Commit | Line | Data |
dd3b58b0 |
1 | package Text::Tradition::Collation; |
d047cd52 |
2 | |
3 | use Graph::Easy; |
8e1394aa |
4 | use IPC::Run qw( run binary ); |
8e1394aa |
5 | use Text::Tradition::Collation::Reading; |
dd3b58b0 |
6 | use Moose; |
7 | |
8 | has 'graph' => ( |
d047cd52 |
9 | is => 'ro', |
10 | isa => 'Graph::Easy', |
11 | handles => { |
8e1394aa |
12 | add_reading => 'add_node', |
13 | del_reading => 'del_node', |
14 | add_path => 'add_edge', |
15 | del_path => 'del_edge', |
16 | reading => 'node', |
17 | path => 'edge', |
18 | readings => 'nodes', |
19 | paths => 'edges', |
d047cd52 |
20 | }, |
21 | default => sub { Graph::Easy->new( undirected => 0 ) }, |
22 | ); |
784877d9 |
23 | |
dd3b58b0 |
24 | |
dd3b58b0 |
25 | has 'tradition' => ( |
8e1394aa |
26 | is => 'rw', |
d047cd52 |
27 | isa => 'Text::Tradition', |
28 | ); |
dd3b58b0 |
29 | |
8e1394aa |
30 | has 'svg' => ( |
31 | is => 'ro', |
32 | isa => 'Str', |
33 | writer => '_save_svg', |
34 | predicate => 'has_svg', |
35 | ); |
36 | |
37 | has 'graphviz' => ( |
38 | is => 'ro', |
39 | isa => 'Str', |
40 | writer => '_save_graphviz', |
41 | predicate => 'has_graphviz', |
42 | ); |
43 | |
44 | has 'graphml' => ( |
45 | is => 'ro', |
46 | isa => 'XML::LibXML::Document', |
47 | writer => '_save_graphml', |
48 | predicate => 'has_graphml', |
49 | ); |
50 | |
3a1f2523 |
51 | # Keeps track of the lemmas within the collation. At most one lemma |
52 | # per position in the graph. |
53 | has 'lemmata' => ( |
54 | is => 'ro', |
55 | isa => 'HashRef[Maybe[Str]]', |
56 | default => sub { {} }, |
57 | ); |
58 | |
4a8828f0 |
59 | has 'wit_list_separator' => ( |
7854e12e |
60 | is => 'rw', |
61 | isa => 'Str', |
62 | default => ', ', |
63 | ); |
64 | |
65 | has 'baselabel' => ( |
66 | is => 'rw', |
67 | isa => 'Str', |
68 | default => 'base text', |
69 | ); |
4a8828f0 |
70 | |
1f563ac3 |
71 | has 'collapsed' => ( |
72 | is => 'rw', |
73 | isa => 'Bool', |
74 | ); |
75 | |
76 | |
dd3b58b0 |
77 | # The collation can be created two ways: |
78 | # 1. Collate a set of witnesses (with CollateX I guess) and process |
79 | # the results as in 2. |
80 | # 2. Read a pre-prepared collation in one of a variety of formats, |
81 | # and make the graph from that. |
82 | |
83 | # The graph itself will (for now) be immutable, and the positions |
84 | # within the graph will also be immutable. We need to calculate those |
85 | # positions upon graph construction. The equivalences between graph |
86 | # nodes will be mutable, entirely determined by the user (or possibly |
87 | # by some semantic pre-processing provided by the user.) So the |
88 | # constructor should just make an empty equivalences object. The |
89 | # constructor will also need to make the witness objects, if we didn't |
90 | # come through option 1. |
91 | |
d047cd52 |
92 | sub BUILD { |
93 | my( $self, $args ) = @_; |
8e1394aa |
94 | $self->graph->use_class('node', 'Text::Tradition::Collation::Reading'); |
d047cd52 |
95 | |
4a8828f0 |
96 | # Pass through any graph-specific options. |
97 | my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse'; |
98 | $self->graph->set_attribute( 'node', 'shape', $shape ); |
d047cd52 |
99 | } |
784877d9 |
100 | |
7854e12e |
101 | # Wrapper around add_path |
102 | |
103 | around add_path => sub { |
104 | my $orig = shift; |
105 | my $self = shift; |
106 | |
107 | # Make sure there are three arguments |
108 | unless( @_ == 3 ) { |
109 | warn "Call add_path with args source, target, witness"; |
110 | return; |
111 | } |
112 | # Make sure the proposed path does not yet exist |
113 | my( $source, $target, $wit ) = @_; |
114 | $source = $self->reading( $source ) |
115 | unless ref( $source ) eq 'Text::Tradition::Collation::Reading'; |
116 | $target = $self->reading( $target ) |
117 | unless ref( $target ) eq 'Text::Tradition::Collation::Reading'; |
118 | foreach my $path ( $source->edges_to( $target ) ) { |
119 | if( $path->label eq $wit ) { |
120 | return; |
121 | } |
122 | } |
123 | # Do the deed |
124 | $self->$orig( @_ ); |
125 | }; |
126 | |
127 | # Wrapper around merge_nodes |
784877d9 |
128 | |
129 | sub merge_readings { |
130 | my $self = shift; |
131 | my $first_node = shift; |
132 | my $second_node = shift; |
