oxc/tasks/benchmark/codspeed/upload.mjs

/*
 * Combine benchmark data from different jobs, and upload to Codspeed.
 */

import assert from 'assert';
import axios from 'axios';
import { createHash } from 'crypto';
import { createReadStream } from 'fs';
import fs from 'fs/promises';
import { join as pathJoin } from 'path';
import { create as createTar } from 'tar';

const METADATA_SUFFIX = '_metadata.json',
  CODSPEED_UPLOAD_URL = 'https://api.codspeed.io/upload';

const dataDir = process.env.DATA_DIR,
  token = process.env.CODSPEED_TOKEN;

// Find profile files and first metadata file
const profileFiles = [];
let metadataPath;
for (const filename of await fs.readdir(dataDir)) {
  const path = pathJoin(dataDir, filename);
  if (filename.endsWith(METADATA_SUFFIX)) {
    if (!metadataPath) metadataPath = path;
  } else {
    const match = filename.match(/_(\d+)\.out$/);
    assert(match, `Unexpected file: ${filename}`);

    const pid = +match[1];
    profileFiles.push({ pid, path });
  }
}

// Move all `.out` files to one directory
console.log('Combining profiles');

const outDir = pathJoin(dataDir, 'out');
await fs.mkdir(outDir);

const pids = new Set(),
  duplicates = [];
let highestPid = -1;
for (const { pid, path } of profileFiles) {
  if (pids.has(pid)) {
    // Duplicate PID
    duplicates.push({ pid, path });
  } else {
    pids.add(pid);
    if (pid > highestPid) highestPid = pid;
    await fs.rename(path, pathJoin(outDir, `${pid}.out`));
  }
}

// Alter PIDs for `.out` files with duplicate filenames
for (let { pid, path } of duplicates) {
  let content = await fs.readFile(path, 'utf8');

  const pidLine = `\npid: ${pid}\n`;
  const index = content.indexOf(pidLine);
  assert(index !== -1, `Could not locate PID in ${path}`);
  const before = content.slice(0, index);
  assert(before.split('\n').length === 3, `Unexpected formatting in ${path}`);

  pid = ++highestPid;
  content = `${before}\npid: ${pid}\n${content.slice(index + pidLine.length)}`;

  await fs.writeFile(pathJoin(outDir, `${pid}.out`), content);
  await fs.rm(path);
}

// ZIP combined profile directory
console.log('Zipping combined profile directory');
const archivePath = pathJoin(dataDir, 'archive.tar.gz');
await createTar({ file: archivePath, gzip: true, cwd: outDir }, ['./']);

// Get size + MD5 hash of archive
console.log('Hashing ZIP');
const { size } = await fs.stat(archivePath);

const hash = createHash('md5');
const inputStream = createReadStream(archivePath);
for await (const chunk of inputStream) {
  hash.update(chunk);
}
const md5 = hash.digest('base64');

// Alter MD5 hash in metadata object
const metadata = JSON.parse(await fs.readFile(metadataPath, 'utf8'));
metadata.profileMd5 = md5;

// If no token, set `metadata.tokenless`, and log hash of metadata JSON.
// For tokenless runs (PRs from forks), `codspeed-runner` logs SHA256 hash of metadata JSON.
// CodSpeed then reads the job logs to find a line matching `CodSpeed Run Hash: "..."`.
// So we used a dummy token for `CodSpeedHQ/action` to prevent it logging the hash,
// so can log the correct hash ourselves here instead.
if (!token) metadata.tokenless = true;
const metadataJson = JSON.stringify(metadata);
if (!token) {
  const metadataHash = createHash('sha256').update(metadataJson).digest('hex');
  console.log(`CodSpeed Run Hash: "${metadataHash}"`);
}

// Upload metadata to CodSpeed
console.log('Uploading metadata to CodSpeed');
const { data } = await axios({
  method: 'post',
  url: CODSPEED_UPLOAD_URL,
  data: metadataJson,
  headers: {
    'Content-Type': 'application/json',
    ...(token ? { Authorization: token } : null),
  },
});
assert(data?.status === 'success', 'Failed to upload metadata to Codspeed');
const { uploadUrl } = data;

// Upload profile ZIP to Codspeed
console.log('Uploading profile ZIP to CodSpeed');
await axios({
  method: 'put',
  url: uploadUrl,
  data: createReadStream(archivePath),
  headers: {
    'Content-Type': 'application/gzip',
    'Content-Length': size,
    'Content-MD5': md5,
  },
});