1 |
# -*- Mode: Perl -*- |
# -*- Mode: Cperl -*- |
2 |
# Table.pm -- |
# Table.pm -- |
3 |
# ITIID : $ITI$ $Header $__Header$ |
# ITIID : $ITI$ $Header $__Header$ |
4 |
# Author : Ulrich Pfeifer |
# Author : Ulrich Pfeifer |
5 |
# Created On : Thu Aug 8 13:05:10 1996 |
# Created On : Thu Aug 8 13:05:10 1996 |
6 |
# Last Modified By: Ulrich Pfeifer |
# Last Modified By: Ulrich Pfeifer |
7 |
# Last Modified On: Sun Nov 22 18:44:37 1998 |
# Last Modified On: Sat Apr 27 17:20:31 2002 |
8 |
# Language : CPerl |
# Language : CPerl |
9 |
# Update Count : 51 |
# Update Count : 172 |
10 |
# Status : Unknown, Use with caution! |
# Status : Unknown, Use with caution! |
11 |
# |
# |
12 |
# Copyright (c) 1996-1997, Ulrich Pfeifer |
# Copyright (c) 1996-1997, Ulrich Pfeifer |
13 |
# |
# |
14 |
|
|
15 |
=head1 NAME |
=head1 NAME |
16 |
|
|
25 |
=cut |
=cut |
26 |
|
|
27 |
package WAIT::Table; |
package WAIT::Table; |
28 |
|
|
29 |
|
use WAIT::Table::Handle (); |
30 |
require WAIT::Parse::Base; |
require WAIT::Parse::Base; |
31 |
|
|
32 |
use strict; |
use strict; |
33 |
use Carp; |
use Carp; |
34 |
use DB_File; |
# use autouse Carp => qw( croak($) ); |
35 |
|
use BerkeleyDB; |
36 |
use Fcntl; |
use Fcntl; |
37 |
|
use LockFile::Simple (); |
38 |
|
|
39 |
my $USE_RECNO = 0; |
my $USE_RECNO = 0; |
40 |
|
|
41 |
=head2 Creating a Table. |
=head2 Creating a Table. |
42 |
|
|
43 |
The constructor WAIT::Table-<gt>new is normally called via the |
The constructor WAIT::Table-E<gt>new is normally called via the |
44 |
create_table method of a database handle. This is not enforced, but |
create_table method of a database handle. This is not enforced, but |
45 |
creating a table doesn not make any sense unless the table is |
creating a table does not make any sense unless the table is |
46 |
registered by the database because the latter implements persistence |
registered by the database because the latter implements persistence |
47 |
of the meta data. Registering is done automatically by letting the |
of the meta data. Registering is done automatically by letting the |
48 |
database handle create a table. |
database handle the creation of a table. |
49 |
|
|
50 |
my $db = create WAIT::Database name => 'sample'; |
my $db = WAIT::Database->create(name => 'sample'); |
51 |
my $tb = $db->create_table (name => 'test', |
my $tb = $db->create_table(name => 'test', |
52 |
attr => ['docid', 'headline'], |
access => $access, |
53 |
layout => $layout, |
layout => $layout, |
54 |
access => $access, |
attr => ['docid', 'headline'], |
55 |
); |
); |
56 |
|
|
57 |
The constructor returns a handle for the table. This handle is hidden by the |
The constructor returns a handle for the table. This handle is hidden by the |
58 |
table module, to prevent direct access if called via Table. |
table module, to prevent direct access if called via Table. |
59 |
|
|
60 |
=over 10 |
=over 10 |
61 |
|
|
62 |
=item C<access> => I<accesobj> |
=item C<access> => I<accessobj> |
63 |
|
|
64 |
A reference to a acces object for the external parts (attributes) of |
A reference to an access object for the external parts (attributes) of |
65 |
tuples. As you may remember, the WAIT System does not enforce that |
tuples. As you may remember, the WAIT System does not enforce that |
66 |
objects are completely stored inside the system to avoid duplication. |
objects are completely stored inside the system to avoid duplication. |
67 |
There is no (strong) point in storing all you HTML-Documents inside |
There is no (strong) point in storing all your HTML documents inside |
68 |
the system when indexing your WWW-Server. |
the system when indexing your WWW-Server. |
69 |
|
|
70 |
|
The access object is designed to work like as a tied hash. You pass |
