utils.ts 14.5 KB
Newer Older
liuzhe-lz's avatar
liuzhe-lz committed
1
2
// Copyright (c) Microsoft Corporation.
// Licensed under the MIT license.
Deshui Yu's avatar
Deshui Yu committed
3
4
5

'use strict';

6
import * as assert from 'assert';
Deshui Yu's avatar
Deshui Yu committed
7
import { randomBytes } from 'crypto';
8
import * as cpp from 'child-process-promise';
9
10
import * as cp from 'child_process';
import { ChildProcess, spawn, StdioOptions } from 'child_process';
Deshui Yu's avatar
Deshui Yu committed
11
import * as fs from 'fs';
J-shang's avatar
J-shang committed
12
import * as net from 'net';
Deshui Yu's avatar
Deshui Yu committed
13
14
import * as os from 'os';
import * as path from 'path';
15
import * as lockfile from 'lockfile';
Deshui Yu's avatar
Deshui Yu committed
16
17
18
import { Deferred } from 'ts-deferred';
import { Container } from 'typescript-ioc';
import * as util from 'util';
19
import * as glob from 'glob';
Deshui Yu's avatar
Deshui Yu committed
20
21

import { Database, DataStore } from './datastore';
22
import { getExperimentStartupInfo, setExperimentStartupInfo } from './experimentStartupInfo';
23
import { ExperimentConfig, Manager } from './manager';
24
import { ExperimentManager } from './experimentManager';
QuanluZhang's avatar
QuanluZhang committed
25
import { HyperParameters, TrainingService, TrialJobStatus } from './trainingService';
Deshui Yu's avatar
Deshui Yu committed
26

27
function getExperimentRootDir(): string {
28
    return getExperimentStartupInfo().logDir;
Deshui Yu's avatar
Deshui Yu committed
29
30
}

31
function getLogDir(): string {
Deshui Yu's avatar
Deshui Yu committed
32
33
34
    return path.join(getExperimentRootDir(), 'log');
}

35
function getLogLevel(): string {
36
    return getExperimentStartupInfo().logLevel;
37
38
}

Deshui Yu's avatar
Deshui Yu committed
39
40
41
42
function getDefaultDatabaseDir(): string {
    return path.join(getExperimentRootDir(), 'db');
}

QuanluZhang's avatar
QuanluZhang committed
43
44
45
46
function getCheckpointDir(): string {
    return path.join(getExperimentRootDir(), 'checkpoint');
}

47
48
49
50
function getExperimentsInfoPath(): string {
    return path.join(os.homedir(), 'nni-experiments', '.experiment');
}

Deshui Yu's avatar
Deshui Yu committed
51
52
53
54
55
56
57
58
function mkDirP(dirPath: string): Promise<void> {
    const deferred: Deferred<void> = new Deferred<void>();
    fs.exists(dirPath, (exists: boolean) => {
        if (exists) {
            deferred.resolve();
        } else {
            const parent: string = path.dirname(dirPath);
            mkDirP(parent).then(() => {
59
                fs.mkdir(dirPath, (err: Error) => {
Deshui Yu's avatar
Deshui Yu committed
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
                    if (err) {
                        deferred.reject(err);
                    } else {
                        deferred.resolve();
                    }
                });
            }).catch((err: Error) => {
                deferred.reject(err);
            });
        }
    });

    return deferred.promise;
}

function mkDirPSync(dirPath: string): void {
    if (fs.existsSync(dirPath)) {
        return;
    }
    mkDirPSync(path.dirname(dirPath));
    fs.mkdirSync(dirPath);
}

const delay: (ms: number) => Promise<void> = util.promisify(setTimeout);

/**
 * Convert index to character
 * @param index index
 * @returns a mapping character
 */
function charMap(index: number): number {
    if (index < 26) {
        return index + 97;
    } else if (index < 52) {
        return index - 26 + 65;
    } else {
        return index - 52 + 48;
    }
}

/**
 * Generate a unique string by length
 * @param len length of string
 * @returns a unique string
 */
function uniqueString(len: number): string {
    if (len === 0) {
        return '';
    }
    const byteLength: number = Math.ceil((Math.log2(52) + Math.log2(62) * (len - 1)) / 8);
    let num: number = randomBytes(byteLength).reduce((a: number, b: number) => a * 256 + b, 0);
    const codes: number[] = [];
    codes.push(charMap(num % 52));
    num = Math.floor(num / 52);
    for (let i: number = 1; i < len; i++) {
        codes.push(charMap(num % 62));
        num = Math.floor(num / 62);
    }

    return String.fromCharCode(...codes);
}

122
123
124
125
function randomInt(max: number): number {
    return Math.floor(Math.random() * max);
}

