donphan/streaming/index.js

1423 lines
44 KiB
JavaScript
Raw Normal View History

// @ts-check
import fs from 'node:fs';
import http from 'node:http';
import path from 'node:path';
import url from 'node:url';
import cors from 'cors';
import dotenv from 'dotenv';
import express from 'express';
import { JSDOM } from 'jsdom';
import { WebSocketServer } from 'ws';
import * as Database from './database.js';
import { AuthenticationError, RequestError, extractStatusAndMessage as extractErrorStatusAndMessage } from './errors.js';
import { logger, httpLogger, initializeLogLevel, attachWebsocketHttpLogger, createWebsocketLogger } from './logging.js';
import { setupMetrics } from './metrics.js';
import * as Redis from './redis.js';
import { isTruthy, normalizeHashtag, firstParam } from './utils.js';
const environment = process.env.NODE_ENV || 'development';
2017-02-02 15:11:36 +00:00
// Correctly detect and load .env or .env.production file based on environment:
const dotenvFile = environment === 'production' ? '.env.production' : '.env';
const dotenvFilePath = path.resolve(
url.fileURLToPath(
new URL(path.join('..', dotenvFile), import.meta.url)
)
);
2017-02-02 15:11:36 +00:00
dotenv.config({
path: dotenvFilePath
});
2017-02-02 00:31:09 +00:00
initializeLogLevel(process.env, environment);
/**
* Declares the result type for accountFromToken / accountFromRequest.
*
* Note: This is here because jsdoc doesn't like importing types that
* are nested in functions
* @typedef ResolvedAccount
* @property {string} accessTokenId
* @property {string[]} scopes
* @property {string} accountId
* @property {string[]} chosenLanguages
*/
/**
* Attempts to safely parse a string as JSON, used when both receiving a message
* from redis and when receiving a message from a client over a websocket
* connection, this is why it accepts a `req` argument.
* @param {string} json
* @param {any?} req
2023-04-30 01:29:54 +01:00
* @returns {Object.<string, any>|null}
*/
const parseJSON = (json, req) => {
try {
return JSON.parse(json);
} catch (err) {
/* FIXME: This logging isn't great, and should probably be done at the
* call-site of parseJSON, not in the method, but this would require changing
* the signature of parseJSON to return something akin to a Result type:
* [Error|null, null|Object<string,any}], and then handling the error
* scenarios.
*/
if (req) {
if (req.accountId) {
req.log.error({ err }, `Error parsing message from user ${req.accountId}`);
} else {
req.log.error({ err }, `Error parsing message from ${req.remoteAddress}`);
}
} else {
logger.error({ err }, `Error parsing message from redis`);
}
return null;
}
};
const PUBLIC_CHANNELS = [
'public',
'public:media',
'public:local',
'public:local:media',
'public:remote',
'public:remote:media',
'hashtag',
'hashtag:local',
];
// Used for priming the counters/gauges for the various metrics that are
// per-channel
const CHANNEL_NAMES = [
'system',
'user',
'user:notification',
'list',
'direct',
...PUBLIC_CHANNELS
];
const startServer = async () => {
const pgPool = Database.getPool(Database.configFromEnv(process.env, environment));
const metrics = setupMetrics(CHANNEL_NAMES, pgPool);
const redisConfig = Redis.configFromEnv(process.env);
const redisClient = Redis.createClient(redisConfig, logger);
const server = http.createServer();
const wss = new WebSocketServer({ noServer: true });
/**
* Adds a namespace to Redis keys or channel names
* Fixes: https://github.com/redis/ioredis/issues/1910
* @param {string} keyOrChannel
* @returns {string}
*/
function redisNamespaced(keyOrChannel) {
if (redisConfig.namespace) {
return `${redisConfig.namespace}:${keyOrChannel}`;
} else {
return keyOrChannel;
}
}
/**
* Removes the redis namespace from a channel name
* @param {string} channel
* @returns {string}
*/
function redisUnnamespaced(channel) {
if (typeof redisConfig.namespace === "string") {
// Note: this removes the configured namespace and the colon that is used
// to separate it:
return channel.slice(redisConfig.namespace.length + 1);
} else {
return channel;
}
}
// Set the X-Request-Id header on WebSockets:
wss.on("headers", function onHeaders(headers, req) {
headers.push(`X-Request-Id: ${req.id}`);
});
const app = express();
app.set('trust proxy', process.env.TRUSTED_PROXY_IP ? process.env.TRUSTED_PROXY_IP.split(/(?:\s*,\s*|\s+)/) : 'loopback,uniquelocal');
app.use(httpLogger);
app.use(cors());
// Handle eventsource & other http requests:
server.on('request', app);
// Handle upgrade requests:
server.on('upgrade', async function handleUpgrade(request, socket, head) {
// Setup the HTTP logger, since websocket upgrades don't get the usual http
// logger. This decorates the `request` object.
attachWebsocketHttpLogger(request);
request.log.info("HTTP Upgrade Requested");
/** @param {Error} err */
const onSocketError = (err) => {
request.log.error({ error: err }, err.message);
};
socket.on('error', onSocketError);
/** @type {ResolvedAccount} */
let resolvedAccount;
try {
resolvedAccount = await accountFromRequest(request);
} catch (err) {
// Unfortunately for using the on('upgrade') setup, we need to manually
// write a HTTP Response to the Socket to close the connection upgrade
// attempt, so the following code is to handle all of that.
