lib/WebPAC/Input.pm

package WebPAC::Input;

use warnings;
use strict;

use blib;

use WebPAC::Common;
use base qw/WebPAC::Common/;
use Text::Iconv;
use Data::Dumper;

=head1 NAME

WebPAC::Input - read different file formats into WebPAC

=head1 VERSION

Version 0.09

=cut

our $VERSION = '0.09';

=head1 SYNOPSIS

This module implements input as database which have fixed and known
I<size> while indexing and single unique numeric identifier for database
position ranging from 1 to I<size>.

Simply, something that is indexed by unmber from 1 .. I<size>.

Examples of such databases are CDS/ISIS files, MARC files, lines in
text file, and so on.

Specific file formats are implemented using low-level interface modules,
located in C<WebPAC::Input::*> namespace which export C<open_db>,
C<fetch_rec> and optional C<init> functions.

Perhaps a little code snippet.

        use WebPAC::Input;

        my $db = WebPAC::Input->new(
                module => 'WebPAC::Input::ISIS',
                low_mem => 1,
        );

        $db->open( path => '/path/to/database' );
        print "database size: ",$db->size,"\n";
        while (my $rec = $db->fetch) {
                # do something with $rec
        }


=head1 FUNCTIONS

=head2 new

Create new input database object.

  my $db = new WebPAC::Input(
        module => 'WebPAC::Input::MARC',
        encoding => 'ISO-8859-2',
        low_mem => 1,
        recode => 'char pairs',
        no_progress_bar => 1,
  );

C<module> is low-level file format module. See L<WebPAC::Input::ISIS> and
L<WebPAC::Input::MARC>.

Optional parametar C<encoding> specify application code page (which will be
used internally). This should probably be your terminal encoding, and by
default, it C<ISO-8859-2>.

Default is not to use C<low_mem> options (see L<MEMORY USAGE> below).

C<recode> is optional string constisting of character or words pairs that
should be replaced in input stream.

C<no_progress_bar> disables progress bar output on C<STDOUT>

This function will also call low-level C<init> if it exists with same
parametars.

=cut

sub new {
        my $class = shift;
        my $self = {@_};
        bless($self, $class);

        my $log = $self->_get_logger;

        $log->logconfess("code_page argument is not suppored any more. change it to encoding") if ($self->{lookup});
        $log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_ref") if ($self->{lookup});

        $log->logconfess("specify low-level file format module") unless ($self->{module});
        my $module = $self->{module};
        $module =~ s#::#/#g;
        $module .= '.pm';
        $log->debug("require low-level module $self->{module} from $module");

        require $module;
        #eval $self->{module} .'->import';

        # check if required subclasses are implemented
        foreach my $subclass (qw/open_db fetch_rec init/) {
                my $n = $self->{module} . '::' . $subclass;
                if (! defined &{ $n }) {
                        my $missing = "missing $subclass in $self->{module}";
                        $self->{$subclass} = sub { $log->logwarn($missing) };
                } else {
                        $self->{$subclass} = \&{ $n };
                }
        }

        if ($self->{init}) {
                $log->debug("calling init");
                $self->{init}->($self, @_);
        }

        $self->{'encoding'} ||= 'ISO-8859-2';

        # running with low_mem flag? well, use DBM::Deep then.
        if ($self->{'low_mem'}) {
                $log->info("running with low_mem which impacts performance (<32 Mb memory usage)");

                my $db_file = "data.db";

                if (-e $db_file) {
                        unlink $db_file or $log->logdie("can't remove '$db_file' from last run");
                        $log->debug("removed '$db_file' from last run");
                }

                require DBM::Deep;

                my $db = new DBM::Deep $db_file;

                $log->logdie("DBM::Deep error: $!") unless ($db);

                if ($db->error()) {
                        $log->logdie("can't open '$db_file' under low_mem: ",$db->error());
                } else {
                        $log->debug("using file '$db_file' for DBM::Deep");
                }

                $self->{'db'} = $db;
        }

        $self ? return $self : return undef;
}

=head2 open

This function will read whole database in memory and produce lookups.

 $input->open(
        path => '/path/to/database/file',
        code_page => '852',
        limit => 500,
        offset => 6000,
        lookup => $lookup_obj,
        stats => 1,
        lookup_ref => sub {
                my ($k,$v) = @_;
                # store lookup $k => $v
        },
        modify_records => {
                900 => { '^a' => { ' : ' => '^b' } },
                901 => { '*' => { '^b' => ' ; ' } },
        },
 );

By default, C<code_page> is assumed to be C<852>.

C<offset> is optional parametar to position at some offset before reading from database.

C<limit> is optional parametar to read just C<limit> records from database

C<stats> create optional report about usage of fields and subfields

C<lookup_coderef> is closure to call when adding C<< key => 'value' >> combinations to
lookup.

C<modify_records> specify mapping from subfields to delimiters or from
delimiters to subfields, as well as oprations on fields (if subfield is
defined as C<*>.

Returns size of database, regardless of C<offset> and C<limit>
parametars, see also C<size>.

