Commit | Line | Data |
6f4946fb |
1 | package Text::Tradition::Parser::CTE; |
2 | |
3 | use strict; |
4 | use warnings; |
a188b944 |
5 | use feature 'say'; |
c9158e60 |
6 | use Encode qw/ decode /; |
00311328 |
7 | use Text::Tradition::Error; |
4d85a60e |
8 | use Text::Tradition::Parser::Util qw/ collate_variants /; |
6f4946fb |
9 | use XML::LibXML; |
10 | use XML::LibXML::XPathContext; |
11 | |
12 | =head1 NAME |
13 | |
14 | Text::Tradition::Parser::CTE |
15 | |
16 | =head1 DESCRIPTION |
17 | |
18 | Parser module for Text::Tradition, given a TEI file exported from |
19 | Classical Text Editor. |
20 | |
21 | =head1 METHODS |
22 | |
a445ce40 |
23 | =head2 parse |
6f4946fb |
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 | |
e9442e1c |
41 | ## DEBUG/TEST |
42 | $opts->{interpret_transposition} = 1; |
43 | |
4d85a60e |
44 | # First, parse the XML. |
c9158e60 |
45 | my( $tei, $xpc ) = _remove_formatting( $opts ); |
46 | return unless $tei; # we have already warned. |
4d85a60e |
47 | |
48 | # CTE uses a DTD rather than any xmlns-based parsing. Thus we |
49 | # need no namespace handling. |
4d85a60e |
50 | # Get the witnesses and create the witness objects. |
51 | foreach my $wit_el ( $xpc->findnodes( '//sourceDesc/listWit/witness' ) ) { |
52 | # The witness xml:id is used internally, and is *not* the sigil name. |
53 | my $id= $wit_el->getAttribute( 'xml:id' ); |
92de40a6 |
54 | # If the witness element has an abbr element, that is the sigil. Otherwise |
55 | # the whole thing is the sigil. |
56 | my $sig = $xpc->findvalue( 'abbr', $wit_el ); |
57 | my $identifier = 'CTE witness'; |
58 | if( $sig ) { |
59 | # The sigil is what is in the <abbr/> tag; the identifier is anything |
60 | # that follows. |
61 | $identifier = _tidy_identifier( |
62 | $xpc->findvalue( 'child::text()', $wit_el ) ); |
63 | } else { |
64 | my @sig_parts = $xpc->findnodes( 'descendant::text()', $wit_el ); |
65 | $sig = _stringify_sigil( @sig_parts ); |
66 | } |
67 | say STDERR "Adding witness $sig ($identifier)"; |
68 | $tradition->add_witness( sigil => $sig, identifier => $identifier, |
69 | sourcetype => 'collation' ); |
b8f262e8 |
70 | $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax |
4d85a60e |
71 | } |
c9158e60 |
72 | |
4d85a60e |
73 | # Now go through the text and find the base tokens, apparatus tags, and |
74 | # anchors. Make a giant array of all of these things in sequence. |
75 | # TODO consider combining this with creation of graph below |
76 | my @base_text; |
77 | foreach my $pg_el ( $xpc->findnodes( '/TEI/text/body/p' ) ) { |
78 | foreach my $xn ( $pg_el->childNodes ) { |
82a45078 |
79 | push( @base_text, _get_base( $xn ) ); |
4d85a60e |
80 | } |
6f4946fb |
81 | } |
4d85a60e |
82 | # We now have to work through this array applying the alternate |
83 | # apparatus readings to the base text. Essentially we will put |
84 | # everything on the graph, from which we will delete the apps and |
85 | # anchors when we are done. |
f6e19c7c |
86 | |
9e0a9786 |
87 | # First, put the base tokens, apps, and anchors in the graph. Save the |
88 | # app siglorum separately as it has to be processed in order. |
89 | my @app_sig; |
4d85a60e |
90 | my $counter = 0; |
91 | my $last = $c->start; |
92 | foreach my $item ( @base_text ) { |
93 | my $r; |
94 | if( $item->{'type'} eq 'token' ) { |
12720144 |
95 | $r = $c->add_reading( { id => 'n'.$counter++, |