133 | $first_node->merge_from( $second_node ); |
134 | unshift( @_, $first_node, $second_node ); |
135 | return $self->graph->merge_nodes( @_ ); |
136 | } |
137 | |
8e1394aa |
138 | =head2 Output method(s) |
139 | |
140 | =over |
141 | |
142 | =item B<as_svg> |
143 | |
144 | print $graph->as_svg( $recalculate ); |
145 | |
146 | Returns an SVG string that represents the graph. Uses GraphViz to do |
4a8828f0 |
147 | this, because Graph::Easy doesn\'t cope well with long graphs. Unless |
8e1394aa |
148 | $recalculate is passed (and is a true value), the method will return a |
149 | cached copy of the SVG after the first call to the method. |
150 | |
151 | =cut |
152 | |
153 | sub as_svg { |
154 | my( $self, $recalc ) = @_; |
155 | return $self->svg if $self->has_svg; |
156 | |
1f563ac3 |
157 | $self->collapse_graph_edges(); |
8e1394aa |
158 | $self->_save_graphviz( $self->graph->as_graphviz() ) |
159 | unless( $self->has_graphviz && !$recalc ); |
160 | |
161 | my @cmd = qw/dot -Tsvg/; |
162 | my( $svg, $err ); |
163 | my $in = $self->graphviz; |
164 | run( \@cmd, \$in, ">", binary(), \$svg ); |
165 | $self->{'svg'} = $svg; |
1f563ac3 |
166 | $self->expand_graph_edges(); |
8e1394aa |
167 | return $svg; |
168 | } |
169 | |
170 | =item B<as_graphml> |
171 | |
172 | print $graph->as_graphml( $recalculate ) |
173 | |
174 | Returns a GraphML representation of the collation graph, with |
175 | transposition information and position information. Unless |
176 | $recalculate is passed (and is a true value), the method will return a |
177 | cached copy of the SVG after the first call to the method. |
178 | |
179 | =cut |
180 | |
181 | sub as_graphml { |
182 | my( $self, $recalc ) = @_; |
183 | return $self->graphml if $self->has_graphml; |
184 | |
185 | # Some namespaces |
186 | my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns'; |
187 | my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance'; |
188 | my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' . |
189 | 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; |
190 | |
191 | # Create the document and root node |
192 | my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" ); |
193 | my $root = $graphml->createElementNS( $graphml_ns, 'graphml' ); |
194 | $graphml->setDocumentElement( $root ); |
195 | $root->setNamespace( $xsi_ns, 'xsi', 0 ); |
196 | $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema ); |
197 | |
198 | # Add the data keys for nodes |
199 | my @node_data = ( 'name', 'token', 'identical', 'position' ); |
200 | foreach my $ndi ( 0 .. $#node_data ) { |
201 | my $key = $root->addNewChild( $graphml_ns, 'key' ); |
202 | $key->setAttribute( 'attr.name', $node_data[$ndi] ); |
203 | $key->setAttribute( 'attr.type', 'string' ); |
204 | $key->setAttribute( 'for', 'node' ); |
205 | $key->setAttribute( 'id', 'd'.$ndi ); |
206 | } |
207 | |
208 | # Add the data keys for edges |
209 | my %wit_hash; |
210 | my $wit_ctr = 0; |
1f563ac3 |
211 | foreach my $wit ( @{$self->tradition->witnesses} ) { |
8e1394aa |
212 | my $wit_key = 'w' . $wit_ctr++; |
213 | $wit_hash{$wit} = $wit_key; |
214 | my $key = $root->addNewChild( $graphml_ns, 'key' ); |
215 | $key->setAttribute( 'attr.name', $wit ); |
216 | $key->setAttribute( 'attr.type', 'string' ); |
217 | $key->setAttribute( 'for', 'edge' ); |
218 | $key->setAttribute( 'id', $wit_key ); |
219 | } |
220 | |
221 | # Add the graph, its nodes, and its edges |
1f563ac3 |
222 | $self->collapse_graph_edges(); |
8e1394aa |
223 | my $graph = $root->addNewChild( $graphml_ns, 'graph' ); |
224 | $graph->setAttribute( 'edgedefault', 'directed' ); |
225 | $graph->setAttribute( 'id', 'g0' ); # TODO make this meaningful |
226 | $graph->setAttribute( 'parse.edgeids', 'canonical' ); |
227 | $graph->setAttribute( 'parse.edges', $self->edges() ); |
228 | $graph->setAttribute( 'parse.nodeids', 'canonical' ); |
229 | $graph->setAttribute( 'parse.nodes', $self->nodes() ); |
230 | $graph->setAttribute( 'parse.order', 'nodesfirst' ); |
231 | |
232 | my $node_ctr = 0; |
233 | my %node_hash; |
234 | foreach my $n ( $self->readings ) { |
235 | my %this_node_data = (); |
236 | foreach my $ndi ( 0 .. $#node_data ) { |
237 | my $value; |
