in reply to Challenge: Fast Common Substrings

Okay here is a benchmark as requested. I doubt it will satisfy everyone.

Some comments on the benchmark and results obtained.

  1. Randomly generated strings of 30 .. 50 lower case characters rarely if ever produce common substrings of 3 characters, never mind longer.

    To this end I used 'A'..'D', which produces a few at 3, 4, & 5 characters long, but never longer that I saw.

  2. Even adjacent pairings of 5000 strings takes a very long time to run, never mind a full combinatorial pairing.

    I used 500 strings instead. It's a command line parameter, if you've the time and inclination to run 5000, go ahead. I doubt it will make any huge difference to the outcome.

    theZip's (TZ) code is horrible to benchmark as it relies upon a global var. Ie. The sub does not return the result.

    I had a half hearted attempt to address this using a helper sub closing over the global, but the result are wildly inaccurate. Possibly? the fault of my adaption. Sorry thezip.

    However, it seems to be far and away the slowest algorithm anyway.

  3. eric256's (eric) code produces numbers that seem not to relate to the inputs as far a I can discern.

    Maybe I screwed up, but I don't think so. It's also the second slowest algorithm.

  4. Mine (Buk), blokhead's (BH1/BH2), moron's (MN), and Limbic~Region's (LR) code all produce the same counts (although Limbic~Region's has been seen to be one shy of the others on a few occasions).

    The level of concurrence between these disparate algorithms and implementations is taken to indicate that they are producing the correct results. I have not manually verified them.

Results (500 random strings of 'A'..'D', tested each against the next for common substrings 3-7 characters 

c:\test>608174-b -N=500 -LENGTH=3
            (warning: too few iterations for a reliable count)

            Rate     TZ   Eric    BH1     MN    BH2    L~R    Buk
TZ   9.06e-002/s     --   -98%   -98%   -99%   -99%   -99%  -100%
Eric      4.15/s  4481%     --   -19%   -41%   -63%   -75%   -85%
BH1       5.12/s  5548%    23%     --   -27%   -55%   -70%   -82%
MN        7.02/s  7642%    69%    37%     --   -38%   -58%   -75%
BH2       11.3/s 12354%   172%   121%    61%     --   -33%   -59%
L~R       16.9/s 18532%   307%   230%   141%    50%     --   -39%
Buk       27.7/s 30457%   567%   441%   295%   145%    64%     --

----

BH2 => 14
L~R => 14
Eric => 31
Buk => 14
BH1 => 14
TZ => 393262
MN => 14

c:\test>608174-b -N=500 -LENGTH=4
            (warning: too few iterations for a reliable count)

            (warning: too few iterations for a reliable count)

            Rate     TZ   Eric    BH1     MN    L~R    BH2    Buk
TZ   9.01e-002/s     --   -97%   -98%   -98%   -99%   -99%  -100%
Eric      2.87/s  3079%     --   -40%   -52%   -83%   -84%   -88%
BH1       4.78/s  5198%    67%     --   -20%   -71%   -73%   -80%
MN        6.00/s  6556%   109%    26%     --   -64%   -65%   -75%
L~R       16.5/s 18153%   474%   245%   174%     --    -5%   -33%
BH2       17.4/s 19167%   506%   264%   189%     6%     --   -29%
Buk       24.5/s 27060%   754%   413%   308%    49%    41%     --

----

BH2 => 5
L~R => 5
Eric => 34
Buk => 5
BH1 => 5
TZ => 396184
MN => 5

c:\test>608174-b -N=500 -LENGTH=5
            (warning: too few iterations for a reliable count)

            (warning: too few iterations for a reliable count)

