Commit | Line | Data |
dd3b58b0 |
1 | package Text::Tradition::Collation; |
d047cd52 |
2 | |
6771a1b1 |
3 | use feature 'say'; |
910a0a6d |
4 | use Encode qw( decode_utf8 ); |
5 | use File::Temp; |
bfcbcecb |
6 | use File::Which; |
c9bf3dbf |
7 | use Graph; |
8e1394aa |
8 | use IPC::Run qw( run binary ); |
82fa4d57 |
9 | use Text::CSV; |
5c0072ef |
10 | use Text::Tradition::Collation::Data; |
b15511bf |
11 | use Text::Tradition::Collation::Reading; |
22222af9 |
12 | use Text::Tradition::Collation::RelationshipStore; |
63778331 |
13 | use Text::Tradition::Error; |
cc31ebaa |
14 | use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx ); |
5c0072ef |
15 | use XML::LibXML; |
16 | use XML::LibXML::XPathContext; |
dd3b58b0 |
17 | use Moose; |
18 | |
5c0072ef |
19 | has _data => ( |
20 | isa => 'Text::Tradition::Collation::Data', |
21 | is => 'ro', |
22 | required => 1, |
23 | handles => [ qw( |
24 | sequence |
25 | paths |
26 | _set_relations |
27 | relations |
28 | _set_start |
29 | _set_end |
30 | ac_label |
31 | has_cached_table |
32 | relationships |
33 | related_readings |
34 | get_relationship |
35 | del_relationship |
36 | equivalence |
37 | equivalence_graph |
38 | readings |
39 | reading |
40 | _add_reading |
41 | del_reading |
42 | has_reading |
43 | wit_list_separator |
44 | baselabel |
45 | linear |
46 | wordsep |
47 | start |
48 | end |
49 | cached_table |
50 | _graphcalc_done |
51 | has_cached_svg |
52 | wipe_table |
53 | )] |
54 | ); |
dd3b58b0 |
55 | |
3a2ebbf4 |
56 | has 'tradition' => ( |
57 | is => 'ro', |
d047cd52 |
58 | isa => 'Text::Tradition', |
8cfd99c4 |
59 | writer => '_set_tradition', |
8d9a1cd8 |
60 | weak_ref => 1, |
d047cd52 |
61 | ); |
dd3b58b0 |
62 | |
4e5a7b2c |
63 | =head1 NAME |
64 | |
65 | Text::Tradition::Collation - a software model for a text collation |
66 | |
67 | =head1 SYNOPSIS |
68 | |
69 | use Text::Tradition; |
70 | my $t = Text::Tradition->new( |
71 | 'name' => 'this is a text', |
72 | 'input' => 'TEI', |
73 | 'file' => '/path/to/tei_parallel_seg_file.xml' ); |
74 | |
75 | my $c = $t->collation; |
76 | my @readings = $c->readings; |
77 | my @paths = $c->paths; |
78 | my @relationships = $c->relationships; |
79 | |
80 | my $svg_variant_graph = $t->collation->as_svg(); |
81 | |
82 | =head1 DESCRIPTION |
83 | |
84 | Text::Tradition is a library for representation and analysis of collated |
85 | texts, particularly medieval ones. The Collation is the central feature of |
86 | a Tradition, where the text, its sequence of readings, and its relationships |
87 | between readings are actually kept. |
88 | |
89 | =head1 CONSTRUCTOR |
90 | |
91 | =head2 new |
92 | |
93 | The constructor. Takes a hash or hashref of the following arguments: |
94 | |
95 | =over |
96 | |
97 | =item * tradition - The Text::Tradition object to which the collation |
98 | belongs. Required. |
99 | |
100 | =item * linear - Whether the collation should be linear; that is, whether |
101 | transposed readings should be treated as two linked readings rather than one, |
102 | and therefore whether the collation graph is acyclic. Defaults to true. |
103 | |
4e5a7b2c |
104 | =item * baselabel - The default label for the path taken by a base text |
105 | (if any). Defaults to 'base text'. |
106 | |
107 | =item * wit_list_separator - The string to join a list of witnesses for |
108 | purposes of making labels in display graphs. Defaults to ', '. |
109 | |
110 | =item * ac_label - The extra label to tack onto a witness sigil when |
111 | representing another layer of path for the given witness - that is, when |
112 | a text has more than one possible reading due to scribal corrections or |
113 | the like. Defaults to ' (a.c.)'. |
114 | |
4e483aa5 |
115 | =item * wordsep - The string used to separate words in the original text. |
116 | Defaults to ' '. |
117 | |
4e5a7b2c |
118 | =back |
119 | |
120 | =head1 ACCESSORS |
121 | |
122 | =head2 tradition |
123 | |
124 | =head2 linear |
125 | |
4e5a7b2c |
126 | =head2 wit_list_separator |
127 | |
128 | =head2 baselabel |
129 | |
130 | =head2 ac_label |
131 | |
4e483aa5 |
132 | =head2 wordsep |
133 | |
4e5a7b2c |
134 | Simple accessors for collation attributes. |
135 | |
136 | =head2 start |
137 | |
138 | The meta-reading at the start of every witness path. |
139 | |
140 | =head2 end |
141 | |
142 | The meta-reading at the end of every witness path. |
143 | |
144 | =head2 readings |
145 | |
146 | Returns all Reading objects in the graph. |
147 | |
148 | =head2 reading( $id ) |
149 | |
150 | Returns the Reading object corresponding to the given ID. |
151 | |
152 | =head2 add_reading( $reading_args ) |
153 | |
154 | Adds a new reading object to the collation. |
155 | See L<Text::Tradition::Collation::Reading> for the available arguments. |
156 | |
157 | =head2 del_reading( $object_or_id ) |
158 | |
159 | Removes the given reading from the collation, implicitly removing its |
160 | paths and relationships. |
161 | |
4e483aa5 |
162 | =head2 merge_readings( $main, $second, $concatenate, $with_str ) |
163 | |
164 | Merges the $second reading into the $main one. If $concatenate is true, then |
165 | the merged node will carry the text of both readings, concatenated with either |
166 | $with_str (if specified) or a sensible default (the empty string if the |
167 | appropriate 'join_*' flag is set on either reading, or else $self->wordsep.) |
4e5a7b2c |
168 | |
4e483aa5 |
169 | The first two arguments may be either readings or reading IDs. |
4e5a7b2c |
170 | |
171 | =head2 has_reading( $id ) |
172 | |
173 | Predicate to see whether a given reading ID is in the graph. |
174 | |
175 | =head2 reading_witnesses( $object_or_id ) |
176 | |
177 | Returns a list of sigils whose witnesses contain the reading. |
178 | |
179 | =head2 paths |
180 | |
181 | Returns all reading paths within the document - that is, all edges in the |
182 | collation graph. Each path is an arrayref of [ $source, $target ] reading IDs. |
183 | |
184 | =head2 add_path( $source, $target, $sigil ) |
185 | |
