#!/usr/bin/env bash
#
# Retrieves MX and A records for 'Alexa Top 1 Million' hosts
# and prints them as pretty formatted JSON objects to stdout.
#
# *Optional* parallelism support with GNU Parallel (recommended):
#     $ sudo apt-get install parallel
#
# Authors: Aaron Zauner <azet@azet.org>
# License: CC0 1.0 (https://creativecommons.org/publicdomain/zero/1.0)
#
set -eo pipefail

readonly top1m_s3l="https://s3.amazonaws.com/alexa-static/top-1m.csv.zip"
readonly top1m_zip=${top1m_s3l##*/}
readonly top1m_csv=${top1m_zip%.*}

function hosts() {
  for line in $(<${top1m_csv}); do
    local host=${line/*,/}
    printf "%s " ${host}
  done
}

function get_mx() {
  # build JSON object from input hostname mapping
  # MX records to the corresponding A records of
  # a given host. including messy in-line formatting.
  local mx_records=($(dig +short +nosearch +keepopen \
                      +time=2 mx "${1}" | sed 's/.*\ //'))
  printf '{\n\t"%s": {\n\t\t"mx_records": [\n' "${1}"
  for mx in "${mx_records[@]}"; do
    local ip=($(getent ahostsv4 ${mx})) # in our case, v4 suffices.
    printf '\t\t\t"%s":   \t"%s",\n' "${mx}" "${ip}"
  done
  printf "\t\t]\n\t}\n}\n"
}

# main
[[ -e ${top1m_csv} ]] || {
  wget "${top1m_s3l}"  &> /dev/null
  unzip "${top1m_zip}" &> /dev/null
}

[[ ${1} == "get_mx" ]] && { 
  get_mx "${2}" && exit 0 
}

if [[ $(which parallel) ]]; then
  printf "<< parallel mode >>\n\n" >&2
  parallel --progress --colsep ',' "${0} get_mx {2}" :::: ${top1m_csv}
else
  printf "<< sequential mode (slow! install \`parallel\`.) >>\n\n" >&2
  for host in $(hosts); do
    get_mx "${host}"
  done
fi
trap '{ rm ${top1m_zip} ${top1m_csv};   \
        printf "\n\n<< finished run. >> \
        $(date --rfc-3339=ns)   \n" >&2 }' EXIT