            Rate     TZ   Eric    BH1     MN    L~R    BH2    Buk
TZ   9.10e-002/s     --   -96%   -98%   -98%   -99%  -100%  -100%
Eric      2.23/s  2352%     --   -54%   -63%   -86%   -90%   -91%
BH1       4.85/s  5227%   117%     --   -20%   -70%   -77%   -80%
MN        6.06/s  6551%   171%    25%     --   -63%   -72%   -75%
L~R       16.4/s 17953%   636%   239%   171%     --   -23%   -33%
BH2       21.3/s 23338%   856%   340%   252%    30%     --   -13%
Buk       24.5/s 26791%   997%   405%   304%    49%    15%     --

----

BH2 => 1
L~R => 1
Eric => 35
Buk => 1
BH1 => 1
TZ => 387220
MN => 1

c:\test>608174-b -N=500 -LENGTH=6
            (warning: too few iterations for a reliable count)

            (warning: too few iterations for a reliable count)

            Rate     TZ   Eric    BH1     MN    L~R    BH2    Buk
TZ   9.04e-002/s     --   -95%   -98%   -99%   -99%  -100%  -100%
Eric      1.73/s  1814%     --   -65%   -72%   -90%   -92%   -93%
BH1       4.92/s  5344%   184%     --   -20%   -70%   -79%   -81%
MN        6.13/s  6687%   255%    25%     --   -63%   -73%   -76%
L~R       16.7/s 18321%   862%   238%   171%     --   -27%   -34%
BH2       22.9/s 25257%  1225%   366%   274%    38%     --    -9%
Buk       25.2/s 27829%  1359%   413%   312%    52%    10%     --

----

BH2 => 0
L~R => 0
Eric => 35
Buk => 0
BH1 => 0
TZ => 391688
MN => 0

c:\test>608174-b -N=500 -LENGTH=7
            (warning: too few iterations for a reliable count)

            (warning: too few iterations for a reliable count)

            Rate     TZ   Eric    BH1     MN    L~R    BH2    Buk
TZ   9.08e-002/s     --   -94%   -98%   -99%   -99%  -100%  -100%
Eric      1.47/s  1521%     --   -70%   -76%   -91%   -94%   -94%
BH1       4.93/s  5327%   235%     --   -21%   -71%   -80%   -81%
MN        6.22/s  6748%   322%    26%     --   -63%   -74%   -76%
L~R       16.9/s 18554%  1051%   244%   172%     --   -30%   -33%
BH2       24.1/s 26469%  1539%   390%   288%    42%     --    -5%
Buk       25.4/s 27899%  1627%   416%   309%    50%     5%     --

----

BH2 => 0
L~R => 0
Eric => 32
Buk => 0
BH1 => 0
TZ => 393468
MN => 0

[download]

The benchmark code. CLI parameters are: -N=nn numbers of strings to generate; -LENGTH=mm: length of common substrings to look for; Interesting challenge Limbic~Region, thanks.

<Reveal this spoiler or all in this thread>
Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.
"Too many [] have been sedated by an oppressive environment of political correctness and risk aversion."

Replies are listed 'Best First'.
Re^2: Challenge: Fast Common Substrings
by BrowserUk (Patriarch) on Apr 05, 2007 at 15:34 UTC
[reply]
[d/l]

      /me benchmarks agian with his in there as a matter of pride.

      C:\Perl\test>substr_bench.pl -N=1000 -LENGTH=2
       Rate  BH2  BH1   MN Eric  L~R  Buk
BH2  3.37/s   -- -20% -53% -70% -78% -87%
BH1  4.21/s  25%   -- -42% -62% -72% -84%
MN   7.22/s 114%  71%   -- -35% -52% -73%
Eric 11.1/s 231% 164%  54%   -- -26% -58%
L~R  15.1/s 347% 258% 109%  35%   -- -44%
Buk  26.7/s 694% 534% 270% 140%  77%   --

