Commit | Line | Data |
6f4946fb |
1 | package Text::Tradition::Parser::CTE; |
2 | |
3 | use strict; |
4 | use warnings; |
c9158e60 |
5 | use Encode qw/ decode /; |
4d85a60e |
6 | use Text::Tradition::Parser::Util qw/ collate_variants /; |
6f4946fb |
7 | use XML::LibXML; |
8 | use XML::LibXML::XPathContext; |
9 | |
10 | =head1 NAME |
11 | |
12 | Text::Tradition::Parser::CTE |
13 | |
14 | =head1 DESCRIPTION |
15 | |
16 | Parser module for Text::Tradition, given a TEI file exported from |
17 | Classical Text Editor. |
18 | |
19 | =head1 METHODS |
20 | |
21 | =over |
22 | |
23 | =item B<parse> |
24 | |
25 | my @apparatus = read( $xml_file ); |
26 | |
27 | Takes a Tradition object and a TEI file exported from Classical Text |
4d85a60e |
28 | Editor using double-endpoint-attachment critical apparatus encoding; |
29 | initializes the Tradition from the file. |
6f4946fb |
30 | |
31 | =cut |
32 | |
4d85a60e |
33 | my %sigil_for; # Save the XML IDs for witnesses. |
34 | my %apps; # Save the apparatus XML for a given ID. |
f6e19c7c |
35 | my %has_ac; # Keep track of witnesses that have corrections. |
6f4946fb |
36 | |
37 | sub parse { |
dfc37e38 |
38 | my( $tradition, $opts ) = @_; |
4d85a60e |
39 | my $c = $tradition->collation; # Some shorthand |
40 | |
41 | # First, parse the XML. |
c9158e60 |
42 | my( $tei, $xpc ) = _remove_formatting( $opts ); |
43 | return unless $tei; # we have already warned. |
4d85a60e |
44 | |
45 | # CTE uses a DTD rather than any xmlns-based parsing. Thus we |
46 | # need no namespace handling. |
4d85a60e |
47 | # Get the witnesses and create the witness objects. |
48 | foreach my $wit_el ( $xpc->findnodes( '//sourceDesc/listWit/witness' ) ) { |
49 | # The witness xml:id is used internally, and is *not* the sigil name. |
50 | my $id= $wit_el->getAttribute( 'xml:id' ); |
c9158e60 |
51 | my @sig_parts = $xpc->findnodes( 'descendant::text()', $wit_el ); |
4d85a60e |
52 | my $sig = _stringify_sigil( @sig_parts ); |
c9158e60 |
53 | print STDERR "Adding witness $sig\n"; |
82fa4d57 |
54 | $tradition->add_witness( sigil => $sig, sourcetype => 'collation' ); |
b8f262e8 |
55 | $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax |
4d85a60e |
56 | } |
c9158e60 |
57 | |
4d85a60e |
58 | # Now go through the text and find the base tokens, apparatus tags, and |
59 | # anchors. Make a giant array of all of these things in sequence. |
60 | # TODO consider combining this with creation of graph below |
61 | my @base_text; |
62 | foreach my $pg_el ( $xpc->findnodes( '/TEI/text/body/p' ) ) { |
63 | foreach my $xn ( $pg_el->childNodes ) { |
64 | push( @base_text, _get_base( $xn ) ); |
65 | } |
6f4946fb |
66 | } |
4d85a60e |
67 | # We now have to work through this array applying the alternate |
68 | # apparatus readings to the base text. Essentially we will put |
69 | # everything on the graph, from which we will delete the apps and |
70 | # anchors when we are done. |
f6e19c7c |
71 | |
72 | # First, put the base tokens, apps, and anchors in the graph. |
4d85a60e |
73 | my $counter = 0; |
74 | my $last = $c->start; |
75 | foreach my $item ( @base_text ) { |
76 | my $r; |
77 | if( $item->{'type'} eq 'token' ) { |
12720144 |
78 | $r = $c->add_reading( { id => 'n'.$counter++, |
79 | text => $item->{'content'} } ); |
4d85a60e |
80 | } elsif ( $item->{'type'} eq 'anchor' ) { |
12720144 |
81 | $r = $c->add_reading( { id => '#ANCHOR_' . $item->{'content'} . '#', |
82 | is_ph => 1 } ); |