238 | $this_node_data{'d'.$ndi} = $n->name if $node_data[$ndi] eq 'name'; |
239 | $this_node_data{'d'.$ndi} = $n->label |
240 | if $node_data[$ndi] eq 'token'; |
241 | $this_node_data{'d'.$ndi} = $n->primary->name if $n->has_primary; |
242 | $this_node_data{'d'.$ndi} = |
243 | $self->{'positions'}->node_position( $n ) |
244 | if $node_data[$ndi] eq 'position'; |
245 | } |
246 | my $node_el = $graph->addNewChild( $graphml_ns, 'node' ); |
247 | my $node_xmlid = 'n' . $node_ctr++; |
248 | $node_hash{ $n->name } = $node_xmlid; |
249 | $node_el->setAttribute( 'id', $node_xmlid ); |
250 | |
251 | foreach my $dk ( keys %this_node_data ) { |
252 | my $d_el = $node_el->addNewChild( $graphml_ns, 'data' ); |
253 | $d_el->setAttribute( 'key', $dk ); |
254 | $d_el->appendTextChild( $this_node_data{$dk} ); |
255 | } |
256 | } |
257 | |
1f563ac3 |
258 | foreach my $e ( $self->paths() ) { |
8e1394aa |
259 | my( $name, $from, $to ) = ( $e->name, |
260 | $node_hash{ $e->from()->name() }, |
261 | $node_hash{ $e->to()->name() } ); |
262 | my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' ); |
263 | $edge_el->setAttribute( 'source', $from ); |
264 | $edge_el->setAttribute( 'target', $to ); |
265 | $edge_el->setAttribute( 'id', $name ); |
266 | # TODO Got to add the witnesses |
267 | } |
268 | |
269 | # Return the thing |
270 | $self->_save_graphml( $graphml ); |
1f563ac3 |
271 | $self->expand_graph_edges(); |
8e1394aa |
272 | return $graphml; |
273 | } |
274 | |
1f563ac3 |
275 | sub collapse_graph_edges { |
276 | my $self = shift; |
277 | # Our collation graph has an edge per witness. This is great for |
278 | # calculation purposes, but terrible for display. Thus we want to |
279 | # display only one edge between any two nodes. |
280 | |
281 | return if $self->collapsed; |
282 | |
283 | print STDERR "Collapsing path edges in graph...\n"; |
284 | |
285 | # Don't list out every witness if we have more than half to list. |
286 | my $majority = int( scalar( @{$self->tradition->witnesses} ) / 2 ) + 1; |
287 | foreach my $node( $self->readings ) { |
288 | my $newlabels = {}; |
289 | # We will visit each node, so we only look ahead. |
290 | foreach my $edge ( $node->outgoing() ) { |
291 | add_hash_entry( $newlabels, $edge->to->name, $edge->name ); |
292 | $self->del_path( $edge ); |
293 | } |
294 | |
295 | foreach my $newdest ( keys %$newlabels ) { |
296 | my $label; |
297 | my @compressed_wits = (); |
298 | if( @{$newlabels->{$newdest}} < $majority ) { |
299 | $label = join( ', ', @{$newlabels->{$newdest}} ); |
300 | } else { |
301 | ## TODO FIX THIS HACK |
302 | my @pclabels; |
303 | foreach my $wit ( @{$newlabels->{$newdest}} ) { |
304 | if( $wit =~ /^(.*?)(\s*\(?p\.\s*c\.\)?)$/ ) { |
305 | push( @pclabels, $wit ); |
306 | } else { |
307 | push( @compressed_wits, $wit ); |
308 | } |
309 | } |
310 | $label = join( ', ', 'majority', @pclabels ); |
311 | } |
312 | |
313 | my $newedge = |
314 | $self->add_path( $node, $self->reading( $newdest ), $label ); |
315 | if( @compressed_wits ) { |
316 | ## TODO fix this hack too. |
317 | $newedge->set_attribute( 'class', |
318 | join( '|', @compressed_wits ) ); |
319 | } |
320 | } |
321 | } |
322 | |
323 | $self->collapsed( 1 ); |
324 | } |
325 | |
326 | sub expand_graph_edges { |
327 | my $self = shift; |
328 | # Our collation graph has only one edge between any two nodes. |
329 | # This is great for display, but not so great for analysis. |
330 | # Expand this so that each witness has its own edge between any |
331 | # two reading nodes. |
332 | return unless $self->collapsed; |
333 | |
334 | print STDERR "Expanding path edges in graph...\n"; |
335 | |
336 | foreach my $edge( $self->paths ) { |
337 | my $from = $edge->from; |
338 | my $to = $edge->to; |
339 | my @wits = split( /, /, $edge->label ); |
340 | if( grep { $_ eq 'majority' } @wits ) { |
341 | push( @wits, split( /\|/, $edge->get_attribute( 'class' ) ) ); |
342 | } |
343 | $self->del_path( $edge ); |
344 | foreach ( @wits ) { |
345 | $self->add_path( $from, $to, $_ ); |
346 | } |
347 | } |
348 | $self->collapsed( 0 ); |
349 | } |
350 | |
8e1394aa |
351 | =back |
352 | |
de51424a |
353 | =head2 Navigation methods |
354 | |
355 | =over |
356 | |
8e1394aa |
357 | =item B<start> |
358 | |