186 | Links the given readings in the collation in sequence, under the given witness |
187 | sigil. The readings may be specified by object or ID. |
188 | |
189 | =head2 del_path( $source, $target, $sigil ) |
190 | |
191 | Links the given readings in the collation in sequence, under the given witness |
192 | sigil. The readings may be specified by object or ID. |
193 | |
194 | =head2 has_path( $source, $target ); |
195 | |
196 | Returns true if the two readings are linked in sequence in any witness. |
197 | The readings may be specified by object or ID. |
198 | |
199 | =head2 relationships |
200 | |
201 | Returns all Relationship objects in the collation. |
202 | |
203 | =head2 add_relationship( $reading, $other_reading, $options ) |
204 | |
205 | Adds a new relationship of the type given in $options between the two readings, |
206 | which may be specified by object or ID. Returns a value of ( $status, @vectors) |
207 | where $status is true on success, and @vectors is a list of relationship edges |
208 | that were ultimately added. |
209 | See L<Text::Tradition::Collation::Relationship> for the available options. |
210 | |
211 | =cut |
dd3b58b0 |
212 | |
5c0072ef |
213 | sub BUILDARGS { |
214 | my ( $class, @args ) = @_; |
215 | my %args = @args == 1 ? %{ $args[0] } : @args; |
216 | # TODO determine these from the Moose::Meta object |
217 | my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel |
218 | linear wordsep start end cached_table _graphcalc_done); |
219 | my %data_args; |
220 | for my $attr (@delegate_attrs) { |
221 | $data_args{$attr} = delete $args{$attr} if exists $args{$attr}; |
222 | } |
223 | $args{_data} = Text::Tradition::Collation::Data->new(%data_args); |
224 | return \%args; |
225 | } |
226 | |
d047cd52 |
227 | sub BUILD { |
3a2ebbf4 |
228 | my $self = shift; |
22222af9 |
229 | $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) ); |
e4b73942 |
230 | $self->_set_start( $self->add_reading( |
231 | { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) ); |
232 | $self->_set_end( $self->add_reading( |
233 | { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) ); |
d047cd52 |
234 | } |
784877d9 |
235 | |
3a2ebbf4 |
236 | ### Reading construct/destruct functions |
237 | |
238 | sub add_reading { |
239 | my( $self, $reading ) = @_; |
240 | unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) { |
241 | my %args = %$reading; |
e4b73942 |
242 | if( $args{'init'} ) { |
243 | # If we are initializing an empty collation, don't assume that we |
244 | # have set a tradition. |
245 | delete $args{'init'}; |
8943ff68 |
246 | } elsif( $self->tradition->can('language') && $self->tradition->has_language |
247 | && !exists $args{'language'} ) { |
6ad2ce78 |
248 | $args{'language'} = $self->tradition->language; |
249 | } |
3a2ebbf4 |
250 | $reading = Text::Tradition::Collation::Reading->new( |
251 | 'collation' => $self, |
252 | %args ); |
253 | } |
254 | # First check to see if a reading with this ID exists. |
255 | if( $self->reading( $reading->id ) ) { |
63778331 |
256 | throw( "Collation already has a reading with id " . $reading->id ); |
3a2ebbf4 |
257 | } |
c1915ab9 |
258 | $self->_graphcalc_done(0); |
3a2ebbf4 |
259 | $self->_add_reading( $reading->id => $reading ); |
260 | # Once the reading has been added, put it in both graphs. |
261 | $self->sequence->add_vertex( $reading->id ); |
cecbe56d |
262 | $self->relations->add_reading( $reading->id ); |
3a2ebbf4 |
263 | return $reading; |
eca16057 |
264 | }; |
265 | |
3a2ebbf4 |
266 | around del_reading => sub { |
267 | my $orig = shift; |
268 | my $self = shift; |
269 | my $arg = shift; |
270 | |
cecbe56d |
271 | if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) { |
272 | $arg = $arg->id; |
3a2ebbf4 |
273 | } |
3a2ebbf4 |
274 | # Remove the reading from the graphs. |
c1915ab9 |
275 | $self->_graphcalc_done(0); |
4e483aa5 |
276 | $self->_clear_cache; # Explicitly clear caches to GC the reading |
cecbe56d |
277 | $self->sequence->delete_vertex( $arg ); |
278 | $self->relations->delete_reading( $arg ); |
3a2ebbf4 |
279 | |
280 | # Carry on. |
cecbe56d |
281 | $self->$orig( $arg ); |
3a2ebbf4 |
282 | }; |
7854e12e |
283 | |
4e483aa5 |
284 | =begin testing |
285 | |
286 | use Text::Tradition; |
287 | |
288 | my $cxfile = 't/data/Collatex-16.xml'; |
289 | my $t = Text::Tradition->new( |
290 | 'name' => 'inline', |
291 | 'input' => 'CollateX', |
292 | 'file' => $cxfile, |
293 | ); |
294 | my $c = $t->collation; |
295 | |
296 | my $rno = scalar $c->readings; |
297 | # Split n21 for testing purposes |
298 | my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } ); |
299 | my $old_r = $c->reading( 'n21' ); |
300 | $old_r->alter_text( 'to' ); |
301 | $c->del_path( 'n20', 'n21', 'A' ); |
302 | $c->add_path( 'n20', 'n21p0', 'A' ); |
303 | $c->add_path( 'n21p0', 'n21', 'A' ); |
304 | $c->flatten_ranks(); |
305 | ok( $c->reading( 'n21p0' ), "New reading exists" ); |
306 | is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" ); |
307 | |
679f17e1 |
308 | # Combine n3 and n4 ( with his ) |
4e483aa5 |
309 | $c->merge_readings( 'n3', 'n4', 1 ); |
310 | ok( !$c->reading('n4'), "Reading n4 is gone" ); |
311 | is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" ); |
312 | |
679f17e1 |
313 | # Collapse n9 and n10 ( rood / root ) |
314 | $c->merge_readings( 'n9', 'n10' ); |
315 | ok( !$c->reading('n10'), "Reading n10 is gone" ); |
316 | is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" ); |
4e483aa5 |
317 | |
318 | # Combine n21 and n21p0 |
319 | my $remaining = $c->reading('n21'); |
320 | $remaining ||= $c->reading('n22'); # one of these should still exist |
321 | $c->merge_readings( 'n21p0', $remaining, 1 ); |
322 | ok( !$c->reading('n21'), "Reading $remaining is gone" ); |
323 | is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" ); |
324 | |
325 | =end testing |
326 | |
327 | =cut |
7854e12e |
328 | |
3a2ebbf4 |
329 | sub merge_readings { |
330 | my $self = shift; |
331 | |
56772e8c |
332 | # Sanity check |
