Gmail mbox parser to find what is using your space

gmail_stats.rb

# This program parses an mbox file from Google Takeout to identify the senders and subjects lines that use the most space in a Gmail account
# Disclaimer: sizes or message detection may not be accurate, takes a few minutes for a 30GB file
# Usage: ruby gmail_stats.rb filename.mbox


# keep track of totals
$senders = {}
$subjects = {}
$labels = {}

def trackStatistic(statistic, key, size)
  statistic[key] ||= {count: 0, size: 0}
  statistic[key][:count] += 1
  statistic[key][:size] += size
end

def displayStatistic(variable, name, count)
  puts "=== Top #{count} #{name} ==="
  variable.sort_by{|k,v| -v[:size] }.first(20).each do |key, data|
    megabytes = data[:size]/(1024.0*1024.0)
    puts "#{key}    Count: #{data[:count]}  Size: #{megabytes.round(2)}mb"
  end
  puts ""
end

def parseMessage(message)
  message_meta = {}
  message.each do |line|
    # parse the headers
    if line.index(": ")
      line_parts = line.split(": ")
      key = line_parts[0]
      value = line_parts[1..-1].join(": ").strip
      message_meta[key] = value
    end

    # stop when we reach the content
    if line.index("mimepart")
      break
    end
  end

  result = {
    from: message_meta["From"],
    subject: message_meta["Subject"],
    size: message.join.bytesize
  }


  # Statistics
  trackStatistic($senders, result[:from], result[:size])
  trackStatistic($subjects, result[:subject], result[:size])

  (message_meta["X-Gmail-Labels"] || "").split(",").each do |label|
    trackStatistic($labels, label, result[:size])
  end


  return result
end

def parseFile(filepath)
  message = []
  File.open(filepath).each do |line|
    if line.index("From ")==0 && line.index("@") && (message==[] || message.last=="")

      if message.any?{|l| l!="" }
        #message to process
        parseMessage(message)
      end

      #start of a new message
      message = []
    end

    message << line.strip
  end

  if message.any?{|l| l!="" }
    # parse the last message
    parseMessage(message)
  end
end


parseFile(ARGV[0])
displayStatistic($senders, "Senders", 20)
displayStatistic($subjects, "Subjects", 20)
displayStatistic($labels, "Labels", 5)