359 | my $beginning = $collation->start(); |
360 | |
361 | Returns the beginning of the collation, a meta-reading with label '#START#'. |
362 | |
363 | =cut |
364 | |
365 | sub start { |
4a8828f0 |
366 | # Return the beginning reading of the graph. |
8e1394aa |
367 | my $self = shift; |
368 | my( $new_start ) = @_; |
369 | if( $new_start ) { |
370 | $self->del_reading( '#START#' ); |
371 | $self->graph->rename_node( $new_start, '#START#' ); |
372 | } |
373 | return $self->reading('#START#'); |
374 | } |
375 | |
e2902068 |
376 | =item B<reading_sequence> |
377 | |
378 | my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] ); |
379 | |
380 | Returns the ordered list of readings, starting with $first and ending |
381 | with $last, along the given witness path. If no path is specified, |
382 | assume that the path is that of the base text (if any.) |
383 | |
384 | =cut |
385 | |
386 | sub reading_sequence { |
387 | my( $self, $start, $end, $witness, $backup ) = @_; |
388 | |
930ff666 |
389 | $witness = $self->baselabel unless $witness; |
e2902068 |
390 | my @readings = ( $start ); |
391 | my %seen; |
392 | my $n = $start; |
930ff666 |
393 | while( $n && $n ne $end ) { |
e2902068 |
394 | if( exists( $seen{$n->name()} ) ) { |
395 | warn "Detected loop at " . $n->name(); |
396 | last; |
397 | } |
398 | $seen{$n->name()} = 1; |
399 | |
400 | my $next = $self->next_reading( $n, $witness, $backup ); |
401 | warn "Did not find any path for $witness from reading " . $n->name |
402 | unless $next; |
403 | push( @readings, $next ); |
404 | $n = $next; |
405 | } |
406 | # Check that the last reading is our end reading. |
407 | my $last = $readings[$#readings]; |
408 | warn "Last reading found from " . $start->label() . |
409 | " for witness $witness is not the end!" |
410 | unless $last eq $end; |
411 | |
412 | return @readings; |
413 | } |
414 | |
4a8828f0 |
415 | =item B<next_reading> |
8e1394aa |
416 | |
4a8828f0 |
417 | my $next_reading = $graph->next_reading( $reading, $witpath ); |
8e1394aa |
418 | |
4a8828f0 |
419 | Returns the reading that follows the given reading along the given witness |
930ff666 |
420 | path. |
8e1394aa |
421 | |
422 | =cut |
423 | |
4a8828f0 |
424 | sub next_reading { |
e2902068 |
425 | # Return the successor via the corresponding path. |
8e1394aa |
426 | my $self = shift; |
4a8828f0 |
427 | return $self->_find_linked_reading( 'next', @_ ); |
8e1394aa |
428 | } |
429 | |
4a8828f0 |
430 | =item B<prior_reading> |
8e1394aa |
431 | |
4a8828f0 |
432 | my $prior_reading = $graph->prior_reading( $reading, $witpath ); |
8e1394aa |
433 | |
4a8828f0 |
434 | Returns the reading that precedes the given reading along the given witness |
930ff666 |
435 | path. |
8e1394aa |
436 | |
437 | =cut |
438 | |
4a8828f0 |
439 | sub prior_reading { |
e2902068 |
440 | # Return the predecessor via the corresponding path. |
8e1394aa |
441 | my $self = shift; |
4a8828f0 |
442 | return $self->_find_linked_reading( 'prior', @_ ); |
8e1394aa |
443 | } |
444 | |
4a8828f0 |
445 | sub _find_linked_reading { |
e2902068 |
446 | my( $self, $direction, $node, $path, $alt_path ) = @_; |
447 | my @linked_paths = $direction eq 'next' |
8e1394aa |
448 | ? $node->outgoing() : $node->incoming(); |
e2902068 |
449 | return undef unless scalar( @linked_paths ); |
8e1394aa |
450 | |
e2902068 |
451 | # We have to find the linked path that contains all of the |
452 | # witnesses supplied in $path. |
453 | my( @path_wits, @alt_path_wits ); |
454 | @path_wits = $self->witnesses_of_label( $path ) if $path; |
455 | @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path; |
456 | my $base_le; |
457 | my $alt_le; |
458 | foreach my $le ( @linked_paths ) { |
930ff666 |
459 | if( $le->name eq $self->baselabel ) { |
e2902068 |
460 | $base_le = $le; |
461 | } else { |
462 | my @le_wits = $self->witnesses_of_label( $le->name ); |
463 | if( _is_within( \@path_wits, \@le_wits ) ) { |
464 | # This is the right path. |
465 | return $direction eq 'next' ? $le->to() : $le->from(); |
466 | } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) { |
467 | $alt_le = $le; |
468 | } |
8e1394aa |
469 | } |
470 | } |
e2902068 |
471 | # Got this far? Return the alternate path if it exists. |