333 | my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ ); |
334 | my $mergemeta = $kept_obj->is_meta; |
335 | throw( "Cannot merge meta and non-meta reading" ) |
336 | unless ( $mergemeta && $del_obj->is_meta ) |
337 | || ( !$mergemeta && !$del_obj->is_meta ); |
338 | if( $mergemeta ) { |
339 | throw( "Cannot merge with start or end node" ) |
340 | if( $kept_obj eq $self->start || $kept_obj eq $self->end |
341 | || $del_obj eq $self->start || $del_obj eq $self->end ); |
a445ce40 |
342 | throw( "Cannot combine text of meta readings" ) if $combine; |
56772e8c |
343 | } |
3a2ebbf4 |
344 | # We only need the IDs for adding paths to the graph, not the reading |
345 | # objects themselves. |
56772e8c |
346 | my $kept = $kept_obj->id; |
347 | my $deleted = $del_obj->id; |
c1915ab9 |
348 | $self->_graphcalc_done(0); |
10943ab0 |
349 | |
3a2ebbf4 |
350 | # The kept reading should inherit the paths and the relationships |
351 | # of the deleted reading. |
352 | foreach my $path ( $self->sequence->edges_at( $deleted ) ) { |
353 | my @vector = ( $kept ); |
354 | push( @vector, $path->[1] ) if $path->[0] eq $deleted; |
355 | unshift( @vector, $path->[0] ) if $path->[1] eq $deleted; |
49d4f2ac |
356 | next if $vector[0] eq $vector[1]; # Don't add a self loop |
3a2ebbf4 |
357 | my %wits = %{$self->sequence->get_edge_attributes( @$path )}; |
358 | $self->sequence->add_edge( @vector ); |
359 | my $fwits = $self->sequence->get_edge_attributes( @vector ); |
360 | @wits{keys %$fwits} = values %$fwits; |
361 | $self->sequence->set_edge_attributes( @vector, \%wits ); |
362 | } |
cecbe56d |
363 | $self->relations->merge_readings( $kept, $deleted, $combine ); |
3a2ebbf4 |
364 | |
365 | # Do the deletion deed. |
4e483aa5 |
366 | if( $combine ) { |
869a1ada |
367 | # Combine the text of the readings |
4e483aa5 |
368 | my $joinstr = $combine_char; |
369 | unless( defined $joinstr ) { |
370 | $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior; |
371 | $joinstr = $self->wordsep unless defined $joinstr; |
372 | } |
a445ce40 |
373 | $kept_obj->_combine( $del_obj, $joinstr ); |
49d4f2ac |
374 | } |
3a2ebbf4 |
375 | $self->del_reading( $deleted ); |
376 | } |
7854e12e |
377 | |
6771a1b1 |
378 | =head2 compress_readings |
379 | |
380 | Where possible in the graph, compresses plain sequences of readings into a |
381 | single reading. The sequences must consist of readings with no |
382 | relationships to other readings, with only a single witness path between |
383 | them and no other witness paths from either that would skip the other. The |
384 | readings must also not be marked as nonsense or bad grammar. |
385 | |
386 | WARNING: This operation cannot be undone. |
387 | |
388 | =cut |
389 | |
390 | sub compress_readings { |
391 | my $self = shift; |
392 | # Anywhere in the graph that there is a reading that joins only to a single |
393 | # successor, and neither of these have any relationships, just join the two |
394 | # readings. |
6771a1b1 |
395 | foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) { |
428bcf0b |
396 | # Now look for readings that can be joined to their successors. |
a445ce40 |
397 | next unless $rdg->is_combinable; |
6771a1b1 |
398 | my %seen; |
399 | while( $self->sequence->successors( $rdg ) == 1 ) { |
400 | my( $next ) = $self->reading( $self->sequence->successors( $rdg ) ); |
401 | throw( "Infinite loop" ) if $seen{$next->id}; |
402 | $seen{$next->id} = 1; |
403 | last if $self->sequence->predecessors( $next ) > 1; |
a445ce40 |
404 | last unless $next->is_combinable; |
6771a1b1 |
405 | say "Joining readings $rdg and $next"; |
406 | $self->merge_readings( $rdg, $next, 1 ); |
407 | } |
408 | } |
409 | # Make sure we haven't screwed anything up |
410 | foreach my $wit ( $self->tradition->witnesses ) { |
411 | my $pathtext = $self->path_text( $wit->sigil ); |
412 | my $origtext = join( ' ', @{$wit->text} ); |
413 | throw( "Text differs for witness " . $wit->sigil ) |
414 | unless $pathtext eq $origtext; |
415 | if( $wit->is_layered ) { |
416 | $pathtext = $self->path_text( $wit->sigil.$self->ac_label ); |
417 | $origtext = join( ' ', @{$wit->layertext} ); |
418 | throw( "Ante-corr text differs for witness " . $wit->sigil ) |
419 | unless $pathtext eq $origtext; |
420 | } |
421 | } |
422 | |
423 | $self->relations->rebuild_equivalence(); |
424 | $self->calculate_ranks(); |
425 | } |
3265b0ce |
426 | |
3a2ebbf4 |
427 | # Helper function for manipulating the graph. |
428 | sub _stringify_args { |
4e483aa5 |
429 | my( $self, $first, $second, @args ) = @_; |
3a2ebbf4 |
430 | $first = $first->id |
431 | if ref( $first ) eq 'Text::Tradition::Collation::Reading'; |
432 | $second = $second->id |
433 | if ref( $second ) eq 'Text::Tradition::Collation::Reading'; |
4e483aa5 |
434 | return( $first, $second, @args ); |
3a2ebbf4 |
435 | } |
df6d9812 |
436 | |
4e5a7b2c |
437 | # Helper function for manipulating the graph. |
438 | sub _objectify_args { |
439 | my( $self, $first, $second, $arg ) = @_; |
440 | $first = $self->reading( $first ) |
441 | unless ref( $first ) eq 'Text::Tradition::Collation::Reading'; |
442 | $second = $self->reading( $second ) |
443 | unless ref( $second ) eq 'Text::Tradition::Collation::Reading'; |
444 | return( $first, $second, $arg ); |
445 | } |
3a2ebbf4 |
446 | ### Path logic |
447 | |
448 | sub add_path { |
449 | my $self = shift; |
450 | |
451 | # We only need the IDs for adding paths to the graph, not the reading |
452 | # objects themselves. |
cecbe56d |
453 | my( $source, $target, $wit ) = $self->_stringify_args( @_ ); |
3a2ebbf4 |
454 | |
c1915ab9 |
455 | $self->_graphcalc_done(0); |
3a2ebbf4 |
456 | # Connect the readings |
359944f7 |
457 | unless( $self->sequence->has_edge( $source, $target ) ) { |
458 | $self->sequence->add_edge( $source, $target ); |
459 | $self->relations->add_equivalence_edge( $source, $target ); |
460 | } |
3a2ebbf4 |
461 | # Note the witness in question |
462 | $self->sequence->set_edge_attribute( $source, $target, $wit, 1 ); |
359944f7 |
463 | } |
b15511bf |