126
127
128
129
130
function randomSelect<T>(a: T[]): T {
    assert(a !== undefined);

    return a[Math.floor(Math.random() * a.length)];
}
131

Deshui Yu's avatar
Deshui Yu committed
132
133
134
135
136
137
138
139
140
141
142
143
function parseArg(names: string[]): string {
    if (process.argv.length >= 4) {
        for (let i: number = 2; i < process.argv.length - 1; i++) {
            if (names.includes(process.argv[i])) {
                return process.argv[i + 1];
            }
        }
    }

    return '';
}

144
function getCmdPy(): string {
145
    let cmd = 'python3';
146
    if (process.platform === 'win32') {
147
148
149
150
151
        cmd = 'python';
    }
    return cmd;
}

152
/**
153
 * Generate command line to start automl algorithm(s),
QuanluZhang's avatar
QuanluZhang committed
154
 * either start advisor or start a process which runs tuner and assessor
155
 *
chicm-ms's avatar
chicm-ms committed
156
 * @param expParams: experiment startup parameters
157
158
 *
 */
159
function getMsgDispatcherCommand(expParams: ExperimentConfig): string {
chicm-ms's avatar
chicm-ms committed
160
161
162
    const clonedParams = Object.assign({}, expParams);
    delete clonedParams.searchSpace;
    return `${getCmdPy()} -m nni --exp_params ${Buffer.from(JSON.stringify(clonedParams)).toString('base64')}`;
163
164
}

165
166
167
168
/**
 * Generate parameter file name based on HyperParameters object
 * @param hyperParameters HyperParameters instance
 */
chicm-ms's avatar
chicm-ms committed
169
function generateParamFileName(hyperParameters: HyperParameters): string {
170
171
172
    assert(hyperParameters !== undefined);
    assert(hyperParameters.index >= 0);

chicm-ms's avatar
chicm-ms committed
173
    let paramFileName: string;
174
    if (hyperParameters.index == 0) {
175
176
177
178
179
180
181
        paramFileName = 'parameter.cfg';
    } else {
        paramFileName = `parameter_${hyperParameters.index}.cfg`
    }
    return paramFileName;
}

Deshui Yu's avatar
Deshui Yu committed
182
183
184
185
186
187
188
189
190
/**
 * Initialize a pseudo experiment environment for unit test.
 * Must be paired with `cleanupUnitTest()`.
 */
function prepareUnitTest(): void {
    Container.snapshot(Database);
    Container.snapshot(DataStore);
    Container.snapshot(TrainingService);
    Container.snapshot(Manager);
191
    Container.snapshot(ExperimentManager);
Deshui Yu's avatar
Deshui Yu committed
192

193
194
195
    const logLevel: string = parseArg(['--log_level', '-ll']);

    setExperimentStartupInfo(true, 'unittest', 8080, 'unittest', undefined, logLevel);
Deshui Yu's avatar
Deshui Yu committed
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
    mkDirPSync(getLogDir());

    const sqliteFile: string = path.join(getDefaultDatabaseDir(), 'nni.sqlite');
    try {
        fs.unlinkSync(sqliteFile);
    } catch (err) {
        // file not exists, good
    }
}

/**
 * Clean up unit test pseudo experiment.
 * Must be paired with `prepareUnitTest()`.
 */
function cleanupUnitTest(): void {
    Container.restore(Manager);
    Container.restore(TrainingService);
    Container.restore(DataStore);
    Container.restore(Database);
215
    Container.restore(ExperimentManager);
216
217
    const logLevel: string = parseArg(['--log_level', '-ll']);
    setExperimentStartupInfo(true, 'unittest', 8080, 'unittest', undefined, logLevel);
Deshui Yu's avatar
Deshui Yu committed
218
219
}

chicm-ms's avatar
chicm-ms committed
220
let cachedipv4Address: string = '';
221
222
223
224
/**
 * Get IPv4 address of current machine
 */
function getIPV4Address(): string {
225
226
227
    if (cachedipv4Address && cachedipv4Address.length > 0) {
        return cachedipv4Address;
    }
228

229
230
231
    const networkInterfaces = os.networkInterfaces();
    if (networkInterfaces.eth0) {
        for (const item of networkInterfaces.eth0) {
232
            if (item.family === 'IPv4') {
233
234
235
                cachedipv4Address = item.address;
                return cachedipv4Address;
            }
236
        }
237
    } else {
238
        throw Error(`getIPV4Address() failed because os.networkInterfaces().eth0 is undefined. Please specify NNI manager IP in config.`);
239
    }
240
241

    throw Error('getIPV4Address() failed because no valid IPv4 address found.')
242
243
}