472 | return $direction eq 'next' ? $alt_le->to() : $alt_le->from() |
473 | if $alt_le; |
474 | |
475 | # Got this far? Return the base path if it exists. |
476 | return $direction eq 'next' ? $base_le->to() : $base_le->from() |
477 | if $base_le; |
478 | |
479 | # Got this far? We have no appropriate path. |
8e1394aa |
480 | warn "Could not find $direction node from " . $node->label |
e2902068 |
481 | . " along path $path"; |
8e1394aa |
482 | return undef; |
483 | } |
484 | |
4a8828f0 |
485 | # Some set logic. |
486 | sub _is_within { |
487 | my( $set1, $set2 ) = @_; |
7854e12e |
488 | my $ret = @$set1; # will be 0, i.e. false, if set1 is empty |
4a8828f0 |
489 | foreach my $el ( @$set1 ) { |
490 | $ret = 0 unless grep { /^\Q$el\E$/ } @$set2; |
491 | } |
492 | return $ret; |
493 | } |
494 | |
de51424a |
495 | |
496 | ## INITIALIZATION METHODS - for use by parsers |
4a8828f0 |
497 | # Walk the paths for each witness in the graph, and return the nodes |
e2902068 |
498 | # that the graph has in common. If $using_base is true, some |
499 | # different logic is needed. |
4a8828f0 |
500 | |
501 | sub walk_witness_paths { |
502 | my( $self, $end ) = @_; |
503 | # For each witness, walk the path through the graph. |
504 | # Then we need to find the common nodes. |
505 | # TODO This method is going to fall down if we have a very gappy |
506 | # text in the collation. |
507 | my $paths = {}; |
3a1f2523 |
508 | my @common_readings; |
4a8828f0 |
509 | foreach my $wit ( @{$self->tradition->witnesses} ) { |
510 | my $curr_reading = $self->start; |
e2902068 |
511 | my @wit_path = $self->reading_sequence( $self->start, $end, |
512 | $wit->sigil ); |
4a8828f0 |
513 | $wit->path( \@wit_path ); |
e2902068 |
514 | |
515 | # Detect the common readings. |
930ff666 |
516 | @common_readings = _find_common( \@common_readings, \@wit_path ); |
4a8828f0 |
517 | } |
518 | |
519 | # Mark all the nodes as either common or not. |
3a1f2523 |
520 | foreach my $cn ( @common_readings ) { |
e2902068 |
521 | print STDERR "Setting " . $cn->name . " / " . $cn->label |
522 | . " as common node\n"; |
4a8828f0 |
523 | $cn->make_common; |
524 | } |
525 | foreach my $n ( $self->readings() ) { |
526 | $n->make_variant unless $n->is_common; |
527 | } |
3a1f2523 |
528 | # Return an array of the common nodes in order. |
529 | return @common_readings; |
4a8828f0 |
530 | } |
531 | |
930ff666 |
532 | sub _find_common { |
533 | my( $common_readings, $new_path ) = @_; |
534 | my @cr; |
535 | if( @$common_readings ) { |
536 | foreach my $n ( @$new_path ) { |
537 | push( @cr, $n ) if grep { $_ eq $n } @$common_readings; |
538 | } |
539 | } else { |
540 | push( @cr, @$new_path ); |
541 | } |
542 | return @cr; |
543 | } |
544 | |
545 | sub _remove_common { |
546 | my( $common_readings, $divergence ) = @_; |
547 | my @cr; |
548 | my %diverged; |
549 | map { $diverged{$_->name} = 1 } @$divergence; |
550 | foreach( @$common_readings ) { |
551 | push( @cr, $_ ) unless $diverged{$_->name}; |
552 | } |
553 | return @cr; |
554 | } |
555 | |
556 | |
e2902068 |
557 | # An alternative to walk_witness_paths, for use when a collation is |
558 | # constructed from a base text and an apparatus. Also modifies the |
559 | # collation graph to remove all 'base text' paths and replace them |
560 | # with real witness paths. |
561 | |
562 | sub walk_and_expand_base { |
563 | my( $self, $end ) = @_; |
564 | |
930ff666 |
565 | my @common_readings; |
e2902068 |
566 | foreach my $wit ( @{$self->tradition->witnesses} ) { |
7854e12e |
567 | my $sig = $wit->sigil; |
1f563ac3 |
568 | $DB::single = 1 if $sig eq 'Vb5'; |
e2902068 |
569 | my $post_sig; |
570 | $post_sig = $wit->post_correctione |
571 | if $wit->has_post_correctione; |
930ff666 |
572 | |
7854e12e |
573 | my @wit_path = $self->reading_sequence( $self->start, $end, $sig ); |
574 | $wit->path( \@wit_path ); |
575 | $self->connect_readings_for_witness( $wit ); |
930ff666 |
576 | @common_readings = _find_common( \@common_readings, \@wit_path ); |
7854e12e |
577 | |
578 | # If there is a post-correctio, get its path and compare. |
579 | # Add a correction range for each divergence. |
580 | if( $post_sig ) { |
7854e12e |
581 | my @corr_wit_path = $self->reading_sequence( $self->start, $end, |
