Chore: dispatch write to workers

This commit is contained in:
SukkaW
2026-08-02 00:09:56 +08:00
parent 6d8070d5f3
commit fa9c02d27d
9 changed files with 528 additions and 199 deletions

View File

@@ -19,6 +19,7 @@ import { OUTPUT_INTERNAL_DIR, SOURCE_DIR } from './constants/dir';
import { DomainsetOutput, AdGuardHomeOutput } from './lib/rules/domainset';
import { foundDebugDomain } from './lib/parse-filter/shared';
import { createWorker } from './lib/worker';
import { endOutputWorkerFarm } from './lib/rules/output-worker-farm';
import type { MaybePromise } from './lib/misc';
import { RulesetOutput } from './lib/rules/ruleset';
import { fetchAssets } from './lib/fetch-assets';
@@ -307,5 +308,8 @@ export const buildRejectDomainSet = task(require.main === module, __filename)(as
.addFromRuleset(readFileIntoProcessedArray(path.join(SOURCE_DIR, 'non_ip/my_reject.conf')))
.write();
await phishingWorker.end();
await Promise.all([
phishingWorker.end(),
endOutputWorkerFarm()
]);
});

View File

@@ -28,6 +28,7 @@ import path from 'node:path';
import { ROOT_DIR } from './constants/dir';
import { isCI } from 'ci-info';
import { printExternalDownloadStats } from './lib/download-stats';
import { endOutputWorkerFarm } from './lib/rules/output-worker-farm';
process.on('uncaughtException', (error) => {
console.error('Uncaught exception:', error);
@@ -130,7 +131,9 @@ const buildFinishedLock = path.join(ROOT_DIR, '.BUILD_FINISHED');
microsoftCdnWorker.end(),
cdnDownloadWorker.end(),
telegramCidrWorker.end(),
mockAssetsWorker.end()
mockAssetsWorker.end(),
// defensive: no-op unless some FileOutput crossed the offload threshold
endOutputWorkerFarm()
]);
// Finish the build to avoid leaking timer/fetch ref

View File

@@ -1,7 +1,6 @@
import { fastStringArrayJoin } from 'foxts/fast-string-array-join';
import { Buffer } from 'node:buffer';
import { createHash } from 'node:crypto';
import fs from 'node:fs';
import fsp from 'node:fs/promises';
/**
@@ -65,21 +64,6 @@ export async function extractContentHashFromFile(filePath: string): Promise<stri
return extractContentHash(await readFileHead(filePath));
}
/** Synchronous variant of {@link extractContentHashFromFile}, for worker threads where blocking is fine */
export function extractContentHashFromFileSync(filePath: string): string | null {
let fd: number | null = null;
try {
fd = fs.openSync(filePath, 'r');
const buf = Buffer.allocUnsafe(FILE_HEAD_CHUNK_SIZE);
const bytesRead = fs.readSync(fd, buf, 0, FILE_HEAD_CHUNK_SIZE, 0);
return extractContentHash(buf.toString('utf8', 0, bytesRead));
} finally {
if (fd !== null) {
fs.closeSync(fd);
}
}
}
function extractContentHash(fileHead: string): string | null {
const markerIndex = fileHead.indexOf(CONTENT_HASH_MARKER_START);
if (markerIndex === -1) {

