I was working on making a cache in bash and I found this link - But this guy is attempting to generate new bash functions by introspecting their source and generating a new one as a cache. This seems to me, way overkill. Here is my implementation and it works just fine:

cache() {
  local file
  file="/tmp/$(printf "%s" ${@})"
  if [[ -f "${file}" && $(($(date +%s) - $(date -r "$file" +%s))) -le 1800 ]]; then
    cat "${file}"
  else
    ${1} ${@:2} | tee "${file}"
  fi
}

mkurls() {
  local url="${1:?You must supply a URL}"
  local batchSize=${2:-100}
  local repoCount
  repoCount=$(curl -fsnL "$url?pagelen=0&page=1" | jq -re .size)
  local pages=$(((repoCount / batchSize) + 1))

  printf "$url?pagelen=${batchSize}&page=%s\n" $(seq 1 "$pages")
}

bbusers() {
  _users() {
    mkurls https://api.bitbucket.org/2.0/teams/twengg/members/ |
      xargs -n 30 -P 20 curl -fsnL | jq -ser '.[].values[] | "\(.display_name), \(.links.self.href)" ' | sort
  }
  cache _users $@
} 

and here are the results:

$ time bbusers &> /dev/null
#output

real    0m5.670s
user    0m0.184s
sys     0m0.099s

$ time bbusers &> /dev/null # seconds later

real    0m0.053s
user    0m0.010s
sys     0m0.017s

As you can see, it's literally 100x faster. If I delete the file behind the cache:

$ rm -fR /tmp/_users

$ time bbusers &> /dev/null

real    0m4.924s
user    0m0.170s
sys     0m0.082s

It goes right back to normal. So, how can this be improved and what am I missing that warrants such a wildly complicated approach as the other guy has?