582 | "$sig$post_sig", $sig ); |
583 | |
584 | # Map ante-corr readings to their indices |
585 | my %in_orig; |
586 | my $i = 0; |
587 | map { $in_orig{$_->name} = $i++ } @wit_path; |
588 | |
589 | # Look for divergences |
590 | my $diverged = 0; |
591 | my $last_common; |
592 | my @correction; |
593 | foreach my $rdg ( @corr_wit_path ) { |
594 | if( exists( $in_orig{$rdg->name} ) && !$diverged ) { |
595 | # We are reading the same here |
596 | $last_common = $in_orig{$rdg->name}; |
7854e12e |
597 | } elsif ( exists( $in_orig{$rdg->name} ) ) { |
598 | # We have been diverging but are reading the same again. |
599 | # Add the correction to the witness. |
930ff666 |
600 | my $offset = $last_common + 1; |
601 | my $length = $in_orig{$rdg->name} - $offset; |
602 | $wit->add_correction( $offset, $length, @correction ); |
7854e12e |
603 | $diverged = 0; |
930ff666 |
604 | @common_readings = _remove_common( \@common_readings, \@correction ); |
7854e12e |
605 | @correction = (); |
606 | $last_common = $in_orig{$rdg->name}; |
607 | } elsif( $diverged ) { |
608 | # We are in the middle of a divergence. |
609 | push( @correction, $rdg ); |
610 | } else { |
611 | # We have started to diverge. Note it. |
612 | $diverged = 1; |
613 | push( @correction, $rdg ); |
614 | } |
e2902068 |
615 | } |
930ff666 |
616 | # Add any divergence that is at the end of the text |
7854e12e |
617 | if( $diverged ) { |
930ff666 |
618 | $wit->add_correction( $last_common+1, $#wit_path, \@correction ); |
e2902068 |
619 | } |
e2902068 |
620 | } |
7854e12e |
621 | } |
e2902068 |
622 | |
7854e12e |
623 | # Remove any 'base text' paths. |
624 | foreach my $path ( $self->paths ) { |
625 | $self->del_path( $path ) |
626 | if $path->label eq $self->baselabel; |
627 | } |
628 | } |
629 | |
630 | sub connect_readings_for_witness { |
631 | my( $self, $wit ) = @_; |
632 | my @chain = @{$wit->path}; |
633 | foreach my $idx ( 0 .. $#chain-1 ) { |
634 | $self->add_path( $chain[$idx], $chain[$idx+1], $wit->sigil ); |
635 | } |
e2902068 |
636 | } |
637 | |
4a8828f0 |
638 | sub common_readings { |
639 | my $self = shift; |
640 | my @common = grep { $_->is_common } $self->readings(); |
de51424a |
641 | return sort { _cmp_position( $a->position, $b->position ) } @common; |
4a8828f0 |
642 | } |
643 | |
644 | # Calculate the relative positions of nodes in the graph, if they |
645 | # were not given to us. |
646 | sub calculate_positions { |
3a1f2523 |
647 | my( $self, @ordered_common ) = @_; |
4a8828f0 |
648 | |
649 | # We have to calculate the position identifiers for each word, |
650 | # keyed on the common nodes. This will be 'fun'. The end result |
651 | # is a hash per witness, whose key is the word node and whose |
652 | # value is its position in the text. Common nodes are always N,1 |
653 | # so have identical positions in each text. |
4a8828f0 |
654 | |
655 | my $node_pos = {}; |
656 | foreach my $wit ( @{$self->tradition->witnesses} ) { |
930ff666 |
657 | print STDERR "Calculating positions in " . $wit->sigil . "\n"; |
658 | _update_positions_from_path( $wit->path, @ordered_common ); |
659 | _update_positions_from_path( $wit->corrected_path, @ordered_common ) |
660 | if $wit->has_post_correctione; |
661 | } |
662 | |
663 | # DEBUG |
664 | foreach my $r ( $self->readings() ) { |
665 | print STDERR "Reading " . $r->name . "/" . $r->label . " has no position\n" |
666 | unless( $r->has_position ); |
4a8828f0 |
667 | } |
3a1f2523 |
668 | |
669 | $self->init_lemmata(); |
4a8828f0 |
670 | } |
671 | |
930ff666 |
672 | sub _update_positions_from_path { |
673 | my( $path, @ordered_common ) = @_; |
674 | |
675 | # First we walk the given path, making a matrix for the witness |
676 | # that corresponds to its eventual position identifier. Common |
677 | # nodes always start a new row, and are thus always in the first |
678 | # column. |
679 | |
680 | my $wit_matrix = []; |
681 | my $cn = 0; # We should hit the common readings in order. |
682 | my $row = []; |
683 | foreach my $wn ( @{$path} ) { |
684 | if( $wn eq $ordered_common[$cn] ) { |
685 | # Set up to look for the next common node, and |
686 | # start a new row of words. |
687 | $cn++; |
688 | push( @$wit_matrix, $row ) if scalar( @$row ); |