96 | text => $item->{'content'} } ); |
4d85a60e |
97 | } elsif ( $item->{'type'} eq 'anchor' ) { |
10e4b1ac |
98 | $r = $c->add_reading( { id => '__ANCHOR_' . $item->{'content'} . '__', |
12720144 |
99 | is_ph => 1 } ); |
4d85a60e |
100 | } elsif ( $item->{'type'} eq 'app' ) { |
10e4b1ac |
101 | my $tag = '__APP_' . $counter++ . '__'; |
12720144 |
102 | $r = $c->add_reading( { id => $tag, is_ph => 1 } ); |
9e0a9786 |
103 | # Apparatus criticus is type a1; app siglorum is type a2 |
104 | if( $item->{'content'}->getAttribute('type') eq 'a1' ) { |
105 | $apps{$tag} = $item->{'content'}; |
106 | } else { |
107 | push( @app_sig, $item->{'content'} ); |
108 | } |
4d85a60e |
109 | } |
f6e19c7c |
110 | $c->add_path( $last, $r, $c->baselabel ); |
4d85a60e |
111 | $last = $r; |
6f4946fb |
112 | } |
f6e19c7c |
113 | $c->add_path( $last, $c->end, $c->baselabel ); |
4d85a60e |
114 | |
115 | # Now we can parse the apparatus entries, and add the variant readings |
116 | # to the graph. |
4d85a60e |
117 | foreach my $app_id ( keys %apps ) { |
e9442e1c |
118 | _add_readings( $c, $app_id, $opts ); |
6f4946fb |
119 | } |
9e0a9786 |
120 | _add_lacunae( $c, @app_sig ); |
4d85a60e |
121 | |
f6e19c7c |
122 | # Finally, add explicit witness paths, remove the base paths, and remove |
123 | # the app/anchor tags. |
dead25ca |
124 | try { |
125 | _expand_all_paths( $c ); |
126 | } catch( $err ) { |
127 | throw( $err ); |
128 | } |
861c3e27 |
129 | |
130 | # Save the text for each witness so that we can ensure consistency |
131 | # later on |
a188b944 |
132 | unless( $opts->{'nocalc'} ) { |
dead25ca |
133 | try { |
134 | $tradition->collation->text_from_paths(); |
135 | $tradition->collation->calculate_ranks(); |
136 | $tradition->collation->flatten_ranks(); |
137 | } catch( $err ) { |
138 | throw( $err ); |
139 | } |
a188b944 |
140 | } |
6f4946fb |
141 | } |
142 | |
4d85a60e |
143 | sub _stringify_sigil { |
144 | my( @nodes ) = @_; |
145 | my @parts = grep { /\w/ } map { $_->data } @nodes; |
222d58f1 |
146 | my $whole = join( '', @parts ); |
147 | $whole =~ s/\W//g; |
148 | return $whole; |
4d85a60e |
149 | } |
6f4946fb |
150 | |
92de40a6 |
151 | sub _tidy_identifier { |
152 | my( $str ) = @_; |
153 | $str =~ s/^\W+//; |
154 | return $str; |
155 | } |
156 | |
c9158e60 |
157 | # Get rid of all the formatting elements that get in the way of tokenization. |
158 | sub _remove_formatting { |
159 | my( $opts ) = @_; |
160 | |
161 | # First, parse the original XML |
162 | my $parser = XML::LibXML->new(); |
163 | my $doc; |
164 | if( exists $opts->{'string'} ) { |
165 | $doc = $parser->parse_string( $opts->{'string'} ); |
166 | } elsif ( exists $opts->{'file'} ) { |
167 | $doc = $parser->parse_file( $opts->{'file'} ); |
dead25ca |
168 | } elsif ( exists $opts->{'xmlobj'} ) { |
169 | $doc = $opts->{'xmlobj'}; |
c9158e60 |
170 | } else { |
171 | warn "Could not find string or file option to parse"; |
172 | return; |
173 | } |
00311328 |
174 | |
c9158e60 |
175 | # Second, remove the formatting |
176 | my $xpc = XML::LibXML::XPathContext->new( $doc->documentElement ); |
177 | my @useless = $xpc->findnodes( '//hi' ); |
178 | foreach my $n ( @useless ) { |
179 | my $parent = $n->parentNode(); |
180 | my @children = $n->childNodes(); |
181 | my $first = shift @children; |
7c2ed85e |
182 | if( $first ) { |
183 | $parent->replaceChild( $first, $n ); |
184 | foreach my $c ( @children ) { |