QuanluZhang's avatar
QuanluZhang committed
244
245
246
247
248
249
250
/**
 * Get the status of canceled jobs according to the hint isEarlyStopped
 */
function getJobCancelStatus(isEarlyStopped: boolean): TrialJobStatus {
    return isEarlyStopped ? 'EARLY_STOPPED' : 'USER_CANCELED';
}

251
252
253
254
/**
 * Utility method to calculate file numbers under a directory, recursively
 * @param directory directory name
 */
chicm-ms's avatar
chicm-ms committed
255
function countFilesRecursively(directory: string): Promise<number> {
256
    if (!fs.existsSync(directory)) {
257
258
259
260
261
        throw Error(`Direcotory ${directory} doesn't exist`);
    }

    const deferred: Deferred<number> = new Deferred<number>();

chicm-ms's avatar
chicm-ms committed
262
263
    let timeoutId: NodeJS.Timer
    const delayTimeout: Promise<number> = new Promise((resolve: Function, reject: Function): void => {
264
265
266
267
268
269
270
        // Set timeout and reject the promise once reach timeout (5 seconds)
        timeoutId = setTimeout(() => {
            reject(new Error(`Timeout: path ${directory} has too many files`));
        }, 5000);
    });

    let fileCount: number = -1;
271
    let cmd: string;
272
    if (process.platform === "win32") {
273
274
        cmd = `powershell "Get-ChildItem -Path ${directory} -Recurse -File | Measure-Object | %{$_.Count}"`
    } else {
275
        cmd = `find ${directory} -type f | wc -l`;
276
277
    }
    cpp.exec(cmd).then((result) => {
278
        if (result.stdout && parseInt(result.stdout)) {
279
            fileCount = parseInt(result.stdout);
280
281
282
283
284
285
286
287
        }
        deferred.resolve(fileCount);
    });
    return Promise.race([deferred.promise, delayTimeout]).finally(() => {
        clearTimeout(timeoutId);
    });
}

288
export function validateFileName(fileName: string): boolean {
chicm-ms's avatar
chicm-ms committed
289
    const pattern: string = '^[a-z0-9A-Z._-]+$';
290
    const validateResult = fileName.match(pattern);
291
    if (validateResult) {
292
293
294
295
296
297
        return true;
    }
    return false;
}

async function validateFileNameRecursively(directory: string): Promise<boolean> {
298
    if (!fs.existsSync(directory)) {
299
300
301
302
303
        throw Error(`Direcotory ${directory} doesn't exist`);
    }

    const fileNameArray: string[] = fs.readdirSync(directory);
    let result = true;
304
    for (const name of fileNameArray) {
305
306
307
308
309
310
311
        const fullFilePath: string = path.join(directory, name);
        try {
            // validate file names and directory names
            result = validateFileName(name);
            if (fs.lstatSync(fullFilePath).isDirectory()) {
                result = result && await validateFileNameRecursively(fullFilePath);
            }
312
            if (!result) {
313
314
                return Promise.reject(new Error(`file name in ${fullFilePath} is not valid!`));
            }
315
        } catch (error) {
316
317
318
            return Promise.reject(error);
        }
    }
319
    return Promise.resolve(result);
320
321
}

322
323
324
325
/**
 * get the version of current package
 */
async function getVersion(): Promise<string> {
chicm-ms's avatar
chicm-ms committed
326
    const deferred: Deferred<string> = new Deferred<string>();
327
    import(path.join(__dirname, '..', 'package.json')).then((pkg) => {
328
        deferred.resolve(pkg.version);
329
330
    }).catch(() => {
        deferred.resolve('999.0.0-developing');
331
332
    });
    return deferred.promise;
333
}
334

335
336
337
/**
 * run command as ChildProcess
 */
J-shang's avatar
J-shang committed
338
function getTunerProc(command: string, stdio: StdioOptions, newCwd: string, newEnv: any, newShell: boolean = true, isDetached: boolean = false): ChildProcess {
339
340
    let cmd: string = command;
    let arg: string[] = [];
341
    if (process.platform === "win32") {
342
        cmd = command.split(" ", 1)[0];
343
        arg = command.substr(cmd.length + 1).split(" ");
344
        newShell = false;
345
        isDetached = true;
346
347
348
349
350
    }
    const tunerProc: ChildProcess = spawn(cmd, arg, {
        stdio,
        cwd: newCwd,
        env: newEnv,
351
352
        shell: newShell,
        detached: isDetached
353
354
355
356
357
358
359
    });
    return tunerProc;
}