View File

@@ -1,5 +1,6 @@
import { fastStringArrayJoin } from 'foxts/fast-string-array-join';
import fs from 'node:fs';
import path from 'node:path';
import picocolors from 'picocolors';
import type { Span } from '../trace';
import { readFileByLine } from './fetch-text-by-line';
@@ -10,16 +11,12 @@ import { extractContentHashFromFile } from './content-hash';
const fileEqual = createCompareSource(fileEqualWithCommentComparator);
/**
* To keep metadata comment `last updated` not change if real content is the same,
* we only write when the actual content differs, and the new `last updated` will
* be written along with new content.
*
* When `contentHash` is provided (and the previous output already embeds a
* content hash marker), the comparison only reads the first chunk of the
* previous file. Otherwise it falls back to a full comment-insensitive
* line-by-line comparison.
* The comparison half shared by {@link compareAndWriteFile} and
* {@link compareAndWriteFileInWorker} -- there is exactly one implementation of
* "is the previous output already up to date", regardless of how the new content
* ends up being written.
*/
export async function compareAndWriteFile(span: Span, linesA: string[], filePath: string, contentHash: string | null = null) {
async function isPreviousOutputEqual(span: Span, linesA: string[], filePath: string, contentHash: string | null) {
// readFileByLine will not include last empty line. So we always pop the linesA for comparison purpose
if (linesA.length > 0 && linesA[linesA.length - 1] === '') {
linesA.pop();
@@ -44,15 +41,53 @@ export async function compareAndWriteFile(span: Span, linesA: string[], filePath
if (isEqual) {
console.log(picocolors.gray(picocolors.dim(`same content, bail out writing: ${filePath}`)));
}
return isEqual;
}
/**
* To keep metadata comment `last updated` not change if real content is the same,
* we only write when the actual content differs, and the new `last updated` will
* be written along with new content.
*
* When `contentHash` is provided (and the previous output already embeds a
* content hash marker), the comparison only reads the first chunk of the
* previous file. Otherwise it falls back to a full comment-insensitive
* line-by-line comparison.
*/
export async function compareAndWriteFile(span: Span, linesA: string[], filePath: string, contentHash: string | null = null) {
if (await isPreviousOutputEqual(span, linesA, filePath, contentHash)) {
return;
}
return writeFileLines(span, linesA, filePath);
}
/**
* Same comparison as {@link compareAndWriteFile}, but the write itself is
* synchronous. For worker threads only: blocking a dedicated worker is free,
* whereas an async write hands the completion back through libuv to an event
* loop we would rather not depend on being idle.
*/
export async function compareAndWriteFileInWorker(span: Span, linesA: string[], filePath: string, contentHash: string | null = null) {
if (await isPreviousOutputEqual(span, linesA, filePath, contentHash)) {
return;
}
writeFileLinesSync(span, linesA, filePath);
}
export function writeFileLines(span: Span, linesA: string[], filePath: string): Promise<void> {
return span.traceChildAsync<void>(
`writing ${filePath}`,
() => writeFile(filePath, fastStringArrayJoin(linesA, '\n') + '\n')
);
}
export function writeFileLinesSync(span: Span, linesA: string[], filePath: string): void {
span.traceChildSync(`writing ${filePath}`, () => {
fs.mkdirSync(path.dirname(filePath), { recursive: true });
fs.writeFileSync(filePath, fastStringArrayJoin(linesA, '\n') + '\n');
});
}

