Annotation of /trunk/script/index_ora

#!/usr/bin/perl -w
#                              -*- Mode: Perl -*- 
# $Basename$
# $Revision: 1.10 $
# Author          : Ulrich Pfeifer
# Created On      : Mon Dec 31 13:57:11 2001
# Last Modified By: Ulrich Pfeifer
# Last Modified On: Fri Jan  4 15:59:20 2002
# Language        : CPerl
# 
# (C) Copyright 2001, Ulrich Pfeifer
# 

use 5.007;

use strict;
use Devel::Peek qw(Dump);

use File::Path;
use DB_File;
use Getopt::Long;
use Cwd;

BEGIN {require WAIT::Config;}
use WAIT::Database;
use WAIT::Parse::Ora;
use WAIT::Document::Ora;
use WAIT::InvertedIndex;


$DB_BTREE->{'cachesize'} = 200_000 ;

my %OPT = (
           database => 'oreilly_de_catalog',
           dir      => '/usr/local/apache/data',
           table    => 'ora',
          );

GetOptions(\%OPT,
           'database=s',
           'dir=s',
           'table=s',
          ) || die "Usage: ...\n";

my @localtime = localtime;
$localtime[5] += 1900;
$localtime[4]++;
my $jobid = sprintf "%04s-%02s-%02s_%02s:%02s_%d", @localtime[5,4,3,2,1], $$;
my $db = WAIT::Database->create(name      => "$OPT{database}-$jobid",
                                directory => $OPT{dir})
    or die "Could not create database $OPT{database}: $@\n";

my $layout = new WAIT::Parse::Ora;

use lib "/usr/local/apache/lib";
use oreilly_de_catalog::wait_handler;

my $stem  = ['OR_tr_20020124', 'OR_lc_20020124', 'split2', 'stop', 'Stem'];
my $text  = [{
              'prefix'    => ['OR_tr_20020124', 'OR_lc_20020124'],
              'intervall' => ['OR_tr_20020124', 'OR_lc_20020124'],
             },
             'OR_tr_20020124', 'OR_lc_20020124', 'split2', 'stop'];
my $wplus = ['split2', 'OR_lc_20020124', 'OR_mixedonly_20020221'];
my $sound = ['OR_tr_20020124', 'OR_lc_20020124', 'split2', 'Soundex'];
my $trigr = ['OR_lc_20020124', 'OR_trigrams_20020125'];
# split6 is better than split13 or split10: it allows them to enter
# shorter sequences when searching.
my $isbn  = ['split6', 'OR_isbn_20020127'];

my $cwd = cwd;

my $droot = shift or die "Usage: $0 <options> Document-Rootdirectories";

my %D;
my $access = tie %D, 'WAIT::Document::Ora', $droot,
  or die "Couldn't tie to dir $droot: $!\n";

my $tb = $db->create_table(name     => $OPT{table},
                           attr     => ['author', 'isbn', 'title',
                                        'headline', 'docid'],
                           layout   => $layout,
                           access   => $access,
                           invindex =>
                           [
                            'title'  => $text,
                            'title'  => $wplus,
                            # 'title'  => $stem,
                            'aboutauthor'  => $text,
                            'aboutauthor'  => $wplus,
                            # 'aboutauthor'  => $stem,
                            'desc'   => $text,
                            'desc'   => $wplus,
                            'abstract' => $text,
                            'abstract' => $wplus,
                            'author' => $text,
                            # 'author' => $sound,
                            'colophon' => $text,
                            'colophon' => $wplus,
                            'isbn'   => $isbn,
                           ]
                          );
die "Couldn't create table $OPT{table}: $@\n" unless $tb;

my ($did, $value);
binmode STDOUT, ":utf8";
while (($did, $value) = each %D) {
  my $record   = $layout->split($value);
  my $headline = $record->{title};
  $headline =~ s/\s+/ /sg;
  printf "%15s %s\n", $record->{isbn}, substr($headline,0,60);
  $tb->insert('docid'  => $did,
              headline => $headline,
              %{$record});
}
$tb->set(top=>1);

my $tritb = $db->create_table(
                              name => "$OPT{table}_fallback",
                              attr => [qw(docid headline)], # name
                                                            # "headline"
                                                            # only for
                                                            # sman
                              invindex => [ headline => $trigr ],
                             );
my %dict;
for my $f ($tb->fields) {
  my(@idx) = @{$tb->table->{inverted}{$f} || []};
  for my $idx (@idx) {
    my $name = $idx->name;
    next if $name =~ /(_|\b)(mixedonly|Stem|Soundex)(\b|_)/;
              # irrelevant for alternatives
    my @keys = $idx->keys;
    @dict{@keys} = ();
  }
}
my @dictkeys = grep s/^p//, keys %dict;
my $maxdebug = 5;
for my $headline (@dictkeys) {
  if ($maxdebug && $headline =~ /[^\040-\177]/) {
    Dump $headline;
    $maxdebug--;
  }
  # printf "%s\n", substr($headline,0,60);
  $tritb->insert(docid => $headline, headline => $headline);
}
$tritb->set(top=>1);
$tritb->close;
$tb->close();
$db->close();