464 | |
3a2ebbf4 |
465 | sub del_path { |
466 | my $self = shift; |
49d4f2ac |
467 | my @args; |
468 | if( ref( $_[0] ) eq 'ARRAY' ) { |
469 | my $e = shift @_; |
470 | @args = ( @$e, @_ ); |
471 | } else { |
472 | @args = @_; |
473 | } |
3a2ebbf4 |
474 | |
475 | # We only need the IDs for adding paths to the graph, not the reading |
476 | # objects themselves. |
49d4f2ac |
477 | my( $source, $target, $wit ) = $self->_stringify_args( @args ); |
3a2ebbf4 |
478 | |
c1915ab9 |
479 | $self->_graphcalc_done(0); |
3a2ebbf4 |
480 | if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) { |
49d4f2ac |
481 | $self->sequence->delete_edge_attribute( $source, $target, $wit ); |
3a2ebbf4 |
482 | } |
483 | unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) { |
484 | $self->sequence->delete_edge( $source, $target ); |
359944f7 |
485 | $self->relations->delete_equivalence_edge( $source, $target ); |
3a2ebbf4 |
486 | } |
784877d9 |
487 | } |
488 | |
3a2ebbf4 |
489 | |
15d2d3df |
490 | # Extra graph-alike utility |
491 | sub has_path { |
3a2ebbf4 |
492 | my $self = shift; |
493 | my( $source, $target, $wit ) = $self->_stringify_args( @_ ); |
494 | return undef unless $self->sequence->has_edge( $source, $target ); |
495 | return $self->sequence->has_edge_attribute( $source, $target, $wit ); |
b15511bf |
496 | } |
497 | |
4e5a7b2c |
498 | =head2 clear_witness( @sigil_list ) |
3a2ebbf4 |
499 | |
4e5a7b2c |
500 | Clear the given witnesses out of the collation entirely, removing references |
501 | to them in paths, and removing readings that belong only to them. Should only |
502 | be called via $tradition->del_witness. |
3a2ebbf4 |
503 | |
504 | =cut |
505 | |
4e5a7b2c |
506 | sub clear_witness { |
507 | my( $self, @sigils ) = @_; |
508 | |
c1915ab9 |
509 | $self->_graphcalc_done(0); |
4e5a7b2c |
510 | # Clear the witness(es) out of the paths |
511 | foreach my $e ( $self->paths ) { |
512 | foreach my $sig ( @sigils ) { |
513 | $self->del_path( $e, $sig ); |
514 | } |
515 | } |
516 | |
517 | # Clear out the newly unused readings |
518 | foreach my $r ( $self->readings ) { |
519 | unless( $self->reading_witnesses( $r ) ) { |
520 | $self->del_reading( $r ); |
521 | } |
522 | } |
523 | } |
3a2ebbf4 |
524 | |
525 | sub add_relationship { |
526 | my $self = shift; |
22222af9 |
527 | my( $source, $target, $opts ) = $self->_stringify_args( @_ ); |
414cc046 |
528 | my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts ); |
c1915ab9 |
529 | $self->_graphcalc_done(0); |
63778331 |
530 | return @vectors; |
22222af9 |
531 | } |
ef9d481f |
532 | |
ca6e6095 |
533 | around qw/ get_relationship del_relationship / => sub { |
534 | my $orig = shift; |
535 | my $self = shift; |
536 | my @args = @_; |
537 | if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) { |
538 | @args = @{$_[0]}; |
539 | } |
540 | my( $source, $target ) = $self->_stringify_args( @args ); |
541 | $self->$orig( $source, $target ); |
542 | }; |
543 | |
22222af9 |
544 | =head2 reading_witnesses( $reading ) |
910a0a6d |
545 | |
22222af9 |
546 | Return a list of sigils corresponding to the witnesses in which the reading appears. |
3265b0ce |
547 | |
22222af9 |
548 | =cut |
1d310495 |
549 | |
1d310495 |
550 | sub reading_witnesses { |
551 | my( $self, $reading ) = @_; |
552 | # We need only check either the incoming or the outgoing edges; I have |
96dc90ec |
553 | # arbitrarily chosen "incoming". Thus, special-case the start node. |
554 | if( $reading eq $self->start ) { |
bed6ce83 |
555 | return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses; |
96dc90ec |
556 | } |
1d310495 |
557 | my %all_witnesses; |
558 | foreach my $e ( $self->sequence->edges_to( $reading ) ) { |
559 | my $wits = $self->sequence->get_edge_attributes( @$e ); |
560 | @all_witnesses{ keys %$wits } = 1; |
561 | } |
c12bb878 |
562 | my $acstr = $self->ac_label; |
563 | foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) { |
564 | delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit}; |
565 | } |
1d310495 |
566 | return keys %all_witnesses; |
910a0a6d |
567 | } |
568 | |
4e5a7b2c |
569 | =head1 OUTPUT METHODS |
8e1394aa |
570 | |
0ecb975c |
571 | =head2 as_svg( \%options ) |
8e1394aa |
572 | |
0068967c |
573 | Returns an SVG string that represents the graph, via as_dot and graphviz. |
bfcbcecb |
574 | See as_dot for a list of options. Must have GraphViz (dot) installed to run. |
8e1394aa |
575 | |
576 | =cut |
577 | |
578 | sub as_svg { |
0ecb975c |
579 | my( $self, $opts ) = @_; |
bfcbcecb |
580 | throw( "Need GraphViz installed to output SVG" ) |
581 | unless File::Which::which( 'dot' ); |
e247aad1 |
582 | my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'}; |
1ff82d4f |
583 | $self->calculate_ranks() |
584 | unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear ); |
be3af600 |
585 | my @cmd = qw/dot -Tsvg/; |
586 | my( $svg, $err ); |
587 | my $dotfile = File::Temp->new(); |
588 | ## USE FOR DEBUGGING |
589 | # $dotfile->unlink_on_destroy(0); |
590 | binmode $dotfile, ':utf8'; |
591 | print $dotfile $self->as_dot( $opts ); |
592 | push( @cmd, $dotfile->filename ); |
593 | run( \@cmd, ">", binary(), \$svg ); |
594 | $svg = decode_utf8( $svg ); |
595 | return $svg; |
8e1394aa |
596 | } |
597 | |
b22576c6 |
598 | |
0ecb975c |
599 | =head2 as_dot( \%options ) |
b22576c6 |
600 | |
0ecb975c |
601 | Returns a string that is the collation graph expressed in dot |
602 | (i.e. GraphViz) format. Options include: |
b22576c6 |
603 | |
0ecb975c |
604 | =over 4 |
b22576c6 |
605 | |
0ecb975c |
606 | =item * from |
b22576c6 |
607 | |
0ecb975c |
608 | =item * to |
df6d9812 |
609 | |
0ecb975c |
610 | =item * color_common |
611 | |
612 | =back |
df6d9812 |
613 | |
614 | =cut |
615 | |
616 | sub as_dot { |
0ecb975c |
617 | my( $self, $opts ) = @_; |
618 | my $startrank = $opts->{'from'} if $opts; |
619 | my $endrank = $opts->{'to'} if $opts; |
620 | my $color_common = $opts->{'color_common'} if $opts; |
b365fbae |
621 | my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank |
622 | && $self->end->rank > 100; |
6648ee3d |
623 | $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs |
b365fbae |
624 | |
b22576c6 |
625 | # Check the arguments |
626 | if( $startrank ) { |
627 | return if $endrank && $startrank > $endrank; |
628 | return if $startrank > $self->end->rank; |
629 | } |
630 | if( defined $endrank ) { |
631 | return if $endrank < 0; |
f1b3b33a |
632 | $endrank = undef if $endrank == $self->end->rank; |
b22576c6 |
633 | } |
634 | |
67da8d6c |
635 | my $graph_name = $self->tradition->name; |
636 | $graph_name =~ s/[^\w\s]//g; |
637 | $graph_name = join( '_', split( /\s+/, $graph_name ) ); |
f13b5582 |
638 | |
639 | my %graph_attrs = ( |
640 | 'rankdir' => 'LR', |
641 | 'bgcolor' => 'none', |
642 | ); |
643 | my %node_attrs = ( |
b8990398 |
644 | 'fontsize' => 14, |
f13b5582 |
645 | 'fillcolor' => 'white', |
646 | 'style' => 'filled', |
647 | 'shape' => 'ellipse' |
648 | ); |
649 | my %edge_attrs = ( |
650 | 'arrowhead' => 'open', |
651 | 'color' => '#000000', |
652 | 'fontcolor' => '#000000', |
653 | ); |
654 | |
67da8d6c |
655 | my $dot = sprintf( "digraph %s {\n", $graph_name ); |
f13b5582 |
656 | $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n"; |
657 | $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n"; |
df6d9812 |
658 | |
b22576c6 |
659 | # Output substitute start/end readings if necessary |
660 | if( $startrank ) { |
43c94341 |
661 | $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n"; |
b22576c6 |
662 | } |
663 | if( $endrank ) { |
43c94341 |
664 | $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n"; |
b22576c6 |
665 | } |
b365fbae |
666 | if( $STRAIGHTENHACK ) { |
667 | ## HACK part 1 |
43c94341 |
668 | my $startlabel = $startrank ? '__SUBSTART__' : '__START__'; |
669 | $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n"; |
b365fbae |
670 | $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];" |
671 | } |
b22576c6 |
672 | my %used; # Keep track of the readings that actually appear in the graph |
30ddc24c |
673 | # Sort the readings by rank if we have ranks; this speeds layout. |
674 | my @all_readings = $self->end->has_rank |
675 | ? sort { $a->rank <=> $b->rank } $self->readings |
676 | : $self->readings; |
4633f9e4 |
677 | # TODO Refrain from outputting lacuna nodes - just grey out the edges. |
30ddc24c |
678 | foreach my $reading ( @all_readings ) { |
b22576c6 |
679 | # Only output readings within our rank range. |
680 | next if $startrank && $reading->rank < $startrank; |
681 | next if $endrank && $reading->rank > $endrank; |
682 | $used{$reading->id} = 1; |
910a0a6d |
683 | # Need not output nodes without separate labels |
3a2ebbf4 |
684 | next if $reading->id eq $reading->text; |
d4b75f44 |
685 | my $rattrs; |
30f0df34 |
686 | my $label = $reading->text; |
629e27b0 |
687 | $label .= '-' if $reading->join_next; |
688 | $label = "-$label" if $reading->join_prior; |
8f9cab7b |
689 | $label =~ s/\"/\\\"/g; |
d4b75f44 |
690 | $rattrs->{'label'} = $label; |
10e4b1ac |
691 | $rattrs->{'id'} = $reading->id; |
0ecb975c |
692 | $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common; |
d4b75f44 |
693 | $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) ); |
df6d9812 |
694 | } |
3a2ebbf4 |
695 | |
30ddc24c |
696 | # Add the real edges. Need to weight one edge per rank jump, in a |
697 | # continuous line. |
b365fbae |
698 | # my $weighted = $self->_add_edge_weights; |
b22576c6 |
699 | my @edges = $self->paths; |
3bdec618 |
700 | my( %substart, %subend ); |
b22576c6 |
701 | foreach my $edge ( @edges ) { |
702 | # Do we need to output this edge? |
508fd430 |
703 | if( $used{$edge->[0]} && $used{$edge->[1]} ) { |
027d819c |
704 | my $label = $self->_path_display_label( $self->path_witnesses( $edge ) ); |
f13b5582 |
705 | my $variables = { %edge_attrs, 'label' => $label }; |
30ddc24c |
706 | |
b22576c6 |
707 | # Account for the rank gap if necessary |
30ddc24c |
708 | my $rank0 = $self->reading( $edge->[0] )->rank |
709 | if $self->reading( $edge->[0] )->has_rank; |
710 | my $rank1 = $self->reading( $edge->[1] )->rank |
711 | if $self->reading( $edge->[1] )->has_rank; |
712 | if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) { |
713 | $variables->{'minlen'} = $rank1 - $rank0; |
714 | } |
715 | |
716 | # Add the calculated edge weights |
b365fbae |
717 | # if( exists $weighted->{$edge->[0]} |
e247aad1 |
718 | # && $weighted->{$edge->[0]} eq $edge->[1] ) { |
719 | # # $variables->{'color'} = 'red'; |
720 | # $variables->{'weight'} = 3.0; |
721 | # } |
30ddc24c |
722 | |
508fd430 |
723 | # EXPERIMENTAL: make edge width reflect no. of witnesses |
724 | my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2; |
725 | $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit |
726 | |
f13b5582 |
727 | my $varopts = _dot_attr_string( $variables ); |
728 | $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n", |
729 | $edge->[0], $edge->[1], $varopts ); |
3bdec618 |
730 | } elsif( $used{$edge->[0]} ) { |
96ba0418 |
731 | $subend{$edge->[0]} = $edge->[1]; |
3bdec618 |
732 | } elsif( $used{$edge->[1]} ) { |
96ba0418 |
733 | $substart{$edge->[1]} = $edge->[0]; |
b22576c6 |
734 | } |
df6d9812 |
735 | } |
bed6ce83 |
736 | |
737 | # If we are asked to, add relationship links |
738 | if( exists $opts->{show_relations} ) { |
739 | my $filter = $opts->{show_relations}; # can be 'transposition' or 'all' |
740 | if( $filter eq 'transposition' ) { |
741 | $filter =~ qr/^transposition$/; |
742 | } |
743 | foreach my $redge ( $self->relationships ) { |
744 | if( $used{$redge->[0]} && $used{$redge->[1]} ) { |
745 | if( $filter ne 'all' ) { |
746 | my $rel = $self->get_relationship( $redge ); |
747 | next unless $rel->type =~ /$filter/; |
748 | my $variables = { |
749 | arrowhead => 'none', |
750 | color => '#FFA14F', |
751 | constraint => 'false', |
752 | label => uc( substr( $rel->type, 0, 4 ) ), |
753 | penwidth => '3', |
754 | }; |
755 | $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n", |
