Commit | Line | Data |
6f4946fb |
1 | package Text::Tradition::Parser::CTE; |
2 | |
3 | use strict; |
4 | use warnings; |
c9158e60 |
5 | use Encode qw/ decode /; |
00311328 |
6 | use Text::Tradition::Error; |
4d85a60e |
7 | use Text::Tradition::Parser::Util qw/ collate_variants /; |
6f4946fb |
8 | use XML::LibXML; |
9 | use XML::LibXML::XPathContext; |
10 | |
11 | =head1 NAME |
12 | |
13 | Text::Tradition::Parser::CTE |
14 | |
15 | =head1 DESCRIPTION |
16 | |
17 | Parser module for Text::Tradition, given a TEI file exported from |
18 | Classical Text Editor. |
19 | |
20 | =head1 METHODS |
21 | |
a445ce40 |
22 | =head2 parse |
6f4946fb |
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. |
c9158e60 |
41 | my( $tei, $xpc ) = _remove_formatting( $opts ); |
42 | return unless $tei; # we have already warned. |
4d85a60e |
43 | |
44 | # CTE uses a DTD rather than any xmlns-based parsing. Thus we |
45 | # need no namespace handling. |
4d85a60e |
46 | # Get the witnesses and create the witness objects. |
47 | foreach my $wit_el ( $xpc->findnodes( '//sourceDesc/listWit/witness' ) ) { |
48 | # The witness xml:id is used internally, and is *not* the sigil name. |
49 | my $id= $wit_el->getAttribute( 'xml:id' ); |
c9158e60 |
50 | my @sig_parts = $xpc->findnodes( 'descendant::text()', $wit_el ); |
4d85a60e |
51 | my $sig = _stringify_sigil( @sig_parts ); |
c9158e60 |
52 | print STDERR "Adding witness $sig\n"; |
82fa4d57 |
53 | $tradition->add_witness( sigil => $sig, sourcetype => 'collation' ); |
b8f262e8 |
54 | $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax |
4d85a60e |
55 | } |
c9158e60 |
56 | |
4d85a60e |
57 | # Now go through the text and find the base tokens, apparatus tags, and |
58 | # anchors. Make a giant array of all of these things in sequence. |
59 | # TODO consider combining this with creation of graph below |
60 | my @base_text; |
61 | foreach my $pg_el ( $xpc->findnodes( '/TEI/text/body/p' ) ) { |
62 | foreach my $xn ( $pg_el->childNodes ) { |
63 | push( @base_text, _get_base( $xn ) ); |
64 | } |
6f4946fb |
65 | } |
4d85a60e |
66 | # We now have to work through this array applying the alternate |
67 | # apparatus readings to the base text. Essentially we will put |
68 | # everything on the graph, from which we will delete the apps and |
69 | # anchors when we are done. |
f6e19c7c |
70 | |
71 | # First, put the base tokens, apps, and anchors in the graph. |
4d85a60e |
72 | my $counter = 0; |
73 | my $last = $c->start; |
74 | foreach my $item ( @base_text ) { |
75 | my $r; |
76 | if( $item->{'type'} eq 'token' ) { |
12720144 |
77 | $r = $c->add_reading( { id => 'n'.$counter++, |
78 | text => $item->{'content'} } ); |
4d85a60e |
79 | } elsif ( $item->{'type'} eq 'anchor' ) { |
10e4b1ac |
80 | $r = $c->add_reading( { id => '__ANCHOR_' . $item->{'content'} . '__', |
12720144 |
81 | is_ph => 1 } ); |
4d85a60e |
82 | } elsif ( $item->{'type'} eq 'app' ) { |
10e4b1ac |