const {statusCode, errorMessage } = extractErrorStatusAndMessage(err);
/** @type {Record<string, string | number | import('pino-http').ReqId>} */
const headers = {
'Connection': 'close',
'Content-Type': 'text/plain',
'Content-Length': 0,
'X-Request-Id': request.id,
'X-Error-Message': errorMessage
};
// Ensure the socket is closed once we've finished writing to it:
socket.once('finish', () => {
socket.destroy();
});
// Write the HTTP response manually:
socket.end(`HTTP/1.1 ${statusCode} ${http.STATUS_CODES[statusCode]}\r\n${Object.keys(headers).map((key) => `${key}: ${headers[key]}`).join('\r\n')}\r\n\r\n`);
// Finally, log the error:
request.log.error({
err,
res: {
statusCode,
headers
}
}, errorMessage);
return;
}
// Remove the error handler, wss.handleUpgrade has its own:
socket.removeListener('error', onSocketError);
wss.handleUpgrade(request, socket, head, function done(ws) {
request.log.info("Authenticated request & upgraded to WebSocket connection");
const wsLogger = createWebsocketLogger(request, resolvedAccount);
// Start the connection:
wss.emit('connection', ws, request, wsLogger);
});
});
/**
* @type {Object.<string, Array.<function(Object<string, any>): void>>}
*/
const subs = {};
const redisSubscribeClient = Redis.createClient(redisConfig, logger);
// When checking metrics in the browser, the favicon is requested this
// prevents the request from falling through to the API Router, which would
// error for this endpoint:
app.get('/favicon.ico', (_req, res) => res.status(404).end());
app.get('/api/v1/streaming/health', (_req, res) => {
res.writeHead(200, { 'Content-Type': 'text/plain' });
res.end('OK');
});
app.get('/metrics', metrics.requestHandler);
/**
* @param {string[]} channels
2023-04-30 01:29:54 +01:00
* @returns {function(): void}
*/
2020-06-02 18:24:53 +01:00
const subscriptionHeartbeat = channels => {
const interval = 6 * 60;
const tellSubscribed = () => {
channels.forEach(channel => redisClient.set(redisNamespaced(`subscribed:${channel}`), '1', 'EX', interval * 3));
};
2020-06-02 18:24:53 +01:00
tellSubscribed();
2020-06-02 18:24:53 +01:00
const heartbeat = setInterval(tellSubscribed, interval * 1000);
return () => {
clearInterval(heartbeat);
};
};
/**
* @param {string} channel
* @param {string} message
*/
const onRedisMessage = (channel, message) => {
metrics.redisMessagesReceived.inc();
logger.debug(`New message on channel ${channel}`);
const key = redisUnnamespaced(channel);
const callbacks = subs[key];
if (!callbacks) {
return;
}
const json = parseJSON(message, null);
if (!json) return;
callbacks.forEach(callback => callback(json));
};
redisSubscribeClient.on("message", onRedisMessage);
/**
* @callback SubscriptionListener
* @param {ReturnType<parseJSON>} json of the message
* @returns void
*/
/**
* @param {string} channel
* @param {SubscriptionListener} callback
*/
const subscribe = (channel, callback) => {
logger.debug(`Adding listener for ${channel}`);
subs[channel] = subs[channel] || [];
if (subs[channel].length === 0) {
logger.debug(`Subscribe ${channel}`);
redisSubscribeClient.subscribe(redisNamespaced(channel), (err, count) => {
if (err) {
logger.error(`Error subscribing to ${channel}`);
} else if (typeof count === 'number') {
metrics.redisSubscriptions.set(count);
}
});
}
subs[channel].push(callback);
};
2017-02-03 17:27:42 +00:00
/**
* @param {string} channel
* @param {SubscriptionListener} callback
*/
const unsubscribe = (channel, callback) => {
logger.debug(`Removing listener for ${channel}`);
if (!subs[channel]) {
return;
}
subs[channel] = subs[channel].filter(item => item !== callback);
if (subs[channel].length === 0) {
logger.debug(`Unsubscribe ${channel}`);
// FIXME: https://github.com/redis/ioredis/issues/1910
redisSubscribeClient.unsubscribe(redisNamespaced(channel), (err, count) => {
if (err) {
logger.error(`Error unsubscribing to ${channel}`);
} else if (typeof count === 'number') {
metrics.redisSubscriptions.set(count);
}
});
delete subs[channel];
}
};
2017-02-03 17:27:42 +00:00
/**
* @param {http.IncomingMessage & ResolvedAccount} req
* @param {string[]} necessaryScopes
2023-04-30 01:29:54 +01:00
* @returns {boolean}
*/
const isInScope = (req, necessaryScopes) =>
req.scopes.some(scope => necessaryScopes.includes(scope));
/**
* @param {string} token
* @param {any} req
* @returns {Promise<ResolvedAccount>}
*/
const accountFromToken = async (token, req) => {
const result = await pgPool.query('SELECT oauth_access_tokens.id, oauth_access_tokens.resource_owner_id, users.account_id, users.chosen_languages, oauth_access_tokens.scopes FROM oauth_access_tokens INNER JOIN users ON oauth_access_tokens.resource_owner_id = users.id WHERE oauth_access_tokens.token = $1 AND oauth_access_tokens.revoked_at IS NULL LIMIT 1', [token]);
2017-02-02 00:31:09 +00:00
if (result.rows.length === 0) {
throw new AuthenticationError('Invalid access token');
}
req.accessTokenId = result.rows[0].id;
req.scopes = result.rows[0].scopes.split(' ');
req.accountId = result.rows[0].account_id;
req.chosenLanguages = result.rows[0].chosen_languages;
return {
accessTokenId: result.rows[0].id,
scopes: result.rows[0].scopes.split(' '),
accountId: result.rows[0].account_id,
chosenLanguages: result.rows[0].chosen_languages,
};
};
/**
* @param {any} req
* @returns {Promise<ResolvedAccount>}
*/
const accountFromRequest = (req) => new Promise((resolve, reject) => {
const authorization = req.headers.authorization;
const location = url.parse(req.url, true);
const accessToken = location.query.access_token || req.headers['sec-websocket-protocol'];
2017-05-21 20:13:11 +01:00
if (!authorization && !accessToken) {
reject(new AuthenticationError('Missing access token'));
return;
}
2017-02-02 16:10:59 +00:00
2017-05-21 20:13:11 +01:00
const token = authorization ? authorization.replace(/^Bearer /, '') : accessToken;
