get rid of common reading calculation for positional fun
[scpubgit/stemmatology.git] / lib / Text / Tradition / Parser / TEI.pm
CommitLineData
f6066bac 1package Text::Tradition::Parser::TEI;
2
3use strict;
4use warnings;
910a0a6d 5use Text::Tradition::Parser::Util qw( collate_variants );
f6066bac 6use XML::LibXML;
7use XML::LibXML::XPathContext;
8
9=head1 NAME
10
11Text::Tradition::Parser::TEI
12
13=head1 DESCRIPTION
14
15Parser module for Text::Tradition, given a TEI parallel-segmentation
16file that describes a text and its variants.
17
18=head1 METHODS
19
20=over
21
22=item B<parse>
23
24parse( $tei_string );
25
26Takes an initialized tradition and a string containing the TEI;
27creates the appropriate nodes and edges on the graph, as well as
28the appropriate witness objects.
29
30=cut
31
910a0a6d 32my $text = {}; # Hash of arrays, one per eventual witness we find.
910a0a6d 33my $substitutions = {}; # Keep track of merged readings
34my $app_anchors = {}; # Track apparatus references
35my $app_ac = {}; # Save a.c. readings
36
37# Create the package variables for tag names.
38
39# Would really like to do this with varname variables, but apparently this
40# is considered a bad idea. The long way round then.
41my( $LISTWIT, $WITNESS, $TEXT, $W, $SEG, $APP, $RDG, $LEM )
42 = ( 'listWit', 'witness', 'text', 'w', 'seg', 'app', 'rdg', 'lem' );
43sub make_tagnames {
44 my( $ns ) = @_;
45 if( $ns ) {
46 $LISTWIT = "$ns:$LISTWIT";
47 $WITNESS = "$ns:$WITNESS";
48 $TEXT = "$ns:$TEXT";
49 $W = "$ns:$W";
50 $SEG = "$ns:$SEG";
51 $APP = "$ns:$APP";
52 $RDG = "$ns:$RDG";
53 $LEM = "$ns:$LEM";
54 }
55}
56
57# Parse the TEI file.
f6066bac 58sub parse {
59 my( $tradition, $xml_str ) = @_;
60
61 # First, parse the XML.
62 my $parser = XML::LibXML->new();
63 my $doc = $parser->parse_string( $xml_str );
64 my $tei = $doc->documentElement();
f2b9605f 65 my $xpc = XML::LibXML::XPathContext->new( $tei );
910a0a6d 66 my $ns;
67 if( $tei->namespaceURI ) {
68 $ns = 'tei';
69 $xpc->registerNs( $ns, $tei->namespaceURI );
70 }
71 make_tagnames( $ns );
72
f6066bac 73 # Then get the witnesses and create the witness objects.
910a0a6d 74 foreach my $wit_el ( $xpc->findnodes( "//$LISTWIT/$WITNESS" ) ) {
75 my $sig = $wit_el->getAttribute( 'xml:id' );
76 my $source = $wit_el->toString();
77 $tradition->add_witness( sigil => $sig, source => $source );
f6066bac 78 }
79
910a0a6d 80 map { $text->{$_->sigil} = [] } $tradition->witnesses;
81 # Look for all word/seg node IDs and note their pre-existence.
82 my @attrs = $xpc->findnodes( "//$W|$SEG/attribute::xml:id" );
83 save_preexisting_nodeids( @attrs );
84
85 # Now go through the children of the text element and pull out the
86 # actual text.
87 foreach my $xml_el ( $xpc->findnodes( "//$TEXT" ) ) {
88 foreach my $xn ( $xml_el->childNodes ) {
89 _get_readings( $tradition, $xn );
90 }
91 }
92 # Our $text global now has lists of readings, one per witness.
93 # Join them up.
94 my $c = $tradition->collation;
95 foreach my $sig ( keys %$text ) {
96 next if $sig eq 'base'; # Skip base text readings with no witnesses.
