1 package Text::Tradition::Parser::CTE;
5 use Encode qw/ decode /;
6 use Text::Tradition::Parser::Util qw/ collate_variants /;
8 use XML::LibXML::XPathContext;
12 Text::Tradition::Parser::CTE
16 Parser module for Text::Tradition, given a TEI file exported from
17 Classical Text Editor.
25 my @apparatus = read( $xml_file );
27 Takes a Tradition object and a TEI file exported from Classical Text
28 Editor using double-endpoint-attachment critical apparatus encoding;
29 initializes the Tradition from the file.
33 my %sigil_for; # Save the XML IDs for witnesses.
34 my %apps; # Save the apparatus XML for a given ID.
35 my %has_ac; # Keep track of witnesses that have corrections.
38 my( $tradition, $opts ) = @_;
39 my $c = $tradition->collation; # Some shorthand
41 # First, parse the XML.
42 my( $tei, $xpc ) = _remove_formatting( $opts );
43 return unless $tei; # we have already warned.
45 # CTE uses a DTD rather than any xmlns-based parsing. Thus we
46 # need no namespace handling.
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' );
51 my @sig_parts = $xpc->findnodes( 'descendant::text()', $wit_el );
52 my $sig = _stringify_sigil( @sig_parts );
53 print STDERR "Adding witness $sig\n";
54 $tradition->add_witness( sigil => $sig, sourcetype => 'collation' );
55 $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax
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
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 ) );
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.
72 # First, put the base tokens, apps, and anchors in the graph.
75 foreach my $item ( @base_text ) {
77 if( $item->{'type'} eq 'token' ) {
78 $r = $c->add_reading( { id => 'n'.$counter++,
79 text => $item->{'content'} } );
80 } elsif ( $item->{'type'} eq 'anchor' ) {
81 $r = $c->add_reading( { id => '__ANCHOR_' . $item->{'content'} . '__',
83 } elsif ( $item->{'type'} eq 'app' ) {
84 my $tag = '__APP_' . $counter++ . '__';
85 $r = $c->add_reading( { id => $tag, is_ph => 1 } );
86 $apps{$tag} = $item->{'content'};
88 $c->add_path( $last, $r, $c->baselabel );
91 $c->add_path( $last, $c->end, $c->baselabel );
93 # Now we can parse the apparatus entries, and add the variant readings
96 foreach my $app_id ( keys %apps ) {
97 _add_readings( $c, $app_id );
100 # Finally, add explicit witness paths, remove the base paths, and remove
101 # the app/anchor tags.
102 expand_all_paths( $c );
104 # Save the text for each witness so that we can ensure consistency
106 $tradition->collation->text_from_paths();
107 $tradition->collation->calculate_ranks();
108 $tradition->collation->flatten_ranks();
111 sub _stringify_sigil {
113 my @parts = grep { /\w/ } map { $_->data } @nodes;
114 my $whole = join( '', @parts );
119 # Get rid of all the formatting elements that get in the way of tokenization.
120 sub _remove_formatting {
123 # First, parse the original XML
124 my $parser = XML::LibXML->new();
126 if( exists $opts->{'string'} ) {
127 $doc = $parser->parse_string( $opts->{'string'} );
128 } elsif ( exists $opts->{'file'} ) {
129 $doc = $parser->parse_file( $opts->{'file'} );
131 warn "Could not find string or file option to parse";
135 # Second, remove the formatting
136 my $xpc = XML::LibXML::XPathContext->new( $doc->documentElement );
137 my @useless = $xpc->findnodes( '//hi' );
138 foreach my $n ( @useless ) {
139 my $parent = $n->parentNode();
140 my @children = $n->childNodes();
141 my $first = shift @children;
142 $parent->replaceChild( $first, $n );
143 foreach my $c ( @children ) {
144 $parent->insertAfter( $c, $first );
149 # Third, write out and reparse to merge the text nodes.
150 my $result = decode( $doc->encoding, $doc->toString() );
151 my $tei = $parser->parse_string( $result )->documentElement;
152 $xpc = XML::LibXML::XPathContext->new( $tei );
153 return( $tei, $xpc );
156 ## Helper function to help us navigate through nested XML, picking out
157 ## the words, the apparatus, and the anchors.
162 if( $xn->nodeType == XML_TEXT_NODE ) {
163 # Base text, just split the words on whitespace and add them
167 my @tokens = split( /\s+/, $str );
168 push( @readings, map { { 'type' => 'token', 'content' => $_ } } @tokens );
169 } elsif( $xn->nodeName eq 'app' ) {
170 # Apparatus, just save the entire XML node.
