Base32 encoding in Ruby (RFC 4648)

base32.rb

# Base32 (RFC 4648) implemented in Ruby
#
# Source: https://ptrchm.com/posts/base32-explained/
class Base32
  ALPHABET = "ABCDEFGHIJKLMNOPQRSTUVWXYZ234567".split("")
  PADDING_CHAR = "="

  BITS_PER_BYTE  = 8  # 1 byte = 8 bits
  BITS_PER_CHAR  = Math.log2(ALPHABET.length).round # 5 = 32 chars = 2^5 number of bits encoded into a single character in the ALPHABET
  BITS_PER_CHUNK = BITS_PER_CHAR.lcm(BITS_PER_BYTE) # 40 (least common mutliple of 5 and 8)

  CHARS_PER_CHUNK = BITS_PER_CHUNK / BITS_PER_CHAR  # 8
  CHUNK_LENGTH    = BITS_PER_CHUNK / BITS_PER_BYTE  # 5

  ENCODING_MASK = ALPHABET.length - 1 # 0x1f
  DECODING_MASK = 0xff

  def self.encode(str)
    str.bytes.each_slice(CHUNK_LENGTH).map do |chunk|
      bits_in_chunk = chunk.length * BITS_PER_BYTE
      number_of_characters = (bits_in_chunk * CHARS_PER_CHUNK / BITS_PER_CHUNK.to_f).ceil
      padding = bits_in_chunk < BITS_PER_CHUNK ? BITS_PER_CHAR - bits_in_chunk % BITS_PER_CHAR : 0

      buf = 0
      chunk.each do |byte|
        buf = (buf << BITS_PER_BYTE) + byte
      end
      buf <<= padding

      encoded = Array.new(CHARS_PER_CHUNK)
      j = number_of_characters - 1

      number_of_characters.times do |i|
        encoded[j] = ALPHABET[(buf >> BITS_PER_CHAR * i) & ENCODING_MASK]
        j -= 1
      end

      (CHARS_PER_CHUNK - number_of_characters).times do |i|
        encoded[number_of_characters + i] = PADDING_CHAR
      end

      encoded
    end.join
  end

  def self.decode(str)
    str = str.delete(PADDING_CHAR)
    bytes = str.each_char.map { |char| ALPHABET.index(char) }

    bytes.each_slice(CHARS_PER_CHUNK).map do |chunk|
      number_of_original_bytes = (chunk.length * BITS_PER_CHAR / BITS_PER_BYTE.to_f).floor
      padding = chunk.length < CHARS_PER_CHUNK ? BITS_PER_CHAR - (number_of_original_bytes * BITS_PER_BYTE) % BITS_PER_CHAR : 0

      buf = 0
      chunk.each do |byte|
        buf = (buf << BITS_PER_CHAR) + byte
      end
      buf >>= padding

      decoded = Array.new(number_of_original_bytes)
      j = decoded.length - 1

      number_of_original_bytes.times do |i|
        decoded[i] = ((buf >> BITS_PER_BYTE * j) & DECODING_MASK).chr
        j -= 1
      end

      decoded
    end.join
  end
end

great write-up. i am finally starting to understand the padding "magic". thanks. fyi: i tried a gem with base32 alphabets some time ago, see here https://github.com/rubycocos/blockchain/tree/master/base32-alphabets greetings from austria.