Commit | Line | Data |
6f4946fb |
1 | package Text::Tradition::Parser::CTE; |
2 | |
3 | use strict; |
4 | use warnings; |
4d85a60e |
5 | use Text::Tradition::Parser::Util qw/ collate_variants /; |
6f4946fb |
6 | use XML::LibXML; |
7 | use XML::LibXML::XPathContext; |
8 | |
9 | =head1 NAME |
10 | |
11 | Text::Tradition::Parser::CTE |
12 | |
13 | =head1 DESCRIPTION |
14 | |
15 | Parser module for Text::Tradition, given a TEI file exported from |
16 | Classical Text Editor. |
17 | |
18 | =head1 METHODS |
19 | |
20 | =over |
21 | |
22 | =item B<parse> |
23 | |
24 | my @apparatus = read( $xml_file ); |
25 | |
26 | Takes a Tradition object and a TEI file exported from Classical Text |
4d85a60e |
27 | Editor using double-endpoint-attachment critical apparatus encoding; |
28 | initializes the Tradition from the file. |
6f4946fb |
29 | |
30 | =cut |
31 | |
4d85a60e |
32 | my %sigil_for; # Save the XML IDs for witnesses. |
33 | my %apps; # Save the apparatus XML for a given ID. |
f6e19c7c |
34 | my %has_ac; # Keep track of witnesses that have corrections. |
6f4946fb |
35 | |
36 | sub parse { |
dfc37e38 |
37 | my( $tradition, $opts ) = @_; |
4d85a60e |
38 | my $c = $tradition->collation; # Some shorthand |
39 | |
40 | # First, parse the XML. |
41 | my $parser = XML::LibXML->new(); |
dfc37e38 |
42 | my $doc; |
43 | if( exists $opts->{'string'} ) { |
44 | $doc = $parser->parse_string( $opts->{'string'} ); |
45 | } elsif ( exists $opts->{'file'} ) { |
46 | $doc = $parser->parse_file( $opts->{'file'} ); |
47 | } else { |
48 | warn "Could not find string or file option to parse"; |
49 | return; |
50 | } |
4d85a60e |
51 | my $tei = $doc->documentElement(); |
52 | my $xpc = XML::LibXML::XPathContext->new( $tei ); |
53 | |
54 | # CTE uses a DTD rather than any xmlns-based parsing. Thus we |
55 | # need no namespace handling. |
56 | |
57 | # Get the witnesses and create the witness objects. |
58 | foreach my $wit_el ( $xpc->findnodes( '//sourceDesc/listWit/witness' ) ) { |
59 | # The witness xml:id is used internally, and is *not* the sigil name. |
60 | my $id= $wit_el->getAttribute( 'xml:id' ); |
61 | my @sig_parts = $xpc->findnodes( './abbr/descendant::text()', $wit_el ); |
62 | my $sig = _stringify_sigil( @sig_parts ); |
63 | $tradition->add_witness( sigil => $sig, source => $wit_el->toString() ); |
64 | $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax |
65 | } |
6f4946fb |
66 | |
4d85a60e |
67 | # Now go through the text and find the base tokens, apparatus tags, and |
68 | # anchors. Make a giant array of all of these things in sequence. |
69 | # TODO consider combining this with creation of graph below |
70 | my @base_text; |
71 | foreach my $pg_el ( $xpc->findnodes( '/TEI/text/body/p' ) ) { |
72 | foreach my $xn ( $pg_el->childNodes ) { |
73 | push( @base_text, _get_base( $xn ) ); |
74 | } |
6f4946fb |
75 | } |
4d85a60e |
76 | |
77 | # We now have to work through this array applying the alternate |
78 | # apparatus readings to the base text. Essentially we will put |
79 | # everything on the graph, from which we will delete the apps and |
80 | # anchors when we are done. |
f6e19c7c |
81 | |
82 | # First, put the base tokens, apps, and anchors in the graph. |
4d85a60e |
83 | my $counter = 0; |
84 | my $last = $c->start; |
85 | foreach my $item ( @base_text ) { |
86 | my $r; |
87 | if( $item->{'type'} eq 'token' ) { |
88 | $r = $c->add_reading( 'n'.$counter++ ); |
89 | $r->text( $item->{'content'} ); |