171 push( @readings, { 'type' => 'app', 'content' => $xn } );
172 } elsif( $xn->nodeName eq 'anchor' ) {
173 # Anchor to mark the end of some apparatus; save its ID.
174 push( @readings, { 'type' => 'anchor',
175 'content' => $xn->getAttribute( 'xml:id' ) } );
176 } elsif ( $xn->nodeName ne 'note' ) { # Any tag we don't know to disregard
177 print STDERR "Unrecognized tag " . $xn->nodeName . "\n";
183 my( $list, @tokens ) = @_;
184 if( @$list && $list->[-1]->{'content'} =~ /\#JOIN\#$/ ) {
185 # The list evidently ended mid-word; join the next token onto it.
186 my $t = shift @tokens;
187 if( ref $t && $t->{'type'} eq 'token' ) {
189 $t = $t->{'content'};
191 # An app or anchor intervened; end the word.
192 unshift( @tokens, $t );
195 $list->[-1]->{'content'} =~ s/\#JOIN\#$/$t/;
197 foreach my $t ( @tokens ) {
198 unless( ref( $t ) ) {
199 $t = { 'type' => 'token', 'content' => $t };
206 my( $c, $app_id ) = @_;
207 my $xn = $apps{$app_id};
208 my $anchor = _anchor_name( $xn->getAttribute( 'to' ) );
209 # Get the lemma, which is all the readings between app and anchor,
210 # excluding other apps or anchors.
211 my @lemma = _return_lemma( $c, $app_id, $anchor );
212 my $lemma_str = join( ' ', grep { $_ !~ /^__/ } map { $_->text } @lemma );
214 # For each reading, send its text to 'interpret' along with the lemma,
215 # and then save the list of witnesses that these tokens belong to.
216 my %wit_rdgs; # Maps from witnesses to the variant text
219 $tag =~ s/^\__APP_(.*)\__$/$1/;
221 foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) {
223 foreach ( $rdg->childNodes ) {
224 push( @text, _get_base( $_ ) );
226 my( $interpreted, $flag ) = ( '', undef );
228 ( $interpreted, $flag ) = interpret(
229 join( ' ', map { $_->{'content'} } @text ), $lemma_str );
231 next if( $interpreted eq $lemma_str ) && !$flag; # Reading is lemma.
234 if( $interpreted eq '#LACUNA#' ) {
235 push( @rdg_nodes, $c->add_reading( { id => 'r'.$tag.".".$ctr++,
236 is_lacuna => 1 } ) );
238 foreach my $w ( split( /\s+/, $interpreted ) ) {
239 my $r = $c->add_reading( { id => 'r'.$tag.".".$ctr++,
241 push( @rdg_nodes, $r );
244 # For each listed wit, save the reading.
245 foreach my $wit ( split( /\s+/, $rdg->getAttribute( 'wit' ) ) ) {
246 $wit .= $flag if $flag;
247 $wit_rdgs{$wit} = \@rdg_nodes;
250 # Does the reading have an ID? If so it probably has a witDetail
251 # attached, and we need to read it.
252 if( $rdg->hasAttribute( 'xml:id' ) ) {
253 warn "Witdetail on meta reading" if $flag; # this could get complicated.
254 my $rid = $rdg->getAttribute( 'xml:id' );
255 my $xpc = XML::LibXML::XPathContext->new( $xn );
256 my @details = $xpc->findnodes( './witDetail[@target="'.$rid.'"]' );
257 foreach my $d ( @details ) {
258 _parse_wit_detail( $d, \%wit_rdgs, \@lemma );
263 # Now collate the variant readings, since it is not done for us.
264 collate_variants( $c, \@lemma, values %wit_rdgs );
266 # Now add the witness paths for each reading.
267 foreach my $wit_id ( keys %wit_rdgs ) {
268 my $witstr = get_sigil( $wit_id, $c );
269 my $rdg_list = $wit_rdgs{$wit_id};
270 _add_wit_path( $c, $rdg_list, $app_id, $anchor, $witstr );
277 return sprintf( "__ANCHOR_%s__", $xmlid );
281 my( $c, $app, $anchor ) = @_;
282 my @nodes = grep { $_->id !~ /^__A(PP|NCHOR)/ }
283 $c->reading_sequence( $c->reading( $app ), $c->reading( $anchor ),
289 # A utility function to change apparatus-ese into a full variant.