185 | $parent->insertAfter( $c, $first ); |
186 | $first = $c; |
187 | } |
188 | } else { |
189 | $parent->removeChild( $n ); |
c9158e60 |
190 | } |
191 | } |
192 | |
193 | # Third, write out and reparse to merge the text nodes. |
00311328 |
194 | my $enc = $doc->encoding || 'UTF-8'; |
195 | my $result = decode( $enc, $doc->toString() ); |
c9158e60 |
196 | my $tei = $parser->parse_string( $result )->documentElement; |
00311328 |
197 | unless( $tei->nodeName =~ /^tei(corpus)?$/i ) { |
198 | throw( "Parsed document has non-TEI root element " . $tei->nodeName ); |
199 | } |
c9158e60 |
200 | $xpc = XML::LibXML::XPathContext->new( $tei ); |
201 | return( $tei, $xpc ); |
202 | } |
203 | |
204 | ## Helper function to help us navigate through nested XML, picking out |
205 | ## the words, the apparatus, and the anchors. |
4d85a60e |
206 | |
207 | sub _get_base { |
208 | my( $xn ) = @_; |
209 | my @readings; |
210 | if( $xn->nodeType == XML_TEXT_NODE ) { |
211 | # Base text, just split the words on whitespace and add them |
212 | # to our sequence. |
213 | my $str = $xn->data; |
214 | $str =~ s/^\s+//; |
c9158e60 |
215 | my @tokens = split( /\s+/, $str ); |
92de40a6 |
216 | push( @readings, map { { type => 'token', content => $_ } } @tokens ); |
4d85a60e |
217 | } elsif( $xn->nodeName eq 'app' ) { |
218 | # Apparatus, just save the entire XML node. |
92de40a6 |
219 | push( @readings, { type => 'app', content => $xn } ); |
4d85a60e |
220 | } elsif( $xn->nodeName eq 'anchor' ) { |
221 | # Anchor to mark the end of some apparatus; save its ID. |
82a45078 |
222 | if( $xn->hasAttribute('xml:id') ) { |
92de40a6 |
223 | push( @readings, { type => 'anchor', |
224 | content => $xn->getAttribute( 'xml:id' ) } ); |
82a45078 |
225 | } # if the anchor has no XML ID, it is not relevant to us. |
92de40a6 |
226 | } elsif( $xn->nodeName !~ /^(note|seg|milestone|emph)$/ ) { # Any tag we don't know to disregard |
a188b944 |
227 | say STDERR "Unrecognized tag " . $xn->nodeName; |
6f4946fb |
228 | } |
4d85a60e |
229 | return @readings; |
6f4946fb |
230 | } |
231 | |
c9158e60 |
232 | sub _append_tokens { |
233 | my( $list, @tokens ) = @_; |
234 | if( @$list && $list->[-1]->{'content'} =~ /\#JOIN\#$/ ) { |
235 | # The list evidently ended mid-word; join the next token onto it. |
236 | my $t = shift @tokens; |
237 | if( ref $t && $t->{'type'} eq 'token' ) { |
238 | # Join the word |
239 | $t = $t->{'content'}; |
240 | } elsif( ref $t ) { |
241 | # An app or anchor intervened; end the word. |
242 | unshift( @tokens, $t ); |
243 | $t = ''; |
244 | } |
245 | $list->[-1]->{'content'} =~ s/\#JOIN\#$/$t/; |
246 | } |
247 | foreach my $t ( @tokens ) { |
248 | unless( ref( $t ) ) { |
249 | $t = { 'type' => 'token', 'content' => $t }; |
250 | } |
251 | push( @$list, $t ); |
252 | } |
253 | } |
254 | |
4d85a60e |
255 | sub _add_readings { |
e9442e1c |
256 | my( $c, $app_id, $opts ) = @_; |
4d85a60e |
257 | my $xn = $apps{$app_id}; |
9e0a9786 |
258 | my $anchor = _anchor_name( $xn->getAttribute( 'to' ) ); |
92de40a6 |
259 | |
4d85a60e |
260 | # Get the lemma, which is all the readings between app and anchor, |
261 | # excluding other apps or anchors. |
9e0a9786 |
262 | my @lemma = _return_lemma( $c, $app_id, $anchor ); |
263 | my $lemma_str = join( ' ', map { $_->text } grep { !$_->is_ph } @lemma ); |
92de40a6 |
264 | |
4d85a60e |
