/[webpac]/branches/ffzg/all2xml.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /branches/ffzg/all2xml.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 81 - (hide annotations)
Tue Jul 8 22:13:56 2003 UTC (20 years, 8 months ago) by dpavlin
Original Path: trunk/all2xml.pl
File MIME type: text/plain
File size: 13117 byte(s)
the great rename: isis2xml.* -> all2xml.*

1 dpavlin 1 #!/usr/bin/perl -w
2    
3     use strict;
4     use OpenIsis;
5     use Getopt::Std;
6     use Data::Dumper;
7     use XML::Simple;
8 dpavlin 5 use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
9 dpavlin 10 use Text::Iconv;
10 dpavlin 13 use Config::IniFiles;
11 dpavlin 40 use Encode;
12 dpavlin 1
13 dpavlin 10 $|=1;
14 dpavlin 9
15 dpavlin 13 my $config_file = $0;
16     $config_file =~ s/\.pl$/.conf/;
17     die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
18    
19 dpavlin 10 my $config;
20    
21 dpavlin 58 #use index_DBI; # default DBI module for index
22     use index_DBI_cache; # faster DBI module using memory cache
23 dpavlin 50 my $index;
24 dpavlin 10
25 dpavlin 1 my %opts;
26    
27 dpavlin 7 # usage:
28     # -d directory name
29     # -m multiple directories
30     # -q quiet
31     # -s run swish
32 dpavlin 1
33 dpavlin 7 getopts('d:m:qs', \%opts);
34    
35 dpavlin 43 my $path; # this is name of database
36 dpavlin 1
37 dpavlin 57 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
38 dpavlin 1
39 dpavlin 40 # this is encoding of all files on disk, including import_xml/*.xml file and
40     # filter/*.pm files! It will be used to store strings in perl internally!
41     my $codepage = 'ISO-8859-2';
42 dpavlin 1
43 dpavlin 40 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
44     # this function will convert data from XML files to local encoding
45     sub x {
46     return $utf2cp->convert($_[0]);
47     }
48 dpavlin 3
49 dpavlin 54 # decode isis/excel or other import codepage
50     my $import2cp;
51 dpavlin 10
52 dpavlin 40 # outgoing xml must be in UTF-8
53     my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
54 dpavlin 29
55 dpavlin 54 # mapping between data type and tag which specify
56     # format in XML file
57     my %type2tag = (
58     'isis' => 'isis',
59 dpavlin 62 'excel' => 'column',
60     'marc' => 'marc',
61 dpavlin 67 'feed' => 'feed'
62 dpavlin 54 );
63 dpavlin 3
64 dpavlin 54 sub data2xml {
65    
66 dpavlin 10 use xmlify;
67    
68 dpavlin 54 my $type = shift @_;
69 dpavlin 3 my $row = shift @_;
70 dpavlin 13 my $add_xml = shift @_;
71 dpavlin 59 # needed to read values from configuration file
72     my $cfg = shift @_;
73     my $database = shift @_;
74 dpavlin 3
75     my $xml;
76    
77 dpavlin 10 use parse_format;
78 dpavlin 3
79 dpavlin 13 my $html = ""; # html formatted display output
80 dpavlin 10
81 dpavlin 13 my %field_usage; # counter for usage of each field
82    
83 dpavlin 32 # sort subrouting using order="" attribute
84     sub by_order {
85     return 0 if (! $config->{indexer}->{$a}->{order});
86     return 0 if (! $config->{indexer}->{$b}->{order});
87 dpavlin 29
88 dpavlin 32 return $config->{indexer}->{$a}->{order} <=>
89     $config->{indexer}->{$b}->{order} ;
90     }
91 dpavlin 3
92 dpavlin 32 foreach my $field (sort by_order keys %{$config->{indexer}}) {
93    
94 dpavlin 40 $field=x($field);
95    
96 dpavlin 13 $field_usage{$field}++;
97    
98 dpavlin 10 my $swish_data = "";
99 dpavlin 3 my $display_data = "";
100 dpavlin 35 my $line_delimiter;
101 dpavlin 3
102 dpavlin 34 my ($swish,$display);
103    
104 dpavlin 54 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
105     foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
106 dpavlin 3
107 dpavlin 40 my $format = x($x->{content});
108     my $delimiter = x($x->{delimiter}) || ' ';
109 dpavlin 3
110 dpavlin 54 my $repeat_off = 0; # repeatable offset
111 dpavlin 29
112     my ($s,$d,$i) = (1,1,0); # swish, display default
113     $s = 0 if (lc($x->{type}) eq "display");
114     $d = 0 if (lc($x->{type}) eq "swish");
115     ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
116    
117 dpavlin 35 # what will separate last line from this one?
