7 |
use XML::Simple; |
use XML::Simple; |
8 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
use Text::Iconv; |
use Text::Iconv; |
10 |
|
use Config::IniFiles; |
11 |
|
use Encode; |
12 |
|
|
13 |
$|=1; |
$|=1; |
14 |
|
|
15 |
my $config; |
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 |
$config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
my $config; |
20 |
|
|
21 |
use index_DBI; # there is no other, right now ;-) |
use index_DBI; # there is no other, right now ;-) |
22 |
|
my $index; |
|
my $index = new index_DBI(); # open index |
|
23 |
|
|
24 |
my %opts; |
my %opts; |
25 |
|
|
31 |
|
|
32 |
getopts('d:m:qs', \%opts); |
getopts('d:m:qs', \%opts); |
33 |
|
|
34 |
my $db_dir = $opts{d} || "ps"; # FIX |
my $path; # this is name of database |
35 |
|
|
36 |
|
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
37 |
|
|
38 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
# this is encoding of all files on disk, including import_xml/*.xml file and |
39 |
|
# filter/*.pm files! It will be used to store strings in perl internally! |
40 |
|
my $codepage = 'ISO-8859-2'; |
41 |
|
|
42 |
|
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
43 |
|
# this function will convert data from XML files to local encoding |
44 |
|
sub x { |
45 |
|
return $utf2cp->convert($_[0]); |
46 |
|
} |
47 |
|
|
48 |
#print Dumper($config->{indexer}); |
# decode isis/excel or other import codepage |
49 |
#print "-" x 70,"\n"; |
my $import2cp; |
50 |
|
|
51 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
# outgoing xml must be in UTF-8 |
52 |
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
53 |
|
|
54 |
my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
# mapping between data type and tag which specify |
55 |
my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
# format in XML file |
56 |
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
my %type2tag = ( |
57 |
|
'isis' => 'isis', |
58 |
|
'excel' => 'column' |
59 |
|
); |
60 |
|
|
61 |
sub isis2xml { |
sub data2xml { |
62 |
|
|
63 |
use xmlify; |
use xmlify; |
64 |
|
|
65 |
|
my $type = shift @_; |
66 |
my $row = shift @_; |
my $row = shift @_; |
67 |
|
my $add_xml = shift @_; |
68 |
|
|
69 |
my $xml; |
my $xml; |
|
$xml .= xmlify('db_dir',$db_dir); # FIX remove? |
|
70 |
|
|
71 |
use parse_format; |
use parse_format; |
72 |
|
|
73 |
|
my $html = ""; # html formatted display output |
74 |
|
|
75 |
|
my %field_usage; # counter for usage of each field |
76 |
|
|
77 |
|
# sort subrouting using order="" attribute |
78 |
|
sub by_order { |
79 |
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
80 |
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
81 |
|
|
82 |
|
return $config->{indexer}->{$a}->{order} <=> |
83 |
|
$config->{indexer}->{$b}->{order} ; |
84 |
|
} |
85 |
|
|
86 |
|
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
87 |
|
|
88 |
|
$field=x($field); |
89 |
|
|
90 |
foreach my $field (keys %{$config->{indexer}}) { |
$field_usage{$field}++; |
91 |
|
|
92 |
my $swish_data = ""; |
my $swish_data = ""; |
93 |
my $display_data = ""; |
my $display_data = ""; |
94 |
my $index_data = ""; |
my $line_delimiter; |
95 |
|
|
96 |
|
my ($swish,$display); |
97 |
|
|
98 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
my $tag = $type2tag{$type} || die "can't find which tag to use for type $type"; |
99 |
|
foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) { |
100 |
|
|
101 |
|
my $format = x($x->{content}); |
102 |
|
my $delimiter = x($x->{delimiter}) || ' '; |
103 |
|
|
104 |
|
my $repeat_off = 0; # repeatable offset |
105 |
|
|
|
my $format = $x->{content}; |
|
106 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