View File

@@ -1,16 +1,24 @@
import type { Span } from '../../trace';
import { HostnameSmolTrie } from 'hntrie/smol';
import { domainToASCII } from 'node:url';
import { not, nullthrow } from 'foxts/guard';
import { fastIpVersion } from 'foxts/fast-ip-version';
import { addArrayElementsToSet } from 'foxts/add-array-elements-to-set';
import type { MaybePromise } from '../misc';
import type { BaseWriteStrategy } from '../writing-strategy/base';
import { merge as mergeCidr } from 'fast-cidr-tools';
import { createRetrieKeywordFilter as createKeywordFilter } from 'foxts/retrie';
import path from 'node:path';
import { SurgeMitmSgmodule } from '../writing-strategy/surge';
import { appendArrayInPlace } from 'foxts/append-array-in-place';
import { isMainThread } from 'node:worker_threads';
import { resolveStrategyOutputPath, serializeStrategy, writeDataToStrategies } from './strategy-write-data';
import type { OutputWorkerPayload, StrategyWriteData } from './strategy-write-data';
import { getOutputWorkerFarm } from './output-worker-farm';
/**
* Below this many dumped domain entries, formatting + hashing + writing inline is
* cheaper than a worker round trip (and avoids tiny outputs queueing behind big
* jobs in the farm). Today only the reject domainsets / adguardhome outputs cross
* this threshold.
*/
const OUTPUT_WORKER_THRESHOLD = 20000;
/**
* Holds the universal rule data (domain, ip, url-regex, etc. etc.)
@@ -399,7 +407,42 @@ export class FileOutput {
// }
private strategiesWritten = false;
private writeToStrategies() {
/** Collect the trie content, '.'-prefix encoded, WITHOUT punycoding (that happens in the fanout) */
private dumpDomains(): string[] {
const domains: string[] = [];
this.domainTrie.dump((domain, includeSubdomain) => {
domains.push(includeSubdomain ? '.' + domain : domain);
});
return domains;
}
private buildStrategyWriteData(domains: string[]): StrategyWriteData {
return {
domains,
domainKeywords: Array.from(this.domainKeywords),
whitelistKeywords: Array.from(this.whitelistKeywords),
wildcards: Array.from(this.wildcardSet),
userAgents: Array.from(this.userAgent),
processNames: Array.from(this.processName),
processPaths: Array.from(this.processPath),
urlRegexes: Array.from(this.urlRegex),
ipcidr: Array.from(this.ipcidr),
ipcidrNoResolve: Array.from(this.ipcidrNoResolve),
ipcidr6: Array.from(this.ipcidr6),
ipcidr6NoResolve: Array.from(this.ipcidr6NoResolve),
ipasn: Array.from(this.ipasn),
ipasnNoResolve: Array.from(this.ipasnNoResolve),
geoip: Array.from(this.geoip),
geoipNoResolve: Array.from(this.groipNoResolve),
sourceIpOrCidr: Array.from(this.sourceIpOrCidr),
sourcePort: Array.from(this.sourcePort),
destPort: Array.from(this.destPort),
protocol: Array.from(this.protocol),
otherRules: this.otherRules
};
}
private guardBeforeWritingToStrategies() {
if (this.pendingPromise) {
throw new Error('You should call done() before calling writeToStrategies()');
}
@@ -412,158 +455,20 @@ export class FileOutput {
if (this.strategies.filter(not(false)).length === 0) {
throw new Error('No strategies to write ' + this.id);
}
// We use both DOMAIN-KEYWORD and whitelisted keyword to whitelist DOMAIN and DOMAIN-SUFFIX
const kwfilter = createKeywordFilter(
Array.from(this.domainKeywords)
.concat(Array.from(this.whitelistKeywords))
);