97 # Determine the list of readings for
98 my $sequence = $text->{$sig};
99 my @real_sequence = ( $c->start );
100 push( @$sequence, $c->end );
101 my $source = $c->start;
102 foreach( _clean_sequence( $sig, $sequence ) ) {
103 my $rdg = _return_rdg( $_ );
104 push( @real_sequence, $rdg );
105 $c->add_path( $source, $rdg, $sig );
106 $source = $rdg;
107 }
108 $tradition->witness( $sig )->path( \@real_sequence );
109 # See if we need to make an a.c. version of the witness.
110 if( exists $app_ac->{$sig} ) {
111 my @uncorrected;
112 push( @uncorrected, @real_sequence );
113 foreach my $app ( keys %{$app_ac->{$sig}} ) {
114 my $start = _return_rdg( $app_anchors->{$app}->{$sig}->{'start'} );
115 my $end = _return_rdg( $app_anchors->{$app}->{$sig}->{'end'} );
116 my @new = map { _return_rdg( $_ ) } @{$app_ac->{$sig}->{$app}};
117 _replace_sequence( \@uncorrected, $start, $end, @new );
118 }
119 my $source = $c->start;
120 foreach my $rdg ( @uncorrected ) {
121 my $has_base = grep { $_->label eq $sig } $source->edges_to( $rdg );
122 if( $rdg ne $c->start && !$has_base ) {
123 print STDERR sprintf( "Adding path %s from %s -> %s\n",
124 $sig.$c->ac_label, $source->name, $rdg->name );
125 $c->add_path( $source, $rdg, $sig.$c->ac_label );
126 }
127 $source = $rdg;
128 }
129 $tradition->witness( $sig )->uncorrected_path( \@uncorrected );
130 }
131 }
132 # Delete readings that are no longer part of the graph.
133 # TODO think this is useless actually
134 foreach ( keys %$substitutions ) {
135 $tradition->collation->del_reading( $tradition->collation->reading( $_ ) );
136 }
7e450e44 137 $tradition->collation->calculate_ranks();
910a0a6d 138}
139
140sub _clean_sequence {
141 my( $wit, $sequence ) = @_;
142 my @clean_sequence;
143 foreach my $rdg ( @$sequence ) {
144 if( $rdg =~ /^PH-(.*)$/ ) {
145 # It is a placeholder. Keep it only if we need it.
146 my $app_id = $1;
147 if( exists $app_ac->{$wit}->{$app_id} ) {
148 print STDERR "Retaining empty placeholder for $app_id\n";
149 push( @clean_sequence, $rdg );
150 }
151 } else {
152 push( @clean_sequence, $rdg );
153 }
f6066bac 154 }
910a0a6d 155 return @clean_sequence;
156}
f6066bac 157
910a0a6d 158sub _replace_sequence {
159 my( $arr, $start, $end, @new ) = @_;
160 my( $start_idx, $end_idx );
161 foreach my $i ( 0 .. $#{$arr} ) {
162 $start_idx = $i if( $arr->[$i]->name eq $start );
163 if( $arr->[$i]->name eq $end ) {
164 $end_idx = $i;
165 last;
166 }
167 }
168 unless( $start_idx && $end_idx ) {
169 warn "Could not find start and end";
170 return;
f2b9605f 171 }
910a0a6d 172 my $length = $end_idx - $start_idx + 1;
173 splice( @$arr, $start_idx, $length, @new );
174}
f6066bac 175
910a0a6d 176sub _return_rdg {
177 my( $rdg ) = @_;
178 # If we were passed a reading name, return the name. If we were
179 # passed a reading object, return the object.
180 my $wantobj = ref( $rdg ) eq 'Text::Tradition::Collation::Reading';
181 my $real = $rdg;
182 if( exists $substitutions->{ $wantobj ? $rdg->name : $rdg } ) {
183 $real = $substitutions->{ $wantobj ? $rdg->name : $rdg };
184 $real = $real->name unless $wantobj;
185 }
186 return $real;
f6066bac 187}
188
910a0a6d 189## Recursive helper function to help us navigate through nested XML,
190## picking out the text. $tradition is the tradition, needed for
191## making readings; $xn is the XML node currently being looked at,
192## $in_var is a flag to say that we are inside a variant, $ac is a
193## flag to say that we are inside an ante-correctionem reading, and
194## @cur_wits is the list of witnesses to which this XML node applies.