756 | $redge->[0], $redge->[1], _dot_attr_string( $variables ) ); |
757 | } |
758 | } |
759 | } |
760 | } |
761 | |
3bdec618 |
762 | # Add substitute start and end edges if necessary |
763 | foreach my $node ( keys %substart ) { |
96ba0418 |
764 | my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) ); |
f13b5582 |
765 | my $variables = { %edge_attrs, 'label' => $witstr }; |
96ba0418 |
766 | my $nrdg = $self->reading( $node ); |
767 | if( $nrdg->has_rank && $nrdg->rank > $startrank ) { |
768 | # Substart is actually one lower than $startrank |
769 | $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 ); |
770 | } |
f13b5582 |
771 | my $varopts = _dot_attr_string( $variables ); |
96ba0418 |
772 | $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n"; |
3bdec618 |
773 | } |
774 | foreach my $node ( keys %subend ) { |
96ba0418 |
775 | my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) ); |
f13b5582 |
776 | my $variables = { %edge_attrs, 'label' => $witstr }; |
777 | my $varopts = _dot_attr_string( $variables ); |
96ba0418 |
778 | $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n"; |
3bdec618 |
779 | } |
b365fbae |
780 | # HACK part 2 |
781 | if( $STRAIGHTENHACK ) { |
43c94341 |
782 | my $endlabel = $endrank ? '__SUBEND__' : '__END__'; |
783 | $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; |
b365fbae |
784 | } |
30ddc24c |
785 | |
df6d9812 |
786 | $dot .= "}\n"; |
787 | return $dot; |
788 | } |
789 | |
f13b5582 |
790 | sub _dot_attr_string { |
791 | my( $hash ) = @_; |
792 | my @attrs; |
793 | foreach my $k ( sort keys %$hash ) { |
794 | my $v = $hash->{$k}; |
795 | push( @attrs, $k.'="'.$v.'"' ); |
796 | } |
797 | return( '[ ' . join( ', ', @attrs ) . ' ]' ); |
798 | } |
799 | |
30ddc24c |
800 | sub _add_edge_weights { |
801 | my $self = shift; |
802 | # Walk the graph from START to END, choosing the successor node with |
803 | # the largest number of witness paths each time. |
804 | my $weighted = {}; |
805 | my $curr = $self->start->id; |
008fc8a6 |
806 | my $ranked = $self->end->has_rank; |
30ddc24c |
807 | while( $curr ne $self->end->id ) { |
008fc8a6 |
808 | my $rank = $ranked ? $self->reading( $curr )->rank : 0; |
30ddc24c |
809 | my @succ = sort { $self->path_witnesses( $curr, $a ) |
810 | <=> $self->path_witnesses( $curr, $b ) } |
811 | $self->sequence->successors( $curr ); |
812 | my $next = pop @succ; |
008fc8a6 |
813 | my $nextrank = $ranked ? $self->reading( $next )->rank : 0; |
30ddc24c |
814 | # Try to avoid lacunae in the weighted path. |
008fc8a6 |
815 | while( @succ && |
816 | ( $self->reading( $next )->is_lacuna || |
817 | $nextrank - $rank > 1 ) ){ |
30ddc24c |
818 | $next = pop @succ; |
819 | } |
820 | $weighted->{$curr} = $next; |
821 | $curr = $next; |
822 | } |
823 | return $weighted; |
824 | } |
825 | |
027d819c |
826 | =head2 path_witnesses( $edge ) |
827 | |
828 | Returns the list of sigils whose witnesses are associated with the given edge. |
829 | The edge can be passed as either an array or an arrayref of ( $source, $target ). |
830 | |
831 | =cut |
832 | |
3a2ebbf4 |
833 | sub path_witnesses { |
834 | my( $self, @edge ) = @_; |
835 | # If edge is an arrayref, cope. |
836 | if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) { |
837 | my $e = shift @edge; |
838 | @edge = @$e; |
839 | } |
840 | my @wits = keys %{$self->sequence->get_edge_attributes( @edge )}; |
508fd430 |
841 | return @wits; |
3a2ebbf4 |
842 | } |
843 | |
7f9f05e8 |
844 | # Helper function. Make a display label for the given witnesses, showing a.c. |
845 | # witnesses only where the main witness is not also in the list. |
027d819c |
846 | sub _path_display_label { |
508fd430 |
847 | my $self = shift; |
7f9f05e8 |
848 | my %wits; |
849 | map { $wits{$_} = 1 } @_; |
850 | |
851 | # If an a.c. wit is listed, remove it if the main wit is also listed. |
852 | # Otherwise keep it for explicit listing. |
853 | my $aclabel = $self->ac_label; |
854 | my @disp_ac; |
855 | foreach my $w ( sort keys %wits ) { |
856 | if( $w =~ /^(.*)\Q$aclabel\E$/ ) { |
857 | if( exists $wits{$1} ) { |
858 | delete $wits{$w}; |
859 | } else { |
860 | push( @disp_ac, $w ); |
861 | } |
862 | } |
863 | } |
864 | |
865 | # See if we are in a majority situation. |
8f9cab7b |
866 | my $maj = scalar( $self->tradition->witnesses ) * 0.6; |
1ff82d4f |
867 | $maj = $maj > 5 ? $maj : 5; |
7f9f05e8 |
868 | if( scalar keys %wits > $maj ) { |
869 | unshift( @disp_ac, 'majority' ); |
870 | return join( ', ', @disp_ac ); |
8f9cab7b |
871 | } else { |
7f9f05e8 |
872 | return join( ', ', sort keys %wits ); |
8f9cab7b |
873 | } |
874 | } |
1dd07bda |
875 | |
bf6e338d |
876 | =head2 readings_at_rank( $rank ) |
1dd07bda |
877 | |
bf6e338d |
878 | Returns a list of readings at a given rank, taken from the alignment table. |
1dd07bda |
879 | |
880 | =cut |
881 | |
bf6e338d |
882 | sub readings_at_rank { |
1dd07bda |
883 | my( $self, $rank ) = @_; |
bf6e338d |
884 | my $table = $self->alignment_table; |
885 | # Table rank is real rank - 1. |
886 | my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}}; |
887 | my %readings; |
888 | foreach my $e ( @elements ) { |
889 | next unless ref( $e ) eq 'HASH'; |
890 | next unless exists $e->{'t'}; |
891 | $readings{$e->{'t'}->id} = $e->{'t'}; |
892 | } |
893 | return values %readings; |
1dd07bda |
894 | } |
8f9cab7b |
895 | |
4e5a7b2c |
896 | =head2 as_graphml |
8e1394aa |
897 | |
4e5a7b2c |
898 | Returns a GraphML representation of the collation. The GraphML will contain |
899 | two graphs. The first expresses the attributes of the readings and the witness |
900 | paths that link them; the second expresses the relationships that link the |
901 | readings. This is the native transfer format for a tradition. |
8e1394aa |
902 | |
56eefa04 |
903 | =begin testing |
904 | |
905 | use Text::Tradition; |
951ddfe8 |
906 | use TryCatch; |
56eefa04 |
907 | |
908 | my $READINGS = 311; |
909 | my $PATHS = 361; |
910 | |
911 | my $datafile = 't/data/florilegium_tei_ps.xml'; |
912 | my $tradition = Text::Tradition->new( 'input' => 'TEI', |