=cut

sub open {
        my $self = shift;
        my $arg = {@_};

        my $log = $self->_get_logger();

        $log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_coderef") if ($arg->{lookup});
        $log->logconfess("lookup_coderef must be CODE, not ",ref($arg->{lookup_coderef}))
                if ($arg->{lookup_coderef} && ref($arg->{lookup_coderef}) ne 'CODE');

        $log->logcroak("need path") if (! $arg->{'path'});
        my $code_page = $arg->{'code_page'} || '852';

        # store data in object
        $self->{'input_code_page'} = $code_page;
        foreach my $v (qw/path offset limit/) {
                $self->{$v} = $arg->{$v} if ($arg->{$v});
        }

        # create Text::Iconv object
        $self->{iconv} = Text::Iconv->new($code_page,$self->{'encoding'});      ## FIXME remove!

        my $filter_ref;
        my $recode_regex;
        my $recode_map;

        if ($self->{recode}) {
                my @r = split(/\s/, $self->{recode});
                if ($#r % 2 != 1) {
                        $log->logwarn("recode needs even number of elements (some number of valid pairs)");
                } else {
                        while (@r) {
                                my $from = shift @r;
                                my $to = shift @r;
                                $recode_map->{$from} = $to;
                        }

                        $recode_regex = join '|' => keys %{ $recode_map };

                        $log->debug("using recode regex: $recode_regex");
                }

        }

        my $rec_regex = $self->modify_record_regexps(%{ $arg->{modify_records} });
        $log->debug("rec_regex: ", Dumper($rec_regex));

        my ($db, $size) = $self->{open_db}->( $self, 
                path => $arg->{path},
                filter => sub {
                                my ($l,$f_nr) = @_;
                                return unless defined($l);

                                ## FIXME remove iconv!
                                $l = $self->{iconv}->convert($l) if ($self->{iconv});
        
                                $l =~ s/($recode_regex)/$recode_map->{$1}/g if ($recode_regex && $recode_map);

                                ## FIXME remove this warning when we are sure that none of API is calling
                                ## this wrongly
                                #warn "filter called without field number" unless ($f_nr);

                                return $l unless ($rec_regex && $f_nr);

                                # apply regexps
                                if ($rec_regex && defined($rec_regex->{$f_nr})) {
                                        $log->logconfess("regexps->{$f_nr} must be ARRAY") if (ref($rec_regex->{$f_nr}) ne 'ARRAY');
                                        my $c = 0;
                                        foreach my $r (@{ $rec_regex->{$f_nr} }) {
                                                while ( eval '$l =~ ' . $r ) { $c++ };
                                        }
                                        warn "## field $f_nr triggered $c regexpes\n" if ($c && $self->{debug});
                                }

                                return $l;
                },
                %{ $arg },
        );

        unless (defined($db)) {
                $log->logwarn("can't open database $arg->{path}, skipping...");
                return;
        }

        unless ($size) {
                $log->logwarn("no records in database $arg->{path}, skipping...");
                return;
        }

        my $from_rec = 1;
        my $to_rec = $size;

        if (my $s = $self->{offset}) {
                $log->debug("skipping to MFN $s");
                $from_rec = $s;
        } else {
                $self->{offset} = $from_rec;
        }

        if ($self->{limit}) {
                $log->debug("limiting to ",$self->{limit}," records");
                $to_rec = $from_rec + $self->{limit} - 1;
                $to_rec = $size if ($to_rec > $size);
        }

        # store size for later
        $self->{size} = ($to_rec - $from_rec) ? ($to_rec - $from_rec + 1) : 0;

        $log->info("processing $self->{size}/$size records [$from_rec-$to_rec] convert $code_page -> $self->{encoding}", $self->{stats} ? ' [stats]' : '');

        # read database
        for (my $pos = $from_rec; $pos <= $to_rec; $pos++) {

                $log->debug("position: $pos\n");

                my $rec = $self->{fetch_rec}->($self, $db, $pos );

                $log->debug(sub { Dumper($rec) });

                if (! $rec) {
                        $log->warn("record $pos empty? skipping...");
                        next;
                }

                # store
                if ($self->{low_mem}) {
                        $self->{db}->put($pos, $rec);
                } else {
                        $self->{data}->{$pos} = $rec;
                }

                # create lookup
                $arg->{'lookup_coderef'}->( $rec ) if ($rec && $arg->{'lookup_coderef'});

                # update counters for statistics
                if ($self->{stats}) {

                        foreach my $fld (keys %{ $rec }) {
                                $self->{_stats}->{fld}->{ $fld }++;

                                $log->logdie("invalid record fild $fld, not ARRAY")
                                        unless (ref($rec->{ $fld }) eq 'ARRAY');
        
                                foreach my $row (@{ $rec->{$fld} }) {

                                        if (ref($row) eq 'HASH') {

                                                foreach my $sf (keys %{ $row }) {
                                                        $self->{_stats}->{sf}->{ $fld }->{ $sf }->{count}++;
                                                        $self->{_stats}->{sf}->{ $fld }->{ $sf }->{repeatable}++
                                                                        if (ref($row->{$sf}) eq 'ARRAY');
                                                }

                                        } else {
                                                $self->{_stats}->{repeatable}->{ $fld }++;
                                        }
                                }
                        }
                }

                $self->progress_bar($pos,$to_rec) unless ($self->{no_progress_bar});

        }

        $self->{pos} = -1;
        $self->{last_pcnt} = 0;

        # store max mfn and return it.
        $self->{max_pos} = $to_rec;
        $log->debug("max_pos: $to_rec");

        return $size;
}

=head2 fetch

Fetch next record from database. It will also displays progress bar.

 my $rec = $isis->fetch;

Record from this function should probably go to C<data_structure> for
normalisation.

