Why splitting if you can point?


Operator overloading in Raku is cool.

I must have done some very good things in a past life, because I received the blessing of getting high quality hints from gfldex from time to time. Keep up with the good work, Flavio!

This time I got a nice interface design suggestion about PWC129 - Add Linked Lists, where my tests were structured like this:

my @tests =
      '1 -> 2 -> 3', # input #1, as a string
      '3 -> 2 -> 1', # input #2, as a string
      '4 -> 4 -> 4', # expected output, as a string

The first two strings in each string have first to be split to only get the values in the list, then the resulting list used to initialize a LinkedList:

my $l1 = LinkedList.create($tl1.split: /\s* '->' \s*/);

gfldex challeges the approach: why the split when you can point? Something like this:

my @tests =
      1 → 2 → 3,
      1 → 2 → 3,
      '4 -> 4 -> 4',

my ($l1, $l2, $sumstr) = @$test;
is ($l1 + $l2).Str, $sumstr, "sum leading to $sumstr";

The result is still expressed as a string ($sumstr), because we want to use the stringification to check against the result!

The good thing is that I also got a hint about the implementation, without getting all of it:

multi sub infix:«→»(List \l, \r) {
    |l, |r

multi sub infix:«→»(\l, \r) {
    l, r;

Of course we’re not dealing with the real LinkedList in the examples above, but the suggestion is easy to follow:

multi sub infix:«→» (*@ls) is assoc<list> {
   return LinkedList.create(@ls) unless @ls[*-1] ~~ LinkedList;
   return @ls.reverse.reduce: -> $t, $h { $t.insert($h); $t };

Well… maybe I’m still a bit attached to my Perl accent, so there is actually no need for multi here, and I’m doing some dispatching inside the sub instead of using the multi mechanism. In this case, though, my point is that I don’t know how to put a slurpy argument element that takes all but the last element 🙄

Here I’m addressing two different cases:

  • adding one or more elements to a pre-existing list;
  • creating a new list from scratch (i.e. a list of elements).

My first implementation was actually leveraging multi:

multi sub infix:«→» ($h, LinkedList $t) is assoc<right> {
   return $t;
multi sub infix:«→» ($h, $t) is assoc<right> {
   return LinkedList.create($h, $t);

but you know… I was thinking about using a whole list all at once and avoid calling the sub too many times.

There’s a slight thing that I’m not entirely happy about, that is we are adding new elements from the left. This might be a bit surprising… in general, I would expect this kind of operation to produce a new list, keeping the old one untouched.

Anyway, with this… tests might be compressed in a single line:

is ((1 → 2 → 3) + (3 → 2 → 1)).Str, '4 -> 4 -> 4', 'single test yay!';

Isn’t this cool?!?

Thanks gfldex for keeping the suggestions flow… and stay safe everyone!

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