118     if ($display_data && $x->{append} && $x->{append} eq "1") {
119 dpavlin 34 $line_delimiter = ' ';
120 dpavlin 35 } elsif ($display_data) {
121 dpavlin 34 $line_delimiter = '<br/>';
122     }
123 dpavlin 29
124 dpavlin 34 # init vars so that we go into while...
125     ($swish,$display) = (1,1);
126 dpavlin 29
127 dpavlin 54 if ($swish || $display) {
128     ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
129 dpavlin 29 # filter="name" ; filter this field through
130     # filter/[name].pm
131     my $filter = $x->{filter};
132 dpavlin 20 if ($filter) {
133 dpavlin 29 require "filter/".$filter.".pm";
134 dpavlin 20 }
135 dpavlin 29 # type="swish" ; field for swish
136     if ($s && $swish) {
137     if ($filter) {
138     no strict 'refs';
139     $swish_data .= join(" ",&$filter($swish));
140     } else {
141     $swish_data .= $swish;
142     }
143     }
144 dpavlin 17
145 dpavlin 29 # type="display" ; field for display
146     if ($d && $display) {
147 dpavlin 35 if ($line_delimiter && $display_data) {
148     $display_data .= $line_delimiter;
149     undef $line_delimiter;
150     }
151 dpavlin 29 if ($filter) {
152     no strict 'refs';
153     $display_data .= join($delimiter,&$filter($display));
154     } else {
155     if ($display_data) {
156     $display_data .= $delimiter.$display;
157     } else {
158     $display_data .= $display;
159     }
160     }
161 dpavlin 20 }
162 dpavlin 29
163     # type="index" ; insert into index
164     if ($i && $display) {
165 dpavlin 40 my $index_data = $display;
166 dpavlin 29 if ($filter) {
167     no strict 'refs';
168     foreach my $d (&$filter($index_data)) {
169 dpavlin 43 $index->insert($field, $d, $path);
170 dpavlin 29 }
171     } else {
172 dpavlin 43 $index->insert($field, $index_data, $path);
173 dpavlin 20 }
174     }
175 dpavlin 17 }
176 dpavlin 3 }
177 dpavlin 9
178 dpavlin 59 # now try to parse variables from configuration file
179     foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
180 dpavlin 13
181 dpavlin 62 my $delimiter = x($x->{delimiter}) || ' ';
182 dpavlin 59 my $val = $cfg->val($database, x($x->{content}));
183    
184     my ($s,$d,$i) = (1,1,0); # swish, display default
185     $s = 0 if (lc($x->{type}) eq "display");
186     $d = 0 if (lc($x->{type}) eq "swish");
187     ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
188    
189     if ($val) {
190 dpavlin 62 $display_data .= $delimiter.$val if ($d);
191 dpavlin 59 $swish_data .= $val if ($s);
192     $index->insert($field, $val, $path) if ($i);
193     }
194    
195     }
196    
197    
198 dpavlin 10 if ($display_data) {
199 dpavlin 29
200 dpavlin 13 if ($field eq "headline") {
201     $xml .= xmlify("headline", $display_data);
202     } else {
203    
204     # find field name (signular, plural)
205     my $field_name = "";
206     if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
207     $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
208     } elsif ($config->{indexer}->{$field}->{name_plural}) {
209     $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
210 dpavlin 20 } elsif ($config->{indexer}->{$field}->{name}) {
211     $field_name = $config->{indexer}->{$field}->{name}."#-#";
212 dpavlin 13 } else {
213 dpavlin 20 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
214 dpavlin 13 }
215     if ($field_name) {
216 dpavlin 40 $html .= x($field_name);
217 dpavlin 13 }
218     $html .= $display_data."###\n";
219     }
220 dpavlin 10 }
221     if ($swish_data) {
222 dpavlin 20 # remove extra spaces
223     $swish_data =~ s/ +/ /g;
224     $swish_data =~ s/ +$//g;
225    
226 dpavlin 40 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
227 dpavlin 10 }
228    
229 dpavlin 9
230 dpavlin 3 }
231 dpavlin 13
232     # dump formatted output in <html>
233     if ($html) {
234 dpavlin 81 #$xml .= xmlify("html",$html);
235     $xml .= "<html><![CDATA[ $html ]]></html>";
236 dpavlin 13 }
237    
238 dpavlin 3 if ($xml) {
239 dpavlin 13 $xml .= $add_xml if ($add_xml);
240 dpavlin 10 return "<xml>\n$xml</xml>\n";
241 dpavlin 3 } else {
242     return;
243     }
244     }
245    
246     ##########################################################################
247    
248 dpavlin 54 # read configuration for this script