71 |
|
the refernce to the object, not the tied hash though. An example |
72 |
|
implementation of an access class that works for manpages is |
73 |
|
WAIT::Document::Nroff. |
74 |
|
|
75 |
|
The implementation needs to take into account that WAIT will keep this |
76 |
|
object in a Data::Dumper or Storable database and re-use it when sman |
77 |
|
is run. So it is not good enough if we can produce the index with it |
78 |
|
now, when we create or actively access the table, WAIT also must be |
79 |
|
able to retrieve documents on its own, when we are in a different |
80 |
|
context. This happens specifically in a retrieval. To get this working |
81 |
|
seemlessly, the access-defining class must implement a close method. |
82 |
|
This method will be called before the Data::Dumper dump takes place. |
83 |
|
In that moment the access-defining class must get rid of all data |
84 |
|
structures that cannot be reconstructed via the Data::Dumper dump, |
85 |
|
such as database handles or C pointers. |
86 |
|
|
87 |
=item C<file> => I<fname> |
=item C<file> => I<fname> |
88 |
|
|
89 |
The filename of the records file. Files for indexes will have I<fname> |
The filename of the records file. Files for indexes will have I<fname> |
90 |
as prefix. I<Mandatory> |
as prefix. I<Mandatory>, but usually taken care of by the |
91 |
|
WAIT::Database handle when the constructor is called via |
92 |
|
WAIT::Database::create_table(). |
93 |
|
|
94 |
=item C<name> => I<name> |
=item C<name> => I<name> |
95 |
|
|
97 |
|
|
98 |
=item C<attr> => [ I<attr> ... ] |
=item C<attr> => [ I<attr> ... ] |
99 |
|
|
100 |
A reference to an array of attribute names. I<Mandatory> |
A reference to an array of attribute names. WAIT will keep the |
101 |
|
contents of these attributes in its table. I<Mandatory> |
102 |
|
|
103 |
=item C<djk> => [ I<attr> ... ] |
=item C<djk> => [ I<attr> ... ] |
104 |
|
|
105 |
A reference to an array of attribute names which make up the |
A reference to an array of attribute names which make up the |
106 |
I<disjointness key>. Don't think about it - i's of no use yet; |
I<disjointness key>. Don't think about it - it's of no use yet; |
107 |
|
|
108 |
=item C<layout> => I<layoutobj> |
=item C<layout> => I<layoutobj> |
109 |
|
|
110 |
A reference to an external parser object. Defaults to anew instance of |
A reference to an external parser object. Defaults to a new instance |
111 |
C<WAIT::Parse::Base> |
of C<WAIT::Parse::Base>. For an example implementation see |
112 |
|
WAIT::Parse::Nroff. A layout class can be implemented as a singleton |
113 |
|
class if you so like. |
114 |
|
|
115 |
|
=item C<keyset> => I<keyset> |
116 |
|
|
117 |
=item C<access> => I<accesobj> |
The set of attributes needed to identify a record. Defaults to all |
118 |
|
attributes. |
119 |
|
|
120 |
A reference to a acces object for the external parts of tuples. |
=item C<invindex> => I<inverted index> |
121 |
|
|
122 |
|
A reference to an anon array defining attributes of each record that |
123 |
|
need to be indexed. See the source of smakewhatis for how to set this |
124 |
|
up. |
125 |
|
|
126 |
=back |
=back |
127 |
|
|
132 |
my %parm = @_; |
my %parm = @_; |
133 |
my $self = {}; |
my $self = {}; |
134 |
|
|
135 |
|
# Check for mandatory attrs early |
136 |
|
$self->{name} = $parm{name} or croak "No name specified"; |
137 |
|
$self->{attr} = $parm{attr} or croak "No attributes specified"; |
138 |
|
|
139 |
# Do that before we eventually add '_weight' to attributes. |
# Do that before we eventually add '_weight' to attributes. |
140 |
$self->{keyset} = $parm{keyset} || [[@{$parm{attr}}]]; |
$self->{keyset} = $parm{keyset} || [[@{$parm{attr}}]]; |
141 |
|
|