resolve(accountFromToken(token, req));
});
/**
* @param {any} req
* @returns {string|undefined}
*/
const channelNameFromPath = req => {
const { path, query } = req;
const onlyMedia = isTruthy(query.only_media);
switch (path) {
case '/api/v1/streaming/user':
return 'user';
case '/api/v1/streaming/user/notification':
return 'user:notification';
case '/api/v1/streaming/public':
return onlyMedia ? 'public:media' : 'public';
case '/api/v1/streaming/public/local':
return onlyMedia ? 'public:local:media' : 'public:local';
case '/api/v1/streaming/public/remote':
return onlyMedia ? 'public:remote:media' : 'public:remote';
case '/api/v1/streaming/hashtag':
return 'hashtag';
case '/api/v1/streaming/hashtag/local':
return 'hashtag:local';
case '/api/v1/streaming/direct':
return 'direct';
case '/api/v1/streaming/list':
return 'list';
default:
return undefined;
}
};
2017-02-05 02:19:04 +00:00
/**
* @param {http.IncomingMessage & ResolvedAccount} req
* @param {import('pino').Logger} logger
* @param {string|undefined} channelName
2023-04-30 01:29:54 +01:00
* @returns {Promise.<void>}
*/
const checkScopes = (req, logger, channelName) => new Promise((resolve, reject) => {
logger.debug(`Checking OAuth scopes for ${channelName}`);
// When accessing public channels, no scopes are needed
if (channelName && PUBLIC_CHANNELS.includes(channelName)) {
resolve();
return;
}
// The `read` scope has the highest priority, if the token has it
// then it can access all streams
const requiredScopes = ['read'];
// When accessing specifically the notifications stream,
// we need a read:notifications, while in all other cases,
// we can allow access with read:statuses. Mind that the
// user stream will not contain notifications unless
// the token has either read or read:notifications scope
// as well, this is handled separately.
if (channelName === 'user:notification') {
requiredScopes.push('read:notifications');
} else {
requiredScopes.push('read:statuses');
}
if (req.scopes && requiredScopes.some(requiredScope => req.scopes.includes(requiredScope))) {
resolve();
return;
}
reject(new AuthenticationError('Access token does not have the required scopes'));
});
/**
* @typedef SystemMessageHandlers
* @property {function(): void} onKill
*/
/**
* @param {any} req
* @param {SystemMessageHandlers} eventHandlers
* @returns {SubscriptionListener}
*/
const createSystemMessageListener = (req, eventHandlers) => {
return message => {
if (!message?.event) {
return;
}
const { event } = message;
req.log.debug(`System message for ${req.accountId}: ${event}`);
if (event === 'kill') {
req.log.debug(`Closing connection for ${req.accountId} due to expired access token`);
eventHandlers.onKill();
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
} else if (event === 'filters_changed') {
req.log.debug(`Invalidating filters cache for ${req.accountId}`);
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
req.cachedFilters = null;
}
};
};
/**
* @param {http.IncomingMessage & ResolvedAccount} req
* @param {http.OutgoingMessage} res
*/
const subscribeHttpToSystemChannel = (req, res) => {
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
const accessTokenChannelId = `timeline:access_token:${req.accessTokenId}`;
const systemChannelId = `timeline:system:${req.accountId}`;
const listener = createSystemMessageListener(req, {
onKill() {
res.end();
},
});
res.on('close', () => {
unsubscribe(accessTokenChannelId, listener);
unsubscribe(systemChannelId, listener);
metrics.connectedChannels.labels({ type: 'eventsource', channel: 'system' }).dec(2);
});
subscribe(accessTokenChannelId, listener);
subscribe(systemChannelId, listener);
metrics.connectedChannels.labels({ type: 'eventsource', channel: 'system' }).inc(2);
};
/**
* @param {any} req
* @param {any} res
* @param {function(Error=): void} next
*/
const authenticationMiddleware = (req, res, next) => {
if (req.method === 'OPTIONS') {
next();
return;
}
const channelName = channelNameFromPath(req);
// If no channelName can be found for the request, then we should terminate
// the connection, as there's nothing to stream back
if (!channelName) {
next(new RequestError('Unknown channel requested'));
return;
}
accountFromRequest(req).then(() => checkScopes(req, req.log, channelName)).then(() => {
subscribeHttpToSystemChannel(req, res);
}).then(() => {
next();
}).catch(err => {
next(err);
});
};
/**
* @param {Error} err
* @param {any} req
* @param {any} res
* @param {function(Error=): void} next
*/
const errorMiddleware = (err, req, res, next) => {
req.log.error({ err }, err.toString());
if (res.headersSent) {
next(err);
return;
}
const {statusCode, errorMessage } = extractErrorStatusAndMessage(err);
res.writeHead(statusCode, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: errorMessage }));
};
2017-02-05 02:19:04 +00:00
/**
* @param {any[]} arr
* @param {number=} shift
2023-04-30 01:29:54 +01:00
* @returns {string}
*/
// @ts-ignore
const placeholders = (arr, shift = 0) => arr.map((_, i) => `$${i + 1 + shift}`).join(', ');
2017-02-02 00:31:09 +00:00
/**
* @param {string} listId
* @param {any} req
2023-04-30 01:29:54 +01:00
* @returns {Promise.<void>}
*/
const authorizeListAccess = async (listId, req) => {
const { accountId } = req;
const result = await pgPool.query('SELECT id, account_id FROM lists WHERE id = $1 AND account_id = $2 LIMIT 1', [listId, accountId]);
if (result.rows.length === 0) {
throw new AuthenticationError('List not found');
}
};
/**
* @param {string[]} channelIds
* @param {http.IncomingMessage & ResolvedAccount} req
* @param {import('pino').Logger} log
* @param {function(string, string): void} output
* @param {undefined | function(string[], SubscriptionListener): void} attachCloseHandler
* @param {'websocket' | 'eventsource'} destinationType