249 dpavlin 13 my $cfg = new Config::IniFiles( -file => $config_file );
250 dpavlin 1
251 dpavlin 54 # read global.conf configuration
252     my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
253    
254 dpavlin 50 # open index
255     $index = new index_DBI(
256 dpavlin 54 $cfg_global->val('global', 'dbi_dbd'),
257     $cfg_global->val('global', 'dbi_dsn'),
258     $cfg_global->val('global', 'dbi_user'),
259     $cfg_global->val('global', 'dbi_passwd') || '',
260 dpavlin 50 );
261    
262 dpavlin 13 foreach my $database ($cfg->Sections) {
263 dpavlin 1
264 dpavlin 54 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
265 dpavlin 40 my $add_xml = $cfg -> val($database, 'xml'); # optional
266 dpavlin 1
267 dpavlin 54 print STDERR "reading ./import_xml/$type.xml\n";
268 dpavlin 1
269 dpavlin 58 # extract just type basic
270     my $type_base = $type;
271     $type_base =~ s/_.+$//g;
272 dpavlin 40
273 dpavlin 59 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
274 dpavlin 58
275 dpavlin 54 # output current progress indicator
276     my $last_p = 0;
277     sub progress {
278     #return if (! $opts{q}); # FIXME
279     my $current = shift;
280 dpavlin 57 my $total = shift || 1;
281 dpavlin 54 my $p = int($current * 100 / $total);
282     if ($p != $last_p) {
283     printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
284     $last_p = $p;
285     }
286 dpavlin 3 }
287    
288 dpavlin 67 my $fake_dir = 1;
289     sub fakeprogress {
290     my $current = shift @_;
291    
292     my @ind = ('-','\\','|','/','-','\\','|','/', '-');
293    
294     $last_p += $fake_dir;
295     $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
296     if ($last_p % 10 == 0) {
297     printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
298     }
299     }
300    
301 dpavlin 54 # now read database
302     print STDERR "using: $type...\n";
303 dpavlin 1
304 dpavlin 58 if ($type_base eq "isis") {
305    
306 dpavlin 54 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
307 dpavlin 1
308 dpavlin 54 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
309     my $db = OpenIsis::open( $isis_db );
310 dpavlin 10
311 dpavlin 54 my $max_rowid = OpenIsis::maxRowid( $db );
312 dpavlin 3
313 dpavlin 54 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
314    
315     my $path = $database;
316    
317     for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
318     my $row = OpenIsis::read( $db, $row_id );
319     if ($row && $row->{mfn}) {
320    
321     progress($row->{mfn}, $max_rowid);
322    
323     my $swishpath = $path."#".int($row->{mfn});
324    
325 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
326 dpavlin 54 $xml = $cp2utf->convert($xml);
327     use bytes; # as opposed to chars
328     print "Path-Name: $swishpath\n";
329     print "Content-Length: ".(length($xml)+1)."\n";
330     print "Document-Type: XML\n\n$xml\n";
331     }
332 dpavlin 3 }
333 dpavlin 54 }
334     print STDERR "\n";
335 dpavlin 3
336 dpavlin 58 } elsif ($type_base eq "excel") {
337 dpavlin 54 use Spreadsheet::ParseExcel;
338     use Spreadsheet::ParseExcel::Utility qw(int2col);
339    
340     $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
341     my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
342 dpavlin 43
343 dpavlin 54 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
344 dpavlin 56 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
345 dpavlin 54
346     my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
347    
348     my $sheet_nr = 0;
349     foreach my $oWks (@{$oBook->{Worksheet}}) {
350     #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
351     last if ($oWks->{Name} eq $sheet);
352     $sheet_nr++;
353     }
354    
355     my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
356    
357     print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
358     defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
359     my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
360    
361 dpavlin 56 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
362 dpavlin 54 my $row;
363     for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
364     my $cell = $oWorksheet->{Cells}[$iR][$iC];
365     if ($cell) {
366     $row->{int2col($iC)} = $cell->Value;