142 |
$self->{mode} = O_CREAT | O_RDWR; |
$self->{mode} = O_CREAT | O_RDWR; |
143 |
|
|
144 |
# Determine and set up subclass |
# Determine and set up subclass |
145 |
$type = ref($type) || $type; |
$type = ref($type) || $type; |
146 |
if (defined $parm{djk}) { |
if (defined $parm{djk}) { |
159 |
} |
} |
160 |
|
|
161 |
$self->{file} = $parm{file} or croak "No file specified"; |
$self->{file} = $parm{file} or croak "No file specified"; |
162 |
if (-d $self->{file} or !mkdir($self->{file}, 0775)) { |
if (-e $self->{file}){ |
163 |
croak "Could not 'mkdir $self->{file}': $!\n"; |
warn "Warning: file '$self->{file}' already exists\n"; |
164 |
} |
} |
165 |
$self->{name} = $parm{name} or croak "No name specified"; |
|
|
$self->{attr} = $parm{attr} or croak "No attributes specified"; |
|
166 |
$self->{djk} = $parm{djk} if defined $parm{djk}; |
$self->{djk} = $parm{djk} if defined $parm{djk}; |
167 |
$self->{layout} = $parm{layout} || new WAIT::Parse::Base; |
$self->{layout} = $parm{layout} || new WAIT::Parse::Base; |
168 |
$self->{access} = $parm{access} if defined $parm{access}; |
$self->{access} = $parm{access} if defined $parm{access}; |
171 |
$self->{indexes} = {}; |
$self->{indexes} = {}; |
172 |
|
|
173 |
bless $self, $type; |
bless $self, $type; |
174 |
|
|
175 |
|
# Checking for readers is not necessary, but let's go with the |
176 |
|
# generic method. |
177 |
|
$self->getlock(O_RDWR|O_CREAT); # dies when failing |
178 |
|
|
179 |
# Call create_index() and create_index() for compatibility |
# Call create_index() and create_index() for compatibility |
180 |
for (@{$self->{keyset}||[]}) { |
for (@{$self->{keyset}||[]}) { |
181 |
#carp "Specification of indexes at table create time is deprecated"; |
#carp "Specification of indexes at table create time is deprecated"; |
186 |
my $att = shift @{$parm{invindex}}; |
my $att = shift @{$parm{invindex}}; |
187 |
my @spec = @{shift @{$parm{invindex}}}; |
my @spec = @{shift @{$parm{invindex}}}; |
188 |
my @opt; |
my @opt; |
189 |
|
|
190 |
if (ref($spec[0])) { |
if (ref($spec[0])) { |
191 |
carp "Secondary pipelines are deprecated\n"; |
carp "Secondary pipelines are deprecated\n"; |
192 |
@opt = %{shift @spec}; |
@opt = %{shift @spec}; |
193 |
} |
} |
194 |
$self->create_inverted_index(attribute => $att, pipeline => \@spec, @opt); |
$self->create_inverted_index(attribute => $att, pipeline => \@spec, @opt); |
195 |
} |
} |
196 |
|
|
197 |
$self; |
$self; |
198 |
# end of backwarn compatibility stuff |
# end of backwarn compatibility stuff |
199 |
} |
} |
213 |
|
|
214 |
sub create_index { |
sub create_index { |
215 |
my $self= shift; |
my $self= shift; |
216 |
|
|
217 |
croak "Cannot create index for table aready populated" |
croak "Cannot create index for table aready populated" |
218 |
if $self->{nextk} > 1; |
if $self->{nextk} > 1; |
219 |
|
|
220 |
require WAIT::Index; |
require WAIT::Index; |
221 |
|
|
222 |
my $name = join '-', @_; |
my $name = join '-', @_; |
223 |
$self->{indexes}->{$name} = |
$self->{indexes}->{$name} = |
224 |
new WAIT::Index file => $self->{file}.'/'.$name, attr => $_; |
new WAIT::Index file => $self->{file}, name => $name, attr => $_; |
225 |
} |
} |
226 |
|
|
227 |
=head2 Creating an inverted index |
=head2 Creating an inverted index |
241 |
|
|
242 |
=item C<pipeline> |
=item C<pipeline> |
243 |
|
|
244 |
A piplines specification is a reference to and array of method names |
A piplines specification is a reference to an array of method names |
245 |
(from package C<WAIT::Filter>) which are to applied in sequence to the |
(from package C<WAIT::Filter>) which are to be applied in sequence to |
246 |
contents of the named attribute. The attribute name may not be in the |