90 | } elsif ( $item->{'type'} eq 'anchor' ) { |
91 | $r = $c->add_reading( '#ANCHOR_' . $item->{'content'} . '#' ); |
f6e19c7c |
92 | $r->is_meta(1); |
4d85a60e |
93 | } elsif ( $item->{'type'} eq 'app' ) { |
94 | my $tag = '#APP_' . $counter++ . '#'; |
95 | $r = $c->add_reading( $tag ); |
f6e19c7c |
96 | $r->is_meta(1); |
4d85a60e |
97 | $apps{$tag} = $item->{'content'}; |
98 | } |
f6e19c7c |
99 | $c->add_path( $last, $r, $c->baselabel ); |
4d85a60e |
100 | $last = $r; |
6f4946fb |
101 | } |
f6e19c7c |
102 | $c->add_path( $last, $c->end, $c->baselabel ); |
4d85a60e |
103 | |
104 | # Now we can parse the apparatus entries, and add the variant readings |
105 | # to the graph. |
106 | |
107 | foreach my $app_id ( keys %apps ) { |
108 | _add_readings( $c, $app_id ); |
6f4946fb |
109 | } |
4d85a60e |
110 | |
f6e19c7c |
111 | # Finally, add explicit witness paths, remove the base paths, and remove |
112 | # the app/anchor tags. |
113 | expand_all_paths( $c ); |
6f4946fb |
114 | } |
115 | |
4d85a60e |
116 | sub _stringify_sigil { |
117 | my( @nodes ) = @_; |
118 | my @parts = grep { /\w/ } map { $_->data } @nodes; |
119 | return join( '', @parts ); |
120 | } |
6f4946fb |
121 | |
122 | ## Recursive little helper function to help us navigate through nested |
4d85a60e |
123 | ## XML, picking out the words, the apparatus, and the anchors. |
124 | |
125 | sub _get_base { |
126 | my( $xn ) = @_; |
127 | my @readings; |
128 | if( $xn->nodeType == XML_TEXT_NODE ) { |
129 | # Base text, just split the words on whitespace and add them |
130 | # to our sequence. |
f6e19c7c |
131 | # TODO consider that XML markup might appear mid-token. |
4d85a60e |
132 | my $str = $xn->data; |
133 | $str =~ s/^\s+//; |
134 | foreach my $w ( split( /\s+/, $str ) ) { |
f6e19c7c |
135 | # HACK to cope with mismatched doublequotes |
136 | $w =~ s/\"//g; |
4d85a60e |
137 | push( @readings, { 'type' => 'token', 'content' => $w } ); |
138 | } |
139 | } elsif( $xn->nodeName eq 'hi' ) { |
140 | # Recurse as if the hi weren't there. |
141 | foreach( $xn->childNodes ) { |
142 | push( @readings, _get_base( $_ ) ); |
143 | } |
144 | } elsif( $xn->nodeName eq 'app' ) { |
145 | # Apparatus, just save the entire XML node. |
146 | push( @readings, { 'type' => 'app', 'content' => $xn } ); |
147 | } elsif( $xn->nodeName eq 'anchor' ) { |
148 | # Anchor to mark the end of some apparatus; save its ID. |
149 | push( @readings, { 'type' => 'anchor', |
150 | 'content' => $xn->getAttribute( 'xml:id' ) } ); |
151 | } elsif ( $xn->nodeName ne 'note' ) { # Any tag we don't know to disregard |
152 | print STDERR "Unrecognized tag " . $xn->nodeName . "\n"; |
6f4946fb |
153 | } |
4d85a60e |
154 | return @readings; |
6f4946fb |
155 | } |
156 | |
4d85a60e |
157 | sub _add_readings { |
158 | my( $c, $app_id ) = @_; |
159 | my $xn = $apps{$app_id}; |
160 | my $anchor = _anchor_name( $xn->getAttribute( 'to' ) ); |
161 | # Get the lemma, which is all the readings between app and anchor, |
162 | # excluding other apps or anchors. |
163 | my @lemma = _return_lemma( $c, $app_id, $anchor ); |
164 | my $lemma_str = join( ' ', grep { $_ !~ /^\#/ } map { $_->text } @lemma ); |
165 | |
166 | # For each reading, send its text to 'interpret' along with the lemma, |
167 | # and then save the list of witnesses that these tokens belong to. |
3a5d151b |
168 | my %wit_rdgs; # Maps from witnesses to the variant text |
4d85a60e |
169 | my $ctr = 0; |
170 | my $tag = $app_id; |
171 | $tag =~ s/^\#APP_(.*)\#$/$1/; |