367     }
368     }
369    
370     progress($iR, $end_row);
371    
372     # print "row[$iR/$end_row] ";
373     # foreach (keys %{$row}) {
374     # print "$_: ",$row->{$_},"\t";
375     # }
376     # print "\n";
377    
378     my $swishpath = $database."#".$iR;
379    
380     next if (! $row);
381    
382 dpavlin 59 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
383 dpavlin 44 $xml = $cp2utf->convert($xml);
384 dpavlin 35 use bytes; # as opposed to chars
385 dpavlin 43 print "Path-Name: $swishpath\n";
386 dpavlin 10 print "Content-Length: ".(length($xml)+1)."\n";
387     print "Document-Type: XML\n\n$xml\n";
388 dpavlin 3 }
389 dpavlin 1 }
390 dpavlin 62 } elsif ($type_base eq "marc") {
391 dpavlin 67
392 dpavlin 62 use MARC;
393    
394     $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
395     my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
396    
397     # optional argument is format
398     my $format = x($config->{format}) || 'usmarc';
399    
400     print STDERR "Reading MARC file '$marc_file'\n";
401    
402     my $marc = new MARC;
403     my $nr = $marc->openmarc({
404     file=>$marc_file, format=>$format
405     }) || die "Can't open MARC file '$marc_file'";
406    
407     my $i=0; # record nr.
408    
409     my $rec;
410    
411     while ($marc->nextmarc(1)) {
412    
413     # XXX
414 dpavlin 67 fakeprogress($i++);
415 dpavlin 62
416     my $swishpath = $database."#".$i;
417    
418     if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
419     $xml = $cp2utf->convert($xml);
420     use bytes; # as opposed to chars
421     print "Path-Name: $swishpath\n";
422     print "Content-Length: ".(length($xml)+1)."\n";
423     print "Document-Type: XML\n\n$xml\n";
424     }
425     }
426 dpavlin 67 } elsif ($type_base eq "feed") {
427    
428     $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
429     my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
430    
431     print STDERR "Reading feed from program '$prog'\n";
432    
433     open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
434    
435 dpavlin 74 my $i=1; # record nr.
436 dpavlin 67
437     my $data;
438     my $line=1;
439    
440     while (<FEED>) {
441     chomp;
442    
443     if (/^$/) {
444 dpavlin 74 my $swishpath = $database."#".$i++;
445 dpavlin 67
446     if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
447     $xml = $cp2utf->convert($xml);
448     use bytes; # as opposed to chars
449     print "Path-Name: $swishpath\n";
450     print "Content-Length: ".(length($xml)+1)."\n";
451     print "Document-Type: XML\n\n$xml\n";
452     }
453     $line = 1;
454     $data = {};
455     next;
456     }
457    
458 dpavlin 74 $line = $1 if (s/^(\d+):\s*//);
459 dpavlin 67 $data->{$line++} = $_;
460    
461 dpavlin 74 fakeprogress($i);
462 dpavlin 67
463     }
464 dpavlin 1 }
465     }
466 dpavlin 3
467 dpavlin 10 # call this to commit index
468     $index->close;
469 dpavlin 3
470     1;
471     __END__
472     ##########################################################################
473    
474     =head1 NAME
475    
476 dpavlin 81 all2xml.pl - read various file formats and dump XML for SWISH-E
477 dpavlin 3
478     =head1 DESCRIPTION
479    
480 dpavlin 81 This command will read ISIS data file using OpenIsis perl module, MARC
481     records using MARC module and optionally Micro$oft Excel files to
482     create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
483     this script B<isn't general xml generator> from isis files (isis allready
484     has something like that). Output of this script is tailor-made for SWISH-E.
485 dpavlin 3
486 dpavlin 81 =head1 BUGS
487    
488     Documentation is really lacking. However, in true Open Source spirit, source
489     is best documentation. I even made considerable effort to comment parts
490     which are not intuitively clear, so...
491    
492 dpavlin 3 =head1 AUTHOR
493    
494     Dobrica Pavlinusic <dpavlin@rot13.org>
495    
496     =head1 COPYRIGHT
497    
498     GNU Public License (GPL) v2 or later
499    
500     =head1 SEE ALSO
501    
502     SWISH-E web site at http://www.swish-e.org
503    
504     =cut

Properties

Name Value
cvs2svn:cvs-rev 1.29
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26