=cut

sub fetch {
        my $self = shift;

        my $log = $self->_get_logger();

        $log->logconfess("it seems that you didn't load database!") unless ($self->{pos});

        if ($self->{pos} == -1) {
                $self->{pos} = $self->{offset};
        } else {
                $self->{pos}++;
        }

        my $mfn = $self->{pos};

        if ($mfn > $self->{max_pos}) {
                $self->{pos} = $self->{max_pos};
                $log->debug("at EOF");
                return;
        }

        $self->progress_bar($mfn,$self->{max_pos}) unless ($self->{no_progress_bar});

        my $rec;

        if ($self->{low_mem}) {
                $rec = $self->{db}->get($mfn);
        } else {
                $rec = $self->{data}->{$mfn};
        }

        $rec ||= 0E0;
}

=head2 pos

Returns current record number (MFN).

 print $isis->pos;

First record in database has position 1.

=cut

sub pos {
        my $self = shift;
        return $self->{pos};
}


=head2 size

Returns number of records in database

 print $isis->size;

Result from this function can be used to loop through all records

 foreach my $mfn ( 1 ... $isis->size ) { ... }

because it takes into account C<offset> and C<limit>.

=cut

sub size {
        my $self = shift;
        return $self->{size};
}

=head2 seek

Seek to specified MFN in file.

 $isis->seek(42);

First record in database has position 1.

=cut

sub seek {
        my $self = shift;
        my $pos = shift || return;

        my $log = $self->_get_logger();

        if ($pos < 1) {
                $log->warn("seek before first record");
                $pos = 1;
        } elsif ($pos > $self->{max_pos}) {
                $log->warn("seek beyond last record");
                $pos = $self->{max_pos};
        }

        return $self->{pos} = (($pos - 1) || -1);
}

=head2 stats

Dump statistics about field and subfield usage

  print $input->stats;

=cut

sub stats {
        my $self = shift;

        my $log = $self->_get_logger();

        my $s = $self->{_stats};
        if (! $s) {
                $log->warn("called stats, but there is no statistics collected");
                return;
        }

        my $max_fld = 0;

        my $out = join("\n",
                map {
                        my $f = $_ || die "no field";
                        my $v = $s->{fld}->{$f} || die "no s->{fld}->{$f}";
                        $max_fld = $v if ($v > $max_fld);

                        my $o = sprintf("%4s %d ~", $f, $v);

                        if (defined($s->{sf}->{$f})) {
                                map {
                                        $o .= sprintf(" %s:%d%s", $_, 
                                                $s->{sf}->{$f}->{$_}->{count},
                                                $s->{sf}->{$f}->{$_}->{repeatable} ? '*' : '',
                                        );
                                } sort keys %{ $s->{sf}->{$f} };
                        }

                        if (my $v_r = $s->{repeatable}->{$f}) {
                                $o .= " ($v_r)" if ($v_r != $v);
                        }

                        $o;
                } sort { $a cmp $b } keys %{ $s->{fld} }
        );

        $log->debug( sub { Dumper($s) } );

        return $out;
}

=head2 modify_record_regexps

Generate hash with regexpes to be applied using L<filter>.

  my $regexpes = $input->modify_record_regexps(
                900 => { '^a' => { ' : ' => '^b' } },
                901 => { '*' => { '^b' => ' ; ' } },
  );

=cut

sub modify_record_regexps {
        my $self = shift;
        my $modify_record = {@_};

        my $regexpes;

        foreach my $f (keys %$modify_record) {
warn "--- f: $f\n";
                foreach my $sf (keys %{ $modify_record->{$f} }) {
warn "---- sf: $sf\n";
                        foreach my $from (keys %{ $modify_record->{$f}->{$sf} }) {
                                my $to = $modify_record->{$f}->{$sf}->{$from};
                                #die "no field?" unless defined($to);
warn "----- transform: |$from| -> |$to|\n";

                                if ($sf =~ /^\^/) {
                                        my $regex = 
                                                's/\Q'. $sf .'\E([^\^]+)\Q'. $from .'\E([^\^]+)/'. $sf .'$1'. $to .'$2/g';
                                        push @{ $regexpes->{$f} }, $regex;
warn ">>>>> $regex [sf]\n";
                                } else {
                                        my $regex =
                                                's/\Q'. $from .'\E/'. $to .'/g';
                                        push @{ $regexpes->{$f} }, $regex;
warn ">>>>> $regex [global]\n";
                                }

                        }
                }
        }

        return $regexpes;
}

=head1 MEMORY USAGE

C<low_mem> options is double-edged sword. If enabled, WebPAC
will run on memory constraint machines (which doesn't have enough
physical RAM to create memory structure for whole source database).