689 | $row = []; |
690 | } |
691 | push( @$row, $wn ); |
692 | } |
693 | push( @$wit_matrix, $row ); # Push the last row onto the matrix |
694 | |
695 | # Now we have a matrix per witness, so that each row in the |
696 | # matrix begins with a common node, and continues with all the |
697 | # variant words that appear in the witness. We turn this into |
698 | # real positions in row,cell format. But we need some |
699 | # trickery in order to make sure that each node gets assigned |
700 | # to only one position. |
701 | |
702 | foreach my $li ( 1..scalar(@$wit_matrix) ) { |
703 | foreach my $di ( 1..scalar(@{$wit_matrix->[$li-1]}) ) { |
704 | my $reading = $wit_matrix->[$li-1]->[$di-1]; |
705 | my $position = "$li,$di"; |
706 | # If we have seen this node before, we need to compare |
707 | # its position with what went before. |
708 | unless( $reading->has_position && |
709 | _cmp_position( $position, $reading->position ) < 1 ) { |
710 | # The new position ID replaces the old one. |
711 | $reading->position( $position ); |
712 | } # otherwise, the old position needs to stay. |
713 | } |
714 | } |
715 | } |
716 | |
4a8828f0 |
717 | sub _cmp_position { |
718 | my( $a, $b ) = @_; |
de51424a |
719 | if ( $a && $b ) { |
720 | my @pos_a = split(/,/, $a ); |
721 | my @pos_b = split(/,/, $b ); |
722 | |
723 | my $big_cmp = $pos_a[0] <=> $pos_b[0]; |
724 | return $big_cmp if $big_cmp; |
725 | # else |
726 | return $pos_a[1] <=> $pos_b[1]; |
727 | } elsif ( $b ) { # a is undefined |
728 | return -1; |
729 | } elsif ( $a ) { # b is undefined |
730 | return 1; |
731 | } |
732 | return 0; # they are both undefined |
8e1394aa |
733 | } |
3a1f2523 |
734 | |
735 | sub all_positions { |
736 | my $self = shift; |
737 | my %positions = (); |
738 | map { $positions{$_->position} = 1 } $self->readings; |
de51424a |
739 | my @answer = sort { _cmp_position( $a, $b ) } keys( %positions ); |
740 | return @answer; |
3a1f2523 |
741 | } |
742 | |
743 | sub readings_at_position { |
744 | my( $self, $pos ) = @_; |
745 | my @answer = grep { $_->position eq $pos } $self->readings; |
746 | return @answer; |
747 | } |
748 | |
749 | ## Lemmatizer functions |
750 | |
751 | sub init_lemmata { |
752 | my $self = shift; |
753 | |
754 | foreach my $position ( $self->all_positions ) { |
755 | $self->lemmata->{$position} = undef; |
756 | } |
757 | |
758 | foreach my $cr ( $self->common_readings ) { |
759 | $self->lemmata->{$cr->position} = $cr->name; |
760 | } |
761 | } |
762 | |
763 | =item B<lemma_readings> |
764 | |
765 | my @state = $graph->lemma_readings( @readings_delemmatized ); |
766 | |
767 | Takes a list of readings that have just been delemmatized, and returns |
768 | a set of tuples of the form ['reading', 'state'] that indicates what |
769 | changes need to be made to the graph. |
770 | |
771 | =over |
772 | |
773 | =item * |
774 | |
775 | A state of 1 means 'lemmatize this reading' |
776 | |
777 | =item * |
778 | |
779 | A state of 0 means 'delemmatize this reading' |
780 | |
781 | =item * |
782 | |
783 | A state of undef means 'an ellipsis belongs in the text here because |
784 | no decision has been made / an earlier decision was backed out' |
785 | |
786 | =back |
787 | |
788 | =cut |
789 | |
790 | sub lemma_readings { |
791 | my( $self, @toggled_off_nodes ) = @_; |
792 | |
793 | # First get the positions of those nodes which have been |
794 | # toggled off. |
795 | my $positions_off = {}; |
796 | map { $positions_off->{ $_->position } = $_->name } @toggled_off_nodes; |
de51424a |
797 | |
3a1f2523 |
798 | # Now for each position, we have to see if a node is on, and we |
799 | # have to see if a node has been turned off. |
800 | my @answer; |
801 | foreach my $pos ( $self->all_positions() ) { |
802 | # Find the state of this position. If there is an active node, |
803 | # its name will be the state; otherwise the state will be 0 |
804 | # (nothing at this position) or undef (ellipsis at this position) |
805 | my $active = $self->lemmata->{$pos}; |
806 | |
807 | # Is there a formerly active node that was toggled off? |
808 | if( exists( $positions_off->{$pos} ) ) { |
809 | my $off_node = $positions_off->{$pos}; |
810 | if( $active && $active ne $off_node) { |