the contents of the named attribute. The attribute name may not be in |
247 |
attribute list. |
the attribute list. |
248 |
|
|
249 |
=item C<predicate> |
=item C<predicate> |
250 |
|
|
251 |
An indication which predicate the index implements. This may be |
An indication which predicate the index implements. This may be |
252 |
e.g. 'plain', 'stemming' or 'soundex'. The indicator will be used for |
e.g. 'plain', 'stemming' or 'soundex'. The indicator will be used for |
253 |
query processing. Currently there is no standard set of predicate |
query processing. Currently there is no standard set of predicate |
254 |
names. The predicate defaults to the last member of the ppline if |
names. The predicate defaults to the last member of the pipeline if |
255 |
omitted. |
omitted. |
256 |
|
|
257 |
=back |
=back |
269 |
croak "No pipeline specified" unless $parm{pipeline}; |
croak "No pipeline specified" unless $parm{pipeline}; |
270 |
|
|
271 |
$parm{predicate} ||= $parm{pipeline}->[-1]; |
$parm{predicate} ||= $parm{pipeline}->[-1]; |
272 |
|
|
273 |
croak "Cannot create index for table aready populated" |
croak "Cannot create index for table aready populated" |
274 |
if $self->{nextk} > 1; |
if $self->{nextk} > 1; |
275 |
|
|
276 |
require WAIT::InvertedIndex; |
require WAIT::InvertedIndex; |
277 |
|
|
278 |
# backward compatibility stuff |
# backward compatibility stuff |
280 |
for (qw(attribute pipeline predicate)) { |
for (qw(attribute pipeline predicate)) { |
281 |
delete $opt{$_}; |
delete $opt{$_}; |
282 |
} |
} |
283 |
|
|
284 |
my $name = join '_', ($parm{attribute}, @{$parm{pipeline}}); |
my $name = join '_', ($parm{attribute}, @{$parm{pipeline}}); |
285 |
my $idx = new WAIT::InvertedIndex(file => $self->{file}.'/'.$name, |
my $idx = new WAIT::InvertedIndex(file => $self->{file}.'/'.$name, |
286 |
filter => [@{$parm{pipeline}}], # clone |
filter => [@{$parm{pipeline}}], # clone |
321 |
|
|
322 |
sub drop { |
sub drop { |
323 |
my $self = shift; |
my $self = shift; |
324 |
|
|
325 |
|
unless ($self->{write_lock}){ |
326 |
|
warn "Cannot drop table without write lock. Nothing done"; |
327 |
|
return; |
328 |
|
} |
329 |
|
|
330 |
if ((caller)[0] eq 'WAIT::Database') { # database knows about this |
if ((caller)[0] eq 'WAIT::Database') { # database knows about this |
331 |
$self->close; # just make sure |
$self->close; # just make sure |
332 |
|
|
333 |
my $file = $self->{file}; |
my $file = $self->{file}; |
334 |
|
|
335 |
for (values %{$self->{indexes}}) { |
for (values %{$self->{indexes}}) { |
336 |
$_->drop; |
$_->drop; |
337 |
} |
} |
338 |
unlink "$file/records"; |
rmdir "$file.read" or warn "Could not rmdir '$file/read'"; |
339 |
! (!-e $file or rmdir $file); |
unlink "$file"; |
340 |
|
|
341 |
} else { |
} else { |
342 |
croak ref($self)."::drop called directly"; |
croak ref($self)."::drop called directly"; |
343 |
} |
} |
377 |
} |
} |
378 |
require WAIT::InvertedIndex; |
require WAIT::InvertedIndex; |
379 |
} |
} |
380 |
|
|
381 |
|
$self->getlock($self->{mode}); |
382 |
|
|
383 |
|
my $dbmode = ($self->{mode} & O_CREAT) ? DB_CREATE : 0; |
384 |
unless (defined $self->{dbh}) { |
unless (defined $self->{dbh}) { |
385 |
if ($USE_RECNO) { |
if ($USE_RECNO) { |
386 |
$self->{dbh} = tie(@{$self->{db}}, 'DB_File', $file, |
tie(%{$self->{db}}, 'BerkeleyDB::Recno', |
387 |
$self->{mode}, 0664, $DB_RECNO); |
-Filename => $self->{file}, |
388 |
|
-Subname => 'records', |
389 |
|
-Flags => $dbmode); |
390 |
} else { |
} else { |
391 |
$self->{dbh} = |
$self->{dbh} = |
392 |
tie(%{$self->{db}}, 'DB_File', $file, |
tie(%{$self->{db}}, 'BerkeleyDB::Btree', |
393 |
$self->{mode}, 0664, $DB_BTREE); |
-Filename => $self->{file}, |
394 |
|
-Subname => 'records', |
395 |
|
