archive-version.js

#!/usr/bin/env node

// [start-readme]
//
// Run this script during the Enterprise deprecation process to download
// static copies of all pages for the oldest supported Enterprise version.
// See the Enterprise deprecation issue template for instructions.
//
// [end-readme]

import { fileURLToPath } from 'url'
import path from 'path'
import fs from 'fs'
import { execSync } from 'child_process'
import createApp from '../../lib/app.js'
import scrape from 'website-scraper'
import program from 'commander'
import rimraf from 'rimraf'
import EnterpriseServerReleases from '../../lib/enterprise-server-releases.js'
import loadRedirects from '../../lib/redirects/precompile.js'
import { loadPageMap } from '../../lib/page-data.js'
const __filename = fileURLToPath(import.meta.url)
const __dirname = path.dirname(fileURLToPath(import.meta.url))

const port = '4001'
const host = `http://localhost:${port}`
const version = EnterpriseServerReleases.oldestSupported
const archivalRepoName = 'help-docs-archived-enterprise-versions'
const archivalRepoUrl = `https://github.com/github/${archivalRepoName}`
const remoteImageStoreBaseURL = 'https://githubdocs.azureedge.net/github-images'

program
  .description(
    'Scrape HTML of the oldest supported Enterprise version and add it to the archival repository.'
  )
  .option('-p, --path-to-archival-repo <PATH>', `path to a local checkout of ${archivalRepoUrl}`)
  .option('-d, --dry-run', 'only scrape the first 10 pages for testing purposes')
  .parse(process.argv)

const pathToArchivalRepo = program.opts().pathToArchivalRepo
const dryRun = program.opts().dryRun

main()

class RewriteAssetPathsPlugin {
  constructor(version, tempDirectory) {
    this.version = version
    this.tempDirectory = tempDirectory
  }

  apply(registerAction) {
    registerAction('onResourceSaved', async ({ resource }) => {
      // Show some activity
      process.stdout.write('.')

      // Only operate on HTML files
      if (!resource.isHtml() && !resource.isCss()) return

      // Get the text contents of the resource
      const text = resource.getText()
      let newBody = ''

      // Rewrite HTML asset paths. Example:
      // ../assets/images/foo/bar.png ->
      // https://githubdocs.azureedge.net/github-images/enterprise/2.17/assets/images/foo/bar.png
      if (resource.isHtml()) {
        newBody = text.replace(
          /(?<attribute>src|href)="(?:\.\.\/|\/)*(?<basepath>_next\/static|javascripts|stylesheets|assets\/fonts|assets\/images|node_modules)/g,
          (match, attribute, basepath) => {
            let replaced = path.join('/enterprise', this.version, basepath)
            if (basepath === 'assets/images') {
              replaced = remoteImageStoreBaseURL + replaced
            }
            const returnValue = `${attribute}="${replaced}`
            return returnValue
          }
        )
      }

      // Rewrite CSS asset paths. Example
      // url("../assets/fonts/alliance/alliance-no-1-regular.woff") ->
      // url("https://githubdocs.azureedge.net/github-images/enterprise/2.20/assets/fonts/alliance/alliance-no-1-regular.woff")
      if (resource.isCss()) {
        newBody = text.replace(
          /(?<attribute>url)\("(?:\.\.\/)*(?<basepath>assets\/fonts|assets\/images)/g,
          (match, attribute, basepath) => {
            const replaced = path.join(
              `${remoteImageStoreBaseURL}/enterprise`,
              this.version,
              basepath
            )
            const returnValue = `${attribute}("${replaced}`
            return returnValue
          }
        )
      }

      const filePath = path.join(this.tempDirectory, resource.getFilename())

      await fs.promises.writeFile(filePath, newBody, 'binary')
    })
  }
}

async function main() {
  if (!pathToArchivalRepo) {
    console.log(`Please specify a path to a local checkout of ${archivalRepoUrl}`)
    const scriptPath = path.relative(process.cwd(), __filename)
    console.log(`Example: ${scriptPath} -p ../${archivalRepoName}`)
    process.exit()
  }

  if (dryRun) {
    console.log(
      'This is a dry run! Creating HTML for redirects and scraping the first 10 pages only.\n'
    )
  }