const strategiesLen = this.strategies.length;
const domainEntries: Array<[domain: string, subdomain: boolean]> = [];
this.domainTrie.dump((domain, includeSubdomain) => {
const d = domainToASCII(domain);
if (d) domainEntries.push([d, includeSubdomain]);
});
domainEntries.sort((a, b) => (a[0].length - b[0].length) || (a[0] < b[0] ? -1 : (a[0] > b[0] ? 1 : 0)));
for (let j = 0, entriesLen = domainEntries.length; j < entriesLen; j++) {
const [domain, includeAllSubdomain] = domainEntries[j];
if (kwfilter(domain)) {
continue;
}
for (let i = 0; i < strategiesLen; i++) {
const strategy = this.strategies[i];
if (includeAllSubdomain) {
strategy.writeDomainSuffix(domain);
} else {
strategy.writeDomain(domain);
}
}
}
// Now, we whitelisted out DOMAIN-KEYWORD
const whiteKwfilter = createKeywordFilter(Array.from(this.whitelistKeywords));
const whitelistedKeywords = Array.from(this.domainKeywords).filter(kw => !whiteKwfilter(kw));
for (let i = 0; i < strategiesLen; i++) {
const strategy = this.strategies[i];
if (whitelistedKeywords.length) {
strategy.writeDomainKeywords(this.domainKeywords);
}
if (this.protocol.size) {
strategy.writeProtocols(this.protocol);
}
}
if (this.wildcardSet.size) {
this.wildcardSet.forEach((wildcard) => {
// Overlapped w/ DOMAIN-kEYWORD
if (kwfilter(wildcard)) {
return;
}
for (let i = 0; i < strategiesLen; i++) {
const strategy = this.strategies[i];
strategy.writeDomainWildcard(wildcard);
}
});
}
const sourceIpOrCidr = Array.from(this.sourceIpOrCidr);
for (let i = 0; i < strategiesLen; i++) {
const strategy = this.strategies[i];
if (this.userAgent.size) {
strategy.writeUserAgents(this.userAgent);
}
if (this.processName.size) {
strategy.writeProcessNames(this.processName);
}
if (this.processPath.size) {
strategy.writeProcessPaths(this.processPath);
}
if (this.sourceIpOrCidr.size) {
strategy.writeSourceIpCidrs(sourceIpOrCidr);
}
if (this.sourcePort.size) {
strategy.writeSourcePorts(this.sourcePort);
}
if (this.destPort.size) {
strategy.writeDestinationPorts(this.destPort);
}
if (this.otherRules.length) {
strategy.writeOtherRules(this.otherRules);
}
if (this.urlRegex.size) {
strategy.writeUrlRegexes(this.urlRegex);
}
}
let ipcidr: string[] | null = null;
let ipcidrNoResolve: string[] | null = null;
let ipcidr6: string[] | null = null;
let ipcidr6NoResolve: string[] | null = null;
if (this.ipcidr.size) {
ipcidr = mergeCidr(Array.from(this.ipcidr), true);
}
if (this.ipcidrNoResolve.size) {
ipcidrNoResolve = mergeCidr(Array.from(this.ipcidrNoResolve), true);
}
if (this.ipcidr6.size) {
ipcidr6 = Array.from(this.ipcidr6);
}
if (this.ipcidr6NoResolve.size) {
ipcidr6NoResolve = Array.from(this.ipcidr6NoResolve);
}
for (let i = 0; i < strategiesLen; i++) {
const strategy = this.strategies[i];
// no-resolve
if (ipcidrNoResolve) {
strategy.writeIpCidrs(ipcidrNoResolve, true);
}
if (ipcidr6NoResolve) {
strategy.writeIpCidr6s(ipcidr6NoResolve, true);
}
if (this.ipasnNoResolve.size) {
strategy.writeIpAsns(this.ipasnNoResolve, true);
}
if (this.groipNoResolve.size) {
strategy.writeGeoip(this.groipNoResolve, true);
}
// triggers DNS resolution
if (ipcidr?.length) {
strategy.writeIpCidrs(ipcidr, false);
}
if (ipcidr6?.length) {
strategy.writeIpCidr6s(ipcidr6, false);
}
if (this.ipasn.size) {
strategy.writeIpAsns(this.ipasn, false);
}
if (this.geoip.size) {
strategy.writeGeoip(this.geoip, false);