-Mode => 0664, |
396 |
|
-Flags => $dbmode); |
397 |
} |
} |
398 |
} |
} |
399 |
|
|
400 |
|
|
401 |
$self; |
$self; |
402 |
} |
} |
403 |
|
|
404 |
sub fetch_extern { |
sub fetch_extern { |
405 |
my $self = shift; |
my $self = shift; |
406 |
|
|
407 |
print "#@_", $self->{'access'}->{Mode}, "\n"; |
# print "#@_", $self->{'access'}->{Mode}, "\n"; # DEBUGGING? |
408 |
if (exists $self->{'access'}) { |
if (exists $self->{'access'}) { |
409 |
mrequire ref($self->{'access'}); |
mrequire ref($self->{'access'}); |
410 |
$self->{'access'}->FETCH(@_); |
$self->{'access'}->FETCH(@_); |
422 |
my (@att) = @_; |
my (@att) = @_; |
423 |
my %att; |
my %att; |
424 |
my $name; |
my $name; |
425 |
|
|
426 |
@att{@att} = @att; |
@att{@att} = @att; |
427 |
|
|
428 |
KEY: for $name (keys %{$self->{indexes}}) { |
KEY: for $name (keys %{$self->{indexes}}) { |
439 |
my $self = shift; |
my $self = shift; |
440 |
my %parm = @_; |
my %parm = @_; |
441 |
|
|
442 |
my $index = $self->_find_index(keys %parm); |
my $index = $self->_find_index(keys %parm) or return; # no index-no have |
443 |
croak "No index found" unless $index; |
|
444 |
defined $self->{db} or $self->open; |
defined $self->{db} or $self->open; |
445 |
return $index->have(@_); |
return $index->have(@_); |
446 |
} |
} |
451 |
|
|
452 |
defined $self->{db} or $self->open; |
defined $self->{db} or $self->open; |
453 |
|
|
454 |
|
# We should move all writing methods to a subclass to check only once |
455 |
|
$self->{mode} & O_RDWR or croak "Cannot insert into table opened in RD_ONLY mode"; |
456 |
|
|
457 |
my $tuple = join($;, map($parm{$_} || '', @{$self->{attr}})); |
my $tuple = join($;, map($parm{$_} || '', @{$self->{attr}})); |
458 |
my $key; |
my $key; |
459 |
my @deleted = keys %{$self->{deleted}}; |
my @deleted = keys %{$self->{deleted}}; |
460 |
|
my $gotkey = 0; |
461 |
|
|
462 |
if (@deleted) { |
if (@deleted) { |
463 |
$key = pop @deleted; |
$key = pop @deleted; |
464 |
delete $self->{deleted}->{$key}; |
delete $self->{deleted}->{$key}; |
465 |
|
# Sanity check |
466 |
|
if ($key && $key>0) { |
467 |
|
$gotkey=1; |
468 |
} else { |
} else { |
469 |
|
warn(sprintf("WAIT database inconsistency during insert ". |
470 |
|
"key[%s]: Please rebuild index\n", |
471 |
|
$key |
472 |
|
)); |
473 |
|
} |
474 |
|
} |
475 |
|
unless ($gotkey) { |
476 |
$key = $self->{nextk}++; |
$key = $self->{nextk}++; |
477 |
} |
} |
478 |
if ($USE_RECNO) { |
if ($USE_RECNO) { |
486 |
if ($key == $self->{nextk}-1) { |
if ($key == $self->{nextk}-1) { |
487 |
$self->{nextk}--; |
$self->{nextk}--; |
488 |
} else { |
} else { |
489 |
|
# warn "setting key[$key] deleted during insert"; |
490 |
$self->{deleted}->{$key}=1; |
$self->{deleted}->{$key}=1; |
491 |
} |
} |
492 |
my $idx; |
my $idx; |
495 |
$idx->remove($key, %parm); |
$idx->remove($key, %parm); |
496 |
} |
} |
497 |
return undef; |
return undef; |
498 |
} |
} |
499 |
} |
} |
500 |
if (defined $self->{inverted}) { |
if (defined $self->{inverted}) { |
501 |
my $att; |
my $att; |
511 |
|
|
512 |
sub sync { |
sub sync { |
513 |
my $self = shift; |
my $self = shift; |
514 |
|
|
515 |
for (values %{$self->{indexes}}) { |
for (values %{$self->{indexes}}) { |
516 |
map $_->sync, $_; |
map $_->sync, $_; |
517 |
} |
} |
528 |
my $key = shift; |
my $key = shift; |
529 |
|
|
530 |
return () if exists $self->{deleted}->{$key}; |
return () if exists $self->{deleted}->{$key}; |
531 |
|
|
532 |
defined $self->{db} or $self->open; |
defined $self->{db} or $self->open; |
533 |
if ($USE_RECNO) { |
if ($USE_RECNO) { |
534 |
$self->unpack($self->{db}->[$key]); |
$self->unpack($self->{db}->[$key]); |
541 |
my $self = shift; |
my $self = shift; |
542 |
my $key = shift; |
my $key = shift; |
543 |
|
|
544 |