83 | my $tag = '__APP_' . $counter++ . '__'; |
12720144 |
84 | $r = $c->add_reading( { id => $tag, is_ph => 1 } ); |
4d85a60e |
85 | $apps{$tag} = $item->{'content'}; |
86 | } |
f6e19c7c |
87 | $c->add_path( $last, $r, $c->baselabel ); |
4d85a60e |
88 | $last = $r; |
6f4946fb |
89 | } |
f6e19c7c |
90 | $c->add_path( $last, $c->end, $c->baselabel ); |
4d85a60e |
91 | |
92 | # Now we can parse the apparatus entries, and add the variant readings |
93 | # to the graph. |
94 | |
95 | foreach my $app_id ( keys %apps ) { |
96 | _add_readings( $c, $app_id ); |
6f4946fb |
97 | } |
4d85a60e |
98 | |
f6e19c7c |
99 | # Finally, add explicit witness paths, remove the base paths, and remove |
100 | # the app/anchor tags. |
a445ce40 |
101 | _expand_all_paths( $c ); |
861c3e27 |
102 | |
103 | # Save the text for each witness so that we can ensure consistency |
104 | # later on |
105 | $tradition->collation->text_from_paths(); |
c9158e60 |
106 | $tradition->collation->calculate_ranks(); |
107 | $tradition->collation->flatten_ranks(); |
6f4946fb |
108 | } |
109 | |
4d85a60e |
110 | sub _stringify_sigil { |
111 | my( @nodes ) = @_; |
112 | my @parts = grep { /\w/ } map { $_->data } @nodes; |
222d58f1 |
113 | my $whole = join( '', @parts ); |
114 | $whole =~ s/\W//g; |
115 | return $whole; |
4d85a60e |
116 | } |
6f4946fb |
117 | |
c9158e60 |
118 | # Get rid of all the formatting elements that get in the way of tokenization. |
119 | sub _remove_formatting { |
120 | my( $opts ) = @_; |
121 | |
122 | # First, parse the original XML |
123 | my $parser = XML::LibXML->new(); |
124 | my $doc; |
125 | if( exists $opts->{'string'} ) { |
126 | $doc = $parser->parse_string( $opts->{'string'} ); |
127 | } elsif ( exists $opts->{'file'} ) { |
128 | $doc = $parser->parse_file( $opts->{'file'} ); |
129 | } else { |
130 | warn "Could not find string or file option to parse"; |
131 | return; |
132 | } |
00311328 |
133 | |
c9158e60 |
134 | # Second, remove the formatting |
135 | my $xpc = XML::LibXML::XPathContext->new( $doc->documentElement ); |
136 | my @useless = $xpc->findnodes( '//hi' ); |
137 | foreach my $n ( @useless ) { |
138 | my $parent = $n->parentNode(); |
139 | my @children = $n->childNodes(); |
140 | my $first = shift @children; |
141 | $parent->replaceChild( $first, $n ); |
142 | foreach my $c ( @children ) { |
143 | $parent->insertAfter( $c, $first ); |
144 | $first = $c; |
145 | } |
146 | } |
147 | |
148 | # Third, write out and reparse to merge the text nodes. |
00311328 |
149 | my $enc = $doc->encoding || 'UTF-8'; |
150 | my $result = decode( $enc, $doc->toString() ); |
c9158e60 |
151 | my $tei = $parser->parse_string( $result )->documentElement; |
00311328 |
152 | unless( $tei->nodeName =~ /^tei(corpus)?$/i ) { |
153 | throw( "Parsed document has non-TEI root element " . $tei->nodeName ); |
154 | } |
c9158e60 |
155 | $xpc = XML::LibXML::XPathContext->new( $tei ); |
156 | return( $tei, $xpc ); |
157 | } |
158 | |