* @param {boolean=} needsFiltering
* @param {boolean=} allowLocalOnly
* @returns {SubscriptionListener}
*/
const streamFrom = (channelIds, req, log, output, attachCloseHandler, destinationType, needsFiltering = false, allowLocalOnly = false) => {
log.info({ channelIds }, `Starting stream`);
/**
* @param {string} event
* @param {object|string} payload
*/
const transmit = (event, payload) => {
// TODO: Replace "string"-based delete payloads with object payloads:
const encodedPayload = typeof payload === 'object' ? JSON.stringify(payload) : payload;
metrics.messagesSent.labels({ type: destinationType }).inc(1);
log.debug({ event, payload }, `Transmitting ${event} to ${req.accountId}`);
output(event, encodedPayload);
};
// The listener used to process each message off the redis subscription,
// message here is an object with an `event` and `payload` property. Some
// events also include a queued_at value, but this is being removed shortly.
/** @type {SubscriptionListener} */
const listener = message => {
if (!message?.event || !message?.payload) {
return;
}
const { event, payload } = message;
// Only send local-only statuses to logged-in users
if ((event === 'update' || event === 'status.update') && payload.local_only && !(req.accountId && allowLocalOnly)) {
log.debug(`Message ${payload.id} filtered because it was local-only`);
return;
}
// Streaming only needs to apply filtering to some channels and only to
// some events. This is because majority of the filtering happens on the
// Ruby on Rails side when producing the event for streaming.
//
// The only events that require filtering from the streaming server are
// `update` and `status.update`, all other events are transmitted to the
// client as soon as they're received (pass-through).
//
// The channels that need filtering are determined in the function
// `channelNameToIds` defined below:
if (!needsFiltering || (event !== 'update' && event !== 'status.update')) {
transmit(event, payload);
return;
}
// The rest of the logic from here on in this function is to handle
// filtering of statuses:
// Filter based on language:
if (Array.isArray(req.chosenLanguages) && payload.language !== null && req.chosenLanguages.indexOf(payload.language) === -1) {
log.debug(`Message ${payload.id} filtered by language (${payload.language})`);
return;
}
// When the account is not logged in, it is not necessary to confirm the block or mute
if (!req.accountId) {
transmit(event, payload);
return;
}
// Filter based on domain blocks, blocks, mutes, or custom filters:
// @ts-ignore
const targetAccountIds = [payload.account.id].concat(payload.mentions.map(item => item.id));
const accountDomain = payload.account.acct.split('@')[1];
// TODO: Move this logic out of the message handling loop
pgPool.connect((err, client, releasePgConnection) => {
if (err) {
log.error(err);
return;
}
const queries = [
// @ts-ignore
client.query(`SELECT 1
FROM blocks
WHERE (account_id = $1 AND target_account_id IN (${placeholders(targetAccountIds, 2)}))
OR (account_id = $2 AND target_account_id = $1)
UNION
SELECT 1
FROM mutes
WHERE account_id = $1
AND target_account_id IN (${placeholders(targetAccountIds, 2)})`, [req.accountId, payload.account.id].concat(targetAccountIds)),
];
if (accountDomain) {
// @ts-ignore
queries.push(client.query('SELECT 1 FROM account_domain_blocks WHERE account_id = $1 AND domain = $2', [req.accountId, accountDomain]));
}
// @ts-ignore
if (!payload.filtered && !req.cachedFilters) {
// @ts-ignore
queries.push(client.query('SELECT filter.id AS id, filter.phrase AS title, filter.context AS context, filter.expires_at AS expires_at, filter.action AS filter_action, keyword.keyword AS keyword, keyword.whole_word AS whole_word FROM custom_filter_keywords keyword JOIN custom_filters filter ON keyword.custom_filter_id = filter.id WHERE filter.account_id = $1 AND (filter.expires_at IS NULL OR filter.expires_at > NOW())', [req.accountId]));
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
}
Promise.all(queries).then(values => {
releasePgConnection();
// Handling blocks & mutes and domain blocks: If one of those applies,
// then we don't transmit the payload of the event to the client
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
if (values[0].rows.length > 0 || (accountDomain && values[1].rows.length > 0)) {
return;
}
// If the payload already contains the `filtered` property, it means
// that filtering has been applied on the ruby on rails side, as
// such, we don't need to construct or apply the filters in streaming:
if (Object.hasOwn(payload, "filtered")) {
transmit(event, payload);
return;
}
// Handling for constructing the custom filters and caching them on the request
// TODO: Move this logic out of the message handling lifecycle
// @ts-ignore
if (!req.cachedFilters) {
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
const filterRows = values[accountDomain ? 2 : 1].rows;
// @ts-ignore
req.cachedFilters = filterRows.reduce((cache, filter) => {
if (cache[filter.id]) {
cache[filter.id].keywords.push([filter.keyword, filter.whole_word]);
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
} else {
cache[filter.id] = {
keywords: [[filter.keyword, filter.whole_word]],
expires_at: filter.expires_at,
filter: {
id: filter.id,
title: filter.title,
context: filter.context,
expires_at: filter.expires_at,
// filter.filter_action is the value from the
// custom_filters.action database column, it is an integer
// representing a value in an enum defined by Ruby on Rails:
//
// enum { warn: 0, hide: 1 }
filter_action: ['warn', 'hide'][filter.filter_action],
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
},
};
}
return cache;
}, {});
// Construct the regular expressions for the custom filters: This
// needs to be done in a separate loop as the database returns one
