Re: Longest String Matching

[AnomalousMonk]

I don't know if this will be as fast as pemungkah's approach, but the Perl regex engine builds a trie for alternations, so even quite large alternations (354K words in Moby file) are very fast. The big time sink is reading all the words and building the alternation string in the first place, but of course this could be done in a separate step and the resulting joined string saved to a file for later quick access. (Also note that the dictionary file I'm using has oddities such as 'words' like "d's" and "e's", and apparently every letter of the alphabet appears as a word.)

>perl -wMstrict -le
"my $wordlist = shift // die 'no wordlist filename given';
 open my $fh_wordlist, '<', $wordlist
   or die qq{opening '$wordlist': $!};
 my $words =
   join '|',
   sort { length($b) <=> length($a) }
   map  { chomp;  $_; }
   <$fh_wordlist>
   ;
 close $fh_wordlist or die qq{closing '$wordlist': $!};
 ;;
 my @lexicon = qw(
   no now know known unknown antidisestablishment
   husband wife husband's wife's xyzzy a
   );
 for my $word (@lexicon) {
   my ($longest_at_end) = $word =~ m{ \B ($words) \z }xms;
   print qq{'$word' -> '$longest_at_end'};
   }
"  ..\..\moby\mwords\354984si.ngl
'no' -> 'o'
'now' -> 'ow'
'know' -> 'now'
'known' -> 'own'
'unknown' -> 'known'
'antidisestablishment' -> 'disestablishment'
'husband' -> 'band'
'wife' -> 'ife'
'husband's' -> 'd's'
'wife's' -> 'e's'
'xyzzy' -> 'y'
Use of uninitialized value $longest_at_end in concatenation (.) or str
+ing at -e line 1.
'a' -> ''

[space_monk]

Someone ought to run Benchmark over the various proposals above and see which work out quickest (and under what circumstances)