4 |
use strict; |
use strict; |
5 |
use warnings; |
use warnings; |
6 |
|
|
7 |
our $VERSION = '0.00'; |
our $VERSION = '0.02'; |
8 |
|
|
9 |
use SWISH::API; |
use SWISH::API; |
10 |
use Text::Iconv; |
use Text::Iconv; |
12 |
use Carp; |
use Carp; |
13 |
use Digest::MD5 qw(md5_hex); |
use Digest::MD5 qw(md5_hex); |
14 |
use Memoize; |
use Memoize; |
|
use IPC::Run qw(start timeout pump finish); |
|
15 |
use File::Which; |
use File::Which; |
16 |
|
|
17 |
use Data::Dumper; |
use Data::Dumper; |
41 |
engines) easier. |
engines) easier. |
42 |
|
|
43 |
In the background, it will fork swish-e binaries (one for each index slice) |
In the background, it will fork swish-e binaries (one for each index slice) |
44 |
and produce UTF-8 encoded XML files for it. So, if your imput charset isn't |
and produce UTF-8 encoded XML files for it. So, if your input charset isn't |
45 |
C<ISO-8859-1> you will have to specify it. |
C<ISO-8859-1> you will have to specify it. |
46 |
|
|
47 |
=head1 Methods used for indexing |
=head1 Methods used for indexing |
48 |
|
|
49 |
=head2 open |
=head2 open_index |
50 |
|
|
51 |
Create new object for index. |
Create new object for index. |
52 |
|
|
53 |
my $i = SWISH::Split->open({ |
my $i = SWISH::Split->open_index({ |
54 |
index => '/path/to/index', |
index => '/path/to/index', |
55 |
slice_name => \&slice_on_path, |
slice_name => \&slice_on_path, |
56 |
slices => 30, |
slices => 30, |
68 |
return shift split(/\//,$_[0]); |
return shift split(/\//,$_[0]); |
69 |
} |
} |
70 |
|
|
71 |
Options to open are following: |
Options to C<open_index> are following: |
72 |
|
|
73 |
=over 5 |
=over 5 |
74 |
|
|
109 |
|
|
110 |
my $iso2utf = Text::Iconv->new('ISO-8859-1','UTF-8'); |
my $iso2utf = Text::Iconv->new('ISO-8859-1','UTF-8'); |
111 |
|
|
112 |
sub open { |
sub open_index { |
113 |
my $class = shift; |
my $class = shift; |
114 |
my $self = {@_}; |
my $self = {@_}; |
115 |
bless($self, $class); |
bless($self, $class); |
148 |
my $swishpath = shift || return; |
my $swishpath = shift || return; |
149 |
my $data = shift || return; |
my $data = shift || return; |
150 |
|
|
151 |
my ($out,$err) = $self->put_slice($swishpath, $self->to_xml($data)); |
my $slice = $self->put_slice($swishpath, $self->to_xml($data)); |
152 |
|
|
153 |
if ($err) { |
return $slice; |
|
carp "$swishpath: $err"; |
|
|
return 0; |
|
|
} |
|
|
|
|
|
return 1; |
|
154 |
} |
} |
155 |
|
|
156 |
=head2 delete |
=head2 delete |
196 |
my $ret = 0; |
my $ret = 0; |
197 |
|
|
198 |
foreach my $s (keys %{$self->{'slice'}}) { |
foreach my $s (keys %{$self->{'slice'}}) { |
199 |
print STDERR "closing slice $s\n"; |
$self->_debug("closing slice $s"); |
200 |
$ret += $self->close_slice($s); |
$ret += $self->close_slice($s); |
201 |
} |
} |
202 |
|
|
274 |
|
|
275 |
my $s = $i->in_slice('path/to/document/in/index'); |
my $s = $i->in_slice('path/to/document/in/index'); |
276 |
|
|
277 |
If there are C<slices> parametar to L<"open"> it will use |
If there are C<slices> parametar to L<"open_index"> it will use |
278 |
MD5 hash to spread documents across slices. That will produce random |
MD5 hash to spread documents across slices. That will produce random |