|
unless ($key) { |
545 |
|
Carp::cluck "Warning: delete_by_key called without key. Looks like a bug in WAIT?"; |
546 |
|
return; |
547 |
|
} |
548 |
|
|
549 |
return $self->{deleted}->{$key} if defined $self->{deleted}->{$key}; |
return $self->{deleted}->{$key} if defined $self->{deleted}->{$key}; |
550 |
my %tuple = $self->fetch($key); |
my %tuple = $self->fetch($key); |
551 |
for (values %{$self->{indexes}}) { |
for (values %{$self->{indexes}}) { |
562 |
} |
} |
563 |
} |
} |
564 |
} |
} |
565 |
|
# warn "setting key[$key] deleted during delete_by_key"; |
566 |
++$self->{deleted}->{$key}; |
++$self->{deleted}->{$key}; |
567 |
} |
} |
568 |
|
|
569 |
sub delete { |
sub delete { |
570 |
my $self = shift; |
my $self = shift; |
571 |
my $tkey = $self->have(@_); |
my $tkey = $self->have(@_); |
572 |
|
# warn "tkey[$tkey]\@_[@_]"; |
573 |
defined $tkey && $self->delete_by_key($tkey, @_); |
defined $tkey && $self->delete_by_key($tkey, @_); |
574 |
} |
} |
575 |
|
|
576 |
sub unpack { |
sub unpack { |
577 |
my $self = shift; |
my($self, $tuple) = @_; |
578 |
my $tuple = shift; |
|
579 |
|
unless (defined $tuple){ |
580 |
|
# require Carp; # unfortunately gives us "bizarre copy...." :-((((( |
581 |
|
warn("Debug: somebody called unpack without argument tuple!"); |
582 |
|
return; |
583 |
|
} |
584 |
|
|
585 |
my $att; |
my $att; |
586 |
my @result; |
my @result; |
592 |
@result; |
@result; |
593 |
} |
} |
594 |
|
|
595 |
|
sub set { |
596 |
|
my ($self, $iattr, $value) = @_; |
597 |
|
|
598 |
|
unless ($self->{write_lock}){ |
599 |
|
warn "Cannot set iattr[$iattr] without write lock. Nothing done"; |
600 |
|
return; |
601 |
|
} |
602 |
|
|
603 |
|
# in the rare case that they haven't written a single record yet, we |
604 |
|
# make sure, the inverted inherits our $self->{mode}: |
605 |
|
defined $self->{db} or $self->open; |
606 |
|
|
607 |
|
for my $att (keys %{$self->{inverted}}) { |
608 |
|
if ($] > 5.003) { # avoid bug in perl up to 5.003_05 |
609 |
|
my $idx; |
610 |
|
for $idx (@{$self->{inverted}->{$att}}) { |
611 |
|
$idx->set($iattr, $value); |
612 |
|
} |
613 |
|
} else { |
614 |
|
map $_->set($iattr, $value), @{$self->{inverted}->{$att}}; |
615 |
|
} |
616 |
|
} |
617 |
|
|
618 |
|
1; |
619 |
|
} |
620 |
|
|
621 |
sub close { |
sub close { |
622 |
my $self = shift; |
my $self = shift; |
623 |
|
|
624 |
if (exists $self->{'access'}) { |
if (exists $self->{'access'}) { |
625 |
eval {$self->{'access'}->close}; # dont bother if not opened |
eval {$self->{'access'}->close}; # dont bother if not opened |
626 |
} |
} |
627 |
for (values %{$self->{indexes}}) { |
if ($WAIT::Index::VERSION) { |
628 |
$_->close(); |
for (values %{$self->{indexes}}) { |
629 |
|
$_->close(); |
630 |
|
} |
631 |
} |
} |
632 |
if (defined $self->{inverted}) { |
if (defined $self->{inverted} && $WAIT::InvertedIndex::VERSION) { |
633 |
|
# require WAIT::InvertedIndex; Uli: we can avoid closing indexes: |
634 |
|
# if WAIT::InvertedIndex has not been loaded, they cannot have |
635 |
|
# been altered so far |
636 |
my $att; |
my $att; |
637 |
for $att (keys %{$self->{inverted}}) { |
for $att (keys %{$self->{inverted}}) { |
638 |
if ($] > 5.003) { # avoid bug in perl up to 5.003_05 |
if ($] > 5.003) { # avoid bug in perl up to 5.003_05 |
656 |
delete $self->{db}; |
delete $self->{db}; |
657 |
} |
} |
658 |
|
|
659 |
|
$self->unlock; |
660 |
|
|
661 |
1; |
1; |
662 |
} |
} |
663 |
|
|
664 |
|
# Locking |
665 |
|
# |
666 |
|
# We allow multiple readers to coexists. But write access excludes |
667 |
|
# all read access and vice versa. In practice read access on tables |
668 |
|
# open for writing will mostly work ;-) |
669 |
|
|
670 |
|
# If a "write" lock is requested, an existing "read" lock will be |
671 |