// filterRow per keyword, so we need all the keywords before
// constructing the regular expression
// @ts-ignore
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
Object.keys(req.cachedFilters).forEach((key) => {
// @ts-ignore
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
req.cachedFilters[key].regexp = new RegExp(req.cachedFilters[key].keywords.map(([keyword, whole_word]) => {
let expr = keyword.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
if (whole_word) {
if (/^[\w]/.test(expr)) {
expr = `\\b${expr}`;
}
if (/[\w]$/.test(expr)) {
expr = `${expr}\\b`;
}
}
return expr;
}).join('|'), 'i');
});
}
// Apply cachedFilters against the payload, constructing a
// `filter_results` array of FilterResult entities
// @ts-ignore
if (req.cachedFilters) {
const status = payload;
// TODO: Calculate searchableContent in Ruby on Rails:
// @ts-ignore
const searchableContent = ([status.spoiler_text || '', status.content].concat((status.poll && status.poll.options) ? status.poll.options.map(option => option.title) : [])).concat(status.media_attachments.map(att => att.description)).join('\n\n').replace(/<br\s*\/?>/g, '\n').replace(/<\/p><p>/g, '\n\n');
const searchableTextContent = JSDOM.fragment(searchableContent).textContent;
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
const now = new Date();
// @ts-ignore
const filter_results = Object.values(req.cachedFilters).reduce((results, cachedFilter) => {
// Check the filter hasn't expired before applying:
if (cachedFilter.expires_at !== null && cachedFilter.expires_at < now) {
return results;
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
}
// Just in-case JSDOM fails to find textContent in searchableContent
if (!searchableTextContent) {
return results;
}
const keyword_matches = searchableTextContent.match(cachedFilter.regexp);
if (keyword_matches) {
// results is an Array of FilterResult; status_matches is always
// null as we only are only applying the keyword-based custom
// filters, not the status-based custom filters.
// https://docs.joinmastodon.org/entities/FilterResult/
results.push({
filter: cachedFilter.filter,
keyword_matches,
status_matches: null
});
}
return results;
}, []);
// Send the payload + the FilterResults as the `filtered` property
// to the streaming connection. To reach this code, the `event` must
// have been either `update` or `status.update`, meaning the
// `payload` is a Status entity, which has a `filtered` property:
//
// filtered: https://docs.joinmastodon.org/entities/Status/#filtered
transmit(event, {
...payload,
filtered: filter_results
});
} else {
transmit(event, payload);
}
}).catch(err => {
log.error(err);
releasePgConnection();
});
});
};
channelIds.forEach(id => {
subscribe(id, listener);
2020-06-02 18:24:53 +01:00
});
if (typeof attachCloseHandler === 'function') {
attachCloseHandler(channelIds, listener);
}
return listener;
};
2017-02-02 00:31:09 +00:00
/**
* @param {any} req
* @param {any} res
2023-04-30 01:29:54 +01:00
* @returns {function(string, string): void}
*/
const streamToHttp = (req, res) => {
const channelName = channelNameFromPath(req);
metrics.connectedClients.labels({ type: 'eventsource' }).inc();
// In theory we'll always have a channel name, but channelNameFromPath can return undefined:
if (typeof channelName === 'string') {
metrics.connectedChannels.labels({ type: 'eventsource', channel: channelName }).inc();
}
res.setHeader('Content-Type', 'text/event-stream');
res.setHeader('Cache-Control', 'no-store');
res.setHeader('Transfer-Encoding', 'chunked');
res.write(':)\n');
const heartbeat = setInterval(() => res.write(':thump\n'), 15000);
req.on('close', () => {
req.log.info({ accountId: req.accountId }, `Ending stream`);
// We decrement these counters here instead of in streamHttpEnd as in that
// method we don't have knowledge of the channel names
metrics.connectedClients.labels({ type: 'eventsource' }).dec();
// In theory we'll always have a channel name, but channelNameFromPath can return undefined:
if (typeof channelName === 'string') {
metrics.connectedChannels.labels({ type: 'eventsource', channel: channelName }).dec();
}
clearInterval(heartbeat);
});
return (event, payload) => {
res.write(`event: ${event}\n`);
res.write(`data: ${payload}\n\n`);
};
};
2017-02-02 00:31:09 +00:00
/**
* @param {any} req
* @param {function(): void} [closeHandler]
* @returns {function(string[], SubscriptionListener): void}
*/
const streamHttpEnd = (req, closeHandler = undefined) => (ids, listener) => {
req.on('close', () => {
2020-06-02 18:24:53 +01:00
ids.forEach(id => {
unsubscribe(id, listener);
2020-06-02 18:24:53 +01:00
});
if (closeHandler) {
closeHandler();
}
});
};
/**
* @param {http.IncomingMessage} req
* @param {import('ws').WebSocket} ws
* @param {string[]} streamName
2023-04-30 01:29:54 +01:00
* @returns {function(string, string): void}
*/
const streamToWs = (req, ws, streamName) => (event, payload) => {
if (ws.readyState !== ws.OPEN) {
req.log.error('Tried writing to closed socket');
return;
}
const message = JSON.stringify({ stream: streamName, event, payload });
ws.send(message, (/** @type {Error|undefined} */ err) => {
if (err) {
req.log.error({err}, `Failed to send to websocket`);
}
});
};
/**
* @param {http.ServerResponse} res
*/
const httpNotFound = res => {
res.writeHead(404, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: 'Not found' }));
};
const api = express.Router();
app.use(api);
api.use(authenticationMiddleware);
api.use(errorMiddleware);
api.get('/api/v1/streaming/*', (req, res) => {
const channelName = channelNameFromPath(req);
// FIXME: In theory we'd never actually reach here due to
// authenticationMiddleware catching this case, however, we need to refactor
// how those middlewares work, so I'm adding the extra check in here.