# Atomically relinking symlink: now we have a new database with a very
# long name like oreilly_de_catalog-2002-01-28_16:12_16467 and we want
# that database to be accessible with the oreilly_de_catalog name.

use File::Spec;
my $dir    = "$OPT{database}-$jobid";
my $slwant = File::Spec->catdir($OPT{dir}, $OPT{database});
my $sltmp  = File::Spec->catdir($OPT{dir}, "$OPT{database}-$$");
unlink $sltmp; # may fail
symlink $dir, $sltmp or die "Couldn't symlink $dir, $sltmp: $!";
rename $sltmp, $slwant or die "Couldn't rename $sltmp, $slwant: $!";
warn "$slwant now points to $dir";
system("chmod 777 $slwant/*/read")==0 or die;

$WAIT::Config = $WAIT::Config; # make perl -w happy


__END__
## ###################################################################
## pod
## ###################################################################

=head1 NAME

index_ora - generate an WAIT index for O'Reilly catalog

=head1 SYNOPSIS

B<index_ora>
[B<-database> I<dbname>]
[B<-dir> I<directory>]
[B<-table> I<table name>]
I<directory>

=head1 DESCRIPTION

=head1 OPTIONS

=over 5

=item B<-database> I<dbname>

Specify database name. Default is F<DB>.

=item B<-dir> I<directory>

Alternate directory where databases are located. Default is the
directory specified during configuration of WAIT.

=item B<-table> I<table name>

Specify an alternate table name. Default is C<ora>.