my ($s,$d,$i) = (1,1,0); # swish, display default |
107 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
108 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
109 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
|
110 |
|
|
111 |
$format = $cludge_codepage->convert($format); |
# what will separate last line from this one? |
112 |
my ($swish,$display) = parse_format($format,$row); |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
113 |
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
$line_delimiter = ' '; |
114 |
|
} elsif ($display_data) { |
115 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
$line_delimiter = '<br/>'; |
116 |
# FIX: this is ugly, UGLY, cludge: OpenIsis return |
} |
117 |
# UTF8 encoding of strings, but as if source charset |
|
118 |
# is ISO8859-1 and not some other. This breaks our |
# init vars so that we go into while... |
119 |
# isis character encoding, so we convert it first |
($swish,$display) = (1,1); |
120 |
# back to ISO8859-1 (which can actually be different |
|
121 |
# encoding in isis) |
if ($swish || $display) { |
122 |
|
($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp); |
123 |
$swish_data .= $swish if ($s && $swish); |
# filter="name" ; filter this field through |
124 |
$display_data .= $display if ($d && $display); |
# filter/[name].pm |
125 |
$index_data .= $display if ($i && $display); |
my $filter = $x->{filter}; |
126 |
|
if ($filter) { |
127 |
|
require "filter/".$filter.".pm"; |
128 |
|
} |
129 |
|
# type="swish" ; field for swish |
130 |
|
if ($s && $swish) { |
131 |
|
if ($filter) { |
132 |
|
no strict 'refs'; |
133 |
|
$swish_data .= join(" ",&$filter($swish)); |
134 |
|
} else { |
135 |
|
$swish_data .= $swish; |
136 |
|
} |
137 |
|
} |
138 |
|
|
139 |
|
# type="display" ; field for display |
140 |
|
if ($d && $display) { |
141 |
|
if ($line_delimiter && $display_data) { |
142 |
|
$display_data .= $line_delimiter; |
143 |
|
undef $line_delimiter; |
144 |
|
} |
145 |
|
if ($filter) { |
146 |
|
no strict 'refs'; |
147 |
|
$display_data .= join($delimiter,&$filter($display)); |
148 |
|
} else { |
149 |
|
if ($display_data) { |
150 |
|
$display_data .= $delimiter.$display; |
151 |
|
} else { |
152 |
|
$display_data .= $display; |
153 |
|
} |
154 |
|
} |
155 |
|
} |
156 |
|
|
157 |
|
# type="index" ; insert into index |
158 |
|
if ($i && $display) { |
159 |
|
my $index_data = $display; |
160 |
|
if ($filter) { |
161 |
|
no strict 'refs'; |
162 |
|
foreach my $d (&$filter($index_data)) { |
163 |
|
$index->insert($field, $d, $path); |
164 |
|
} |
165 |
|
} else { |
166 |
|
$index->insert($field, $index_data, $path); |
167 |
|
} |
168 |
|
} |
169 |
|
} |
170 |
} |
} |
171 |
|
|
172 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
|
173 |
if ($display_data) { |
if ($display_data) { |
174 |
$display_data = $isis_codepage->convert($display_data)."##" || $display_data; |
|
175 |
$xml .= xmlify($field."_display", $display_data); |
if ($field eq "headline") { |
176 |
|
$xml .= xmlify("headline", $display_data); |
177 |
|
} else { |
178 |
|
|
179 |
|
# find field name (signular, plural) |
180 |
|
my $field_name = ""; |
181 |
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
182 |
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
183 |
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
184 |
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
185 |
|
} elsif ($config->{indexer}->{$field}->{name}) { |
186 |
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
187 |
|
} else { |
188 |
|
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
189 |
|
} |
190 |
|
if ($field_name) { |
191 |
|
$html .= x($field_name); |
192 |
|
} |
193 |
|
$html .= $display_data."###\n"; |