}
}
private writeToStrategies(domains: string[]) {
this.guardBeforeWritingToStrategies();
writeDataToStrategies(this.buildStrategyWriteData(domains), this.strategies);
}
write(): Promise<unknown> {
return this.span.traceChildAsync('write all', async (childSpan) => {
await childSpan.traceChildAsync('done', () => this.done());
childSpan.traceChildSync('write to strategies', () => this.writeToStrategies());
return childSpan.traceChildAsync('output to disk', (childSpan) => {
const promises: Array<Promise<void>> = [];
const domains = childSpan.traceChildSync('dump domain trie', () => this.dumpDomains());
const title = nullthrow(this.title, 'Missing title');
const descriptions = nullthrow(this.description, 'Missing description');
if (this.dataSource.size) {
@@ -574,24 +479,49 @@ export class FileOutput {
appendArrayInPlace(descriptions, Array.from(this.dataSource).sort().map((source) => ` - ${source}`));
}
// Big outputs offload everything from punycode to the write onto a worker
// thread: the payload crosses in a few ms (flat string arrays structured-clone
// cheaply), the main-thread event loop stays free, and the worker writes
// synchronously so no completion ever waits on a busy main thread.
//
// Only worth doing from the main thread -- tasks that already run entirely on
// a worker (build-microsoft-cdn, build-telegram-cidr, build-cdn-download-conf)
// are not contending with anything, and must not spawn a nested worker farm.
if (isMainThread && domains.length >= OUTPUT_WORKER_THRESHOLD) {
this.guardBeforeWritingToStrategies();
const payload: OutputWorkerPayload = {
id: this.id,
title,
description: descriptions,
dateMs: this.date.getTime(),
strategies: this.strategies.map(serializeStrategy),
data: this.buildStrategyWriteData(domains)
};
return childSpan.traceWorkerChild(
'write via output worker',
rawSpan => getOutputWorkerFarm().writeOutput(rawSpan, payload)
);
}
childSpan.traceChildSync('write to strategies', () => this.writeToStrategies(domains));
return childSpan.traceChildAsync('output to disk', (childSpan) => {
const promises: Array<Promise<void>> = [];
for (let i = 0, len = this.strategies.length; i < len; i++) {
const strategy = this.strategies[i];
const basename = (strategy.overwriteFilename || this.id) + '.' + strategy.fileExtension;
const filePath = resolveStrategyOutputPath(strategy, this.id);
promises.push(
childSpan.traceChildAsync('write ' + strategy.name, (childSpan) => Promise.resolve(strategy.output(
childSpan,
nullthrow(this.title, 'Missing title'),
descriptions,
this.date,
path.join(
strategy.outputDir,
strategy.type
? path.join(strategy.type, basename)
: basename
)
)))
childSpan.traceChildAsync('write ' + strategy.name, (childSpan) => Promise.resolve(
// Already off the main thread: block on the write instead of handing
// the completion back through libuv.
isMainThread
? strategy.output(childSpan, title, descriptions, this.date, filePath)
: strategy.outputInWorker(childSpan, title, descriptions, this.date, filePath)
))
);
}
@@ -602,7 +532,7 @@ export class FileOutput {
async compile(): Promise<Array<string[] | null>> {
await this.done();
this.writeToStrategies();
this.writeToStrategies(this.dumpDomains());
return this.strategies.reduce<Array<string[] | null>>((acc, strategy) => {
acc.push(strategy.content);