if (!channelName) {
httpNotFound(res);
return;
}
channelNameToIds(req, channelName, req.query).then(({ channelIds, options }) => {
const onSend = streamToHttp(req, res);
const onEnd = streamHttpEnd(req, subscriptionHeartbeat(channelIds));
// @ts-ignore
streamFrom(channelIds, req, req.log, onSend, onEnd, 'eventsource', options.needsFiltering, options.allowLocalOnly);
}).catch(err => {
const {statusCode, errorMessage } = extractErrorStatusAndMessage(err);
res.log.info({ err }, 'Eventsource subscription error');
res.writeHead(statusCode, { 'Content-Type': 'application/json' });
res.end(JSON.stringify({ error: errorMessage }));
});
});
/**
* @typedef StreamParams
* @property {string} [tag]
* @property {string} [list]
* @property {string} [only_media]
*/
/**
* @param {any} req
2023-04-30 01:29:54 +01:00
* @returns {string[]}
*/
const channelsForUserStream = req => {
const arr = [`timeline:${req.accountId}`];
if (isInScope(req, ['read', 'read:notifications'])) {
arr.push(`timeline:${req.accountId}:notifications`);
}
return arr;
};
/**
* @param {any} req
* @param {string} name
* @param {StreamParams} params
2023-04-30 01:29:54 +01:00
* @returns {Promise.<{ channelIds: string[], options: { needsFiltering: boolean } }>}
*/
const channelNameToIds = (req, name, params) => new Promise((resolve, reject) => {
switch (name) {
case 'user':
resolve({
channelIds: channelsForUserStream(req),
options: { needsFiltering: false, allowLocalOnly: true },
});
2020-06-02 18:24:53 +01:00
break;
case 'user:notification':
resolve({
channelIds: [`timeline:${req.accountId}:notifications`],
options: { needsFiltering: false, allowLocalOnly: true },
});
break;
case 'public':
resolve({
channelIds: ['timeline:public'],
options: { needsFiltering: true, allowLocalOnly: isTruthy(params.allow_local_only) },
});
break;
case 'public:allow_local_only':
resolve({
channelIds: ['timeline:public'],
options: { needsFiltering: true, allowLocalOnly: true },
});
break;
case 'public:local':
resolve({
channelIds: ['timeline:public:local'],
options: { needsFiltering: true, allowLocalOnly: true },
});
break;
case 'public:remote':
resolve({
channelIds: ['timeline:public:remote'],
options: { needsFiltering: true, allowLocalOnly: false },
});
break;
2018-05-21 11:43:38 +01:00
case 'public:media':
resolve({
channelIds: ['timeline:public:media'],
options: { needsFiltering: true, allowLocalOnly: isTruthy(params.allow_local_only) },
});
2018-05-21 11:43:38 +01:00
break;
case 'public:allow_local_only:media':
resolve({
channelIds: ['timeline:public:media'],
options: { needsFiltering: true, allowLocalOnly: true },
});
break;
2018-05-21 11:43:38 +01:00
case 'public:local:media':
resolve({
channelIds: ['timeline:public:local:media'],
options: { needsFiltering: true, allowLocalOnly: true },
});
2018-05-21 11:43:38 +01:00
break;
case 'public:remote:media':
resolve({
channelIds: ['timeline:public:remote:media'],
options: { needsFiltering: true, allowLocalOnly: false },
});
break;
case 'direct':
resolve({
channelIds: [`timeline:direct:${req.accountId}`],
options: { needsFiltering: false, allowLocalOnly: true },
});
break;
case 'hashtag':
if (!params.tag) {
reject(new RequestError('Missing tag name parameter'));
} else {
resolve({
channelIds: [`timeline:hashtag:${normalizeHashtag(params.tag)}`],
options: { needsFiltering: true, allowLocalOnly: true },
});
}
break;
case 'hashtag:local':
if (!params.tag) {
reject(new RequestError('Missing tag name parameter'));
} else {
resolve({
channelIds: [`timeline:hashtag:${normalizeHashtag(params.tag)}:local`],
options: { needsFiltering: true, allowLocalOnly: true },
});
}
break;
case 'list':
if (!params.list) {
reject(new RequestError('Missing list name parameter'));
return;
}
authorizeListAccess(params.list, req).then(() => {
resolve({
channelIds: [`timeline:list:${params.list}`],
options: { needsFiltering: false, allowLocalOnly: true },
});
}).catch(() => {
reject(new AuthenticationError('Not authorized to stream this list'));
});
break;
default:
reject(new RequestError('Unknown stream type'));
}
});
/**
* @param {string} channelName
* @param {StreamParams} params
2023-04-30 01:29:54 +01:00
* @returns {string[]}
*/
const streamNameFromChannelName = (channelName, params) => {
if (channelName === 'list' && params.list) {
return [channelName, params.list];
} else if (['hashtag', 'hashtag:local'].includes(channelName) && params.tag) {
return [channelName, params.tag];
} else {
return [channelName];
}
};
/**
* @typedef WebSocketSession
* @property {import('ws').WebSocket & { isAlive: boolean}} websocket
* @property {http.IncomingMessage & ResolvedAccount} request
* @property {import('pino').Logger} logger
* @property {Object.<string, { channelName: string, listener: SubscriptionListener, stopHeartbeat: function(): void }>} subscriptions
*/
/**
* @param {WebSocketSession} session
* @param {string} channelName
* @param {StreamParams} params
* @returns {void}
*/
const subscribeWebsocketToChannel = ({ websocket, request, logger, subscriptions }, channelName, params) => {
checkScopes(request, logger, channelName).then(() => channelNameToIds(request, channelName, params)).then(({
channelIds,
options,
}) => {
if (subscriptions[channelIds.join(';')]) {
return;
}
const onSend = streamToWs(request, websocket, streamNameFromChannelName(channelName, params));
const stopHeartbeat = subscriptionHeartbeat(channelIds);
const listener = streamFrom(channelIds, request, logger, onSend, undefined, 'websocket', options.needsFiltering, options.allowLocalOnly);
metrics.connectedChannels.labels({ type: 'websocket', channel: channelName }).inc();
subscriptions[channelIds.join(';')] = {
channelName,
listener,
stopHeartbeat,
};