194 |
|
} |
195 |
} |
} |
196 |
if ($swish_data) { |
if ($swish_data) { |
197 |
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
# remove extra spaces |
198 |
$swish_data = $i->convert($swish_data); |
$swish_data =~ s/ +/ /g; |
199 |
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
$swish_data =~ s/ +$//g; |
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
|
|
} |
|
200 |
|
|
201 |
# index |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
|
if ($index_data && $index_data ne "") { |
|
|
$index_data = $index_codepage->convert($index_data) || $index_data; |
|
|
$index->insert($field, $index_data, $db_dir); |
|
202 |
} |
} |
203 |
|
|
204 |
|
|
205 |
|
} |
206 |
|
|
207 |
|
# dump formatted output in <html> |
208 |
|
if ($html) { |
209 |
|
$xml .= xmlify("html",$html); |
210 |
} |
} |
211 |
|
|
212 |
if ($xml) { |
if ($xml) { |
213 |
#print STDERR "x: $xml\n"; |
$xml .= $add_xml if ($add_xml); |
214 |
return "<xml>\n$xml</xml>\n"; |
return "<xml>\n$xml</xml>\n"; |
215 |
} else { |
} else { |
216 |
return; |
return; |
219 |
|
|
220 |
########################################################################## |
########################################################################## |
221 |
|
|
222 |
my $last_tell=0; |
# read configuration for this script |
223 |
|
my $cfg = new Config::IniFiles( -file => $config_file ); |
224 |
|
|
225 |
my @isis_dirs = ( '.' ); # use dirname as database name |
# read global.conf configuration |
226 |
|
my $cfg_global = new Config::IniFiles( -file => 'global.conf' ); |
227 |
|
|
228 |
if ($opts{m}) { |
# open index |
229 |
@isis_dirs = split(/,/,$opts{m}); |
$index = new index_DBI( |
230 |
} |
$cfg_global->val('global', 'dbi_dbd'), |
231 |
|
$cfg_global->val('global', 'dbi_dsn'), |
232 |
|
$cfg_global->val('global', 'dbi_user'), |
233 |
|
$cfg_global->val('global', 'dbi_passwd') || '', |
234 |
|
); |
235 |
|
|
236 |
my @isis_dbs; |
foreach my $database ($cfg->Sections) { |
237 |
|
|
238 |
foreach (@isis_dirs) { |
my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined"; |
239 |
if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") { |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
240 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI"; |
|
241 |
} |
print STDERR "reading ./import_xml/$type.xml\n"; |
242 |
if (-e $config->{isis_data}."/$db_dir/$_/PERI") { |
|
243 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI"; |
$config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type} ], forcecontent => 1); |
244 |
} |
|
245 |
if (-e $config->{isis_data}."/$db_dir/$_/AMS") { |
# output current progress indicator |
246 |
push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS"; |
my $last_p = 0; |
247 |
} |
sub progress { |
248 |
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
#return if (! $opts{q}); # FIXME |
249 |
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
my $current = shift; |
250 |
|
my $total = shift; |
251 |
|
my $p = int($current * 100 / $total); |
252 |
|
if ($p != $last_p) { |
253 |
|
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p ); |
254 |
|
$last_p = $p; |
255 |
|
} |
256 |
} |
} |
|
} |
|
257 |
|
|
258 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
# now read database |
259 |
|
print STDERR "using: $type...\n"; |
260 |
|
|
261 |
my $db; |
if ($type eq "isis") { |
262 |
|
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
263 |
|
|
264 |
foreach my $isis_db (@isis_dbs) { |
$import2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
265 |
|
my $db = OpenIsis::open( $isis_db ); |
266 |
|
|
267 |
|
my $max_rowid = OpenIsis::maxRowid( $db ); |
268 |
|
|
269 |
