PWC180 - First Unique Character

TL;DR

Here we are with TASK #1 from The Weekly Challenge #180. Enjoy!

The challenge

You are given a string, $s.

Write a script to find out the first unique character in the given string and print its index (0-based).

Example 1

Input: $s = "Perl Weekly Challenge"
Output: 0 as 'P' is the first unique character

Example 2

Input: $s = "Long Live Perl"
Output: 1 as 'o' is the first unique character

The questions

I guess that we have to decide how we get the string, taking care of whatever encoding stuff we can think of.

And, of course, that the character we mean whatever Unicode character we can think of, and more.

And, oh-by-the-way, that the whole string is going to fit nicely in memory, while still leaving us space to do our stuff.

The solution

A few days ago I read a tweet like this:

Q: what’s s double linked list?

A: a data structure only used in interviews

I had a good laugh and thought that, in my case, it was sort of true, in the sense that I could not remember using linked lists (or their grown-up sister, the doubly-linked list) anywhere else than university excercises in the nineties.

But, of course, the wheel goes round and round and here I find a challenge (mind you, not an interview question!) where a doubly-linked list fits perfectly. Never say never.

This said, here’s the plan:

  • we will have to iterate over all input characters. With the whole power of Unicode at our input builder’s disposal, it would not make sense to try and optimize for the case well I’ve already seen all of them at least twice, I guess I can just bail out. (This was, by the way, my original plan when I believed that ASCII letters were at play only).
  • we keep track of whatever we have already encountered in a hash, where the key is the character and the value can be one of:
    • the index in the string at which we first saw the character, OR
    • the undefined value (Nil or undef, depending on the language) if we meet the character beyond the first time.
  • we also keep track of any new character’s index in a shiny doubly-linked list, with the convention that we remove the slot for a character as soon as we meet it the second time.

At the end of the iteration, the doubly-linked list can be either empty (i.e. every character was repeated at least once) or not, so we just have to consider the first element and we’re done.

Let’s start with Raku first, where we go all-in with classes, using even a class within another class. I am moderately surprised to have encountered very few roadblocks while coding this.

#!/usr/bin/env raku
use v6;

class DoubleLinkedList { ... }
sub MAIN ($string = 'Perl Weekly Challenge') {
   my $order = DoubleLinkedList.new;
   my %dll-element-for;
   my $i = 0;
   for $string.comb -> $character {
      if %dll-element-for{$character}:exists {
         if defined %dll-element-for{$character} {
            $order.remove(%dll-element-for{$character});
            %dll-element-for{$character} = Nil;
         }
      }
      else {
         %dll-element-for{$character} = $order.push($i).tail;
      }
      ++$i;
   }

   die 'no result, sorry!' unless defined $order.head;
   put $order.head.value;
}

class DoubleLinkedList {
   class Element {
      has $.value;
      has $.pred is rw is built = Nil;
      has $.succ is rw is built = Nil;
   }

   has $.head is rw is built = Nil;
   has $.tail is rw is built = Nil;

   method push ($value) {
      my $element = Element.new(
         value => $value, pred => $.tail, succ => Nil);
      $.tail.succ = $element if defined $.tail;
      $.tail = $element;
      $.head //= $element;
      return self;
   }

   method remove ($element) {
      if (defined $element.pred) {
         $element.pred.succ = $element.succ;
      }
      else {
         $.head = $element.succ;
      }
      if (defined $element.succ) {
         $element.succ.pred = $element.pred;
      }
      else {
         $.tail = $element.pred;
      }
      return self;
   }
}

The Perl alternative is very minimalistic. I’m avoiding the Element class to just use an anonymous hash, at the end of the day there’s no real business logic in it.

#!/usr/bin/env perl
use v5.24;
use warnings;
use experimental 'signatures';
no warnings 'experimental::signatures';
use I18N::Langinfo qw(langinfo CODESET);
use Encode qw(decode);

{
   package DoubleLinkedList;
   sub new ($p) { return bless { head => undef, tail => undef }, $p }
   sub push ($self, $value) {
      my $e = { value => $value, pred => $self->{tail}, succ => undef };
      $self->{tail}{succ} = $e if defined $self->{tail};
      $self->{tail} = $e;
      $self->{head} //= $e;
      return $self;
   }
   sub remove ($self, $e) {
      if (defined $e->{pred}) { $e->{pred}{succ} = $e->{succ} }
      else                    { $self->{head} = $e->{succ}    }
      if (defined $e->{succ}) { $e->{succ}{pred} = $e->{pred} }
      else                    { $self->{tail} = $e->{pred}    }
      return $self;
   }
}

my $codeset = langinfo(CODESET);
my $string = decode($codeset, shift(@ARGV) // 'Perl Weekly Challenge');

my $order = DoubleLinkedList->new;
my %dll_element_for;
my $i = 0;
for my $character (split m{}mxs, $string) {
   if (exists $dll_element_for{$character}) {
      if (defined $dll_element_for{$character}) {
         $order->remove($dll_element_for{$character});
         $dll_element_for{$character} = undef;
      }
   }
   else {
      $dll_element_for{$character} = $order->push($i)->{tail};
   }
   ++$i;
}

die "no result, sorry!\n" unless defined $order->{head};
say $order->{head}{value};

I know that the Perl “object system” is probably too bare-bones, but it’s amazing in these small programs.

Stay safe and unique!


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