159 | ## Helper function to help us navigate through nested XML, picking out |
160 | ## the words, the apparatus, and the anchors. |
4d85a60e |
161 | |
162 | sub _get_base { |
163 | my( $xn ) = @_; |
164 | my @readings; |
165 | if( $xn->nodeType == XML_TEXT_NODE ) { |
166 | # Base text, just split the words on whitespace and add them |
167 | # to our sequence. |
168 | my $str = $xn->data; |
169 | $str =~ s/^\s+//; |
c9158e60 |
170 | my @tokens = split( /\s+/, $str ); |
171 | push( @readings, map { { 'type' => 'token', 'content' => $_ } } @tokens ); |
4d85a60e |
172 | } elsif( $xn->nodeName eq 'app' ) { |
173 | # Apparatus, just save the entire XML node. |
174 | push( @readings, { 'type' => 'app', 'content' => $xn } ); |
175 | } elsif( $xn->nodeName eq 'anchor' ) { |
176 | # Anchor to mark the end of some apparatus; save its ID. |
177 | push( @readings, { 'type' => 'anchor', |
178 | 'content' => $xn->getAttribute( 'xml:id' ) } ); |
179 | } elsif ( $xn->nodeName ne 'note' ) { # Any tag we don't know to disregard |
180 | print STDERR "Unrecognized tag " . $xn->nodeName . "\n"; |
6f4946fb |
181 | } |
4d85a60e |
182 | return @readings; |
6f4946fb |
183 | } |
184 | |
c9158e60 |
185 | sub _append_tokens { |
186 | my( $list, @tokens ) = @_; |
187 | if( @$list && $list->[-1]->{'content'} =~ /\#JOIN\#$/ ) { |
188 | # The list evidently ended mid-word; join the next token onto it. |
189 | my $t = shift @tokens; |
190 | if( ref $t && $t->{'type'} eq 'token' ) { |
191 | # Join the word |
192 | $t = $t->{'content'}; |
193 | } elsif( ref $t ) { |
194 | # An app or anchor intervened; end the word. |
195 | unshift( @tokens, $t ); |
196 | $t = ''; |
197 | } |
198 | $list->[-1]->{'content'} =~ s/\#JOIN\#$/$t/; |
199 | } |
200 | foreach my $t ( @tokens ) { |
201 | unless( ref( $t ) ) { |
202 | $t = { 'type' => 'token', 'content' => $t }; |
203 | } |
204 | push( @$list, $t ); |
205 | } |
206 | } |
207 | |
4d85a60e |
208 | sub _add_readings { |
209 | my( $c, $app_id ) = @_; |
210 | my $xn = $apps{$app_id}; |
211 | my $anchor = _anchor_name( $xn->getAttribute( 'to' ) ); |
212 | # Get the lemma, which is all the readings between app and anchor, |
213 | # excluding other apps or anchors. |
214 | my @lemma = _return_lemma( $c, $app_id, $anchor ); |
10e4b1ac |
215 | my $lemma_str = join( ' ', grep { $_ !~ /^__/ } map { $_->text } @lemma ); |
4d85a60e |
216 | |
217 | # For each reading, send its text to 'interpret' along with the lemma, |
218 | # and then save the list of witnesses that these tokens belong to. |
3a5d151b |
219 | my %wit_rdgs; # Maps from witnesses to the variant text |
4d85a60e |
220 | my $ctr = 0; |
221 | my $tag = $app_id; |
10e4b1ac |
222 | $tag =~ s/^\__APP_(.*)\__$/$1/; |
c9158e60 |
223 | |
4d85a60e |
224 | foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) { |
225 | my @text; |
4d85a60e |
226 | foreach ( $rdg->childNodes ) { |
227 | push( @text, _get_base( $_ ) ); |
228 | } |
12720144 |