f6e19c7c |
172 | $DB::single = 1 if $tag < 2; |
4d85a60e |
173 | foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) { |
174 | my @text; |
4d85a60e |
175 | foreach ( $rdg->childNodes ) { |
176 | push( @text, _get_base( $_ ) ); |
177 | } |
178 | my $interpreted = @text |
179 | ? interpret( join( ' ', map { $_->{'content'} } @text ), $lemma_str ) |
180 | : ''; |
181 | my @rdg_nodes; |
182 | foreach my $w ( split( /\s+/, $interpreted ) ) { |
183 | my $r = $c->add_reading( $tag . "/" . $ctr++ ); |
184 | $r->text( $w ); |
185 | push( @rdg_nodes, $r ); |
186 | } |
f6e19c7c |
187 | |
188 | # For each listed wit, save the reading. |
189 | foreach my $wit ( split( /\s+/, $rdg->getAttribute( 'wit' ) ) ) { |
190 | $wit_rdgs{$wit} = \@rdg_nodes; |
191 | } |
3a5d151b |
192 | # Does the reading have an ID? If so it probably has a witDetail |
f6e19c7c |
193 | # attached, and we need to read it. |
3a5d151b |
194 | if( $rdg->hasAttribute( 'xml:id' ) ) { |
f6e19c7c |
195 | my $rid = $rdg->getAttribute( 'xml:id' ); |
196 | my $xpc = XML::LibXML::XPathContext->new( $xn ); |
197 | my @details = $xpc->findnodes( './witDetail[@target="'.$rid.'"]' ); |
198 | foreach my $d ( @details ) { |
199 | _parse_wit_detail( $d, \%wit_rdgs, \@lemma ); |
200 | } |
3a5d151b |
201 | } |
f6e19c7c |
202 | } |
203 | |
4d85a60e |
204 | # Now collate the variant readings, since it is not done for us. |
f6e19c7c |
205 | collate_variants( $c, \@lemma, values %wit_rdgs ); |
4d85a60e |
206 | |
207 | # Now add the witness paths for each reading. |
f6e19c7c |
208 | foreach my $wit_id ( keys %wit_rdgs ) { |
209 | my $witstr = get_sigil( $wit_id, $c ); |
210 | my $rdg_list = $wit_rdgs{$wit_id}; |
211 | _add_wit_path( $c, $rdg_list, $app_id, $anchor, $witstr ); |
6f4946fb |
212 | } |
4d85a60e |
213 | } |
6f4946fb |
214 | |
4d85a60e |
215 | sub _anchor_name { |
216 | my $xmlid = shift; |
217 | $xmlid =~ s/^\#//; |
218 | return sprintf( "#ANCHOR_%s#", $xmlid ); |
6f4946fb |
219 | } |
220 | |
4d85a60e |
221 | sub _return_lemma { |
222 | my( $c, $app, $anchor ) = @_; |
223 | my $app_node = $c->graph->node( $app ); |
224 | my $anchor_node = $c->graph->node( $anchor ); |
f6e19c7c |
225 | my @nodes = grep { $_->name !~ /^\#A(PP|NCHOR)/ } |
226 | $c->reading_sequence( $app_node, $anchor_node, $c->baselabel ); |
4d85a60e |
227 | return @nodes; |
228 | } |
6f4946fb |
229 | |
230 | sub interpret { |
4d85a60e |
231 | # A utility function to change apparatus-ese into a full variant. |
232 | my( $reading, $lemma ) = @_; |
233 | return $reading if $reading eq $lemma; |
234 | my $oldreading = $reading; |
235 | # $lemma =~ s/\s+[[:punct:]]+$//; |
236 | # $reading =~ s/\s*\(?sic([\s\w.]+)?\)?$//; |
237 | my @words = split( /\s+/, $lemma ); |
238 | if( $reading =~ /^(.*) praem.$/ ) { |
239 | $reading = "$1 $lemma"; |
240 | } elsif( $reading =~ /^(.*) add.$/ ) { |
241 | $reading = "$lemma $1"; |
f6e19c7c |
242 | } elsif( $reading eq 'om.' |
243 | || $reading =~ /locus [uv]acuus/ |
244 | || $reading =~ /inscriptionem compegi e/ # TODO huh? |
245 | || $reading eq 'def.' # TODO huh? |
246 | ) { |
4d85a60e |
247 | $reading = ''; |
248 | } elsif( $reading eq 'inv.' ) { |
249 | # Hope it is two words. |
250 | print STDERR "WARNING: want to invert a lemma that is not two words\n" |
251 | unless scalar( @words ) == 2; |
252 | $reading = join( ' ', reverse( @words ) ); |
253 | } elsif( $reading eq 'iter.' ) { |
254 | # Repeat the lemma |
255 | $reading = "$lemma $lemma"; |