my $db = OpenIsis::open( $isis_db ); |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
|
if (0) { |
|
|
# # FIX |
|
|
# if (! $db ) { |
|
|
print STDERR "WARNING: can't open '$isis_db'\n"; |
|
|
next ; |
|
|
} |
|
270 |
|
|
271 |
my $max_rowid = OpenIsis::maxRowid( $db ); |
my $path = $database; |
272 |
|
|
273 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
274 |
|
my $row = OpenIsis::read( $db, $row_id ); |
275 |
|
if ($row && $row->{mfn}) { |
276 |
|
|
277 |
|
progress($row->{mfn}, $max_rowid); |
278 |
|
|
279 |
my $path = $isis_db; |
my $swishpath = $path."#".int($row->{mfn}); |
|
$path =~ s#$config->{isis_data}/*##g; |
|
280 |
|
|
281 |
my $last_p = 0; |
if (my $xml = data2xml($type,$row,$add_xml)) { |
282 |
|
$xml = $cp2utf->convert($xml); |
283 |
|
use bytes; # as opposed to chars |
284 |
|
print "Path-Name: $swishpath\n"; |
285 |
|
print "Content-Length: ".(length($xml)+1)."\n"; |
286 |
|
print "Document-Type: XML\n\n$xml\n"; |
287 |
|
} |
288 |
|
} |
289 |
|
} |
290 |
|
print STDERR "\n"; |
291 |
|
|
292 |
|
} elsif ($type eq "excel") { |
293 |
|
use Spreadsheet::ParseExcel; |
294 |
|
use Spreadsheet::ParseExcel::Utility qw(int2col); |
295 |
|
|
296 |
|
$import2cp = Text::Iconv->new($config->{excel_codepage},$codepage); |
297 |
|
my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!"; |
298 |
|
|
299 |
|
my $sheet = x($config->{sheet}) || die "no sheet in $type.xml"; |
300 |
|
my $start_row = x($config->{start_row}) || die "no start_row in $type.xml"; |
301 |
|
|
302 |
|
my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'"; |
303 |
|
|
304 |
|
my $sheet_nr = 0; |
305 |
|
foreach my $oWks (@{$oBook->{Worksheet}}) { |
306 |
|
#print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n"; |
307 |
|
last if ($oWks->{Name} eq $sheet); |
308 |
|
$sheet_nr++; |
309 |
|
} |
310 |
|
|
311 |
|
my $oWorksheet = $oBook->{Worksheet}[$sheet_nr]; |
312 |
|
|
313 |
|
print STDERR "using sheet: ",$oWorksheet->{Name},"\n"; |
314 |
|
defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file"; |
315 |
|
my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow}; |
316 |
|
|
317 |
|
for(my $iR = $oWorksheet->{MinRow} ; defined $end_row && $iR <= $end_row ; $iR++) { |
318 |
|
my $row; |
319 |
|
for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) { |
320 |
|
my $cell = $oWorksheet->{Cells}[$iR][$iC]; |
321 |
|
if ($cell) { |
322 |
|
$row->{int2col($iC)} = $cell->Value; |
323 |
|
} |
324 |
|
} |
325 |
|
|
326 |
|
progress($iR, $end_row); |
327 |
|
|
328 |
# { my $row_id = 4514; |
# print "row[$iR/$end_row] "; |
329 |
# FIX |
# foreach (keys %{$row}) { |
330 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
# print "$_: ",$row->{$_},"\t"; |
331 |
my $row = OpenIsis::read( $db, $row_id ); |
# } |
332 |
if ($row && $row->{mfn}) { |
# print "\n"; |
333 |
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
334 |
# output current process indicator |
my $swishpath = $database."#".$iR; |
335 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
|
336 |
if ($p != $last_p) { |
next if (! $row); |
337 |
printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); |
|
338 |
$last_p = $p; |
if (my $xml = data2xml($type,$row,$add_xml)) { |
339 |
} |
$xml = $cp2utf->convert($xml); |
340 |
|
use bytes; # as opposed to chars |
341 |
if (my $xml = isis2xml($row)) { |
print "Path-Name: $swishpath\n"; |
|
#print STDERR "--ret-->$xml\n"; |
|
|
print "Path-Name: $path#".int($row->{mfn})."\n"; |
|
342 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |
343 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
344 |
} |
} |
345 |
} |
} |
346 |
} |
} |
|
print STDERR "\n"; |
|
347 |
} |
} |
348 |
|
|
349 |
# call this to commit index |
# call this to commit index |