279 |
distribution of your documents in slices, which might or might not be best |
distribution of your documents in slices, which might or might not be best |
280 |
for your data. If you have to re-index large number of slices on each |
for your data. If you have to re-index large number of slices on each |
298 |
# first, pass path through slice_name function |
# first, pass path through slice_name function |
299 |
my $slice = &{$self->{'slice_name'}}($path); |
my $slice = &{$self->{'slice_name'}}($path); |
300 |
# then calculate MD5 hash |
# then calculate MD5 hash |
301 |
$slice = md5_hex($slice); |
my $hash = md5_hex($slice); |
302 |
# take first 8 chars to produce number |
# take first 8 chars to produce number |
303 |
# FIXME how random is this? |
# FIXME how random is this? |
304 |
$slice = hex(substr($slice,0,8)); |
$hash = hex(substr($hash,0,8)); |
305 |
|
|
306 |
$slice = ($slice % $self->{'slices'}) + 1; |
$slice = ($hash % $self->{'slices'}) + 1; |
307 |
print "hash: $slice / ",$self->{'slices'}," => $slice\n"; |
$self->_debug("hash: $hash / ",$self->{'slices'}," => $slice"); |
308 |
return $slice; |
return $slice; |
309 |
} else { |
} else { |
310 |
return &{$self->{'split'}}($path); |
return &{$self->{'split'}}($path); |
335 |
my $config_filename = $i->make_config('slice name'); |
my $config_filename = $i->make_config('slice name'); |
336 |
|
|
337 |
It returns configuration filename. If no C<swish_config> was defined in |
It returns configuration filename. If no C<swish_config> was defined in |
338 |
L<"open">, default swish-e configuration will be used. It will index all data for |
L<"open_index">, default swish-e configuration will be used. It will index all data for |
339 |
searching, but none for properties. |
searching, but none for properties. |
340 |
|
|
341 |
If you want to see what is allready defined for swish-e in configuration |
If you want to see what is allready defined for swish-e in configuration |
342 |
take a look at source code for C<DEFAULT_SWISH_CONF>. |
take a look at source code for C<DEFAULT_SWISH_CONF>. |
343 |
|
|
344 |
It uses C<cat> utility to comunicate with C<swish-e>. Path is provided |
It uses C<stdin> as C<IndexDir> to comunicate with C<swish-e>. |
|
by C<File::Which>. Do Windows users have to change that to C<COPY /B> |
|
|
or something similar? |
|
345 |
|
|
346 |
=cut |
=cut |
347 |
|
|
360 |
print $tmp_fh <<"DEFAULT_SWISH_CONF"; |
print $tmp_fh <<"DEFAULT_SWISH_CONF"; |
361 |
# swish-e config file |
# swish-e config file |
362 |
|
|
363 |
IndexDir cat |
IndexDir stdin |
|
#SwishProgParameters - |
|
364 |
|
|
365 |
# input file definition |
# input file definition |
366 |
DefaultContents XML* |
DefaultContents XML* |
395 |
|
|
396 |
=head2 create_slice |
=head2 create_slice |
397 |
|
|
398 |
On first run, starts C<swish-e> using L<IPC::Run>. On subsequent calls just return |
On first run, starts C<swish-e>. On subsequent calls just return |
399 |
it's handles using L<Memoize>. |
it's handles using L<Memoize>. |
400 |
|
|
401 |
my $s = create_slice('/path/to/document'); |
my $s = create_slice('/path/to/document'); |
416 |
|
|
417 |
