Re: Search Algorithm

[maverick]

If you're going to to be doing a lot of searches through a lot of files and the files that you're searching through don't change very often, then it may be worth your time to build an index.

The basic idea is this. Using a dbm, make the key a word out of one of the files, and the value a list of files that contain that word. Thus you trade time in building the index for the speed of hash lookups.

You'll end up with two programs. One that builds the index and one that does the search.

This one will build the index.

#!/usr/bin/perl

use strict;

# Berkeley DBMs are my fav.
use DB_File;

my %Index;

# remove the old index and start fresh
unlink("/home/maverick/tmp/index_dbm");
tie (%Index,'DB_File',"/home/maverick/tmp/index_dbm",O_RDWR|O_CREAT,06
+40,$DB_BTREE) || die "Tie Failed: $!";

foreach my $file (glob("/home/maverick/tmp/*.txt")) {

        open(F,$file) || die "Can't open $file: $!";

        # slirp up the file and make a list of words
        my @words = map { split(/\W+/,$_) } <F>;

        # add this file to the list of matches for this word
        my %uniq;
        foreach (@words) {
                if (!defined($uniq{$_})) {
                        # we've not seen this word before, so we add i
+t.
                        # I'm also assuming that ~ is safe to use as a
+ seperator.
                        if (!defined($Index{$_})) {
                                # it's the first additon of this word,
+ so I don't need to prepend a '~'
                                $Index{$_} = $file;
                        }
                        else {
                                $Index{$_} .= "~$file";
                        }
                        $uniq{$_} = 1;
                }
        }
    close(F);
}
untie %Index;

This one searchs using it.

#!/usr/bin/perl

use strict;

use DB_File;

my %Index;
tie (%Index,'DB_File',"/home/maverick/tmp/index_dbm",O_RDWR,0640,$DB_B
+TREE) || die "Tie Failed: $!";

print ">";
while(<>) {
    # chop off the newline
        $_ =~ s/[\r\n]//go;
        if (defined($Index{$_})) {
                print "$_ found in:\n";
        # replace all the ~ with \n (without modifying the index)
                print join("\n",split(/~/,$Index{$_})),"\n";
        }
        else {
                print "Not Found\n";
        }
        print ">";
}
untie(%Index);

file1.txt contains:

here's a file that contains a bunch of random
keywords on many different lines that we can use for the sake of examp
+le.

file2.txt contains:

here's another file that contains even more random text
for the sake of example.  I hope this helps solve the problem
presented by tenfourty.

and if you run the index maker and then search:

darkstar:~/tmp>./mkindex.pl 
darkstar:~/tmp>./search.pl 
>tenfourty
tenfourty found in:
/home/maverick/tmp/file2.txt
>maverick
Not Found
>example
example found in:
/home/maverick/tmp/file1.txt
/home/maverick/tmp/file2.txt
>text
text found in:
/home/maverick/tmp/file2.txt
>

Hope this helps :)

/\/\averick