}).catch(err => {
const {statusCode, errorMessage } = extractErrorStatusAndMessage(err);
logger.error({ err }, 'Websocket subscription error');
// If we have a socket that is alive and open still, send the error back to the client:
if (websocket.isAlive && websocket.readyState === websocket.OPEN) {
websocket.send(JSON.stringify({
error: errorMessage,
status: statusCode
}));
}
});
};
/**
* @param {WebSocketSession} session
* @param {string[]} channelIds
*/
const removeSubscription = ({ request, logger, subscriptions }, channelIds) => {
logger.info({ channelIds, accountId: request.accountId }, `Ending stream`);
const subscription = subscriptions[channelIds.join(';')];
if (!subscription) {
return;
}
channelIds.forEach(channelId => {
unsubscribe(channelId, subscription.listener);
});
metrics.connectedChannels.labels({ type: 'websocket', channel: subscription.channelName }).dec();
subscription.stopHeartbeat();
delete subscriptions[channelIds.join(';')];
};
/**
* @param {WebSocketSession} session
* @param {string} channelName
* @param {StreamParams} params
* @returns {void}
*/
const unsubscribeWebsocketFromChannel = (session, channelName, params) => {
const { websocket, request, logger } = session;
channelNameToIds(request, channelName, params).then(({ channelIds }) => {
removeSubscription(session, channelIds);
}).catch(err => {
logger.error({err}, 'Websocket unsubscribe error');
// If we have a socket that is alive and open still, send the error back to the client:
if (websocket.isAlive && websocket.readyState === websocket.OPEN) {
// TODO: Use a better error response here
websocket.send(JSON.stringify({ error: "Error unsubscribing from channel" }));
}
});
};
/**
* @param {WebSocketSession} session
*/
const subscribeWebsocketToSystemChannel = ({ websocket, request, subscriptions }) => {
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
const accessTokenChannelId = `timeline:access_token:${request.accessTokenId}`;
const systemChannelId = `timeline:system:${request.accountId}`;
const listener = createSystemMessageListener(request, {
onKill() {
websocket.close();
},
});
subscribe(accessTokenChannelId, listener);
subscribe(systemChannelId, listener);
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
subscriptions[accessTokenChannelId] = {
channelName: 'system',
Revamp post filtering system (#18058) * Add model for custom filter keywords * Use CustomFilterKeyword internally Does not change the API * Fix /filters/edit and /filters/new * Add migration tests * Remove whole_word column from custom_filters (covered by custom_filter_keywords) * Redesign /filters Instead of a list, present a card that displays more information and handles multiple keywords per filter. * Redesign /filters/new and /filters/edit to add and remove keywords This adds a new gem dependency: cocoon, as well as a npm dependency: cocoon-js-vanilla. Those are used to easily populate and remove form fields from the user interface when manipulating multiple keyword filters at once. * Add /api/v2/filters to edit filter with multiple keywords Entities: - `Filter`: `id`, `title`, `filter_action` (either `hide` or `warn`), `context` `keywords` - `FilterKeyword`: `id`, `keyword`, `whole_word` API endpoits: - `GET /api/v2/filters` to list filters (including keywords) - `POST /api/v2/filters` to create a new filter `keywords_attributes` can also be passed to create keywords in one request - `GET /api/v2/filters/:id` to read a particular filter - `PUT /api/v2/filters/:id` to update a new filter `keywords_attributes` can also be passed to edit, delete or add keywords in one request - `DELETE /api/v2/filters/:id` to delete a particular filter - `GET /api/v2/filters/:id/keywords` to list keywords for a filter - `POST /api/v2/filters/:filter_id/keywords/:id` to add a new keyword to a filter - `GET /api/v2/filter_keywords/:id` to read a particular keyword - `PUT /api/v2/filter_keywords/:id` to edit a particular keyword - `DELETE /api/v2/filter_keywords/:id` to delete a particular keyword * Change from `irreversible` boolean to `action` enum * Remove irrelevent `irreversible_must_be_within_context` check * Fix /filters/new and /filters/edit with update for filter_action * Fix Rubocop/Codeclimate complaining about task names * Refactor FeedManager#phrase_filtered? This moves regexp building and filter caching to the `CustomFilter` class. This does not change the functional behavior yet, but this changes how the cache is built, doing per-custom_filter regexps so that filters can be matched independently, while still offering caching. * Perform server-side filtering and output result in REST API * Fix numerous filters_changed events being sent when editing multiple keywords at once * Add some tests * Use the new API in the WebUI - use client-side logic for filters we have fetched rules for. This is so that filter changes can be retroactively applied without reloading the UI. - use server-side logic for filters we haven't fetched rules for yet (e.g. network error, or initial timeline loading) * Minor optimizations and refactoring * Perform server-side filtering on the streaming server * Change the wording of filter action labels * Fix issues pointed out by linter * Change design of “Show anyway” link in accordence to review comments * Drop “irreversible” filtering behavior * Move /api/v2/filter_keywords to /api/v1/filters/keywords * Rename `filter_results` attribute to `filtered` * Rename REST::LegacyFilterSerializer to REST::V1::FilterSerializer * Fix systemChannelId value in streaming server * Simplify code by removing client-side filtering code The simplifcation comes at a cost though: filters aren't retroactively applied anymore.