/**
 * judge whether the process is alive
 */
Yuge Zhang's avatar
Yuge Zhang committed
360
async function isAlive(pid: any): Promise<boolean> {
chicm-ms's avatar
chicm-ms committed
361
    const deferred: Deferred<boolean> = new Deferred<boolean>();
362
    let alive: boolean = false;
Yuge Zhang's avatar
Yuge Zhang committed
363
    if (process.platform === 'win32') {
364
365
366
367
368
369
370
        try {
            const str = cp.execSync(`powershell.exe Get-Process -Id ${pid} -ErrorAction SilentlyContinue`).toString();
            if (str) {
                alive = true;
            }
        }
        catch (error) {
chicm-ms's avatar
chicm-ms committed
371
            //ignore
372
373
        }
    }
Yuge Zhang's avatar
Yuge Zhang committed
374
    else {
375
376
377
378
379
380
381
382
383
384
385
386
        try {
            await cpp.exec(`kill -0 ${pid}`);
            alive = true;
        } catch (error) {
            //ignore
        }
    }
    deferred.resolve(alive);
    return deferred.promise;
}

/**
387
 * kill process
388
 */
Yuge Zhang's avatar
Yuge Zhang committed
389
async function killPid(pid: any): Promise<void> {
chicm-ms's avatar
chicm-ms committed
390
    const deferred: Deferred<void> = new Deferred<void>();
391
392
    try {
        if (process.platform === "win32") {
Yuge Zhang's avatar
Yuge Zhang committed
393
            await cpp.exec(`cmd.exe /c taskkill /PID ${pid} /F`);
394
        }
395
        else {
396
397
398
399
400
401
402
403
404
            await cpp.exec(`kill -9 ${pid}`);
        }
    } catch (error) {
        // pid does not exist, do nothing here
    }
    deferred.resolve();
    return deferred.promise;
}

405
function getNewLine(): string {
406
407
408
    if (process.platform === "win32") {
        return "\r\n";
    }
409
    else {
410
411
412
413
        return "\n";
    }
}

414
415
/**
 * Use '/' to join path instead of '\' for all kinds of platform
416
 * @param path
417
418
419
420
421
422
423
 */
function unixPathJoin(...paths: any[]): string {
    const dir: string = paths.filter((path: any) => path !== '').join('/');
    if (dir === '') return '.';
    return dir;
}

424
425
426
427
428
429
430
431
432
/**
 * lock a file sync
 */
function withLockSync(func: Function, filePath: string, lockOpts: {[key: string]: any}, ...args: any): any {
    const lockName = path.join(path.dirname(filePath), path.basename(filePath) + `.lock.${process.pid}`);
    if (typeof lockOpts.stale === 'number'){
        const lockPath = path.join(path.dirname(filePath), path.basename(filePath) + '.lock.*');
        const lockFileNames: string[] = glob.sync(lockPath);
        const canLock: boolean = lockFileNames.map((fileName) => {
433
434
            return fs.existsSync(fileName) && Date.now() - fs.statSync(fileName).mtimeMs < lockOpts.stale;
        }).filter(unexpired=>unexpired === true).length === 0;
435
436
437
438
439
440
441
442
443
444
        if (!canLock) {
            throw new Error('File has been locked.');
        }
    }
    lockfile.lockSync(lockName, lockOpts);
    const result = func(...args);
    lockfile.unlockSync(lockName);
    return result;
}

J-shang's avatar
J-shang committed
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
async function isPortOpen(host: string, port: number): Promise<boolean> {
    return new Promise<boolean>((resolve, reject) => {
        try{
            const stream = net.createConnection(port, host);
            const id = setTimeout(() => {
                stream.destroy();
                resolve(false);
            }, 1000);

            stream.on('connect', () => {
                clearTimeout(id);
                stream.destroy();
                resolve(true);
            });

            stream.on('error', () => {
                clearTimeout(id);
                stream.destroy();
                resolve(false);
            });
        } catch (error) {
            reject(error);
        }
    });
}

async function getFreePort(host: string, start: number, end: number): Promise<number> {
    if (start > end) {
        throw new Error(`no more free port`);
    }
    if (await isPortOpen(host, start)) {
        return await getFreePort(host, start + 1, end);
    } else {
        return start;
    }
}

482
483
484
485
486
export function importModule(modulePath: string): any {
    module.paths.unshift(path.dirname(modulePath));
    return require(path.basename(modulePath));
}

487
export {
488
    countFilesRecursively, validateFileNameRecursively, generateParamFileName, getMsgDispatcherCommand, getCheckpointDir, getExperimentsInfoPath,
J-shang's avatar
J-shang committed
489
    getLogDir, getExperimentRootDir, getJobCancelStatus, getDefaultDatabaseDir, getIPV4Address, unixPathJoin, withLockSync, getFreePort, isPortOpen,
490
491
    mkDirP, mkDirPSync, delay, prepareUnitTest, parseArg, cleanupUnitTest, uniqueString, randomInt, randomSelect, getLogLevel, getVersion, getCmdPy, getTunerProc, isAlive, killPid, getNewLine
};