If your machine has 512Mb or more of RAM and database is around 10000 records,
memory shouldn't be an issue. If you don't have enough physical RAM, you
might consider using virtual memory (if your operating system is handling it
well, like on FreeBSD or Linux) instead of dropping to L<DBM::Deep> to handle
parsed structure of ISIS database (this is what C<low_mem> option does).

Hitting swap at end of reading source database is probably o.k. However,
hitting swap before 90% will dramatically decrease performance and you will
be better off with C<low_mem> and using rest of availble memory for
operating system disk cache (Linux is particuallary good about this).
However, every access to database record will require disk access, so
generation phase will be slower 10-100 times.

Parsed structures are essential - you just have option to trade RAM memory
(which is fast) for disk space (which is slow). Be sure to have planty of
disk space if you are using C<low_mem> and thus L<DBM::Deep>.

However, when WebPAC is running on desktop machines (or laptops :-), it's
highly undesireable for system to start swapping. Using C<low_mem> option can
reduce WecPAC memory usage to around 64Mb for same database with lookup
fields and sorted indexes which stay in RAM. Performance will suffer, but
memory usage will really be minimal. It might be also more confortable to
run WebPAC reniced on those machines.


=head1 AUTHOR

Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>

=head1 COPYRIGHT & LICENSE

Copyright 2005-2006 Dobrica Pavlinusic, All Rights Reserved.

This program is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.

