#!/usr/bin/env ruby

require File.expand_path('../lib/archive_bot', __FILE__)

include ArchiveBot::Redis

r = make_redis

trim_threshold = 500

def flush_logs(r, log_key, job_key, last_trimmed, min_last)
  r.zrangebyscore(log_key, last_trimmed, min_last).each do |l|
    $stdout.puts "#{job_key} #{l}"
  end

  $stdout.flush

  r.zremrangebyscore(log_key, last_trimmed, min_last)
  r.hset(job_key, 'last_trimmed_log_entry', min_last)
end

$stdin.each_line do |line|
  job_key = line.chomp

  data = r.hmget(job_key,
                 'log_key',
                 'finished_at',
                 'last_analyzed_log_entry',
                 'last_broadcasted_log_entry',
                 'last_trimmed_log_entry')

  log_key = data.shift
  finished_at = data.shift

  # NB: Redis zset scores are technically floats, not ints, so might as well
  # play along
  data.map!(&:to_f)

  # Find the longest run of messages outstanding.
  min_last = [data[0], data[1]].min
  last_trimmed = data[2]

  # If this job is marked as finished, then we want to trim the whole log -- no
  # point in keep everything around.
  #
  # Otherwise, is the last time we trimmed greater than our threshold?  If so,
  # spool out the log to stdout and trim it from Redis.
  if finished_at
    flush_logs(r, log_key, job_key, last_trimmed, '+inf')
  else
    if min_last - last_trimmed >= trim_threshold
      flush_logs(r, log_key, job_key, last_trimmed, min_last)
    end
  end
end