http://www.perlmonks.org?node_id=11138449


in reply to Suffix-prefix matching done right

Without addressing parallelization nor performing any benchmarks, here's naive implementation. Hopefully, fuzzy matching implemented and optimized in C is fast. Maybe, re-writing this with String::Approx instead of re::engine::TRE (subroutine call instead of Perl regexp engine overhead) would be faster. Re-visiting fuzzy string matching was fun :). Solution below was more readable/clear before I tried to get to (perceived, no tests) optimizations like dropping blocks, reversing loop, + remembering that to use a reference to substr result is efficient (no idea if it holds in this case), etc., but here's FWIW.

use strict; use warnings; use feature 'say'; use constant ERR => 1; use re::engine::TRE max_cost => ERR, cost_ins => -1, # no insertions cost_del => -1, # no deletions ; my $suffix_source = 'abbaba'; my $prefix_source = 'babbaaaa'; my $max_len = 0; $prefix_source =~ /^${ \substr $suffix_source, -$_ }/ and $max_len = $_ and last for reverse ERR + 1 .. length $suffix_source; say substr( $suffix_source, -$max_len ), ' ', substr( $prefix_source, 0, $max_len ) if $max_len; __END__ baba babb

Update: crude benchmarks (~200 chars strings, ~10 errors allowed) reveal that, for this task, both modules I mentioned are hugely (some 10s of times) slower than "classic" Perl implementation by tybalt89 :)