=cut

1; # End of WebPAC::Input
1	dpavlin	1	package WebPAC::Input;
2
3			use warnings;
4			use strict;
5
6	dpavlin	507	use blib;
7
8	dpavlin	487	use WebPAC::Common;
9	dpavlin	285	use base qw/WebPAC::Common/;
10			use Text::Iconv;
11	dpavlin	308	use Data::Dumper;
12	dpavlin	285
13	dpavlin	1	=head1 NAME
14
15	dpavlin	286	WebPAC::Input - read different file formats into WebPAC
16	dpavlin	1
17			=head1 VERSION
18
19	dpavlin	599	Version 0.09
20	dpavlin	1
21			=cut
22
23	dpavlin	599	our $VERSION = '0.09';
24	dpavlin	1
25			=head1 SYNOPSIS
26
27	dpavlin	286	This module implements input as database which have fixed and known
28			I<size> while indexing and single unique numeric identifier for database
29			position ranging from 1 to I<size>.
30	dpavlin	1
31	dpavlin	286	Simply, something that is indexed by unmber from 1 .. I<size>.
32
33			Examples of such databases are CDS/ISIS files, MARC files, lines in
34			text file, and so on.
35
36			Specific file formats are implemented using low-level interface modules,
37			located in C<WebPAC::Input::*> namespace which export C<open_db>,
38			C<fetch_rec> and optional C<init> functions.
39
40	dpavlin	1	Perhaps a little code snippet.
41
42	dpavlin	597	use WebPAC::Input;
43	dpavlin	1
44	dpavlin	597	my $db = WebPAC::Input->new(
45			module => 'WebPAC::Input::ISIS',
46	dpavlin	286	low_mem => 1,
47	dpavlin	597	);
48	dpavlin	1
49	dpavlin	597	$db->open( path => '/path/to/database' );
50	dpavlin	416	print "database size: ",$db->size,"\n";
51			while (my $rec = $db->fetch) {
52			# do something with $rec
53			}
54	dpavlin	1
55	dpavlin	286
56
57	dpavlin	1	=head1 FUNCTIONS
58
59	dpavlin	3	=head2 new
60	dpavlin	1
61	dpavlin	3	Create new input database object.
62
63	dpavlin	9	my $db = new WebPAC::Input(
64	dpavlin	286	module => 'WebPAC::Input::MARC',
65	dpavlin	585	encoding => 'ISO-8859-2',
66	dpavlin	10	low_mem => 1,
67	dpavlin	416	recode => 'char pairs',
68	dpavlin	483	no_progress_bar => 1,
69	dpavlin	9	);
70	dpavlin	3
71	dpavlin	597	C<module> is low-level file format module. See L<WebPAC::Input::ISIS> and
72	dpavlin	286	L<WebPAC::Input::MARC>.
73
74	dpavlin	585	Optional parametar C<encoding> specify application code page (which will be
75	dpavlin	9	used internally). This should probably be your terminal encoding, and by
76			default, it C<ISO-8859-2>.
77
78	dpavlin	10	Default is not to use C<low_mem> options (see L<MEMORY USAGE> below).
79
80	dpavlin	483	C<recode> is optional string constisting of character or words pairs that
81			should be replaced in input stream.
82
83			C<no_progress_bar> disables progress bar output on C<STDOUT>
84
85	dpavlin	285	This function will also call low-level C<init> if it exists with same
86			parametars.
87
88	dpavlin	1	=cut
89
90	dpavlin	3	sub new {
91	dpavlin	285	my $class = shift;
92			my $self = {@_};
93	dpavlin	3	bless($self, $class);
94
95	dpavlin	285	my $log = $self->_get_logger;
96
97	dpavlin	585	$log->logconfess("code_page argument is not suppored any more. change it to encoding") if ($self->{lookup});
98			$log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_ref") if ($self->{lookup});
99
100	dpavlin	286	$log->logconfess("specify low-level file format module") unless ($self->{module});
101			my $module = $self->{module};
102			$module =~ s#::#/#g;
103			$module .= '.pm';
104			$log->debug("require low-level module $self->{module} from $module");
105	dpavlin	289
106	dpavlin	286	require $module;
107	dpavlin	289	#eval $self->{module} .'->import';
108	dpavlin	286
109	dpavlin	285	# check if required subclasses are implemented
110	dpavlin	289	foreach my $subclass (qw/open_db fetch_rec init/) {
111			my $n = $self->{module} . '::' . $subclass;
112			if (! defined &{ $n }) {
113	dpavlin	290	my $missing = "missing $subclass in $self->{module}";
114	dpavlin	301	$self->{$subclass} = sub { $log->logwarn($missing) };
115	dpavlin	286	} else {
116	dpavlin	289	$self->{$subclass} = \&{ $n };
117	dpavlin	286	}
118	dpavlin	285	}
119
120	dpavlin	289	if ($self->{init}) {
121	dpavlin	285	$log->debug("calling init");
122	dpavlin	289	$self->{init}->($self, @_);
123	dpavlin	285	}
124
125	dpavlin	585	$self->{'encoding'} \|\|= 'ISO-8859-2';
126	dpavlin	9
127	dpavlin	10	# running with low_mem flag? well, use DBM::Deep then.
128			if ($self->{'low_mem'}) {
129			$log->info("running with low_mem which impacts performance (<32 Mb memory usage)");
130
131			my $db_file = "data.db";
132
133			if (-e $db_file) {
134			unlink $db_file or $log->logdie("can't remove '$db_file' from last run");
135			$log->debug("removed '$db_file' from last run");
136			}
137
138			require DBM::Deep;
139
140			my $db = new DBM::Deep $db_file;
141
142			$log->logdie("DBM::Deep error: $!") unless ($db);
143
144			if ($db->error()) {
145			$log->logdie("can't open '$db_file' under low_mem: ",$db->error());
146			} else {