913 | 'name' => 'test0', |
914 | 'file' => $datafile, |
915 | 'linear' => 1 ); |
916 | |
917 | ok( $tradition, "Got a tradition object" ); |
918 | is( scalar $tradition->witnesses, 13, "Found all witnesses" ); |
919 | ok( $tradition->collation, "Tradition has a collation" ); |
920 | |
921 | my $c = $tradition->collation; |
922 | is( scalar $c->readings, $READINGS, "Collation has all readings" ); |
923 | is( scalar $c->paths, $PATHS, "Collation has all paths" ); |
924 | is( scalar $c->relationships, 0, "Collation has all relationships" ); |
925 | |
926 | # Add a few relationships |
927 | $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } ); |
928 | $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } ); |
929 | $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } ); |
930 | |
931 | # Now write it to GraphML and parse it again. |
932 | |
933 | my $graphml = $c->as_graphml; |
934 | my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml ); |
935 | is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" ); |
936 | is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" ); |
937 | is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" ); |
938 | |
9fef629b |
939 | # Now add a stemma, write to GraphML, and look at the output. |
951ddfe8 |
940 | SKIP: { |
37bf09f4 |
941 | skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' ); |
951ddfe8 |
942 | my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' ); |
943 | is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" ); |
944 | is( $tradition->stemmata, 1, "Tradition now has the stemma" ); |
945 | $graphml = $c->as_graphml; |
946 | like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" ); |
947 | } |
2a812726 |
948 | |
56eefa04 |
949 | =end testing |
950 | |
8e1394aa |
951 | =cut |
952 | |
a445ce40 |
953 | ## TODO MOVE this to Tradition.pm and modularize it better |
8e1394aa |
954 | sub as_graphml { |
a30ca502 |
955 | my( $self, $options ) = @_; |
3d14b48e |
956 | $self->calculate_ranks unless $self->_graphcalc_done; |
957 | |
a30ca502 |
958 | my $start = $options->{'from'} |
959 | ? $self->reading( $options->{'from'} ) : $self->start; |
960 | my $end = $options->{'to'} |
961 | ? $self->reading( $options->{'to'} ) : $self->end; |
962 | if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) { |
963 | throw( 'Start node must be before end node' ); |
964 | } |
965 | # The readings need to be ranked for this to work. |
966 | $start = $self->start unless $start->has_rank; |
967 | $end = $self->end unless $end->has_rank; |
414cc046 |
968 | my $rankoffset = 0; |
969 | unless( $start eq $self->start ) { |
970 | $rankoffset = $start->rank - 1; |
971 | } |
a30ca502 |
972 | my %use_readings; |
973 | |
8e1394aa |
974 | # Some namespaces |
975 | my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns'; |
976 | my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance'; |
977 | my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' . |
910a0a6d |
978 | 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; |
8e1394aa |
979 | |
980 | # Create the document and root node |
428bcf0b |
981 | require XML::LibXML; |
8e1394aa |
982 | my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" ); |
983 | my $root = $graphml->createElementNS( $graphml_ns, 'graphml' ); |
984 | $graphml->setDocumentElement( $root ); |
985 | $root->setNamespace( $xsi_ns, 'xsi', 0 ); |
986 | $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema ); |
bbd064a9 |
987 | |
988 | # List of attribute types to save on our objects and their corresponding |
989 | # GraphML types |
990 | my %save_types = ( |
991 | 'Str' => 'string', |
992 | 'Int' => 'int', |
993 | 'Bool' => 'boolean', |
10e4b1ac |
994 | 'ReadingID' => 'string', |
bbd064a9 |
995 | 'RelationshipType' => 'string', |
996 | 'RelationshipScope' => 'string', |
997 | ); |
998 | |
bbd064a9 |
999 | # Add the data keys for the graph. Include an extra key 'version' for the |
1000 | # GraphML output version. |
e309421a |
1001 | my %graph_data_keys; |
1002 | my $gdi = 0; |
bbd064a9 |
1003 | my %graph_attributes = ( 'version' => 'string' ); |
1004 | # Graph attributes include those of Tradition and those of Collation. |
1005 | my %gattr_from; |
1006 | my $tmeta = $self->tradition->meta; |
1007 | my $cmeta = $self->meta; |
1008 | map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes; |
1009 | map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes; |
1010 | foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) { |
1011 | next if $attr->name =~ /^_/; |
bbd064a9 |
1012 | next unless $save_types{$attr->type_constraint->name}; |
1013 | $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; |
1014 | } |
9fef629b |
1015 | # Extra custom keys for complex objects that should be saved in some form. |
1016 | # The subroutine should return a string, or undef/empty. |
951ddfe8 |
1017 | if( $tmeta->has_method('stemmata') ) { |
1018 | $graph_attributes{'stemmata'} = sub { |
1019 | my @stemstrs; |
1020 | map { push( @stemstrs, $_->editable( {linesep => ''} ) ) } |
1021 | $self->tradition->stemmata; |
1022 | join( "\n", @stemstrs ); |
1023 | }; |
1024 | } |
1025 | |
8943ff68 |
1026 | if( $tmeta->has_method('user') ) { |
1027 | $graph_attributes{'user'} = sub { |
1028 | $self->tradition->user ? $self->tradition->user->id : undef |
1029 | }; |
1030 | } |
bbd064a9 |
1031 | |
1032 | foreach my $datum ( sort keys %graph_attributes ) { |
e309421a |
1033 | $graph_data_keys{$datum} = 'dg'.$gdi++; |
1034 | my $key = $root->addNewChild( $graphml_ns, 'key' ); |
9fef629b |
1035 | my $dtype = ref( $graph_attributes{$datum} ) ? 'string' |
1036 | : $graph_attributes{$datum}; |
e309421a |
1037 | $key->setAttribute( 'attr.name', $datum ); |
9fef629b |
1038 | $key->setAttribute( 'attr.type', $dtype ); |
e309421a |
1039 | $key->setAttribute( 'for', 'graph' ); |
1040 | $key->setAttribute( 'id', $graph_data_keys{$datum} ); |
1041 | } |
f6066bac |
1042 | |
bbd064a9 |
1043 | # Add the data keys for reading nodes |
1044 | my %reading_attributes; |