229 | my( $interpreted, $flag ) = ( '', undef ); |
230 | if( @text ) { |
231 | ( $interpreted, $flag ) = interpret( |
232 | join( ' ', map { $_->{'content'} } @text ), $lemma_str ); |
233 | } |
b8f262e8 |
234 | next if( $interpreted eq $lemma_str ) && !$flag; # Reading is lemma. |
235 | |
4d85a60e |
236 | my @rdg_nodes; |
b8f262e8 |
237 | if( $interpreted eq '#LACUNA#' ) { |
10e4b1ac |
238 | push( @rdg_nodes, $c->add_reading( { id => 'r'.$tag.".".$ctr++, |
b8f262e8 |
239 | is_lacuna => 1 } ) ); |
240 | } else { |
241 | foreach my $w ( split( /\s+/, $interpreted ) ) { |
10e4b1ac |
242 | my $r = $c->add_reading( { id => 'r'.$tag.".".$ctr++, |
b8f262e8 |
243 | text => $w } ); |
244 | push( @rdg_nodes, $r ); |
245 | } |
4d85a60e |
246 | } |
f6e19c7c |
247 | # For each listed wit, save the reading. |
248 | foreach my $wit ( split( /\s+/, $rdg->getAttribute( 'wit' ) ) ) { |
12720144 |
249 | $wit .= $flag if $flag; |
f6e19c7c |
250 | $wit_rdgs{$wit} = \@rdg_nodes; |
251 | } |
12720144 |
252 | |
3a5d151b |
253 | # Does the reading have an ID? If so it probably has a witDetail |
f6e19c7c |
254 | # attached, and we need to read it. |
3a5d151b |
255 | if( $rdg->hasAttribute( 'xml:id' ) ) { |
12720144 |
256 | warn "Witdetail on meta reading" if $flag; # this could get complicated. |
f6e19c7c |
257 | my $rid = $rdg->getAttribute( 'xml:id' ); |
258 | my $xpc = XML::LibXML::XPathContext->new( $xn ); |
259 | my @details = $xpc->findnodes( './witDetail[@target="'.$rid.'"]' ); |
260 | foreach my $d ( @details ) { |
261 | _parse_wit_detail( $d, \%wit_rdgs, \@lemma ); |
262 | } |
3a5d151b |
263 | } |
f6e19c7c |
264 | } |
265 | |
4d85a60e |
266 | # Now collate the variant readings, since it is not done for us. |
12720144 |
267 | collate_variants( $c, \@lemma, values %wit_rdgs ); |
b8f262e8 |
268 | |
4d85a60e |
269 | # Now add the witness paths for each reading. |
a445ce40 |
270 | my $aclabel = $c->ac_label; |
f6e19c7c |
271 | foreach my $wit_id ( keys %wit_rdgs ) { |
a445ce40 |
272 | my $witstr = _get_sigil( $wit_id, $aclabel ); |
f6e19c7c |
273 | my $rdg_list = $wit_rdgs{$wit_id}; |
274 | _add_wit_path( $c, $rdg_list, $app_id, $anchor, $witstr ); |
6f4946fb |
275 | } |
4d85a60e |
276 | } |
6f4946fb |
277 | |
4d85a60e |
278 | sub _anchor_name { |
279 | my $xmlid = shift; |
280 | $xmlid =~ s/^\#//; |
10e4b1ac |
281 | return sprintf( "__ANCHOR_%s__", $xmlid ); |
6f4946fb |
282 | } |
283 | |
4d85a60e |
284 | sub _return_lemma { |
285 | my( $c, $app, $anchor ) = @_; |
10e4b1ac |
286 | my @nodes = grep { $_->id !~ /^__A(PP|NCHOR)/ } |
12720144 |
287 | $c->reading_sequence( $c->reading( $app ), $c->reading( $anchor ), |
288 | $c->baselabel ); |
4d85a60e |
289 | return @nodes; |
290 | } |
6f4946fb |
291 | |
a445ce40 |
292 | =head2 interpret( $reading, $lemma ) |
293 | |
294 | Given a string in $reading and a corresponding lemma in $lemma, interpret what |
295 | the actual reading should be. Used to deal with apparatus-ese shorthands for |