----

BH2 => 14
L~R => 12
Eric => 14
Buk => 14
BH1 => 14
MN => 14

C:\Perl\test>substr_bench.pl -N=1000 -LENGTH=3
       Rate  BH1   MN  BH2 Eric  L~R  Buk
BH1  3.88/s   -- -23% -39% -46% -73% -80%
MN   5.05/s  30%   -- -21% -30% -65% -74%
BH2  6.40/s  65%  27%   -- -11% -56% -67%
Eric 7.21/s  86%  43%  13%   -- -50% -63%
L~R  14.4/s 272% 185% 125% 100%   -- -26%
Buk  19.4/s 400% 284% 203% 169%  35%   --

----

BH2 => 9
L~R => 9
Eric => 9
Buk => 9
BH1 => 9
MN => 9

C:\Perl\test>substr_bench.pl -N=1000 -LENGTH=4
       Rate  BH1   MN Eric  BH2  L~R  Buk
BH1  3.56/s   -- -19% -43% -57% -74% -79%
MN   4.39/s  23%   -- -30% -47% -68% -74%
Eric 6.22/s  75%  42%   -- -25% -54% -63%
BH2  8.35/s 135%  90%  34%   -- -38% -51%
L~R  13.5/s 281% 209% 118%  62%   -- -20%
Buk  16.9/s 375% 285% 171% 102%  25%   --

----

BH2 => 4
L~R => 4
Eric => 4
Buk => 4
BH1 => 4
MN => 4

C:\Perl\test>substr_bench.pl -N=1000 -LENGTH=5
       Rate  BH1   MN Eric  BH2  L~R  Buk
BH1  3.82/s   -- -14% -34% -59% -72% -78%
MN   4.44/s  16%   -- -24% -52% -67% -74%
Eric 5.82/s  52%  31%   -- -37% -57% -66%
BH2  9.27/s 143% 109%  59%   -- -32% -46%
L~R  13.6/s 255% 205% 133%  46%   -- -21%
Buk  17.1/s 348% 285% 194%  85%  26%   --

----

BH2 => 0
L~R => 0
Eric => 0
Buk => 0
BH1 => 0
MN => 0

C:\Perl\test>substr_bench.pl -N=1000 -LENGTH=6
            (warning: too few iterations for a reliable count)

       Rate  BH1   MN Eric  BH2  L~R  Buk
BH1  3.56/s   -- -21% -38% -64% -73% -80%
MN   4.50/s  27%   -- -21% -54% -66% -74%
Eric 5.73/s  61%  27%   -- -42% -57% -67%
BH2  9.84/s 177% 119%  72%   -- -25% -43%
L~R  13.2/s 271% 193% 130%  34%   -- -24%
Buk  17.4/s 389% 286% 203%  77%  32%   --

----

BH2 => 1
L~R => 1
Eric => 1
Buk => 1
BH1 => 1
MN => 1

      [download]

      ___________
      Eric Hodges
[reply]
[d/l]
Re^2: Challenge: Fast Common Substrings
by eric256 (Parson) on Apr 06, 2007 at 20:07 UTC

    Hey

    I misread the definition of substring to mean all substrings of that length or lower...quick easy fix and it starts placing decent. 

    sub common_sub {
    my ($s1,$s2,$len) = @_;

    my $len_s1 = length($s1);
    my $len_s2 = length($s2);

    my $match_s1 = {};
    my $match_s2 = {};
    
    for my $start (0..length($s1)-1) {
        #for my $l (1..$len) {
            next if $start+$len > $len_s1;
            $match_s1->{substr($s1, $start, $len)} ||= 1;
        #}
    }

    for my $start (0..length($s2)-1) {
        #for my $l (1..$len) {
            next if $start+$len > $len_s2;
            $match_s2->{substr($s2, $start, $len)} ||= 1;
        #}
    }
    $match_s1->{$_}++ for keys %$match_s2;
    return grep { $match_s1->{$_} == 2 } keys %$match_s1;
}

    [download]

    ___________
    Eric Hodges
[reply]
[d/l]

In Section Seekers of Perl Wisdom