265 | # For each reading, send its text to 'interpret' along with the lemma, |
266 | # and then save the list of witnesses that these tokens belong to. |
3a5d151b |
267 | my %wit_rdgs; # Maps from witnesses to the variant text |
4d85a60e |
268 | my $ctr = 0; |
269 | my $tag = $app_id; |
10e4b1ac |
270 | $tag =~ s/^\__APP_(.*)\__$/$1/; |
c9158e60 |
271 | |
4d85a60e |
272 | foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) { |
e9442e1c |
273 | my @witlist = split( /\s+/, $rdg->getAttribute( 'wit' ) ); |
4d85a60e |
274 | my @text; |
4d85a60e |
275 | foreach ( $rdg->childNodes ) { |
276 | push( @text, _get_base( $_ ) ); |
277 | } |
12720144 |
278 | my( $interpreted, $flag ) = ( '', undef ); |
279 | if( @text ) { |
280 | ( $interpreted, $flag ) = interpret( |
e9442e1c |
281 | join( ' ', map { $_->{'content'} } @text ), $lemma_str, $anchor, $opts ); |
12720144 |
282 | } |
e9442e1c |
283 | next if( $interpreted eq $lemma_str ) && !keys %$flag; # Reading is lemma. |
b8f262e8 |
284 | |
4d85a60e |
285 | my @rdg_nodes; |
b8f262e8 |
286 | if( $interpreted eq '#LACUNA#' ) { |
10e4b1ac |
287 | push( @rdg_nodes, $c->add_reading( { id => 'r'.$tag.".".$ctr++, |
b8f262e8 |
288 | is_lacuna => 1 } ) ); |
e9442e1c |
289 | } elsif( $flag->{'TR'} ) { |
290 | # Our reading is transposed to after the given string. Look |
291 | # down the collation base text and try to find it. |
292 | # The @rdg_nodes should remain blank here, so that the correct |
293 | # omission goes into the graph. |
f9ffe014 |
294 | my @transp_nodes; |
e9442e1c |
295 | foreach my $w ( split( /\s+/, $interpreted ) ) { |
296 | my $r = $c->add_reading( { id => 'r'.$tag.".".$ctr++, |
297 | text => $w } ); |
298 | push( @transp_nodes, $r ); |
299 | } |
300 | if( $anchor && @lemma ) { |
f9ffe014 |
301 | my $success = _attach_transposition( $c, \@lemma, $anchor, |
302 | \@transp_nodes, \@witlist, $flag->{'TR'} ); |
303 | unless( $success ) { |
304 | # If we didn't manage to insert the displaced reading, |
305 | # then restore it here rather than silently deleting it. |
306 | push( @rdg_nodes, @transp_nodes ); |
307 | } |
e9442e1c |
308 | } |
b8f262e8 |
309 | } else { |
310 | foreach my $w ( split( /\s+/, $interpreted ) ) { |
10e4b1ac |
311 | my $r = $c->add_reading( { id => 'r'.$tag.".".$ctr++, |
b8f262e8 |
312 | text => $w } ); |
313 | push( @rdg_nodes, $r ); |
314 | } |
4d85a60e |
315 | } |
92de40a6 |
316 | |
f6e19c7c |
317 | # For each listed wit, save the reading. |
a5978ac8 |
318 | # If an A.C. or P.C. reading is implied rather than explicitly noted, |
319 | # this is where it will be dealt with. |
e9442e1c |
320 | foreach my $wit ( @witlist ) { |
321 | $wit .= '_ac' if $flag->{'AC'}; |
f6e19c7c |
322 | $wit_rdgs{$wit} = \@rdg_nodes; |
a5978ac8 |
323 | # If the PC flag is set, there is a corresponding AC that |
324 | # follows the lemma and has to be explicitly declared. |
325 | if( $flag->{'PC'} ) { |
326 | $wit_rdgs{$wit.'_ac'} = \@lemma; |
327 | } |
f6e19c7c |
328 | } |
12720144 |
329 | |
3a5d151b |
330 | # Does the reading have an ID? If so it probably has a witDetail |
a5978ac8 |
331 | # attached, and we need to read it. If an A.C. or P.C. reading is |
332 | # declared explicity, this is where it will be dealt with. |
3a5d151b |
333 | if( $rdg->hasAttribute( 'xml:id' ) ) { |
12720144 |
334 | warn "Witdetail on meta reading" if $flag; # this could get complicated. |
f6e19c7c |
335 | my $rid = $rdg->getAttribute( 'xml:id' ); |