147			$log->debug("using file '$db_file' for DBM::Deep");
148			}
149
150			$self->{'db'} = $db;
151			}
152
153	dpavlin	3	$self ? return $self : return undef;
154	dpavlin	1	}
155
156	dpavlin	285	=head2 open
157
158			This function will read whole database in memory and produce lookups.
159
160	dpavlin	286	$input->open(
161	dpavlin	285	path => '/path/to/database/file',
162			code_page => '852',
163	dpavlin	286	limit => 500,
164			offset => 6000,
165	dpavlin	285	lookup => $lookup_obj,
166	dpavlin	506	stats => 1,
167	dpavlin	585	lookup_ref => sub {
168			my ($k,$v) = @_;
169			# store lookup $k => $v
170			},
171	dpavlin	597	modify_records => {
172			900 => { '^a' => { ' : ' => '^b' } },
173			901 => { '*' => { '^b' => ' ; ' } },
174			},
175	dpavlin	285	);
176
177			By default, C<code_page> is assumed to be C<852>.
178
179	dpavlin	286	C<offset> is optional parametar to position at some offset before reading from database.
180	dpavlin	285
181	dpavlin	286	C<limit> is optional parametar to read just C<limit> records from database
182	dpavlin	285
183	dpavlin	506	C<stats> create optional report about usage of fields and subfields
184
185	dpavlin	597	C<lookup_coderef> is closure to call when adding C<< key => 'value' >> combinations to
186	dpavlin	585	lookup.
187
188	dpavlin	597	C<modify_records> specify mapping from subfields to delimiters or from
189			delimiters to subfields, as well as oprations on fields (if subfield is
190			defined as C<*>.
191
192	dpavlin	286	Returns size of database, regardless of C<offset> and C<limit>
193			parametars, see also C<size>.
194	dpavlin	285
195			=cut
196
197			sub open {
198			my $self = shift;
199			my $arg = {@_};
200
201			my $log = $self->_get_logger();
202
203	dpavlin	585	$log->logconfess("lookup argument is not suppored any more. rewrite call to lookup_coderef") if ($arg->{lookup});
204			$log->logconfess("lookup_coderef must be CODE, not ",ref($arg->{lookup_coderef}))
205			if ($arg->{lookup_coderef} && ref($arg->{lookup_coderef}) ne 'CODE');
206
207	dpavlin	285	$log->logcroak("need path") if (! $arg->{'path'});
208			my $code_page = $arg->{'code_page'} \|\| '852';
209
210			# store data in object
211	dpavlin	292	$self->{'input_code_page'} = $code_page;
212	dpavlin	286	foreach my $v (qw/path offset limit/) {
213	dpavlin	285	$self->{$v} = $arg->{$v} if ($arg->{$v});
214			}
215
216			# create Text::Iconv object
217	dpavlin	585	$self->{iconv} = Text::Iconv->new($code_page,$self->{'encoding'}); ## FIXME remove!
218	dpavlin	285
219	dpavlin	416	my $filter_ref;
220	dpavlin	597	my $recode_regex;
221			my $recode_map;
222	dpavlin	416
223			if ($self->{recode}) {
224			my @r = split(/\s/, $self->{recode});
225			if ($#r % 2 != 1) {
226			$log->logwarn("recode needs even number of elements (some number of valid pairs)");
227			} else {
228			while (@r) {
229			my $from = shift @r;
230			my $to = shift @r;
231	dpavlin	597	$recode_map->{$from} = $to;
232	dpavlin	416	}
233
234	dpavlin	597	$recode_regex = join '\|' => keys %{ $recode_map };
235	dpavlin	416
236	dpavlin	597	$log->debug("using recode regex: $recode_regex");
237	dpavlin	416	}
238
239			}
240
241	dpavlin	597	my $rec_regex = $self->modify_record_regexps(%{ $arg->{modify_records} });
242			$log->debug("rec_regex: ", Dumper($rec_regex));
243
244	dpavlin	289	my ($db, $size) = $self->{open_db}->( $self,
245	dpavlin	285	path => $arg->{path},
246	dpavlin	597	filter => sub {
247			my ($l,$f_nr) = @_;
248			return unless defined($l);
249
250			## FIXME remove iconv!
251			$l = $self->{iconv}->convert($l) if ($self->{iconv});
252
253			$l =~ s/($recode_regex)/$recode_map->{$1}/g if ($recode_regex && $recode_map);
254
255	dpavlin	599	## FIXME remove this warning when we are sure that none of API is calling
256			## this wrongly
257	dpavlin	606	#warn "filter called without field number" unless ($f_nr);
258	dpavlin	597
259	dpavlin	599	return $l unless ($rec_regex && $f_nr);
260
261	dpavlin	597	# apply regexps
262			if ($rec_regex && defined($rec_regex->{$f_nr})) {
263			$log->logconfess("regexps->{$f_nr} must be ARRAY") if (ref($rec_regex->{$f_nr}) ne 'ARRAY');
264			my $c = 0;
265			foreach my $r (@{ $rec_regex->{$f_nr} }) {
266			while ( eval '$l =~ ' . $r ) { $c++ };
267			}
268			warn "## field $f_nr triggered $c regexpes\n" if ($c && $self->{debug});
269			}
270
271			return $l;
272			},
273	dpavlin	523	%{ $arg },
274	dpavlin	285	);
275
276	dpavlin	496	unless (defined($db)) {
277	dpavlin	285	$log->logwarn("can't open database $arg->{path}, skipping...");
278			return;
279			}
280
281			unless ($size) {
282			$log->logwarn("no records in database $arg->{path}, skipping...");
283			return;
284			}
285
286	dpavlin	339	my $from_rec = 1;
287			my $to_rec = $size;
288	dpavlin	285
289	dpavlin	286	if (my $s = $self->{offset}) {
290	dpavlin	513	$log->debug("skipping to MFN $s");