|
# released. If a "read" lock ist requested, an existing "write" lock |
672 |
|
# will be released. Requiring a lock already hold has no effect. |
673 |
|
|
674 |
|
sub getlock { |
675 |
|
my ($self, $mode) = @_; |
676 |
|
|
677 |
|
# autoclean cleans on DESTROY, stale sends SIGZERO to the owner |
678 |
|
# |
679 |
|
my $lockmgr = LockFile::Simple->make(-autoclean => 1, -stale => 1); |
680 |
|
my $file = $self->{file}; |
681 |
|
my $lockdir = $self->{file} . '.read'; |
682 |
|
|
683 |
|
unless (-d $lockdir) { |
684 |
|
mkdir $lockdir, 0755 or die "Could not mkdir $lockdir: $!"; |
685 |
|
} |
686 |
|
|
687 |
|
if ($mode & O_RDWR) { # Get a write lock. Release it again |
688 |
|
# and die if there is any valid |
689 |
|
# readers. |
690 |
|
|
691 |
|
# Have a write lock already |
692 |
|
return $self if $self->{write_lock}; |
693 |
|
|
694 |
|
if ($self->{read_lock}) { # We are a becoming a writer now. So |
695 |
|
# we release the read lock to avoid |
696 |
|
# blocking ourselves. |
697 |
|
$self->{read_lock}->release; |
698 |
|
delete $self->{read_lock}; |
699 |
|
} |
700 |
|
|
701 |
|
# Get the preliminary write lock |
702 |
|
$self->{write_lock} = $lockmgr->lock($self->{file} . '.write') |
703 |
|
or die "Can't lock '$self->{file}.write'"; |
704 |
|
|
705 |
|
# If we actually want to write we must check if there are any |
706 |
|
# readers. The write lock is confirmed if wen cannot find any |
707 |
|
# valid readers. |
708 |
|
|
709 |
|
local *DIR; |
710 |
|
opendir DIR, $lockdir or |
711 |
|
die "Could not opendir '$lockdir': $!"; |
712 |
|
for my $lockfile (grep { -f "$lockdir/$_" } readdir DIR) { |
713 |
|
# Check if the locks are still valid. Since we are protected by |
714 |
|
# a write lock, we could use a plain file. But we want to use |
715 |
|
# the stale testing from LockFile::Simple. |
716 |
|
if (my $lck = $lockmgr->trylock("$lockdir/$lockfile")) { |
717 |
|
warn "Removing stale lockfile '$lockdir/$lockfile'"; |
718 |
|
$lck->release; |
719 |
|
} else { # Found an active reader, rats! |
720 |
|
$self->{write_lock}->release; |
721 |
|
die "Cannot write table '$file' while it's in use"; |
722 |
|
} |
723 |
|
} |
724 |
|
closedir DIR; |
725 |
|
} else { |
726 |
|
# Have a read lock already |
727 |
|
return $self if $self->{read_lock}; |
728 |
|
|
729 |
|
# Get the preliminary write lock to protect the directory |
730 |
|
# operations. |
731 |
|
|
732 |
|
my $write_lock = $lockmgr->lock($self->{file} . '.read/write') |
733 |
|
or die "Can't lock '$self->{file}.read/write'"; |
734 |
|
|
735 |
|
# Find a new read slot. Maybe the plain file would be better? |
736 |
|
my $id = time; |
737 |
|
while (-f "$lockdir/$id.lock") { # here assume ".lock" format! |
738 |
|
$id++; |
739 |
|
} |
740 |
|
|
741 |
|
$self->{read_lock} = $lockmgr->lock("$lockdir/$id") |
742 |
|
or die "Can't lock '$lockdir/$id'"; |
743 |
|
|
744 |
|
# We are a reader now. So we release the write lock |
745 |
|
$write_lock->release; |
746 |
|
} |
747 |
|
return $self; |
748 |
|
} |
749 |
|
|
750 |
|
sub unlock { |
751 |
|
my $self = shift; |
752 |
|
|
753 |
|
# Either we have a read or a write lock (or we close the table already) |
754 |
|
# unless ($self->{read_lock} || $self->{write_lock}) { |
755 |
|
# warn "WAIT::Table::unlock: Table aparently hold's no lock" |
756 |
|
# } |
757 |
|
if ($self->{write_lock}) { |
758 |
|
$self->{write_lock}->release(); |
759 |
|
delete $self->{write_lock}; |
760 |
|
} |
761 |
|
if ($self->{read_lock}) { |
762 |
|
$self->{read_lock}->release(); |
763 |
|
delete $self->{read_lock}; |
764 |
|
} |
765 |
|
|
766 |
|
} |
767 |
|
|
768 |
|
sub DESTROY { |
769 |
|
my $self = shift; |
770 |
|
|
771 |
|
if ($self->{write_lock} || $self->{read_lock}) { |
