xmpl - the identity API


The identity API in xmpl.

Considering that several instances of xmpl might be chained together, it’s useful to have an API to figure out who is who. This is where the endpoint /identity comes from:

get '/identity' => sub ($c) { $c->render(json => {id => identity()}) };
put '/identity' => sub ($c) {
    identity($c->req->text // '');

The API provides both a facility to get the identity, as well as one to set it to a string, This input string will be trimmed and kept if it contains at least one non-space character.

The actual workhorse is the identity() function, which works somehow like a method in disguise in that it exposes a double interface, i.e. a getter and a setter in one single function. The value is kept in a state variable:

sub identity ($new_value = undef) {
   state $id = default_identity();
   $id = trim_or_default($new_value, \&default_identity)
       if defined $new_value;
   return $id;

In pure getter/setter spirit, the value of state variable $id is set to a $new_value only if this is defined. Moreover, the function also makes sure to remove leading and trailing spaces, as well as ensuring that the final result contains some non-spacing character (i.e. it’s not empty).

The default_identity can be set externally through environment variable IDENTITY:

sub default_identity {
   state $id = trim_or_default($ENV{IDENTITY}, \&random_identity);

Again, the code makes sure to fallback to a default if there’s no meaningful IDENTITY set in the environment, by means of random_identity:

sub random_identity {
   return trim_or_default(
      eval { require Sys::Hostname; Sys::Hostname::hostname() },
      sub { sha1_sum(Time::HiRes::time() . '-' . rand()) },

As a first attempt, this fallback identity is not random at all, attempting to take the hostname. As a further fallback, a random code is generated from the time and a random draw. This should eventually do.

The trim_or_default function is a masterpiece of overengineering:

sub trim_or_default ($string, $default) {
   ($string //= '') =~ s{\A\s+|\s+\z}{}gmxs;
   return $string if length $string;
   $default = $default->() if ref($default) eq 'CODE';
   return trim_or_default($default, Time::HiRes::time());

The input string might be undef but it’s OK. The trimming happens, then if the string is not empty it is returned. Otherwise…

  • … if the default value is a code reference, it is executed. This allows delaying the execution of the associated code only at the moment when it’s strictly necessary;
  • … afterwards, the outcome might use some trimming itself. But wait! trim_or_default does exactly this… hence the recursive call!

I guess it’s enough boredom for today, have a good one!

Comments? Octodon, , GitHub, Reddit, or drop me a line!