=head1 AUTHOR

Ulrich Pfeifer E<lt>F<pfeifer@wait.de>E<gt>

1	ulpfr	55	#!/usr/bin/perl -w
2			# -- Mode: Perl --
3			# $Basename$
4	laperla	73	# $Revision: 1.10 $
5	ulpfr	55	# Author : Ulrich Pfeifer
6			# Created On : Mon Dec 31 13:57:11 2001
7			# Last Modified By: Ulrich Pfeifer
8	ulpfr	62	# Last Modified On: Fri Jan 4 15:59:20 2002
9	ulpfr	55	# Language : CPerl
10			#
11	laperla	73	# (C) Copyright 2001, Ulrich Pfeifer
12	ulpfr	55	#
13
14	laperla	69	use 5.007;
15
16	ulpfr	55	use strict;
17	laperla	71	use Devel::Peek qw(Dump);
18	laperla	69
19	ulpfr	55	use File::Path;
20			use DB_File;
21			use Getopt::Long;
22			use Cwd;
23
24	laperla	69	BEGIN {require WAIT::Config;}
25			use WAIT::Database;
26			use WAIT::Parse::Ora;
27			use WAIT::Document::Ora;
28			use WAIT::InvertedIndex;
29	ulpfr	55
30
31			$DB_BTREE->{'cachesize'} = 200_000 ;
32
33	laperla	67	my %OPT = (
34	laperla	73	database => 'oreilly_de_catalog',
35			dir => '/usr/local/apache/data',
36	ulpfr	55	table => 'ora',
37			);
38
39			GetOptions(\%OPT,
40			'database=s',
41			'dir=s',
42			'table=s',
43			) \|\| die "Usage: ...\n";
44
45	laperla	68	my @localtime = localtime;
46			$localtime[5] += 1900;
47			$localtime[4]++;
48			my $jobid = sprintf "%04s-%02s-%02s_%02s:%02s_%d", @localtime[5,4,3,2,1], $$;
49			my $db = WAIT::Database->create(name => "$OPT{database}-$jobid",
50			directory => $OPT{dir})
51			or die "Could not create database $OPT{database}: $@\n";
52	ulpfr	55
53			my $layout = new WAIT::Parse::Ora;
54
55	laperla	69	use lib "/usr/local/apache/lib";
56			use oreilly_de_catalog::wait_handler;
57
58			my $stem = ['OR_tr_20020124', 'OR_lc_20020124', 'split2', 'stop', 'Stem'];
59	ulpfr	55	my $text = [{
60	laperla	69	'prefix' => ['OR_tr_20020124', 'OR_lc_20020124'],
61			'intervall' => ['OR_tr_20020124', 'OR_lc_20020124'],
62	ulpfr	55	},
63	laperla	69	'OR_tr_20020124', 'OR_lc_20020124', 'split2', 'stop'];
64	laperla	73	my $wplus = ['split2', 'OR_lc_20020124', 'OR_mixedonly_20020221'];
65	laperla	71	my $sound = ['OR_tr_20020124', 'OR_lc_20020124', 'split2', 'Soundex'];
66			my $trigr = ['OR_lc_20020124', 'OR_trigrams_20020125'];
67	laperla	73	# split6 is better than split13 or split10: it allows them to enter
68			# shorter sequences when searching.
69	laperla	72	my $isbn = ['split6', 'OR_isbn_20020127'];
70	ulpfr	55
71			my $cwd = cwd;
72
73	laperla	73	my $droot = shift or die "Usage: $0 <options> Document-Rootdirectories";
74
75	ulpfr	55	my %D;
76	laperla	73	my $access = tie %D, 'WAIT::Document::Ora', $droot,
77			or die "Couldn't tie to dir $droot: $!\n";
78	ulpfr	55
79			my $tb = $db->create_table(name => $OPT{table},
80	ulpfr	62	attr => ['author', 'isbn', 'title',
81	ulpfr	55	'headline', 'docid'],
82			layout => $layout,
83			access => $access,
84			invindex =>
85			[
86	laperla	72	'title' => $text,
87	laperla	73	'title' => $wplus,
88	laperla	72	# 'title' => $stem,
89			'aboutauthor' => $text,
90	laperla	73	'aboutauthor' => $wplus,
91	laperla	72	# 'aboutauthor' => $stem,
92			'desc' => $text,
93	laperla	73	'desc' => $wplus,
94	laperla	72	'abstract' => $text,
95	laperla	73	'abstract' => $wplus,
96	ulpfr	55	'author' => $text,
97	laperla	72	# 'author' => $sound,
98	laperla	65	'colophon' => $text,
99	laperla	73	'colophon' => $wplus,
100	laperla	72	'isbn' => $isbn,
101	ulpfr	55	]
102			);
103			die "Couldn't create table $OPT{table}: $@\n" unless $tb;
104
105			my ($did, $value);
106	laperla	69	binmode STDOUT, ":utf8";
107	ulpfr	55	while (($did, $value) = each %D) {
108			my $record = $layout->split($value);
109			my $headline = $record->{title};
110			$headline =~ s/\s+/ /sg;
111	ulpfr	62	printf "%15s %s\n", $record->{isbn}, substr($headline,0,60);
112	ulpfr	55	$tb->insert('docid' => $did,
113			headline => $headline,
114			%{$record});
115			}
116			$tb->set(top=>1);
117	laperla	71
118			my $tritb = $db->create_table(
119			name => "$OPT{table}_fallback",
120	laperla	72	attr => [qw(docid headline)], # name
121			# "headline"
122			# only for
123			# sman
124	laperla	71	invindex => [ headline => $trigr ],
125			);
126			my %dict;
127			for my $f ($tb->fields) {
128			my(@idx) = @{$tb->table->{inverted}{$f} \|\| []};
129			for my $idx (@idx) {
130	laperla	72	my $name = $idx->name;
131	laperla	73	next if $name =~ /(_\|\b)(mixedonly\|Stem\|Soundex)(\b\|_)/;
132			# irrelevant for alternatives
133	laperla	71	my @keys = $idx->keys;
134			@dict{@keys} = ();
135			}
136			}
137			my @dictkeys = grep s/^p//, keys %dict;
138			my $maxdebug = 5;
139			for my $headline (@dictkeys) {
140			if ($maxdebug && $headline =~ /[^\040-\177]/) {
141			Dump $headline;
142			$maxdebug--;
143			}
144	laperla	72	# printf "%s\n", substr($headline,0,60);
145	laperla	71	$tritb->insert(docid => $headline, headline => $headline);
146			}
147			$tritb->set(top=>1);
148			$tritb->close;
149	ulpfr	55	$tb->close();
150			$db->close();
151
152	laperla	73	# Atomically relinking symlink: now we have a new database with a very
153			# long name like oreilly_de_catalog-2002-01-28_16:12_16467 and we want
154			# that database to be accessible with the oreilly_de_catalog name.
155	laperla	68
156			use File::Spec;
157	laperla	73	my $dir = "$OPT{database}-$jobid";
158			my $slwant = File::Spec->catdir($OPT{dir}, $OPT{database});
159			my $sltmp = File::Spec->catdir($OPT{dir}, "$OPT{database}-$$");
160			unlink $sltmp; # may fail
161			symlink $dir, $sltmp or die "Couldn't symlink $dir, $sltmp: $!";
162			rename $sltmp, $slwant or die "Couldn't rename $sltmp, $slwant: $!";
163			warn "$slwant now points to $dir";
164			system("chmod 777 $slwant/*/read")==0 or die;
165	laperla	68
166	ulpfr	55	$WAIT::Config = $WAIT::Config; # make perl -w happy
167
168
169			__END__
170			## ###################################################################
171			## pod
172			## ###################################################################
173
174			=head1 NAME
175
176			index_ora - generate an WAIT index for O'Reilly catalog
177
178			=head1 SYNOPSIS
179
180			B<index_ora>
181			[B<-database> I<dbname>]
182			[B<-dir> I<directory>]
183			[B<-table> I<table name>]
184			I<directory>
185
186			=head1 DESCRIPTION
187
188			=head1 OPTIONS
189
190			=over 5
191
192			=item B<-database> I<dbname>
193
194			Specify database name. Default is F<DB>.
195
196			=item B<-dir> I<directory>
197
198			Alternate directory where databases are located. Default is the
199			directory specified during configuration of WAIT.
200
201			=item B<-table> I<table name>
202
203			Specify an alternate table name. Default is C<ora>.
204
205			=head1 AUTHOR
206
207			Ulrich Pfeifer E<lt>F<pfeifer@wait.de>E<gt>
208