290 my( $reading, $lemma ) = @_;
291 return $reading if $reading eq $lemma;
292 my $oldreading = $reading;
293 # $lemma =~ s/\s+[[:punct:]]+$//;
294 my $flag; # In case of p.c. indications
295 my @words = split( /\s+/, $lemma );
296 if( $reading =~ /^(.*) praem.$/ ) {
297 $reading = "$1 $lemma";
298 } elsif( $reading =~ /^(.*) add.$/ ) {
299 $reading = "$lemma $1";
300 } elsif( $reading =~ /add. alia manu/
301 || $reading =~ /inscriptionem compegi e/ # TODO huh?
302 || $reading eq 'inc.' # TODO huh?
306 } elsif( $reading =~ /locus [uv]acuus/
307 || $reading eq 'def.'
308 || $reading eq 'illeg.'
309 || $reading eq 'onleesbar'
311 $reading = '#LACUNA#';
312 } elsif( $reading eq 'om.' ) {
314 } elsif( $reading =~ /^in[uv]\.$/
315 || $reading eq 'transp.' ) {
316 # Hope it is two words.
317 print STDERR "WARNING: want to invert a lemma that is not two words\n"
318 unless scalar( @words ) == 2;
319 $reading = join( ' ', reverse( @words ) );
320 } elsif( $reading =~ /^iter(\.|at)$/ ) {
322 $reading = "$lemma $lemma";
323 } elsif( $reading eq 'in marg.' ) {
324 # There was nothing before a correction.
327 } elsif( $reading =~ /^(.*?)\s*\(?sic([\s\w.]+)?\)?$/ ) {
328 # Discard any 'sic' notation; indeed, indeed.
330 } elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) {
331 # The first and last N words captured should replace the first and
332 # last N words of the lemma.
333 my @begin = split( /\s+/, $1 );
334 my @end = split( /\s+/, $2 );
335 if( scalar( @begin ) + scalar ( @end ) > scalar( @words ) ) {
336 # Something is wrong and we can't do the splice.
337 print STDERR "ERROR: $lemma is too short to accommodate $oldreading\n";
339 splice( @words, 0, scalar @begin, @begin );
340 splice( @words, -(scalar @end), scalar @end, @end );
341 $reading = join( ' ', @words );
344 if( $oldreading ne $reading || $flag || $oldreading =~ /\./ ) {
346 $int .= " ($flag)" if $flag;
347 print STDERR "Interpreted $oldreading as $int given $lemma\n";
349 return( $reading, $flag );
352 sub _parse_wit_detail {
353 my( $detail, $readings, $lemma ) = @_;
354 my $wit = $detail->getAttribute( 'wit' );
355 my $content = $detail->textContent;
356 if( $content =~ /a\.\s*c\./ ) {
357 # Replace the key in the $readings hash
358 my $rdg = delete $readings->{$wit};
359 $readings->{$wit.'_ac'} = $rdg;
360 $has_ac{$sigil_for{$wit}} = 1;
361 } elsif( $content =~ /p\.\s*c\./ ) {
362 # If no key for the wit a.c. exists, add one pointing to the lemma
363 unless( exists $readings->{$wit.'_ac'} ) {
364 $readings->{$wit.'_ac'} = $lemma;
366 $has_ac{$sigil_for{$wit}} = 1;
367 } # else don't bother just yet
371 my( $xml_id, $c ) = @_;
372 if( $xml_id =~ /^(.*)_ac$/ ) {
374 return $sigil_for{$real_id} . $c->ac_label;
376 return $sigil_for{$xml_id};
380 sub expand_all_paths {
383 # Walk the collation and fish out the paths for each witness
384 foreach my $wit ( $c->tradition->witnesses ) {
385 my $sig = $wit->sigil;
386 my @path = grep { !$_->is_ph }
387 $c->reading_sequence( $c->start, $c->end, $sig );
388 $wit->path( \@path );
389 if( $has_ac{$sig} ) {
390 my @ac_path = grep { !$_->is_ph }
391 $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label );
392 $wit->uncorrected_path( \@ac_path );
397 foreach my $anchor ( grep { $_->is_ph } $c->readings ) {
398 $c->del_reading( $anchor );
400 # Delete the base edges
401 map { $c->del_path( $_, $c->baselabel ) } $c->paths;
403 # Make the path edges
404 $c->make_witness_paths();
408 my( $c, $rdg, $app, $anchor, $wit ) = @_;
410 push( @nodes, $c->reading( $anchor ) );
412 my $cur = $c->reading( $app );
413 foreach my $n ( @nodes ) {
414 $c->add_path( $cur, $n, $wit );
423 This package is free software and is provided "as is" without express
424 or implied warranty. You can redistribute it and/or modify it under
425 the same terms as Perl itself.
429 Tara L Andrews, aurum@cpan.org