View File

@@ -0,0 +1,30 @@
import type { JestWorkerFarm } from 'jest-worker';
import { createWorker } from '../worker';
type OutputWorkerModule = typeof import('./output.worker');
type OutputWorkerFarm = JestWorkerFarm<Pick<OutputWorkerModule, 'writeOutput'>>;
let farm: OutputWorkerFarm | null = null;
/**
* Lazily boot the output worker farm. Only FileOutput#write dispatches here, and
* only when an output crosses the offload threshold -- today that is exclusively
* the reject domainsets / adguardhome outputs of build-reject-domainset.
*
* IMPORTANT: whoever triggers the lazy boot is responsible for a matching
* endOutputWorkerFarm() (idempotent, safe to call unconditionally), otherwise the
* worker threads keep a standalone task run alive.
*/
export function getOutputWorkerFarm(): OutputWorkerFarm {
// 3 workers: reject, reject_extra and reject_phishing format & write in parallel
farm ??= createWorker<OutputWorkerModule>(require.resolve('./output.worker'), 3)(['writeOutput']);
return farm;
}
export async function endOutputWorkerFarm(): Promise<void> {
if (farm) {
const f = farm;
farm = null;
await f.end();
}
}

View File

@@ -0,0 +1,37 @@
import { workerJob } from '../../trace';
import type { RawSpan, WorkerJobResult } from '../../trace';
import { resolveStrategyOutputPath, reviveStrategy, writeDataToStrategies } from './strategy-write-data';
import type { OutputWorkerPayload } from './strategy-write-data';
/**
* The off-main-thread half of FileOutput#write. Everything from "raw collections"
* to "bytes on disk" happens here: punycode, keyword filtering, sorting, the
* per-strategy format fanout (incl. the sing-box JSON stringify), banner and
* content hash, compare and write. All fs calls are synchronous on purpose --
* blocking this worker thread is free, and it avoids bouncing libuv completions
* off a busy main-thread event loop.
*/
export function writeOutput(rawSpan: RawSpan | undefined, payload: OutputWorkerPayload): Promise<WorkerJobResult<void>> {
return workerJob(rawSpan, (span) => {
const strategies = payload.strategies.map(reviveStrategy);
span.traceChildSync('write to strategies', () => writeDataToStrategies(payload.data, strategies));
const date = new Date(payload.dateMs);
return span.traceChildAsync('output to disk', async (childSpan) => {
// Sequential on purpose: the writes are synchronous, so there is nothing to
// overlap, and this keeps the emitted trace in strategy order.
for (let i = 0, len = strategies.length; i < len; i++) {
const strategy = strategies[i];
const filePath = resolveStrategyOutputPath(strategy, payload.id);
// eslint-disable-next-line no-await-in-loop -- see above
await childSpan.traceChildAsync(
'write ' + strategy.name,
(strategySpan) => strategy.outputInWorker(strategySpan, payload.title, payload.description, date, filePath)
);
}
});
});
}