195## Returns the list of readings, if any, created on the run.
196
197{
198 my @active_wits;
199 my $current_app;
200
201 sub _get_readings {
202 my( $tradition, $xn, $in_var, $ac, @cur_wits ) = @_;
203 @cur_wits = @active_wits unless $in_var;
204
205 my @new_readings;
206 if( $xn->nodeType == XML_TEXT_NODE ) {
207 # Some words, thus make some readings.
208 my $str = $xn->data;
209 return unless $str =~ /\S/; # skip whitespace-only text nodes
210 #print STDERR "Handling text node " . $str . "\n";
211 # Check that all the witnesses we have are active.
212 foreach my $c ( @cur_wits ) {
213 warn "Could not find $c in active wits"
214 unless grep { $c eq $_ } @active_wits;
215 }
216 $str =~ s/^\s+//;
217 my $final = $str =~ s/\s+$//;
218 foreach my $w ( split( /\s+/, $str ) ) {
219 # For now, skip punctuation.
220 next if $w !~ /[[:alnum:]]/;
221 my $rdg = make_reading( $tradition->collation, $w );
222 push( @new_readings, $rdg );
223 unless( $in_var ) {
910a0a6d 224 $rdg->make_common;
225 }
226 foreach ( @cur_wits ) {
227 warn "Empty wit!" unless $_;
228 warn "Empty reading!" unless $rdg;
229 push( @{$text->{$_}}, $rdg ) unless $ac;
230 }
231 }
232 } elsif( $xn->nodeName eq 'w' ) {
233 # Everything in this tag is one word. Also save any original XML ID.
234 #print STDERR "Handling word " . $xn->toString . "\n";
235 # Check that all the witnesses we have are active.
236 foreach my $c ( @cur_wits ) {
237 warn "Could not find $c in active wits"
238 unless grep { $c eq $_ } @active_wits;
239 }
240 my $xml_id = $xn->getAttribute( 'xml:id' );
241 my $rdg = make_reading( $tradition->collation, $xn->textContent, $xml_id );
242 push( @new_readings, $rdg );
243 unless( $in_var ) {
910a0a6d 244 $rdg->make_common;
245 }
246 foreach( @cur_wits ) {
247 warn "Empty wit!" unless $_;
248 warn "Empty reading!" unless $rdg;
249 push( @{$text->{$_}}, $rdg ) unless $ac;
250 }
251 } elsif ( $xn->nodeName eq 'app' ) {
252 $current_app = $xn->getAttribute( 'xml:id' );
253 # print STDERR "Handling app $current_app\n";
254 # Keep the reading sets in this app.
255 my @sets;
256 # Recurse through all children (i.e. rdgs) for sets of words.
257 foreach ( $xn->childNodes ) {
258 my @rdg_set = _get_readings( $tradition, $_, $in_var, $ac, @cur_wits );
259 push( @sets, \@rdg_set ) if @rdg_set;
260 }
261 # Now collate these sets if we have more than one.
262 my $subs = collate_variants( $tradition->collation, @sets ) if @sets > 1;
263 map { $substitutions->{$_} = $subs->{$_} } keys %$subs;
264 # TODO Look through substitutions to see if we can make anything common now.
265 # Return the entire set of unique readings.
266 my %unique;
267 foreach my $s ( @sets ) {
268 map { $unique{$_->name} = $_ } @$s;
269 }
270 push( @new_readings, values( %unique ) );
271 # Exit the current app.
272 $current_app = '';
273 } elsif ( $xn->nodeName eq 'lem' || $xn->nodeName eq 'rdg' ) {
274 # Alter the current witnesses and recurse.
275 #print STDERR "Handling reading for " . $xn->getAttribute( 'wit' ) . "\n";
276 $ac = $xn->getAttribute( 'type' ) && $xn->getAttribute( 'type' ) eq 'a.c.';
277 my @rdg_wits = get_sigla( $xn );
278 @rdg_wits = ( 'base' ) unless @rdg_wits; # Allow for editorially-supplied readings
279 my @words;
280 foreach ( $xn->childNodes ) {
281 my @rdg_set = _get_readings( $tradition, $_, 1, $ac, @rdg_wits );
282 push( @words, @rdg_set ) if @rdg_set;
283 }
284 # If we have more than one word in a reading, it should become a segment.