291	dpavlin	339	$from_rec = $s;
292	dpavlin	285	} else {
293	dpavlin	339	$self->{offset} = $from_rec;
294	dpavlin	285	}
295
296	dpavlin	286	if ($self->{limit}) {
297	dpavlin	301	$log->debug("limiting to ",$self->{limit}," records");
298	dpavlin	339	$to_rec = $from_rec + $self->{limit} - 1;
299			$to_rec = $size if ($to_rec > $size);
300	dpavlin	285	}
301
302			# store size for later
303	dpavlin	339	$self->{size} = ($to_rec - $from_rec) ? ($to_rec - $from_rec + 1) : 0;
304	dpavlin	285
305	dpavlin	585	$log->info("processing $self->{size}/$size records [$from_rec-$to_rec] convert $code_page -> $self->{encoding}", $self->{stats} ? ' [stats]' : '');
306	dpavlin	285
307			# read database
308	dpavlin	339	for (my $pos = $from_rec; $pos <= $to_rec; $pos++) {
309	dpavlin	285
310	dpavlin	286	$log->debug("position: $pos\n");
311	dpavlin	285
312	dpavlin	289	my $rec = $self->{fetch_rec}->($self, $db, $pos );
313	dpavlin	285
314	dpavlin	308	$log->debug(sub { Dumper($rec) });
315
316	dpavlin	285	if (! $rec) {
317	dpavlin	286	$log->warn("record $pos empty? skipping...");
318	dpavlin	285	next;
319			}
320
321			# store
322	dpavlin	286	if ($self->{low_mem}) {
323			$self->{db}->put($pos, $rec);
324	dpavlin	285	} else {
325	dpavlin	286	$self->{data}->{$pos} = $rec;
326	dpavlin	285	}
327
328			# create lookup
329	dpavlin	585	$arg->{'lookup_coderef'}->( $rec ) if ($rec && $arg->{'lookup_coderef'});
330	dpavlin	285
331	dpavlin	506	# update counters for statistics
332			if ($self->{stats}) {
333	dpavlin	593
334			foreach my $fld (keys %{ $rec }) {
335	dpavlin	506	$self->{_stats}->{fld}->{ $fld }++;
336	dpavlin	593
337			$log->logdie("invalid record fild $fld, not ARRAY")
338			unless (ref($rec->{ $fld }) eq 'ARRAY');
339
340			foreach my $row (@{ $rec->{$fld} }) {
341
342			if (ref($row) eq 'HASH') {
343
344			foreach my $sf (keys %{ $row }) {
345			$self->{_stats}->{sf}->{ $fld }->{ $sf }->{count}++;
346			$self->{_stats}->{sf}->{ $fld }->{ $sf }->{repeatable}++
347			if (ref($row->{$sf}) eq 'ARRAY');
348	dpavlin	506	}
349	dpavlin	593
350			} else {
351			$self->{_stats}->{repeatable}->{ $fld }++;
352			}
353	dpavlin	506	}
354	dpavlin	593	}
355	dpavlin	506	}
356
357	dpavlin	483	$self->progress_bar($pos,$to_rec) unless ($self->{no_progress_bar});
358	dpavlin	285
359			}
360
361	dpavlin	286	$self->{pos} = -1;
362			$self->{last_pcnt} = 0;
363	dpavlin	285
364			# store max mfn and return it.
365	dpavlin	339	$self->{max_pos} = $to_rec;
366			$log->debug("max_pos: $to_rec");
367	dpavlin	285
368			return $size;
369			}
370
371			=head2 fetch
372
373			Fetch next record from database. It will also displays progress bar.
374
375			my $rec = $isis->fetch;
376
377			Record from this function should probably go to C<data_structure> for
378			normalisation.
379
380			=cut
381
382			sub fetch {
383			my $self = shift;
384
385			my $log = $self->_get_logger();
386
387	dpavlin	286	$log->logconfess("it seems that you didn't load database!") unless ($self->{pos});
388	dpavlin	285
389	dpavlin	286	if ($self->{pos} == -1) {
390			$self->{pos} = $self->{offset};
391	dpavlin	285	} else {
392	dpavlin	286	$self->{pos}++;
393	dpavlin	285	}
394
395	dpavlin	286	my $mfn = $self->{pos};
396	dpavlin	285
397	dpavlin	286	if ($mfn > $self->{max_pos}) {
398			$self->{pos} = $self->{max_pos};
399	dpavlin	285	$log->debug("at EOF");
400			return;
401			}
402
403	dpavlin	483	$self->progress_bar($mfn,$self->{max_pos}) unless ($self->{no_progress_bar});
404	dpavlin	285
405			my $rec;
406
407	dpavlin	286	if ($self->{low_mem}) {
408			$rec = $self->{db}->get($mfn);
409	dpavlin	285	} else {
410	dpavlin	286	$rec = $self->{data}->{$mfn};
411	dpavlin	285	}
412
413			$rec \|\|= 0E0;
414			}
415
416			=head2 pos
417
418			Returns current record number (MFN).
419
420			print $isis->pos;
421
422			First record in database has position 1.
423
424			=cut
425
426			sub pos {
427			my $self = shift;
428	dpavlin	286	return $self->{pos};
429	dpavlin	285	}
430
431
432			=head2 size
433
434			Returns number of records in database
435
436			print $isis->size;
437
438			Result from this function can be used to loop through all records
439
440			foreach my $mfn ( 1 ... $isis->size ) { ... }
441
442	dpavlin	286	because it takes into account C<offset> and C<limit>.
443	dpavlin	285
444			=cut
445
446			sub size {
447			my $self = shift;
448	dpavlin	286	return $self->{size};
449	dpavlin	285	}
450
451			=head2 seek
452
453			Seek to specified MFN in file.
454
455			$isis->seek(42);
456
457			First record in database has position 1.
458
459			=cut
460
461			sub seek {
462			my $self = shift;
463			my $pos = shift \|\| return;
464
465			my $log = $self->_get_logger();
466
467			if ($pos < 1) {
468			$log->warn("seek before first record");
469			$pos = 1;
470	dpavlin	286	} elsif ($pos > $self->{max_pos}) {
471	dpavlin	285	$log->warn("seek beyond last record");