View File

@@ -0,0 +1,265 @@
import path from 'node:path';
import { domainToASCII } from 'node:url';
import { merge as mergeCidr } from 'fast-cidr-tools';
import { createRetrieKeywordFilter as createKeywordFilter } from 'foxts/retrie';
import type { BaseWriteStrategy } from '../writing-strategy/base';
import { AdGuardHome } from '../writing-strategy/adguardhome';
import { ClashClassicRuleSet, ClashDomainSet, ClashIPSet } from '../writing-strategy/clash';
import { LegacyClashPremiumClassicRuleSet } from '../writing-strategy/legacy-clash-premium';
import { SingboxSource } from '../writing-strategy/singbox';
import { SurfboardRuleSet } from '../writing-strategy/surfboard';
import { SurgeDomainSet, SurgeMitmSgmodule, SurgeRuleSet } from '../writing-strategy/surge';
/**
* A structured-clone friendly snapshot of everything FileOutput feeds into its
* strategies. Only `domains` is big (the '.'-prefixed trie dump, still unicode);
* everything else is at most a few thousand short strings. Plain arrays are used
* (not Set) so the contract survives any worker serialization mode.
*/
export interface StrategyWriteData {
/** Raw trie dump, '.'-prefix means include-all-subdomain, NOT yet punycoded */
domains: string[],
domainKeywords: string[],
whitelistKeywords: string[],
wildcards: string[],
userAgents: string[],
processNames: string[],
processPaths: string[],
urlRegexes: string[],
ipcidr: string[],
ipcidrNoResolve: string[],
ipcidr6: string[],
ipcidr6NoResolve: string[],
ipasn: string[],
ipasnNoResolve: string[],
geoip: string[],
geoipNoResolve: string[],
sourceIpOrCidr: string[],
sourcePort: string[],
destPort: string[],
protocol: string[],
otherRules: string[]
}
/**
* Write a StrategyWriteData snapshot into strategies. This is the extracted body
* of FileOutput#writeToStrategies and MUST keep the exact same write order, since
* the order of writeXxx calls determines the line order of the final output.
*/
export function writeDataToStrategies(data: StrategyWriteData, strategies: BaseWriteStrategy[]): void {
// We use both DOMAIN-KEYWORD and whitelisted keyword to whitelist DOMAIN and DOMAIN-SUFFIX
const kwfilter = createKeywordFilter(data.domainKeywords.concat(data.whitelistKeywords));
const strategiesLen = strategies.length;
const domainEntries: Array<[domain: string, subdomain: boolean]> = [];
for (let j = 0, len = data.domains.length; j < len; j++) {
const line = data.domains[j];
const includeSubdomain = line.codePointAt(0) === 46; /* '.' */
const d = domainToASCII(includeSubdomain ? line.slice(1) : line);
if (d) domainEntries.push([d, includeSubdomain]);
}
domainEntries.sort((a, b) => (a[0].length - b[0].length) || (a[0] < b[0] ? -1 : (a[0] > b[0] ? 1 : 0)));
for (let j = 0, entriesLen = domainEntries.length; j < entriesLen; j++) {
const [domain, includeAllSubdomain] = domainEntries[j];
if (kwfilter(domain)) {
continue;
}
for (let i = 0; i < strategiesLen; i++) {
const strategy = strategies[i];
if (includeAllSubdomain) {
strategy.writeDomainSuffix(domain);
} else {
strategy.writeDomain(domain);
}
}
}
// Now, we whitelisted out DOMAIN-KEYWORD
const whiteKwfilter = createKeywordFilter(data.whitelistKeywords);
const whitelistedKeywords = data.domainKeywords.filter(kw => !whiteKwfilter(kw));
const domainKeywords = new Set(data.domainKeywords);
const protocol = new Set(data.protocol);
for (let i = 0; i < strategiesLen; i++) {
const strategy = strategies[i];
if (whitelistedKeywords.length) {
strategy.writeDomainKeywords(domainKeywords);
}
if (protocol.size) {
strategy.writeProtocols(protocol);
}
}
if (data.wildcards.length) {
data.wildcards.forEach((wildcard) => {
// Overlapped w/ DOMAIN-kEYWORD
if (kwfilter(wildcard)) {
return;
}
for (let i = 0; i < strategiesLen; i++) {
const strategy = strategies[i];
strategy.writeDomainWildcard(wildcard);
}
});
}
const userAgent = new Set(data.userAgents);
const processName = new Set(data.processNames);
const processPath = new Set(data.processPaths);
const sourcePort = new Set(data.sourcePort);
const destPort = new Set(data.destPort);
const urlRegex = new Set(data.urlRegexes);
for (let i = 0; i < strategiesLen; i++) {
const strategy = strategies[i];
if (userAgent.size) {
strategy.writeUserAgents(userAgent);
}
if (processName.size) {
strategy.writeProcessNames(processName);
}
if (processPath.size) {
strategy.writeProcessPaths(processPath);
}
if (data.sourceIpOrCidr.length) {
strategy.writeSourceIpCidrs(data.sourceIpOrCidr);
}
if (sourcePort.size) {
strategy.writeSourcePorts(sourcePort);
}
if (destPort.size) {
strategy.writeDestinationPorts(destPort);
}
if (data.otherRules.length) {
strategy.writeOtherRules(data.otherRules);
}
if (urlRegex.size) {
strategy.writeUrlRegexes(urlRegex);
}
}
let ipcidr: string[] | null = null;
let ipcidrNoResolve: string[] | null = null;