811 | push( @answer, [ $off_node, 0 ], [ $active, 1 ] ); |
812 | } else { |
813 | push( @answer, [ $off_node, $active ] ); |
814 | } |
815 | |
816 | # No formerly active node, so we just see if there is a currently |
817 | # active one. |
818 | } elsif( $active ) { |
819 | # Push the active node, whatever it is. |
820 | push( @answer, [ $active, 1 ] ); |
821 | } else { |
822 | # Push the state that is there. Arbitrarily use the first node |
823 | # at that position. |
824 | my @pos_nodes = $self->readings_at_position( $pos ); |
de51424a |
825 | push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] ); |
3a1f2523 |
826 | } |
827 | } |
828 | |
829 | return @answer; |
830 | } |
831 | |
de51424a |
832 | =item B<toggle_reading> |
833 | |
834 | my @readings_delemmatized = $graph->toggle_reading( $reading_name ); |
835 | |
836 | Takes a reading node name, and either lemmatizes or de-lemmatizes |
837 | it. Returns a list of all readings that are de-lemmatized as a result |
838 | of the toggle. |
839 | |
840 | =cut |
841 | |
842 | sub toggle_reading { |
843 | my( $self, $rname ) = @_; |
844 | |
845 | return unless $rname; |
846 | my $reading = $self->reading( $rname ); |
847 | if( !$reading || $reading->is_common() ) { |
848 | # Do nothing, it's a common node. |
849 | return; |
850 | } |
851 | |
852 | my $pos = $reading->position; |
853 | my $old_state = $self->lemmata->{$pos}; |
854 | my @readings_off; |
855 | if( $old_state && $old_state eq $rname ) { |
856 | # Turn off the node. We turn on no others by default. |
857 | push( @readings_off, $reading ); |
858 | } else { |
859 | # Turn on the node. |
860 | $self->lemmata->{$pos} = $rname; |
861 | # Any other 'on' readings in the same position should be off. |
862 | push( @readings_off, $self->same_position_as( $reading ) ); |
863 | # Any node that is an identical transposed one should be off. |
864 | push( @readings_off, $reading->identical_readings ); |
865 | } |
866 | @readings_off = unique_list( @readings_off ); |
867 | |
868 | # Turn off the readings that need to be turned off. |
869 | my @readings_delemmatized; |
870 | foreach my $n ( @readings_off ) { |
871 | my $state = $self->lemmata->{$n->position}; |
872 | if( $state && $state eq $n->name ) { |
873 | # this reading is still on, so turn it off |
874 | push( @readings_delemmatized, $n ); |
875 | my $new_state = undef; |
876 | if( $n eq $reading ) { |
877 | # This is the reading that was clicked, so if there are no |
878 | # other readings there, turn off the position. In all other |
879 | # cases, restore the ellipsis. |
880 | my @other_n = $self->same_position_as( $n ); |
881 | $new_state = 0 unless @other_n; |
882 | } |
883 | $self->lemmata->{$n->position} = $new_state; |
884 | } elsif( $old_state && $old_state eq $n->name ) { |
885 | # another reading has already been turned on here |
886 | push( @readings_delemmatized, $n ); |
887 | } # else some other reading was on anyway, so pass. |
888 | } |
889 | return @readings_delemmatized; |
890 | } |
891 | |
892 | sub same_position_as { |
893 | my( $self, $reading ) = @_; |
894 | my $pos = $reading->position; |
895 | my @same = grep { $_ ne $reading } $self->readings_at_position( $reading->position ); |
896 | return @same; |
897 | } |
3a1f2523 |
898 | |
4a8828f0 |
899 | # Return the string that joins together a list of witnesses for |
900 | # display on a single path. |
901 | sub path_label { |
902 | my $self = shift; |
903 | return join( $self->wit_list_separator, @_ ); |
904 | } |
905 | |
906 | sub witnesses_of_label { |
de51424a |
907 | my( $self, $label ) = @_; |
4a8828f0 |
908 | my $regex = $self->wit_list_separator; |
de51424a |
909 | my @answer = split( /\Q$regex\E/, $label ); |
910 | return @answer; |
4a8828f0 |
911 | } |
8e1394aa |
912 | |
de51424a |
913 | sub unique_list { |
914 | my( @list ) = @_; |
915 | my %h; |
916 | map { $h{$_->name} = $_ } @list; |
917 | return values( %h ); |
918 | } |
919 | |
1f563ac3 |
920 | sub add_hash_entry { |
921 | my( $hash, $key, $entry ) = @_; |
922 | if( exists $hash->{$key} ) { |
923 | push( @{$hash->{$key}}, $entry ); |
924 | } else { |
925 | $hash->{$key} = [ $entry ]; |
926 | } |
927 | } |
928 | |
dd3b58b0 |
929 | no Moose; |
930 | __PACKAGE__->meta->make_immutable; |