336 | my $xpc = XML::LibXML::XPathContext->new( $xn ); |
337 | my @details = $xpc->findnodes( './witDetail[@target="'.$rid.'"]' ); |
338 | foreach my $d ( @details ) { |
339 | _parse_wit_detail( $d, \%wit_rdgs, \@lemma ); |
340 | } |
3a5d151b |
341 | } |
f6e19c7c |
342 | } |
343 | |
4d85a60e |
344 | # Now collate the variant readings, since it is not done for us. |
12720144 |
345 | collate_variants( $c, \@lemma, values %wit_rdgs ); |
b8f262e8 |
346 | |
92de40a6 |
347 | # Now add the witness paths for each reading. If we don't have an anchor |
348 | # (e.g. with an initial witStart) there was no witness path to speak of. |
9e0a9786 |
349 | foreach my $wit_id ( keys %wit_rdgs ) { |
350 | my $witstr = _get_sigil( $wit_id, $c->ac_label ); |
351 | my $rdg_list = $wit_rdgs{$wit_id}; |
352 | _add_wit_path( $c, $rdg_list, $app_id, $anchor, $witstr ); |
92de40a6 |
353 | } |
4d85a60e |
354 | } |
6f4946fb |
355 | |
4d85a60e |
356 | sub _anchor_name { |
357 | my $xmlid = shift; |
358 | $xmlid =~ s/^\#//; |
10e4b1ac |
359 | return sprintf( "__ANCHOR_%s__", $xmlid ); |
6f4946fb |
360 | } |
361 | |
4d85a60e |
362 | sub _return_lemma { |
363 | my( $c, $app, $anchor ) = @_; |
10e4b1ac |
364 | my @nodes = grep { $_->id !~ /^__A(PP|NCHOR)/ } |
12720144 |
365 | $c->reading_sequence( $c->reading( $app ), $c->reading( $anchor ), |
366 | $c->baselabel ); |
4d85a60e |
367 | return @nodes; |
368 | } |
6f4946fb |
369 | |
e9442e1c |
370 | # Make a best-effort attempt to attach a transposition farther down the line. |
371 | # $lemmaseq contains the Reading objects of the lemma |
372 | # $anchor contains the point at which we should start scanning for a match |
373 | # $rdgseq contains the Reading objects of the transposed reading |
374 | # (should be identical to the lemma) |
375 | # $witlist contains the list of applicable witnesses |
376 | # $reftxt contains the text to match, after which the $rdgseq should go. |
377 | sub _attach_transposition { |
378 | my( $c, $lemmaseq, $anchor, $rdgseq, $witlist, $reftxt ) = @_; |
379 | my @refwords = split( /\s+/, $reftxt ); |
380 | my $checked = $c->reading( $anchor ); |
381 | my $found; |
f9ffe014 |
382 | my $success; |
e9442e1c |
383 | while( $checked ne $c->end && !$found ) { |
384 | my $next = $c->next_reading( $checked, $c->baselabel ); |
385 | if( $next->text eq $refwords[0] ) { |
386 | # See if the entire sequence of words matches. |
387 | $found = $next; |
388 | foreach my $w ( 1..$#refwords ) { |
389 | $found = $c->next_reading( $next, $c->baselabel ); |
390 | unless( $found->text eq $refwords[$w] ) { |
391 | $found = undef; |
392 | last; |
393 | } |
394 | } |
395 | } |
396 | $checked = $next; |
397 | } |
398 | if( $found ) { |
399 | # The $found variable should now contain the reading after which we |
400 | # should stick the transposition. |
401 | my $fnext = $c->next_reading( $found, $c->baselabel ); |
402 | my $aclabel = $c->ac_label; |
403 | foreach my $wit_id ( @$witlist ) { |
404 | my $witstr = _get_sigil( $wit_id, $aclabel ); |
405 | _add_wit_path( $c, $rdgseq, $found->id, $fnext->id, $witstr ); |
406 | } |
407 | # ...and add the transposition relationship between lemma and rdgseq. |
408 | if( @$lemmaseq == @$rdgseq ) { |
409 | foreach my $i ( 0..$#{$lemmaseq} ) { |
410 | $c->add_relationship( $lemmaseq->[$i], $rdgseq->[$i], |
411 | { type => 'transposition', annotation => 'Detected by CTE' } ); |