4d85a60e |
83 | } elsif ( $item->{'type'} eq 'app' ) { |
84 | my $tag = '#APP_' . $counter++ . '#'; |
12720144 |
85 | $r = $c->add_reading( { id => $tag, is_ph => 1 } ); |
4d85a60e |
86 | $apps{$tag} = $item->{'content'}; |
87 | } |
f6e19c7c |
88 | $c->add_path( $last, $r, $c->baselabel ); |
4d85a60e |
89 | $last = $r; |
6f4946fb |
90 | } |
f6e19c7c |
91 | $c->add_path( $last, $c->end, $c->baselabel ); |
4d85a60e |
92 | |
93 | # Now we can parse the apparatus entries, and add the variant readings |
94 | # to the graph. |
95 | |
96 | foreach my $app_id ( keys %apps ) { |
97 | _add_readings( $c, $app_id ); |
6f4946fb |
98 | } |
4d85a60e |
99 | |
f6e19c7c |
100 | # Finally, add explicit witness paths, remove the base paths, and remove |
101 | # the app/anchor tags. |
102 | expand_all_paths( $c ); |
861c3e27 |
103 | |
104 | # Save the text for each witness so that we can ensure consistency |
105 | # later on |
106 | $tradition->collation->text_from_paths(); |
c9158e60 |
107 | $tradition->collation->calculate_ranks(); |
108 | $tradition->collation->flatten_ranks(); |
6f4946fb |
109 | } |
110 | |
4d85a60e |
111 | sub _stringify_sigil { |
112 | my( @nodes ) = @_; |
113 | my @parts = grep { /\w/ } map { $_->data } @nodes; |
114 | return join( '', @parts ); |
115 | } |
6f4946fb |
116 | |
c9158e60 |
117 | # Get rid of all the formatting elements that get in the way of tokenization. |
118 | sub _remove_formatting { |
119 | my( $opts ) = @_; |
120 | |
121 | # First, parse the original XML |
122 | my $parser = XML::LibXML->new(); |
123 | my $doc; |
124 | if( exists $opts->{'string'} ) { |
125 | $doc = $parser->parse_string( $opts->{'string'} ); |
126 | } elsif ( exists $opts->{'file'} ) { |
127 | $doc = $parser->parse_file( $opts->{'file'} ); |
128 | } else { |
129 | warn "Could not find string or file option to parse"; |
130 | return; |
131 | } |
132 | |
133 | # Second, remove the formatting |
134 | my $xpc = XML::LibXML::XPathContext->new( $doc->documentElement ); |
135 | my @useless = $xpc->findnodes( '//hi' ); |
136 | foreach my $n ( @useless ) { |
137 | my $parent = $n->parentNode(); |
138 | my @children = $n->childNodes(); |
139 | my $first = shift @children; |
140 | $parent->replaceChild( $first, $n ); |
141 | foreach my $c ( @children ) { |
142 | $parent->insertAfter( $c, $first ); |
143 | $first = $c; |
144 | } |
145 | } |
146 | |
147 | # Third, write out and reparse to merge the text nodes. |
148 | my $result = decode( $doc->encoding, $doc->toString() ); |
149 | my $tei = $parser->parse_string( $result )->documentElement; |
150 | $xpc = XML::LibXML::XPathContext->new( $tei ); |
151 | return( $tei, $xpc ); |
152 | } |
153 | |
154 | ## Helper function to help us navigate through nested XML, picking out |
155 | ## the words, the apparatus, and the anchors. |
4d85a60e |
156 | |
157 | sub _get_base { |
158 | my( $xn ) = @_; |
159 | my @readings; |
160 | if( $xn->nodeType == XML_TEXT_NODE ) { |
161 | # Base text, just split the words on whitespace and add them |
162 | # to our sequence. |
163 | my $str = $xn->data; |
164 | $str =~ s/^\s+//; |
c9158e60 |
165 | my @tokens = split( /\s+/, $str ); |
166 | push( @readings, map { { 'type' => 'token', 'content' => $_ } } @tokens ); |
4d85a60e |
167 | } elsif( $xn->nodeName eq 'app' ) { |
168 | # Apparatus, just save the entire XML node. |
169 | push( @readings, { 'type' => 'app', 'content' => $xn } ); |
