without arming the gate the send would fail
// without ever backing off.
if (message.startsWith('rate-limited:')) {
_rateLimitGate.activate(parseRateLimitRetrySeconds(message));
}
if (!pending.completer.isCompleted) {
pending.completer.completeError(
Exception(message.isNotEmpty ? message : 'Event rejected'),
);
}
}
}
void _scheduleFlush() {
_flushTimer ??= Timer(
const Duration(milliseconds: _eventBatchMs),
_flushEventBuffer,
);
}
void _flushBufferedEventsNow() {
_flushTimer?.cancel();
_flushTimer = null;
_flushEventBuffer();
}
void _flushEventBuffer() {
_flushTimer = null;
if (_eventBuffer.isEmpty) return;
final batch = List<_BufferedEvent>.from(_eventBuffer);
_eventBuffer.clear();
for (final buffered in batch) {
final sub = _liveSubscriptions[buffered.subId];
if (sub == null) continue;
// Deduplicate per subscription. The same relay event can legitimately
// match multiple live subscriptions, e.g. the channel list unread listener
// and the open channel message listener.
final deliveryKey = '${buffered.subId}:${buffered.event.id}';
if (_recentDeliveryKeys.contains(deliveryKey)) continue;
// Cap the dedup set to prevent unbounded memory growth.
if (_recentDeliveryKeys.length >= _maxRecentDeliveryKeys) {
_recentDeliveryKeys.clear();
}
_recentDeliveryKeys.add(deliveryKey);
sub.onEvent(buffered.event);
}
}
String _nextSubId(String prefix) {
_subIdCounter++;
return '$prefix-$_subIdCounter';
}
void _sendLiveReq(String subId, _LiveSubscription subscription) {
if (subscription.reqGeneration == _connectionGeneration) {
subscription.ambiguousGeneration = _connectionGeneration;
}
subscription.reqGeneration = _connectionGeneration;
_sendReq(subId, _replayFilter(subscription));
}
void _sendReq(String subId, NostrFilter filter) {
_socket?.send(['REQ', subId, filter.toJson()]);
}
void _sendHeldHistory() {
for (final MapEntry(key: subId, value: sub)
in _historySubscriptions.entries) {
final filter = sub.heldFilter;
if (filter == null) continue;
sub.heldFilter = null;
_sendReq(subId, filter);
}
}
void _sendClose(String subId) {
_socket?.send(['CLOSE', subId]);
}
void _unsubscribe(String subId) {
final subscription = _liveSubscriptions[subId];
if (subscription != null) {
_removeLiveSubscription(subId, subscription);
}
_sendClose(subId);
}
void _removeLiveSubscription(String subId, _LiveSubscription subscription) {
if (_liveSubscriptions[subId] != subscription) return;
_liveSubscriptions.remove(subId);
_pendingClosedRetries.remove(subId);
subscription.closedRetryTimer?.cancel();
subscription.closedRetryTimer = null;
_recentDeliveryKeys.removeWhere((key) => key.startsWith('$subId:'));
}
void _resetClosedRetry(_LiveSubscription subscription) {
subscription.closedRetryAttempt = 0;
subscription.closedRetryTimer?.cancel();
subscription.closedRetryTimer = null;
}
void _cancelAllClosedRetries() {
_pendingClosedRetries.clear();
for (final subscription in _liveSubscriptions.values) {
subscription.closedRetryTimer?.cancel();
subscription.closedRetryTimer = null;
}
}
void _resetAllClosedRetries() {
_pendingClosedRetries.clear();
for (final subscription in _liveSubscriptions.values) {
_resetClosedRetry(subscription);
}
}
void _cancelAllHistory(Object? error) {
for (final entry in _historySubscriptions.values) {
entry.timeout.cancel();
if (!entry.completer.isCompleted) {
entry.completer.completeError(error ?? Exception('Connection lost'));
}
}
_historySubscriptions.clear();
}
void _rejectAllPending(Object? error) {
for (final entry in _pendingEvents.values) {
entry.timeout.cancel();
if (!entry.completer.isCompleted) {
entry.completer.completeError(error ?? Exception('Connection lost'));
}
}
_pendingEvents.clear();
}
void _dispose() {
_disposed = true;
_contextGeneration++;
_beforePauseCallbacks.clear();
_connectionGeneration++;
_reconnectTimer?.cancel();
_flushTimer?.cancel();
_backgroundGraceTimer?.cancel();
_backgroundedAt = null;
_cancelAllClosedRetries();
_rateLimitGate.reset();
_visibleChannelsByOwner.clear();
_socketConnected = false;
_cancelAllHistory(null);
_rejectAllPending(null);
final subscriptions = _liveSubscriptions.values.toList();
_liveSubscriptions.clear();
for (final subscription in subscriptions) {
subscription.closedRetryTimer?.cancel();
subscription.closedRetryTimer = null;
}
_recentDeliveryKeys.clear();
_socket?.dispose();
_socket = null;
_httpQueryClient.close();
}
}
final relaySessionProvider =
NotifierProvider<RelaySessionNotifier, SessionState>(
RelaySessionNotifier.new,
);