412 | } |
f9ffe014 |
413 | $success = 1; |
e9442e1c |
414 | } else { |
dead25ca |
415 | throw( "Lemma at $found and transposed sequence different lengths?!" ); |
e9442e1c |
416 | } |
417 | } else { |
418 | say STDERR "WARNING: Unable to find $reftxt in base text for transposition"; |
e9442e1c |
419 | } |
f9ffe014 |
420 | return $success; |
e9442e1c |
421 | } |
422 | |
a445ce40 |
423 | =head2 interpret( $reading, $lemma ) |
424 | |
425 | Given a string in $reading and a corresponding lemma in $lemma, interpret what |
426 | the actual reading should be. Used to deal with apparatus-ese shorthands for |
427 | marking transpositions, prefixed or suffixed words, and the like. |
428 | |
429 | =cut |
430 | |
6f4946fb |
431 | sub interpret { |
4d85a60e |
432 | # A utility function to change apparatus-ese into a full variant. |
e9442e1c |
433 | my( $reading, $lemma, $anchor, $opts ) = @_; |
4d85a60e |
434 | return $reading if $reading eq $lemma; |
435 | my $oldreading = $reading; |
436 | # $lemma =~ s/\s+[[:punct:]]+$//; |
e9442e1c |
437 | my $flag = {}; # To pass back extra info about the interpretation |
4d85a60e |
438 | my @words = split( /\s+/, $lemma ); |
f9ffe014 |
439 | # Discard any 'sic' notation - that rather goes without saying. |
a5978ac8 |
440 | $reading =~ s/([[:punct:]]+)?sic([[:punct:]]+)?//g; |
f9ffe014 |
441 | |
442 | # Now look for common jargon. |
443 | if( $reading =~ /^(.*) praem.$/ || $reading =~ /^praem\. (.*)$/ ) { |
4d85a60e |
444 | $reading = "$1 $lemma"; |
f9ffe014 |
445 | } elsif( $reading =~ /^(.*) add.$/ || $reading =~ /^add\. (.*)$/ ) { |
4d85a60e |
446 | $reading = "$lemma $1"; |
b8f262e8 |
447 | } elsif( $reading =~ /locus [uv]acuus/ |
448 | || $reading eq 'def.' |
c9158e60 |
449 | || $reading eq 'illeg.' |
a5978ac8 |
450 | || $reading eq 'desunt' |
f6e19c7c |
451 | ) { |
b8f262e8 |
452 | $reading = '#LACUNA#'; |
453 | } elsif( $reading eq 'om.' ) { |
4d85a60e |
454 | $reading = ''; |
c9158e60 |
455 | } elsif( $reading =~ /^in[uv]\.$/ |
a5978ac8 |
456 | || $reading =~ /^tr(ans(p)?)?\.$/ ) { |
4d85a60e |
457 | # Hope it is two words. |
a188b944 |
458 | say STDERR "WARNING: want to invert a lemma that is not two words" |
4d85a60e |
459 | unless scalar( @words ) == 2; |
460 | $reading = join( ' ', reverse( @words ) ); |
12720144 |
461 | } elsif( $reading =~ /^iter(\.|at)$/ ) { |
4d85a60e |
462 | # Repeat the lemma |
463 | $reading = "$lemma $lemma"; |
a5978ac8 |
464 | } elsif( $reading =~ /^(.*?)\s*\(?in marg\.\)?$/ ) { |
465 | $reading = $1; |
466 | if( $reading ) { |
467 | # The given text is a correction. |
468 | $flag->{'PC'} = 1; |
469 | } else { |
470 | # The lemma itself was the correction; the witness carried |
471 | # no reading pre-correction. |
472 | $flag->{'AC'} = 1; |
473 | } |
12720144 |
474 | } elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) { |
4d85a60e |
475 | # The first and last N words captured should replace the first and |
476 | # last N words of the lemma. |
477 | my @begin = split( /\s+/, $1 ); |
478 | my @end = split( /\s+/, $2 ); |
479 | if( scalar( @begin ) + scalar ( @end ) > scalar( @words ) ) { |
480 | # Something is wrong and we can't do the splice. |
dead25ca |
481 | throw( "$lemma is too short to accommodate $oldreading" ); |
4d85a60e |
482 | } else { |
483 | splice( @words, 0, scalar @begin, @begin ); |
484 | splice( @words, -(scalar @end), scalar @end, @end ); |