  // Build the production assets, to simulate a production deployment
  console.log('Running `npm run build` for production assets')
  execSync('npm run build', { stdio: 'inherit' })
  console.log('Finish building production assets')

  const fullPathToArchivalRepo = path.join(process.cwd(), pathToArchivalRepo)

  if (!fs.existsSync(fullPathToArchivalRepo)) {
    console.log(`archival repo path does not exist: ${fullPathToArchivalRepo}`)
    process.exit()
  }

  console.log(`Enterprise version to archive: ${version}`)
  const pageMap = await loadPageMap()
  const permalinksPerVersion = Object.keys(pageMap).filter((key) =>
    key.includes(`/enterprise-server@${version}`)
  )

  const urls = dryRun
    ? permalinksPerVersion.slice(0, 10).map((href) => `${host}${href}`)
    : permalinksPerVersion.map((href) => `${host}${href}`)

  console.log(`found ${urls.length} pages for version ${version}`)

  if (dryRun) {
    console.log(`\nscraping html for these pages only:\n${urls.join('\n')}\n`)
  }

  const finalDirectory = path.join(fullPathToArchivalRepo, version)
  const tempDirectory = path.join(__dirname, '../website-scraper-temp')

  // remove temp directory
  rimraf.sync(tempDirectory)

  // remove and recreate empty target directory
  rimraf.sync(finalDirectory)
  fs.mkdirSync(finalDirectory, { recursive: true })

  const scraperOptions = {
    urls,
    urlFilter: (url) => {
      // Do not download assets from other hosts like S3 or octodex.github.com
      // (this will keep them as remote references in the downloaded pages)
      return url.startsWith(`http://localhost:${port}/`)
    },
    directory: tempDirectory,
    filenameGenerator: 'bySiteStructure',
    requestConcurrency: 6,
    plugins: [new RewriteAssetPathsPlugin(version, tempDirectory)],
  }

  createApp().listen(port, async () => {
    console.log(`started server on ${host}`)

    await scrape(scraperOptions).catch((err) => {
      console.error('scraping error')
      console.error(err)
    })

    fs.renameSync(path.join(tempDirectory, `/localhost_${port}`), path.join(finalDirectory))
    rimraf.sync(tempDirectory)

    console.log(
      `\n\ndone scraping! added files to ${path.relative(process.cwd(), finalDirectory)}\n`
    )

    // create redirect html files to preserve frontmatter redirects
    await createRedirectsFile(permalinksPerVersion, pageMap, finalDirectory)

    console.log(`next step: deprecate ${version} in lib/enterprise-server-releases.js`)

    process.exit()
  })
}

async function createRedirectsFile(permalinks, pageMap, finalDirectory) {
  const pagesPerVersion = permalinks.map((permalink) => pageMap[permalink])
  const redirects = await loadRedirects(pagesPerVersion, pageMap)

  const redirectsPerVersion = {}

  Object.entries(redirects).forEach(([oldPath, newPath]) => {
    // remove any liquid variables that sneak in
    oldPath = oldPath.replace('/{{ page.version }}', '').replace('/{{ currentVersion }}', '')
    // ignore any old paths that are not in this version
    if (
      !(
        oldPath.includes(`/enterprise-server@${version}`) ||
        oldPath.includes(`/enterprise/${version}`)
      )
    )
      return

    redirectsPerVersion[oldPath] = newPath
  })

  fs.writeFileSync(
    path.posix.join(finalDirectory, 'redirects.json'),
    JSON.stringify(redirectsPerVersion, null, 2)
  )
}