256 | } elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) { |
257 | # The first and last N words captured should replace the first and |
258 | # last N words of the lemma. |
259 | my @begin = split( /\s+/, $1 ); |
260 | my @end = split( /\s+/, $2 ); |
261 | if( scalar( @begin ) + scalar ( @end ) > scalar( @words ) ) { |
262 | # Something is wrong and we can't do the splice. |
263 | print STDERR "ERROR: $lemma is too short to accommodate $oldreading\n"; |
264 | } else { |
265 | splice( @words, 0, scalar @begin, @begin ); |
266 | splice( @words, -(scalar @end), scalar @end, @end ); |
267 | $reading = join( ' ', @words ); |
268 | } |
6f4946fb |
269 | } |
4d85a60e |
270 | print STDERR "Interpreted $oldreading as $reading given $lemma\n"; |
271 | return $reading; |
272 | } |
273 | |
f6e19c7c |
274 | sub _parse_wit_detail { |
275 | my( $detail, $readings, $lemma ) = @_; |
276 | my $wit = $detail->getAttribute( 'wit' ); |
277 | my $content = $detail->textContent; |
278 | if( $content =~ /a\.\s*c\./ ) { |
279 | # Replace the key in the $readings hash |
280 | my $rdg = delete $readings->{$wit}; |
281 | $readings->{$wit.'_ac'} = $rdg; |
282 | $has_ac{$sigil_for{$wit}} = 1; |
283 | } elsif( $content =~ /p\.\s*c\./ ) { |
284 | # If no key for the wit a.c. exists, add one pointing to the lemma |
285 | unless( exists $readings->{$wit.'_ac'} ) { |
286 | $readings->{$wit.'_ac'} = $lemma; |
287 | } |
288 | $has_ac{$sigil_for{$wit}} = 1; |
289 | } # else don't bother just yet |
290 | } |
291 | |
292 | sub get_sigil { |
293 | my( $xml_id, $c ) = @_; |
294 | if( $xml_id =~ /^(.*)_ac$/ ) { |
295 | my $real_id = $1; |
296 | return $sigil_for{$real_id} . $c->ac_label; |
297 | } else { |
298 | return $sigil_for{$xml_id}; |
299 | } |
300 | } |
301 | |
302 | sub expand_all_paths { |
303 | my( $c ) = @_; |
304 | |
305 | # Walk the collation and fish out the paths for each witness |
306 | foreach my $wit ( $c->tradition->witnesses ) { |
307 | my $sig = $wit->sigil; |
308 | my @path = grep { $_->name !~ /(APP|ANCHOR)/ } |
309 | $c->reading_sequence( $c->start, $c->end, $sig ); |
310 | $wit->path( \@path ); |
311 | if( $has_ac{$sig} ) { |
312 | my @ac_path = grep { $_->name !~ /(APP|ANCHOR)/ } |
313 | $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label, $sig ); |
314 | $wit->uncorrected_path( \@ac_path ); |
315 | } |
316 | } |
317 | |
318 | # Delete the anchors |
319 | foreach my $anchor ( grep { $_->name =~ /(APP|ANCHOR)/ } $c->readings ) { |
320 | $c->del_reading( $anchor ); |
321 | } |
322 | # Delete all edges |
323 | map { $c->del_path( $_ ) } $c->paths; |
324 | |
325 | # Make the path edges |
326 | $c->make_witness_paths(); |
4d85a60e |
327 | } |
328 | |
329 | sub _add_wit_path { |
f6e19c7c |
330 | my( $c, $rdg, $app, $anchor, $wit ) = @_; |
4d85a60e |
331 | my @nodes = @$rdg; |
f6e19c7c |
332 | push( @nodes, $c->reading( $anchor ) ); |
4d85a60e |
333 | |
f6e19c7c |
334 | my $cur = $c->reading( $app ); |
4d85a60e |
335 | foreach my $n ( @nodes ) { |
f6e19c7c |
336 | $c->add_path( $cur, $n, $wit ); |
4d85a60e |
337 | $cur = $n; |
6f4946fb |
338 | } |
6f4946fb |
339 | } |
340 | |
341 | =back |
342 | |
343 | =head1 LICENSE |
344 | |
345 | This package is free software and is provided "as is" without express |
346 | or implied warranty. You can redistribute it and/or modify it under |
347 | the same terms as Perl itself. |
348 | |
349 | =head1 AUTHOR |
350 | |
351 | Tara L Andrews, aurum@cpan.org |
352 | |
353 | =cut |
354 | |
355 | 1; |
356 | |