170 | } elsif( $xn->nodeName eq 'anchor' ) { |
171 | # Anchor to mark the end of some apparatus; save its ID. |
172 | push( @readings, { 'type' => 'anchor', |
173 | 'content' => $xn->getAttribute( 'xml:id' ) } ); |
174 | } elsif ( $xn->nodeName ne 'note' ) { # Any tag we don't know to disregard |
175 | print STDERR "Unrecognized tag " . $xn->nodeName . "\n"; |
6f4946fb |
176 | } |
4d85a60e |
177 | return @readings; |
6f4946fb |
178 | } |
179 | |
c9158e60 |
180 | sub _append_tokens { |
181 | my( $list, @tokens ) = @_; |
182 | if( @$list && $list->[-1]->{'content'} =~ /\#JOIN\#$/ ) { |
183 | # The list evidently ended mid-word; join the next token onto it. |
184 | my $t = shift @tokens; |
185 | if( ref $t && $t->{'type'} eq 'token' ) { |
186 | # Join the word |
187 | $t = $t->{'content'}; |
188 | } elsif( ref $t ) { |
189 | # An app or anchor intervened; end the word. |
190 | unshift( @tokens, $t ); |
191 | $t = ''; |
192 | } |
193 | $list->[-1]->{'content'} =~ s/\#JOIN\#$/$t/; |
194 | } |
195 | foreach my $t ( @tokens ) { |
196 | unless( ref( $t ) ) { |
197 | $t = { 'type' => 'token', 'content' => $t }; |
198 | } |
199 | push( @$list, $t ); |
200 | } |
201 | } |
202 | |
4d85a60e |
203 | sub _add_readings { |
204 | my( $c, $app_id ) = @_; |
205 | my $xn = $apps{$app_id}; |
206 | my $anchor = _anchor_name( $xn->getAttribute( 'to' ) ); |
207 | # Get the lemma, which is all the readings between app and anchor, |
208 | # excluding other apps or anchors. |
209 | my @lemma = _return_lemma( $c, $app_id, $anchor ); |
210 | my $lemma_str = join( ' ', grep { $_ !~ /^\#/ } map { $_->text } @lemma ); |
211 | |
212 | # For each reading, send its text to 'interpret' along with the lemma, |
213 | # and then save the list of witnesses that these tokens belong to. |
3a5d151b |
214 | my %wit_rdgs; # Maps from witnesses to the variant text |
4d85a60e |
215 | my $ctr = 0; |
216 | my $tag = $app_id; |
217 | $tag =~ s/^\#APP_(.*)\#$/$1/; |
c9158e60 |
218 | |
4d85a60e |
219 | foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) { |
220 | my @text; |
4d85a60e |
221 | foreach ( $rdg->childNodes ) { |
222 | push( @text, _get_base( $_ ) ); |
223 | } |
12720144 |
224 | my( $interpreted, $flag ) = ( '', undef ); |
225 | if( @text ) { |
226 | ( $interpreted, $flag ) = interpret( |
227 | join( ' ', map { $_->{'content'} } @text ), $lemma_str ); |
228 | } |
b8f262e8 |
229 | next if( $interpreted eq $lemma_str ) && !$flag; # Reading is lemma. |
230 | |
4d85a60e |
231 | my @rdg_nodes; |
b8f262e8 |
232 | if( $interpreted eq '#LACUNA#' ) { |
233 | push( @rdg_nodes, $c->add_reading( { id => $tag . "/" . $ctr++, |
234 | is_lacuna => 1 } ) ); |
235 | } else { |
236 | foreach my $w ( split( /\s+/, $interpreted ) ) { |
237 | my $r = $c->add_reading( { id => $tag . "/" . $ctr++, |
238 | text => $w } ); |
239 | push( @rdg_nodes, $r ); |
240 | } |
4d85a60e |
241 | } |
f6e19c7c |
242 | # For each listed wit, save the reading. |
243 | foreach my $wit ( split( /\s+/, $rdg->getAttribute( 'wit' ) ) ) { |
12720144 |
244 | $wit .= $flag if $flag; |
f6e19c7c |
245 | $wit_rdgs{$wit} = \@rdg_nodes; |
246 | } |
12720144 |
247 | |
3a5d151b |
248 | # Does the reading have an ID? If so it probably has a witDetail |
f6e19c7c |
249 | # attached, and we need to read it. |
3a5d151b |
250 | if( $rdg->hasAttribute( 'xml:id' ) ) { |
12720144 |
251 | warn "Witdetail on meta reading" if $flag; # this could get complicated. |
f6e19c7c |