772 |
|
warn "Table handle destroyed without closing it first"; |
773 |
|
$self->unlock; |
774 |
|
} |
775 |
|
} |
776 |
|
|
777 |
sub open_scan { |
sub open_scan { |
778 |
my $self = shift; |
my $self = shift; |
779 |
my $code = shift; |
my $code = shift; |
829 |
} |
} |
830 |
|
|
831 |
sub search { |
sub search { |
832 |
my $self = shift; |
my $self = shift; |
833 |
my $attr = shift; |
my ($query, $attr, $cont, $raw); |
834 |
my $cont = shift; |
if (ref $_[0]) { |
835 |
my $raw = shift; |
$query = shift; |
836 |
|
|
837 |
|
$attr = $query->{attr}; |
838 |
|
$cont = $query->{cont}; |
839 |
|
$raw = $query->{raw}; |
840 |
|
} else { |
841 |
|
require Carp; |
842 |
|
Carp::cluck("Using three argument search interface is deprecated, use hashref interface instead"); |
843 |
|
$attr = shift; |
844 |
|
$cont = shift; |
845 |
|
$raw = shift; |
846 |
|
$query = { |
847 |
|
attr => $attr, |
848 |
|
cont => $cont, |
849 |
|
raw => $raw, |
850 |
|
}; |
851 |
|
} |
852 |
|
|
853 |
my %result; |
my %result; |
854 |
|
|
855 |
defined $self->{db} or $self->open; # require layout |
defined $self->{db} or $self->open; # require layout |
859 |
my $name = $_->name; |
my $name = $_->name; |
860 |
if (exists $raw->{$name} and @{$raw->{$name}}) { |
if (exists $raw->{$name} and @{$raw->{$name}}) { |
861 |
my $scale = 1/scalar(@{$raw->{$name}}); |
my $scale = 1/scalar(@{$raw->{$name}}); |
862 |
my %r = $_->search_raw(@{$raw->{$name}}); |
my %r = $_->search_raw($query, @{$raw->{$name}}); |
863 |
my ($key, $val); |
my ($key, $val); |
864 |
while (($key, $val) = each %r) { |
while (($key, $val) = each %r) { |
865 |
if (exists $result{$key}) { |
if (exists $result{$key}) { |
873 |
} |
} |
874 |
if (defined $cont and $cont ne '') { |
if (defined $cont and $cont ne '') { |
875 |
for (@{$self->{inverted}->{$attr}}) { |
for (@{$self->{inverted}->{$attr}}) { |
876 |
my %r = $_->search($cont); |
my %r = $_->search($query, $cont); |
877 |
my ($key, $val); |
my ($key, $val); |
878 |
while (($key, $val) = each %r) { |
while (($key, $val) = each %r) { |
879 |
if (exists $result{$key}) { |
if (exists $result{$key}) { |
897 |
my %pos; |
my %pos; |
898 |
|
|
899 |
if (defined $raw) { |
if (defined $raw) { |
900 |
for (@{$self->{inverted}->{$attr}}) { |
for (@{$self->{inverted}->{$attr}}) { # objects of type |
901 |
|
# WAIT::InvertedIndex for |
902 |
|
# this index field $attr |
903 |
my $name = $_->name; |
my $name = $_->name; |
904 |
if (exists $raw->{$name}) { |
if (exists $raw->{$name}) { |
905 |
my %qt; |
my %qt; |
933 |
} |
} |
934 |
|
|
935 |
sub hilight { |
sub hilight { |
936 |
my ($tb, $text, $query, $raw) = @_; |
my ($tb, $buf, $qplain, $qraw) = @_; |
937 |
my $type = $tb->layout(); |
my $layout = $tb->layout(); |
938 |
|
|
939 |
my @result; |
my @result; |
940 |
|
|
941 |
$query ||= {}; |
$qplain ||= {}; |
942 |
$raw ||= {}; |
$qraw ||= {}; |
943 |
my @ttxt = $type->tag($text); |
my @ttxt = $layout->tag($buf); |
944 |
while (@ttxt) { |
while (@ttxt) { |
945 |
no strict 'refs'; |
no strict 'refs'; |
946 |
my %tag = %{shift @ttxt}; |
my %tag = %{shift @ttxt}; |
948 |
my $fld; |
my $fld; |
949 |
|
|
950 |
my %hl; |
my %hl; |
951 |
for $fld (grep defined $tag{$_}, keys %$query, keys %$raw) { |
for $fld (grep defined $tag{$_}, keys %$qplain, keys %$qraw) { |
952 |
my $hp = $tb->hilight_positions($fld, $txt, |
my $hp = $tb->hilight_positions($fld, $txt, |
953 |
$query->{$fld}, $raw->{$fld}); |
$qplain->{$fld}, $qraw->{$fld}); |
954 |
for (keys %$hp) { |
for (keys %$hp) { |
955 |
if (exists $hl{$_}) { # -w ;-( |
if (exists $hl{$_}) { # -w ;-( |
956 |
$hl{$_} = max($hl{$_}, $hp->{$_}); |
$hl{$_} = max($hl{$_}, $hp->{$_}); |
976 |
} |
} |
977 |
|
|
978 |
1; |
1; |
|
|
|