472	dpavlin	286	$pos = $self->{max_pos};
473	dpavlin	285	}
474
475	dpavlin	286	return $self->{pos} = (($pos - 1) \|\| -1);
476	dpavlin	285	}
477
478	dpavlin	506	=head2 stats
479	dpavlin	285
480	dpavlin	506	Dump statistics about field and subfield usage
481
482	dpavlin	507	print $input->stats;
483	dpavlin	506
484			=cut
485
486			sub stats {
487			my $self = shift;
488	dpavlin	507
489			my $log = $self->_get_logger();
490
491			my $s = $self->{_stats};
492			if (! $s) {
493			$log->warn("called stats, but there is no statistics collected");
494			return;
495			}
496
497			my $max_fld = 0;
498
499			my $out = join("\n",
500			map {
501			my $f = $_ \|\| die "no field";
502			my $v = $s->{fld}->{$f} \|\| die "no s->{fld}->{$f}";
503			$max_fld = $v if ($v > $max_fld);
504
505	dpavlin	519	my $o = sprintf("%4s %d ~", $f, $v);
506	dpavlin	507
507			if (defined($s->{sf}->{$f})) {
508			map {
509	dpavlin	593	$o .= sprintf(" %s:%d%s", $_,
510			$s->{sf}->{$f}->{$_}->{count},
511			$s->{sf}->{$f}->{$_}->{repeatable} ? '*' : '',
512			);
513	dpavlin	507	} sort keys %{ $s->{sf}->{$f} };
514			}
515
516			if (my $v_r = $s->{repeatable}->{$f}) {
517			$o .= " ($v_r)" if ($v_r != $v);
518			}
519
520			$o;
521	dpavlin	519	} sort { $a cmp $b } keys %{ $s->{fld} }
522	dpavlin	507	);
523
524			$log->debug( sub { Dumper($s) } );
525
526			return $out;
527	dpavlin	506	}
528
529	dpavlin	598	=head2 modify_record_regexps
530	dpavlin	597
531			Generate hash with regexpes to be applied using L<filter>.
532
533			my $regexpes = $input->modify_record_regexps(
534			900 => { '^a' => { ' : ' => '^b' } },
535			901 => { '*' => { '^b' => ' ; ' } },
536			);
537
538			=cut
539
540			sub modify_record_regexps {
541			my $self = shift;
542			my $modify_record = {@_};
543
544			my $regexpes;
545
546			foreach my $f (keys %$modify_record) {
547			warn "--- f: $f\n";
548			foreach my $sf (keys %{ $modify_record->{$f} }) {
549			warn "---- sf: $sf\n";
550			foreach my $from (keys %{ $modify_record->{$f}->{$sf} }) {
551			my $to = $modify_record->{$f}->{$sf}->{$from};
552			#die "no field?" unless defined($to);
553			warn "----- transform: \|$from\| -> \|$to\|\n";
554
555			if ($sf =~ /^\^/) {
556			my $regex =
557			's/\Q'. $sf .'\E([^\^]+)\Q'. $from .'\E([^\^]+)/'. $sf .'$1'. $to .'$2/g';
558			push @{ $regexpes->{$f} }, $regex;
559			warn ">>>>> $regex [sf]\n";
560			} else {
561			my $regex =
562			's/\Q'. $from .'\E/'. $to .'/g';
563			push @{ $regexpes->{$f} }, $regex;
564			warn ">>>>> $regex [global]\n";
565			}
566
567			}
568			}
569			}
570
571			return $regexpes;
572			}
573
574	dpavlin	3	=head1 MEMORY USAGE
575	dpavlin	1
576	dpavlin	3	C<low_mem> options is double-edged sword. If enabled, WebPAC
577			will run on memory constraint machines (which doesn't have enough
578			physical RAM to create memory structure for whole source database).
579	dpavlin	1
580	dpavlin	3	If your machine has 512Mb or more of RAM and database is around 10000 records,
581			memory shouldn't be an issue. If you don't have enough physical RAM, you
582			might consider using virtual memory (if your operating system is handling it
583			well, like on FreeBSD or Linux) instead of dropping to L<DBM::Deep> to handle
584			parsed structure of ISIS database (this is what C<low_mem> option does).
585	dpavlin	1
586	dpavlin	3	Hitting swap at end of reading source database is probably o.k. However,
587			hitting swap before 90% will dramatically decrease performance and you will
588			be better off with C<low_mem> and using rest of availble memory for
589			operating system disk cache (Linux is particuallary good about this).
590			However, every access to database record will require disk access, so
591			generation phase will be slower 10-100 times.
592
593			Parsed structures are essential - you just have option to trade RAM memory
594			(which is fast) for disk space (which is slow). Be sure to have planty of
595			disk space if you are using C<low_mem> and thus L<DBM::Deep>.
596
597			However, when WebPAC is running on desktop machines (or laptops :-), it's
598			highly undesireable for system to start swapping. Using C<low_mem> option can
599			reduce WecPAC memory usage to around 64Mb for same database with lookup
600			fields and sorted indexes which stay in RAM. Performance will suffer, but
601			memory usage will really be minimal. It might be also more confortable to
602			run WebPAC reniced on those machines.
603
604
605			=head1 AUTHOR
606
607			Dobrica Pavlinusic, C<< <dpavlin@rot13.org> >>
608
609	dpavlin	1	=head1 COPYRIGHT & LICENSE
610
611	dpavlin	599	Copyright 2005-2006 Dobrica Pavlinusic, All Rights Reserved.
612	dpavlin	1
613			This program is free software; you can redistribute it and/or modify it
614			under the same terms as Perl itself.
615
616			=cut
617
618			1; # End of WebPAC::Input