485 | $reading = join( ' ', @words ); |
486 | } |
e9442e1c |
487 | } elsif( $opts->{interpret_transposition} && |
a5978ac8 |
488 | ( $reading =~ /^post\s*(?<lem>.*?)\s+tr(ans(p)?)?\.$/ || |
489 | $reading =~ /^tr(ans(p)?)?\. post\s*(?<lem>.*)$/) ) { |
e9442e1c |
490 | # Try to deal with transposed readings |
491 | ## DEBUG |
492 | say STDERR "Will attempt transposition: $reading at $anchor"; |
493 | $reading = $lemma; |
a5978ac8 |
494 | $flag->{'TR'} = $+{lem}; |
12720144 |
495 | } |
496 | return( $reading, $flag ); |
4d85a60e |
497 | } |
498 | |
f6e19c7c |
499 | sub _parse_wit_detail { |
500 | my( $detail, $readings, $lemma ) = @_; |
501 | my $wit = $detail->getAttribute( 'wit' ); |
502 | my $content = $detail->textContent; |
a5978ac8 |
503 | if( $content =~ /^a\.?\s*c(orr)?\.$/ ) { |
f6e19c7c |
504 | # Replace the key in the $readings hash |
505 | my $rdg = delete $readings->{$wit}; |
506 | $readings->{$wit.'_ac'} = $rdg; |
507 | $has_ac{$sigil_for{$wit}} = 1; |
a5978ac8 |
508 | } elsif( $content =~ /^p\.?\s*c(orr)?\.$/ || $content =~ /^s\.?\s*l\.$/ ) { |
f6e19c7c |
509 | # If no key for the wit a.c. exists, add one pointing to the lemma |
510 | unless( exists $readings->{$wit.'_ac'} ) { |
511 | $readings->{$wit.'_ac'} = $lemma; |
512 | } |
513 | $has_ac{$sigil_for{$wit}} = 1; |
a5978ac8 |
514 | } else { #...not sure what it is? |
515 | say STDERR "WARNING: Unrecognized sigil addendum $content"; |
516 | } |
f6e19c7c |
517 | } |
518 | |
9e0a9786 |
519 | sub _add_lacunae { |
520 | my( $c, @apps ) = @_; |
521 | # Go through the apparatus entries in order, noting where to start and stop our |
522 | # various witnesses. |
523 | my %lacunose; |
524 | my $ctr = 0; |
525 | foreach my $app ( @apps ) { |
526 | # Find the anchor, if any. This marks the point where the text starts |
527 | # or ends. |
528 | my $anchor = $app->getAttribute( 'to' ); |
529 | my $aname; |
530 | if( $anchor ) { |
531 | $aname = _anchor_name( $anchor ); |
532 | } |
533 | |
534 | foreach my $rdg ( $app->getChildrenByTagName( 'rdg' ) ) { |
535 | my @witlist = map { _get_sigil( $_, $c->ac_label ) } |
536 | split( /\s+/, $rdg->getAttribute( 'wit' ) ); |
537 | my @start = $rdg->getChildrenByTagName( 'witStart' ); |
538 | my @end = $rdg->getChildrenByTagName( 'witEnd' ); |
539 | if( @start && @end ) { |
540 | throw( "App sig entry at $anchor has both witStart and witEnd!" ); |
541 | } |
542 | if( @start && $anchor && |
543 | $c->prior_reading( $aname, $c->baselabel ) ne $c->start ) { |
544 | # We are picking back up after a hiatus. Find the last end and |
545 | # add a lacuna link between there and here. |
546 | foreach my $wit ( @witlist ) { |
547 | my $stoppoint = delete $lacunose{$wit}; |
548 | $stoppoint = $c->start unless $stoppoint; |
549 | my $stopname = _anchor_name( $stoppoint ); |
550 | say STDERR "Adding lacuna for $wit between $stoppoint and $anchor"; |
551 | my $lacuna = $c->add_reading( { id => "as_$anchor.".$ctr++, |
552 | is_lacuna => 1 } ); |
553 | _add_wit_path( $c, [ $lacuna ], $stopname, $aname, $wit ); |
554 | } |
555 | } elsif( @end && $anchor && |
556 | $c->next_reading( $aname, $c->baselabel ) ne $c->end ) { |
557 | # We are stopping. If we've already stopped for the given witness, |
558 | # flag an error; otherwise record the stopping point. |
559 | foreach my $wit ( @witlist ) { |
560 | if( $lacunose{$wit} ) { |