285 # $tradition->collation->add_segment( @words ) if @words > 1;
286
287 if( $ac ) {
288 # Add the reading set to the a.c. readings.
289 foreach ( @rdg_wits ) {
290 $app_ac->{$_}->{$current_app} = \@words;
291 }
292 } else {
293 # Add the reading set to the app anchors for each witness
294 # or put in placeholders for empty p.c. readings
295 foreach ( @rdg_wits ) {
296 my $start = @words ? $words[0]->name : "PH-$current_app";
297 my $end = @words ? $words[-1]->name : "PH-$current_app";
298 $app_anchors->{$current_app}->{$_}->{'start'} = $start;
299 $app_anchors->{$current_app}->{$_}->{'end'} = $end;
300 push( @{$text->{$_}}, $start ) unless @words;
301 }
302 }
303 push( @new_readings, @words );
304 } elsif( $xn->nodeName eq 'witStart' ) {
305 # Add the relevant wit(s) to the active list.
306 #print STDERR "Handling witStart\n";
307 push( @active_wits, @cur_wits );
308 } elsif( $xn->nodeName eq 'witEnd' ) {
309 # Take the relevant wit(s) out of the list.
310 #print STDERR "Handling witEnd\n";
311 my $regexp = '^(' . join( '|', @cur_wits ) . ')$';
312 @active_wits = grep { $_ !~ /$regexp/ } @active_wits;
313 } elsif( $xn->nodeName eq 'witDetail' ) {
314 # Ignore these for now.
315 return;
316 } else {
317 # Recurse as if this tag weren't there.
318 #print STDERR "Recursing on tag " . $xn->nodeName . "\n";
319 foreach( $xn->childNodes ) {
320 push( @new_readings, _get_readings( $tradition, $_, $in_var, $ac, @cur_wits ) );
321 }
322 }
323 return @new_readings;
324 }
325
326}
327
328# Helper to extract a list of witness sigla from a reading element.
f6066bac 329sub get_sigla {
330 my( $rdg ) = @_;
331 # Cope if we have been handed a NodeList. There is only
332 # one reading here.
333 if( ref( $rdg ) eq 'XML::LibXML::NodeList' ) {
910a0a6d 334 $rdg = $rdg->shift;
f6066bac 335 }
336
337 my @wits;
338 if( ref( $rdg ) eq 'XML::LibXML::Element' ) {
910a0a6d 339 my $witstr = $rdg->getAttribute( 'wit' );
340 $witstr =~ s/^\s+//;
341 $witstr =~ s/\s+$//;
342 @wits = split( /\s+/, $witstr );
343 map { $_ =~ s/^\#// } @wits;
f6066bac 344 }
345 return @wits;
346}
347
910a0a6d 348# Helper with its counters to actually make the readings.
f2b9605f 349{
350 my $word_ctr = 0;
351 my %used_nodeids;
352
910a0a6d 353 sub save_preexisting_nodeids {
354 foreach( @_ ) {
355 $used_nodeids{$_->getValue()} = 1;
356 }
357 }
358
f2b9605f 359 sub make_reading {
910a0a6d 360 my( $graph, $word, $xml_id ) = @_;
361 if( $xml_id ) {
362 if( exists $used_nodeids{$xml_id} ) {
363 if( $used_nodeids{$xml_id} != 1 ) {
364 warn "Already used assigned XML ID somewhere else!";
365 $xml_id = undef;
366 }
367 } else {
368 warn "Undetected pre-existing XML ID";
369 }
370 }
371 if( !$xml_id ) {
372 until( $xml_id ) {
373 my $try_id = 'w'.$word_ctr++;
374 next if exists $used_nodeids{$try_id};
375 $xml_id = $try_id;
376 }
377 }
378 my $rdg = $graph->add_reading( $xml_id );
379 $rdg->text( $word );
380 $used_nodeids{$xml_id} = $rdg;
381 return $rdg;
f2b9605f 382 }
383}
384
f6066bac 3851;