252 | my $rid = $rdg->getAttribute( 'xml:id' ); |
253 | my $xpc = XML::LibXML::XPathContext->new( $xn ); |
254 | my @details = $xpc->findnodes( './witDetail[@target="'.$rid.'"]' ); |
255 | foreach my $d ( @details ) { |
256 | _parse_wit_detail( $d, \%wit_rdgs, \@lemma ); |
257 | } |
3a5d151b |
258 | } |
f6e19c7c |
259 | } |
260 | |
4d85a60e |
261 | # Now collate the variant readings, since it is not done for us. |
12720144 |
262 | collate_variants( $c, \@lemma, values %wit_rdgs ); |
b8f262e8 |
263 | |
4d85a60e |
264 | # Now add the witness paths for each reading. |
f6e19c7c |
265 | foreach my $wit_id ( keys %wit_rdgs ) { |
266 | my $witstr = get_sigil( $wit_id, $c ); |
267 | my $rdg_list = $wit_rdgs{$wit_id}; |
268 | _add_wit_path( $c, $rdg_list, $app_id, $anchor, $witstr ); |
6f4946fb |
269 | } |
4d85a60e |
270 | } |
6f4946fb |
271 | |
4d85a60e |
272 | sub _anchor_name { |
273 | my $xmlid = shift; |
274 | $xmlid =~ s/^\#//; |
275 | return sprintf( "#ANCHOR_%s#", $xmlid ); |
6f4946fb |
276 | } |
277 | |
4d85a60e |
278 | sub _return_lemma { |
279 | my( $c, $app, $anchor ) = @_; |
12720144 |
280 | my @nodes = grep { $_->id !~ /^\#A(PP|NCHOR)/ } |
281 | $c->reading_sequence( $c->reading( $app ), $c->reading( $anchor ), |
282 | $c->baselabel ); |
4d85a60e |
283 | return @nodes; |
284 | } |
6f4946fb |
285 | |
286 | sub interpret { |
4d85a60e |
287 | # A utility function to change apparatus-ese into a full variant. |
288 | my( $reading, $lemma ) = @_; |
289 | return $reading if $reading eq $lemma; |
290 | my $oldreading = $reading; |
291 | # $lemma =~ s/\s+[[:punct:]]+$//; |
12720144 |
292 | my $flag; # In case of p.c. indications |
4d85a60e |
293 | my @words = split( /\s+/, $lemma ); |
294 | if( $reading =~ /^(.*) praem.$/ ) { |
295 | $reading = "$1 $lemma"; |
296 | } elsif( $reading =~ /^(.*) add.$/ ) { |
297 | $reading = "$lemma $1"; |
b8f262e8 |
298 | } elsif( $reading =~ /add. alia manu/ |
299 | || $reading =~ /inscriptionem compegi e/ # TODO huh? |
300 | || $reading eq 'inc.' # TODO huh? |
301 | ) { |
12720144 |
302 | # Ignore it. |
303 | $reading = $lemma; |
b8f262e8 |
304 | } elsif( $reading =~ /locus [uv]acuus/ |
305 | || $reading eq 'def.' |
c9158e60 |
306 | || $reading eq 'illeg.' |
307 | || $reading eq 'onleesbar' |
f6e19c7c |
308 | ) { |
b8f262e8 |
309 | $reading = '#LACUNA#'; |
310 | } elsif( $reading eq 'om.' ) { |
4d85a60e |
311 | $reading = ''; |
c9158e60 |
312 | } elsif( $reading =~ /^in[uv]\.$/ |
313 | || $reading eq 'transp.' ) { |
4d85a60e |
314 | # Hope it is two words. |
315 | print STDERR "WARNING: want to invert a lemma that is not two words\n" |
316 | unless scalar( @words ) == 2; |
317 | $reading = join( ' ', reverse( @words ) ); |
12720144 |
318 | } elsif( $reading =~ /^iter(\.|at)$/ ) { |
4d85a60e |
319 | # Repeat the lemma |
320 | $reading = "$lemma $lemma"; |
12720144 |
321 | } elsif( $reading eq 'in marg.' ) { |
322 | # There was nothing before a correction. |
323 | $reading = ''; |
324 | $flag = '_ac'; |
c9158e60 |
325 | } elsif( $reading =~ /^(.*?)\s*\(?sic([\s\w.]+)?\)?$/ ) { |
326 | # Discard any 'sic' notation; indeed, indeed. |
327 | $reading = $1; |
12720144 |
328 | } elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) { |
4d85a60e |
329 | # The first and last N words captured should replace the first and |
330 | # last N words of the lemma. |
331 | my @begin = split( /\s+/, $1 ); |
332 | my @end = split( /\s+/, $2 ); |