my $swish_config = $self->make_config($s); |
my $swish_config = $self->make_config($s); |
418 |
|
|
419 |
print STDERR "creating slice $s\n"; # FIXME |
my $swish = qq{| swish-e }; |
420 |
|
$swish .= qq{ -u } if (-f $self->{'index'}.'/'.$s); |
421 |
|
$swish .= qq{ -S prog -c } . $swish_config; |
422 |
|
|
423 |
my @swish = qw(swish-e -S prog -c); |
$self->_debug("creating slice $s using $swish"); |
|
push @swish, $swish_config; |
|
424 |
|
|
425 |
## Build the harness, open all pipes, and launch the subprocesses |
## Build the harness, open all pipes, and launch the subprocesses |
426 |
$self->{'slice'}->{$s}->{'h'} = start \@swish, |
open(my $fh, $swish) || croak "can't open $swish: $!"; |
|
\$self->{'slice'}->{$s}->{'in'}, |
|
|
\$self->{'slice'}->{$s}->{'out'}, |
|
|
\$self->{'slice'}->{$s}->{'err'}, |
|
|
timeout( 90 ); # FIXME |
|
427 |
|
|
428 |
$self->{'slice'}->{$s}->{'out_len'} = 0; |
$self->{'slice'}->{$s}->{'h'} = $fh; |
|
$self->{'slice'}->{$s}->{'err_len'} = 0; |
|
429 |
|
|
430 |
$self->slice_output($s); |
$self->slice_output($s); |
431 |
|
|
434 |
|
|
435 |
=head2 put_slice |
=head2 put_slice |
436 |
|
|
437 |
Pass XML data to swish and receive output and errors. |
Pass XML data to swish. |
438 |
|
|
439 |
|
my $slice = $i->put_slice('/swish/path', '<xml>data</xml>'); |
440 |
|
|
441 |
my ($out,$err) = $i->put_slice('/swish/path', '<xml>data</xml>'); |
Returns slice in which XML ended up. |
442 |
|
|
443 |
=cut |
=cut |
444 |
|
|
453 |
my $s = $self->create_slice($path) || confess "create_slice returned null"; |
my $s = $self->create_slice($path) || confess "create_slice returned null"; |
454 |
|
|
455 |
confess "no slice $s" unless(exists($self->{'slice'}) && exists($self->{'slice'}->{$s})); |
confess "no slice $s" unless(exists($self->{'slice'}) && exists($self->{'slice'}->{$s})); |
|
confess "no 'in' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'in'})); |
|
456 |
confess "no 'h' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'h'})); |
confess "no 'h' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'h'})); |
457 |
|
|
458 |
$self->slice_output($s); |
$self->slice_output($s); |
459 |
|
|
460 |
use bytes; # as opposed to chars |
use bytes; # as opposed to chars |
461 |
$self->{'slice'}->{$s}->{'in'} .= |
my $fh = $self->{'slice'}->{$s}->{'h'} || confess "handle for slice $s undefined"; |
462 |
"Path-Name: $path\n". |
print { $fh } "Path-Name: $path\n". |
463 |
"Content-Length: ".(length($xml)+1)."\n". |
"Content-Length: ".(length($xml)+1)."\n". |
464 |
|
"Update-Mode: Index\n". |
465 |
"Document-Type: XML\n\n$xml\n"; |
"Document-Type: XML\n\n$xml\n"; |
466 |
|
|
|
# do I/O |
|
|
$self->{'slice'}->{$s}->{'h'}->pump while length $self->{'slice'}->{$s}->{'in'} ; # wait for all input to go |
|
|
|
|
467 |
$self->slice_output($s); |
$self->slice_output($s); |
468 |
|
|
469 |
|
$self->_debug("dumping in slice $s: $path"); |
470 |
|
|
471 |
$self->{'paths'}->{$path} = ADDED; |
$self->{'paths'}->{$path} = ADDED; |
472 |
|
|
477 |
|
|
478 |
Prints to STDERR output and errors from C<swish-e>. |
Prints to STDERR output and errors from C<swish-e>. |
479 |
|
|
480 |
$i->slice_output($s); |
my $slice = $i->slice_output($s); |