561 | throw( "Trying to end $wit at $anchor when already ended at " |
562 | . $lacunose{$wit} ); |
563 | } |
564 | $lacunose{$wit} = $anchor; |
565 | } |
566 | } |
567 | } |
568 | } |
569 | |
570 | # For whatever remains in the %lacunose hash, add a lacuna between that spot and |
571 | # $c->end for each of the witnesses. |
572 | foreach my $wit ( keys %lacunose ) { |
573 | next unless $lacunose{$wit}; |
574 | my $aname = _anchor_name( $lacunose{$wit} ); |
575 | say STDERR "Adding lacuna for $wit from $aname to end"; |
576 | my $lacuna = $c->add_reading( { id => 'as_'.$lacunose{$wit}.'.'.$ctr++, |
577 | is_lacuna => 1 } ); |
578 | _add_wit_path( $c, [ $lacuna ], $aname, $c->end, $wit ); |
579 | } |
580 | } |
581 | |
a445ce40 |
582 | sub _get_sigil { |
583 | my( $xml_id, $layerlabel ) = @_; |
f6e19c7c |
584 | if( $xml_id =~ /^(.*)_ac$/ ) { |
585 | my $real_id = $1; |
a445ce40 |
586 | return $sigil_for{$real_id} . $layerlabel; |
f6e19c7c |
587 | } else { |
588 | return $sigil_for{$xml_id}; |
589 | } |
590 | } |
591 | |
a445ce40 |
592 | sub _expand_all_paths { |
f6e19c7c |
593 | my( $c ) = @_; |
594 | |
595 | # Walk the collation and fish out the paths for each witness |
596 | foreach my $wit ( $c->tradition->witnesses ) { |
597 | my $sig = $wit->sigil; |
12720144 |
598 | my @path = grep { !$_->is_ph } |
f6e19c7c |
599 | $c->reading_sequence( $c->start, $c->end, $sig ); |
600 | $wit->path( \@path ); |
601 | if( $has_ac{$sig} ) { |
12720144 |
602 | my @ac_path = grep { !$_->is_ph } |
861c3e27 |
603 | $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label ); |
f6e19c7c |
604 | $wit->uncorrected_path( \@ac_path ); |
605 | } |
606 | } |
607 | |
608 | # Delete the anchors |
12720144 |
609 | foreach my $anchor ( grep { $_->is_ph } $c->readings ) { |
f6e19c7c |
610 | $c->del_reading( $anchor ); |
611 | } |
12720144 |
612 | # Delete the base edges |
613 | map { $c->del_path( $_, $c->baselabel ) } $c->paths; |
f6e19c7c |
614 | |
615 | # Make the path edges |
616 | $c->make_witness_paths(); |
7c2ed85e |
617 | |
618 | # Now remove any orphan nodes, and warn that we are doing so. |
82a45078 |
619 | while( $c->sequence->predecessorless_vertices > 1 ) { |
620 | foreach my $v ( $c->sequence->predecessorless_vertices ) { |
621 | my $r = $c->reading( $v ); |
622 | next if $r->is_start; |
623 | say STDERR "Deleting orphan reading $r / " . $r->text; |
624 | $c->del_reading( $r ); |
625 | } |
7c2ed85e |
626 | } |
4d85a60e |
627 | } |
628 | |
629 | sub _add_wit_path { |
f6e19c7c |
630 | my( $c, $rdg, $app, $anchor, $wit ) = @_; |
4d85a60e |
631 | my @nodes = @$rdg; |
f6e19c7c |
632 | push( @nodes, $c->reading( $anchor ) ); |
4d85a60e |
633 | |
f6e19c7c |
634 | my $cur = $c->reading( $app ); |
4d85a60e |
635 | foreach my $n ( @nodes ) { |
f6e19c7c |
636 | $c->add_path( $cur, $n, $wit ); |
4d85a60e |
637 | $cur = $n; |
6f4946fb |
638 | } |
6f4946fb |
639 | } |
640 | |
00311328 |
641 | sub throw { |
642 | Text::Tradition::Error->throw( |
643 | 'ident' => 'Parser::CTE error', |
644 | 'message' => $_[0], |
645 | ); |
646 | } |
647 | |
6f4946fb |
648 | =head1 LICENSE |
649 | |
650 | This package is free software and is provided "as is" without express |
651 | or implied warranty. You can redistribute it and/or modify it under |
652 | the same terms as Perl itself. |
653 | |
654 | =head1 AUTHOR |
655 | |
656 | Tara L Andrews, aurum@cpan.org |
657 | |
658 | =cut |
659 | |
660 | 1; |
661 | |