296 | marking transpositions, prefixed or suffixed words, and the like. |
297 | |
298 | =cut |
299 | |
6f4946fb |
300 | sub interpret { |
4d85a60e |
301 | # A utility function to change apparatus-ese into a full variant. |
302 | my( $reading, $lemma ) = @_; |
303 | return $reading if $reading eq $lemma; |
304 | my $oldreading = $reading; |
305 | # $lemma =~ s/\s+[[:punct:]]+$//; |
12720144 |
306 | my $flag; # In case of p.c. indications |
4d85a60e |
307 | my @words = split( /\s+/, $lemma ); |
308 | if( $reading =~ /^(.*) praem.$/ ) { |
309 | $reading = "$1 $lemma"; |
310 | } elsif( $reading =~ /^(.*) add.$/ ) { |
311 | $reading = "$lemma $1"; |
b8f262e8 |
312 | } elsif( $reading =~ /add. alia manu/ |
313 | || $reading =~ /inscriptionem compegi e/ # TODO huh? |
314 | || $reading eq 'inc.' # TODO huh? |
315 | ) { |
12720144 |
316 | # Ignore it. |
317 | $reading = $lemma; |
b8f262e8 |
318 | } elsif( $reading =~ /locus [uv]acuus/ |
319 | || $reading eq 'def.' |
c9158e60 |
320 | || $reading eq 'illeg.' |
321 | || $reading eq 'onleesbar' |
f6e19c7c |
322 | ) { |
b8f262e8 |
323 | $reading = '#LACUNA#'; |
324 | } elsif( $reading eq 'om.' ) { |
4d85a60e |
325 | $reading = ''; |
c9158e60 |
326 | } elsif( $reading =~ /^in[uv]\.$/ |
327 | || $reading eq 'transp.' ) { |
4d85a60e |
328 | # Hope it is two words. |
329 | print STDERR "WARNING: want to invert a lemma that is not two words\n" |
330 | unless scalar( @words ) == 2; |
331 | $reading = join( ' ', reverse( @words ) ); |
12720144 |
332 | } elsif( $reading =~ /^iter(\.|at)$/ ) { |
4d85a60e |
333 | # Repeat the lemma |
334 | $reading = "$lemma $lemma"; |
12720144 |
335 | } elsif( $reading eq 'in marg.' ) { |
336 | # There was nothing before a correction. |
337 | $reading = ''; |
338 | $flag = '_ac'; |
c9158e60 |
339 | } elsif( $reading =~ /^(.*?)\s*\(?sic([\s\w.]+)?\)?$/ ) { |
340 | # Discard any 'sic' notation; indeed, indeed. |
341 | $reading = $1; |
12720144 |
342 | } elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) { |
4d85a60e |
343 | # The first and last N words captured should replace the first and |
344 | # last N words of the lemma. |
345 | my @begin = split( /\s+/, $1 ); |
346 | my @end = split( /\s+/, $2 ); |
347 | if( scalar( @begin ) + scalar ( @end ) > scalar( @words ) ) { |
348 | # Something is wrong and we can't do the splice. |
349 | print STDERR "ERROR: $lemma is too short to accommodate $oldreading\n"; |
350 | } else { |
351 | splice( @words, 0, scalar @begin, @begin ); |
352 | splice( @words, -(scalar @end), scalar @end, @end ); |
353 | $reading = join( ' ', @words ); |
354 | } |
6f4946fb |
355 | } |
12720144 |
356 | if( $oldreading ne $reading || $flag || $oldreading =~ /\./ ) { |
357 | my $int = $reading; |
358 | $int .= " ($flag)" if $flag; |
359 | print STDERR "Interpreted $oldreading as $int given $lemma\n"; |
360 | } |
361 | return( $reading, $flag ); |
4d85a60e |
362 | } |
363 | |
f6e19c7c |
364 | sub _parse_wit_detail { |