481 |
|
|
482 |
Normally, you don't need to call it. |
Normally, you don't need to call it. |
483 |
|
|
484 |
|
B<This is dummy placeholder function for very old code that assumes this |
485 |
|
module is using C<IPC::Run> which it isn't any more.> |
486 |
|
|
487 |
=cut |
=cut |
488 |
|
|
489 |
sub slice_output { |
sub slice_output { |
492 |
my $s = shift || confess "slice_output needs slice"; |
my $s = shift || confess "slice_output needs slice"; |
493 |
|
|
494 |
confess "no slice $s" unless(exists($self->{'slice'}) && exists($self->{'slice'}->{$s})); |
confess "no slice $s" unless(exists($self->{'slice'}) && exists($self->{'slice'}->{$s})); |
|
confess "no 'in' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'in'})); |
|
|
confess "no 'out' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'out'})); |
|
495 |
|
|
496 |
if (length $self->{'slice'}->{$s}->{'out'} > $self->{'slice'}->{$s}->{'out_len'}) { |
# FIXME |
|
#print STDERR "swish-e OUT: ",$self->{'slice'}->{$s}->{'out'},"\n" if ($self->{'slice'}->{$s}->{'out'}); |
|
|
$self->{'slice'}->{$s}->{'out_len'} = length $self->{'slice'}->{$s}->{'out'}; |
|
|
return 1; |
|
|
} elsif (length $self->{'slice'}->{$s}->{'err'} > $self->{'slice'}->{$s}->{'err_len'}) { |
|
|
print STDERR "swish-e ERR: ",$self->{'slice'}->{$s}->{'err'},"\n" if ($self->{'slice'}->{$s}->{'err'}); |
|
|
$self->{'slice'}->{$s}->{'err_len'} = length $self->{'slice'}->{$s}->{'err'}; |
|
|
# this is fatal |
|
|
return 0; |
|
|
} |
|
497 |
|
|
498 |
return 1; |
return $s; |
499 |
} |
} |
500 |
|
|
501 |
=head2 close_slice |
=head2 close_slice |
517 |
confess "no 'h' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'h'})); |
confess "no 'h' in slice $s: ".Dumper($s) unless (exists($self->{'slice'}->{$s}->{'h'})); |
518 |
|
|
519 |
# pump rest of content (if any) |
# pump rest of content (if any) |
520 |
$self->{'slice'}->{$s}->{'h'}->pump while length $self->{'slice'}->{$s}->{'in'}; |
close $self->{'slice'}->{$s}->{'h'} || carp "can't close slice $s: $!"; |
521 |
|
|
522 |
$self->slice_output($s); |
$self->slice_output($s); |
523 |
|
|
524 |
# clean up |
undef $self->{'slice'}->{$s}->{'h'}; |
|
$self->{'slice'}->{$s}->{'h'}->finish or confess "finish on slice $s returned: $?"; |
|
525 |
|
|
526 |
delete($self->{'slice'}->{$s}) && return 1; |
delete($self->{'slice'}->{$s}) && return 1; |
527 |
return 0; |
return 0; |
536 |
|
|
537 |
This function is extracted from L<"add"> method so that you can L<Memoize> it. |
This function is extracted from L<"add"> method so that you can L<Memoize> it. |
538 |
If your data set has a lot of repeatable data, and memory is not a problem, you |
If your data set has a lot of repeatable data, and memory is not a problem, you |
539 |
can add C<memoize_to_xml> option to L<"open">. |
can add C<memoize_to_xml> option to L<"open_index">. |
540 |
|
|
541 |
=cut |
=cut |
542 |
|
|
561 |
$xml .= qq{</xml>}; |
$xml .= qq{</xml>}; |
562 |
} |
} |
563 |
|
|
564 |
|
sub _debug { |
565 |
|
my $self = shift; |
566 |
|
print STDERR "## ",@_,"\n" if ($self->{'debug'}); |
567 |
|
return; |
568 |
|
} |
569 |
|
|
570 |
1; |
1; |
571 |
__END__ |
__END__ |
572 |
|
|