ndk_info.rb

require 'httparty'
require 'nokogiri'
require 'zip_tricks'
require 'uri'

NDK = Struct.new(:version, :url, :sha)

def get_ndk_info
  ndks = []

  # Current LTS version
  current_url = 'https://github.com/android/ndk/wiki'

  response = HTTParty.get(current_url)
  document = Nokogiri.HTML(response.body)

  h3_current_lts = document.xpath('//*[@id="wiki-body"]/div/h3[text() = "Current LTS Release"]')
  p_current_lts = h3_current_lts.xpath('following-sibling::p')
  table_current_lts = h3_current_lts.xpath('following-sibling::table')

  _, col1, _, col3 = table_current_lts.xpath('tbody/tr[td[1]="Linux"]/td')

  current_lts_version = p_current_lts.first.text.split.first
  current_lts_url = col1.xpath('a/@href').first.value
  current_lts_sha = col3.text

  ndks << NDK.new(current_lts_version, current_lts_url, current_lts_sha)

  # Older versions
  older_url = 'https://github.com/android/ndk/wiki/Unsupported-Downloads'

  response = HTTParty.get(older_url)
  document = Nokogiri.HTML(response.body)

  h3_versions = document.xpath('//*[@id="wiki-body"]/div/h3')

  h3_versions.each do |h3_version|
    version = h3_version.text

    table = h3_version.xpath('following-sibling::table').first

    if table.nil?
      # The r10e and r9d are listed in an unordered list element, without a sha,
      # so they have been manually calculated and hardcoded here

      missing_shas = {
        'r10e' => "f692681b007071103277f6edc6f91cb5c5494a32",
        'r9d' => "6d0cdb0b06eeafaa89890d05627aee89122b143f",
      }

      list = h3_version.xpath('following-sibling::ul')
      item = list.xpath('li[starts-with(.,"Linux")]')
      url = item.xpath('a/@href').first.value

      if version == 'r9d'
        # We don't want to support this version, because it's not packaged as a
        # .zip file, which means we can't peak into it's contents as easily
        next
      end

      ndks << NDK.new(version, url, missing_shas[version])
    else
      # Grab the url and sha from the table row that has Linux as platform
      _, col1, _, col3 = table.xpath('tbody/tr[td[1]="Linux"]/td')

      url = col1.xpath('a/@href').first.value
      sha = col3.text

      ndks << NDK.new(version, url, sha)
    end
  end

  ndks
end

def ndk_info
  @ndk_info ||= get_ndk_info
end

# Trick described in
# https://rhardih.io/2021/04/listing-the-contents-of-a-remote-zip-archive-without-downloading-the-entire-file/
def get_file_names(zip_url)
  # ZIP file format: https://en.wikipedia.org/wiki/ZIP_(file_format)

  # 1. Request just enough bytes from the end of the file, to get the End of
  # central directory record (EOCD)
  response = HTTParty.get(zip_url, { headers: { 'Range' => "bytes=-100" } })

  # 2. Extract the central directory byte offset, size and comment length
  ss = StringScanner.new(response.body)
  ss.scan_until(/\x50\x4b\x05\x06/) # end of central directory signature

  ss.pos += 12 - 4 # pos is index of first byte after a match
  cd_size, cd_offset, comment_length = ss.peek(10).unpack("L2S")

  throw "Not enough bytes requested for EOCD" if comment_length > 80

  # 3. Use the offset and size to request just the bytes that contain the
  # Central directory file header
  response = HTTParty.get(zip_url, {
    headers: {
      'Range' => "bytes=#{cd_offset}-#{cd_offset + cd_size}"
    }
  })

  # 4. Go through each entry in the file and collect the filenames
  file_names = []

  ss = StringScanner.new(response.body)

  until ss.scan_until(/\x50\x4b\x01\x02/).nil? # central directory signature
    ss.pos += 28 - 4
    file_name_length =  ss.peek(2).unpack("S").first

    ss.pos += 18
    file_name = ss.peek(file_name_length)

    file_names << file_name
  end

  file_names
end