let ipcidr6: string[] | null = null;
let ipcidr6NoResolve: string[] | null = null;
if (data.ipcidr.length) {
ipcidr = mergeCidr(data.ipcidr, true);
}
if (data.ipcidrNoResolve.length) {
ipcidrNoResolve = mergeCidr(data.ipcidrNoResolve, true);
}
if (data.ipcidr6.length) {
ipcidr6 = data.ipcidr6;
}
if (data.ipcidr6NoResolve.length) {
ipcidr6NoResolve = data.ipcidr6NoResolve;
}
const ipasn = new Set(data.ipasn);
const ipasnNoResolve = new Set(data.ipasnNoResolve);
const geoip = new Set(data.geoip);
const geoipNoResolve = new Set(data.geoipNoResolve);
for (let i = 0; i < strategiesLen; i++) {
const strategy = strategies[i];
// no-resolve
if (ipcidrNoResolve) {
strategy.writeIpCidrs(ipcidrNoResolve, true);
}
if (ipcidr6NoResolve) {
strategy.writeIpCidr6s(ipcidr6NoResolve, true);
}
if (ipasnNoResolve.size) {
strategy.writeIpAsns(ipasnNoResolve, true);
}
if (geoipNoResolve.size) {
strategy.writeGeoip(geoipNoResolve, true);
}
// triggers DNS resolution
if (ipcidr?.length) {
strategy.writeIpCidrs(ipcidr, false);
}
if (ipcidr6?.length) {
strategy.writeIpCidr6s(ipcidr6, false);
}
if (ipasn.size) {
strategy.writeIpAsns(ipasn, false);
}
if (geoip.size) {
strategy.writeGeoip(geoip, false);
}
}
}
/** Where a strategy's output for a given output id lands on disk */
export function resolveStrategyOutputPath(strategy: BaseWriteStrategy, id: string): string {
const basename = (strategy.overwriteFilename || id) + '.' + strategy.fileExtension;
return path.join(
strategy.outputDir,
strategy.type
? path.join(strategy.type, basename)
: basename
);
}
/** Serializable stand-in for a strategy class instance crossing the worker boundary */
export interface StrategyDescriptor {
name: string,
type: string,
outputDir: string,
overwriteFilename: string | null
}
export function serializeStrategy(strategy: BaseWriteStrategy): StrategyDescriptor {
return {
name: strategy.name,
type: strategy.type,
outputDir: strategy.outputDir,
overwriteFilename: strategy.overwriteFilename
};
}
const strategyRegistry: Record<string, (d: StrategyDescriptor) => BaseWriteStrategy> = {
'surge domainset': (d) => new SurgeDomainSet(d.outputDir),
'surge ruleset': (d) => new SurgeRuleSet(d.type, d.outputDir),
'surge sgmodule': (d) => new SurgeMitmSgmodule(d.overwriteFilename ?? '', d.outputDir),
'clash domainset': (d) => new ClashDomainSet(d.outputDir),
'clash ipcidr': (d) => new ClashIPSet(d.outputDir),
'clash classic ruleset': (d) => new ClashClassicRuleSet(d.type, d.outputDir),
'legacy clash premium classic ruleset': (d) => new LegacyClashPremiumClassicRuleSet(d.type as 'ip' | 'non_ip', d.outputDir),
'surfboard for android ruleset': (d) => new SurfboardRuleSet(d.type as 'ip' | 'non_ip', d.outputDir),
singbox: (d) => new SingboxSource(d.type as 'domainset' | 'non_ip' | 'ip', d.outputDir),
adguardhome: (d) => new AdGuardHome(d.outputDir)
};
export function reviveStrategy(descriptor: StrategyDescriptor): BaseWriteStrategy {
if (!(descriptor.name in strategyRegistry)) {
throw new TypeError(`Unknown strategy "${descriptor.name}", add it to strategyRegistry before offloading it to the output worker`);
}
const strategy = strategyRegistry[descriptor.name](descriptor);
if (descriptor.overwriteFilename) {
strategy.withFilename(descriptor.overwriteFilename);
}
return strategy;
}
/** The payload of a single FileOutput#write offloaded to the output worker */
export interface OutputWorkerPayload {
id: string,
title: string,
description: string[],
dateMs: number,
strategies: StrategyDescriptor[],
data: StrategyWriteData
}

View File

@@ -1,7 +1,7 @@
import { isCI } from 'ci-info';
import type { Span } from '../../trace';
import { calculateContentHash } from '../content-hash';
import { compareAndWriteFile, writeFileLines } from '../create-file';
import { compareAndWriteFile, compareAndWriteFileInWorker, writeFileLines, writeFileLinesSync } from '../create-file';
/**
* The class is not about holding rule data, instead it determines how the
@@ -98,6 +98,47 @@ export abstract class BaseWriteStrategy {
);
};
/**
* Worker-thread twin of {@link output}: identical comparison, but the write is
* synchronous since blocking a dedicated worker costs nothing.
*/
public async outputInWorker(
span: Span,
title: string,
description: string[] | readonly string[],
date: Date,
filePath: string
): Promise<void> {
const result = this.result;
if (!result) {
return;
}
if (isCI && this.skipCompareOnCI) {
writeFileLinesSync(
span,
this.withPadding(title, description, date, result, null),
filePath
);
return;
}
const contentHash = calculateContentHash(title, description, result);
await compareAndWriteFileInWorker(
span,
this.withPadding(
title,
description,
date,
result,
contentHash
),
filePath,
contentHash
);
}
public get content() {
return this.result;
}