333 | if( scalar( @begin ) + scalar ( @end ) > scalar( @words ) ) { |
334 | # Something is wrong and we can't do the splice. |
335 | print STDERR "ERROR: $lemma is too short to accommodate $oldreading\n"; |
336 | } else { |
337 | splice( @words, 0, scalar @begin, @begin ); |
338 | splice( @words, -(scalar @end), scalar @end, @end ); |
339 | $reading = join( ' ', @words ); |
340 | } |
6f4946fb |
341 | } |
12720144 |
342 | if( $oldreading ne $reading || $flag || $oldreading =~ /\./ ) { |
343 | my $int = $reading; |
344 | $int .= " ($flag)" if $flag; |
345 | print STDERR "Interpreted $oldreading as $int given $lemma\n"; |
346 | } |
347 | return( $reading, $flag ); |
4d85a60e |
348 | } |
349 | |
f6e19c7c |
350 | sub _parse_wit_detail { |
351 | my( $detail, $readings, $lemma ) = @_; |
352 | my $wit = $detail->getAttribute( 'wit' ); |
353 | my $content = $detail->textContent; |
354 | if( $content =~ /a\.\s*c\./ ) { |
355 | # Replace the key in the $readings hash |
356 | my $rdg = delete $readings->{$wit}; |
357 | $readings->{$wit.'_ac'} = $rdg; |
358 | $has_ac{$sigil_for{$wit}} = 1; |
359 | } elsif( $content =~ /p\.\s*c\./ ) { |
360 | # If no key for the wit a.c. exists, add one pointing to the lemma |
361 | unless( exists $readings->{$wit.'_ac'} ) { |
362 | $readings->{$wit.'_ac'} = $lemma; |
363 | } |
364 | $has_ac{$sigil_for{$wit}} = 1; |
365 | } # else don't bother just yet |
366 | } |
367 | |
368 | sub get_sigil { |
369 | my( $xml_id, $c ) = @_; |
370 | if( $xml_id =~ /^(.*)_ac$/ ) { |
371 | my $real_id = $1; |
372 | return $sigil_for{$real_id} . $c->ac_label; |
373 | } else { |
374 | return $sigil_for{$xml_id}; |
375 | } |
376 | } |
377 | |
378 | sub expand_all_paths { |
379 | my( $c ) = @_; |
380 | |
381 | # Walk the collation and fish out the paths for each witness |
382 | foreach my $wit ( $c->tradition->witnesses ) { |
383 | my $sig = $wit->sigil; |
12720144 |
384 | my @path = grep { !$_->is_ph } |
f6e19c7c |
385 | $c->reading_sequence( $c->start, $c->end, $sig ); |
386 | $wit->path( \@path ); |
387 | if( $has_ac{$sig} ) { |
12720144 |
388 | my @ac_path = grep { !$_->is_ph } |
861c3e27 |
389 | $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label ); |
f6e19c7c |
390 | $wit->uncorrected_path( \@ac_path ); |
391 | } |
392 | } |
393 | |
394 | # Delete the anchors |
12720144 |
395 | foreach my $anchor ( grep { $_->is_ph } $c->readings ) { |
f6e19c7c |
396 | $c->del_reading( $anchor ); |
397 | } |
12720144 |
398 | # Delete the base edges |
399 | map { $c->del_path( $_, $c->baselabel ) } $c->paths; |
f6e19c7c |
400 | |
401 | # Make the path edges |
402 | $c->make_witness_paths(); |
4d85a60e |
403 | } |
404 | |
405 | sub _add_wit_path { |
f6e19c7c |
406 | my( $c, $rdg, $app, $anchor, $wit ) = @_; |
4d85a60e |
407 | my @nodes = @$rdg; |
f6e19c7c |
408 | push( @nodes, $c->reading( $anchor ) ); |
4d85a60e |
409 | |
f6e19c7c |
410 | my $cur = $c->reading( $app ); |
4d85a60e |
411 | foreach my $n ( @nodes ) { |
f6e19c7c |
412 | $c->add_path( $cur, $n, $wit ); |
4d85a60e |
413 | $cur = $n; |
6f4946fb |
414 | } |
6f4946fb |
415 | } |
416 | |
417 | =back |
418 | |
419 | =head1 LICENSE |
420 | |
421 | This package is free software and is provided "as is" without express |
422 | or implied warranty. You can redistribute it and/or modify it under |
423 | the same terms as Perl itself. |
424 | |
425 | =head1 AUTHOR |
426 | |
427 | Tara L Andrews, aurum@cpan.org |
428 | |
429 | =cut |
430 | |
431 | 1; |
432 | |