2022-06-28 08:42:13 +01:00
listener,
stopHeartbeat: () => {
},
};
subscriptions[systemChannelId] = {
channelName: 'system',
listener,
stopHeartbeat: () => {
},
};
metrics.connectedChannels.labels({ type: 'websocket', channel: 'system' }).inc(2);
};
/**
* @param {import('ws').WebSocket & { isAlive: boolean }} ws
* @param {http.IncomingMessage & ResolvedAccount} req
* @param {import('pino').Logger} log
*/
function onConnection(ws, req, log) {
// Note: url.parse could throw, which would terminate the connection, so we
// increment the connected clients metric straight away when we establish
// the connection, without waiting:
metrics.connectedClients.labels({ type: 'websocket' }).inc();
// Setup connection keep-alive state:
ws.isAlive = true;
ws.on('pong', () => {
ws.isAlive = true;
});
/**
* @type {WebSocketSession}
*/
const session = {
websocket: ws,
request: req,
logger: log,
subscriptions: {},
};
ws.on('close', function onWebsocketClose() {
const subscriptions = Object.keys(session.subscriptions);
subscriptions.forEach(channelIds => {
removeSubscription(session, channelIds.split(';'));
});
// Decrement the metrics for connected clients:
metrics.connectedClients.labels({ type: 'websocket' }).dec();
// We need to unassign the session object as to ensure it correctly gets
// garbage collected, without doing this we could accidentally hold on to
// references to the websocket, the request, and the logger, causing
// memory leaks.
// This is commented out because `delete` only operated on object properties
// It needs to be replaced by `session = undefined`, but it requires every calls to
// `session` to check for it, thus a significant refactor
// delete session;
});
// Note: immediately after the `error` event is emitted, the `close` event
// is emitted. As such, all we need to do is log the error here.
ws.on('error', (/** @type {Error} */ err) => {
log.error(err);
});
ws.on('message', (data, isBinary) => {
if (isBinary) {
log.warn('Received binary data, closing connection');
ws.close(1003, 'The mastodon streaming server does not support binary messages');
return;
}
const message = data.toString('utf8');
const json = parseJSON(message, session.request);
if (!json) return;
const { type, stream, ...params } = json;
if (type === 'subscribe') {
subscribeWebsocketToChannel(session, firstParam(stream), params);
} else if (type === 'unsubscribe') {
unsubscribeWebsocketFromChannel(session, firstParam(stream), params);
} else {
// Unknown action type
}
});
subscribeWebsocketToSystemChannel(session);
// Parse the URL for the connection arguments (if supplied), url.parse can throw:
const location = req.url && url.parse(req.url, true);
if (location && location.query.stream) {
subscribeWebsocketToChannel(session, firstParam(location.query.stream), location.query);
}
}
wss.on('connection', onConnection);
setInterval(() => {
wss.clients.forEach(ws => {
// @ts-ignore
if (ws.isAlive === false) {
ws.terminate();
return;
}
// @ts-ignore
ws.isAlive = false;
ws.ping('', false);
});
}, 30000);
attachServerWithConfig(server, address => {
logger.info(`Streaming API now listening on ${address}`);
});
const onExit = () => {
server.close();
process.exit(0);
};
/** @param {Error} err */
const onError = (err) => {
logger.error(err);
server.close();
process.exit(0);
};
process.on('SIGINT', onExit);
process.on('SIGTERM', onExit);
process.on('exit', onExit);
process.on('uncaughtException', onError);
};
/**
* @param {any} server
* @param {function(string): void} [onSuccess]
*/
const attachServerWithConfig = (server, onSuccess) => {
if (process.env.SOCKET) {
server.listen(process.env.SOCKET, () => {
if (onSuccess) {
fs.chmodSync(server.address(), 0o666);
onSuccess(server.address());
}
});
} else {
const port = +(process.env.PORT || 4000);
let bind = process.env.BIND ?? '127.0.0.1';
// Web uses the URI syntax for BIND, which means IPv6 addresses may
// be wrapped in square brackets:
if (bind.startsWith('[') && bind.endsWith(']')) {
bind = bind.slice(1, -1);
}
server.listen(port, bind, () => {
if (onSuccess) {
onSuccess(`${server.address().address}:${server.address().port}`);
}
});
}
};
startServer();