365 | my( $detail, $readings, $lemma ) = @_; |
366 | my $wit = $detail->getAttribute( 'wit' ); |
367 | my $content = $detail->textContent; |
368 | if( $content =~ /a\.\s*c\./ ) { |
369 | # Replace the key in the $readings hash |
370 | my $rdg = delete $readings->{$wit}; |
371 | $readings->{$wit.'_ac'} = $rdg; |
372 | $has_ac{$sigil_for{$wit}} = 1; |
373 | } elsif( $content =~ /p\.\s*c\./ ) { |
374 | # If no key for the wit a.c. exists, add one pointing to the lemma |
375 | unless( exists $readings->{$wit.'_ac'} ) { |
376 | $readings->{$wit.'_ac'} = $lemma; |
377 | } |
378 | $has_ac{$sigil_for{$wit}} = 1; |
379 | } # else don't bother just yet |
380 | } |
381 | |
a445ce40 |
382 | sub _get_sigil { |
383 | my( $xml_id, $layerlabel ) = @_; |
f6e19c7c |
384 | if( $xml_id =~ /^(.*)_ac$/ ) { |
385 | my $real_id = $1; |
a445ce40 |
386 | return $sigil_for{$real_id} . $layerlabel; |
f6e19c7c |
387 | } else { |
388 | return $sigil_for{$xml_id}; |
389 | } |
390 | } |
391 | |
a445ce40 |
392 | sub _expand_all_paths { |
f6e19c7c |
393 | my( $c ) = @_; |
394 | |
395 | # Walk the collation and fish out the paths for each witness |
396 | foreach my $wit ( $c->tradition->witnesses ) { |
397 | my $sig = $wit->sigil; |
12720144 |
398 | my @path = grep { !$_->is_ph } |
f6e19c7c |
399 | $c->reading_sequence( $c->start, $c->end, $sig ); |
400 | $wit->path( \@path ); |
401 | if( $has_ac{$sig} ) { |
12720144 |
402 | my @ac_path = grep { !$_->is_ph } |
861c3e27 |
403 | $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label ); |
f6e19c7c |
404 | $wit->uncorrected_path( \@ac_path ); |
405 | } |
406 | } |
407 | |
408 | # Delete the anchors |
12720144 |
409 | foreach my $anchor ( grep { $_->is_ph } $c->readings ) { |
f6e19c7c |
410 | $c->del_reading( $anchor ); |
411 | } |
12720144 |
412 | # Delete the base edges |
413 | map { $c->del_path( $_, $c->baselabel ) } $c->paths; |
f6e19c7c |
414 | |
415 | # Make the path edges |
416 | $c->make_witness_paths(); |
4d85a60e |
417 | } |
418 | |
419 | sub _add_wit_path { |
f6e19c7c |
420 | my( $c, $rdg, $app, $anchor, $wit ) = @_; |
4d85a60e |
421 | my @nodes = @$rdg; |
f6e19c7c |
422 | push( @nodes, $c->reading( $anchor ) ); |
4d85a60e |
423 | |
f6e19c7c |
424 | my $cur = $c->reading( $app ); |
4d85a60e |
425 | foreach my $n ( @nodes ) { |
f6e19c7c |
426 | $c->add_path( $cur, $n, $wit ); |
4d85a60e |
427 | $cur = $n; |
6f4946fb |
428 | } |
6f4946fb |
429 | } |
430 | |
00311328 |
431 | sub throw { |
432 | Text::Tradition::Error->throw( |
433 | 'ident' => 'Parser::CTE error', |
434 | 'message' => $_[0], |
435 | ); |
436 | } |
437 | |
6f4946fb |
438 | =head1 LICENSE |
439 | |
440 | This package is free software and is provided "as is" without express |
441 | or implied warranty. You can redistribute it and/or modify it under |
442 | the same terms as Perl itself. |
443 | |
444 | =head1 AUTHOR |
445 | |
446 | Tara L Andrews, aurum@cpan.org |
447 | |
448 | =cut |
449 | |
450 | 1; |
451 | |