1045 | my $rmeta = Text::Tradition::Collation::Reading->meta; |
1046 | foreach my $attr( $rmeta->get_all_attributes ) { |
1047 | next if $attr->name =~ /^_/; |
bbd064a9 |
1048 | next unless $save_types{$attr->type_constraint->name}; |
1049 | $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; |
1050 | } |
a445ce40 |
1051 | if( $self->start->does('Text::Tradition::Morphology' ) ) { |
1052 | # Extra custom key for the reading morphology |
1053 | $reading_attributes{'lexemes'} = 'string'; |
1054 | } |
7cd9f181 |
1055 | |
ef9d481f |
1056 | my %node_data_keys; |
1057 | my $ndi = 0; |
bbd064a9 |
1058 | foreach my $datum ( sort keys %reading_attributes ) { |
910a0a6d |
1059 | $node_data_keys{$datum} = 'dn'.$ndi++; |
1060 | my $key = $root->addNewChild( $graphml_ns, 'key' ); |
1061 | $key->setAttribute( 'attr.name', $datum ); |
bbd064a9 |
1062 | $key->setAttribute( 'attr.type', $reading_attributes{$datum} ); |
910a0a6d |
1063 | $key->setAttribute( 'for', 'node' ); |
1064 | $key->setAttribute( 'id', $node_data_keys{$datum} ); |
8e1394aa |
1065 | } |
1066 | |
bbd064a9 |
1067 | # Add the data keys for edges, that is, paths and relationships. Path |
1068 | # data does not come from a Moose class so is here manually. |
ef9d481f |
1069 | my $edi = 0; |
1070 | my %edge_data_keys; |
bbd064a9 |
1071 | my %edge_attributes = ( |
3a2ebbf4 |
1072 | witness => 'string', # ID/label for a path |
3a2ebbf4 |
1073 | extra => 'boolean', # Path key |
3a2ebbf4 |
1074 | ); |
bbd064a9 |
1075 | my @path_attributes = keys %edge_attributes; # track our manual additions |
1076 | my $pmeta = Text::Tradition::Collation::Relationship->meta; |
1077 | foreach my $attr( $pmeta->get_all_attributes ) { |
1078 | next if $attr->name =~ /^_/; |
bbd064a9 |
1079 | next unless $save_types{$attr->type_constraint->name}; |
1080 | $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; |
1081 | } |
1082 | foreach my $datum ( sort keys %edge_attributes ) { |
3a2ebbf4 |
1083 | $edge_data_keys{$datum} = 'de'.$edi++; |
910a0a6d |
1084 | my $key = $root->addNewChild( $graphml_ns, 'key' ); |
3a2ebbf4 |
1085 | $key->setAttribute( 'attr.name', $datum ); |
bbd064a9 |
1086 | $key->setAttribute( 'attr.type', $edge_attributes{$datum} ); |
910a0a6d |
1087 | $key->setAttribute( 'for', 'edge' ); |
3a2ebbf4 |
1088 | $key->setAttribute( 'id', $edge_data_keys{$datum} ); |
8e1394aa |
1089 | } |
3a2ebbf4 |
1090 | |
cc31ebaa |
1091 | # Add the collation graph itself. First, sanitize the name to a valid XML ID. |
1092 | my $xmlidname = $self->tradition->name; |
1093 | $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g; |
1094 | if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) { |
1095 | $xmlidname = '_'.$xmlidname; |
1096 | } |
2c669bca |
1097 | my $sgraph = $root->addNewChild( $graphml_ns, 'graph' ); |
1098 | $sgraph->setAttribute( 'edgedefault', 'directed' ); |
cc31ebaa |
1099 | $sgraph->setAttribute( 'id', $xmlidname ); |
2c669bca |
1100 | $sgraph->setAttribute( 'parse.edgeids', 'canonical' ); |
cc31ebaa |
1101 | $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later |
2c669bca |
1102 | $sgraph->setAttribute( 'parse.nodeids', 'canonical' ); |
cc31ebaa |
1103 | $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later |
2c669bca |
1104 | $sgraph->setAttribute( 'parse.order', 'nodesfirst' ); |
22222af9 |
1105 | |
2a812726 |
1106 | # Tradition/collation attribute data |
bbd064a9 |
1107 | foreach my $datum ( keys %graph_attributes ) { |
1108 | my $value; |
1109 | if( $datum eq 'version' ) { |
2a812726 |
1110 | $value = '3.2'; |
9fef629b |
1111 | } elsif( ref( $graph_attributes{$datum} ) ) { |
1112 | my $sub = $graph_attributes{$datum}; |
1113 | $value = &$sub(); |
bbd064a9 |
1114 | } elsif( $gattr_from{$datum} eq 'Tradition' ) { |
1115 | $value = $self->tradition->$datum; |
1116 | } else { |
1117 | $value = $self->$datum; |
1118 | } |
2c669bca |
1119 | _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value ); |
e309421a |
1120 | } |
8e1394aa |
1121 | |
1122 | my $node_ctr = 0; |
1123 | my %node_hash; |
22222af9 |
1124 | # Add our readings to the graph |
3a2ebbf4 |
1125 | foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) { |
a30ca502 |
1126 | next if $n->has_rank && $n ne $self->start && $n ne $self->end && |
1127 | ( $n->rank < $start->rank || $n->rank > $end->rank ); |
1128 | $use_readings{$n->id} = 1; |
2c669bca |
1129 | # Add to the main graph |
1130 | my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' ); |
910a0a6d |
1131 | my $node_xmlid = 'n' . $node_ctr++; |
3a2ebbf4 |
1132 | $node_hash{ $n->id } = $node_xmlid; |
910a0a6d |
1133 | $node_el->setAttribute( 'id', $node_xmlid ); |
bbd064a9 |
1134 | foreach my $d ( keys %reading_attributes ) { |
255875b8 |
1135 | my $nval = $n->$d; |
7cd9f181 |
1136 | # Custom serialization |
1137 | if( $d eq 'lexemes' ) { |
1138 | # If nval is a true value, we have lexemes so we need to |
1139 | # serialize them. Otherwise set nval to undef so that the |
1140 | # key is excluded from this reading. |
1141 | $nval = $nval ? $n->_serialize_lexemes : undef; |
18c64d55 |
1142 | } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) { |
1143 | $nval = undef; |
7cd9f181 |
1144 | } |
cc31ebaa |
1145 | if( $rankoffset && $d eq 'rank' && $n ne $self->start ) { |
414cc046 |
1146 | # Adjust the ranks within the subgraph. |
cc31ebaa |
1147 | $nval = $n eq $self->end ? $end->rank - $rankoffset + 1 |
1148 | : $nval - $rankoffset; |
414cc046 |
1149 | } |
255875b8 |
1150 | _add_graphml_data( $node_el, $node_data_keys{$d}, $nval ) |
1151 | if defined $nval; |
1152 | } |
b15511bf |
1153 | } |
1154 | |
2c669bca |
1155 | # Add the path edges to the sequence graph |
df6d9812 |
1156 | my $edge_ctr = 0; |
3a2ebbf4 |
1157 | foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) { |
1158 | # We add an edge in the graphml for every witness in $e. |
a30ca502 |
1159 | next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} ); |
1160 | my @edge_wits = sort $self->path_witnesses( $e ); |
cc31ebaa |
1161 | $e->[0] = $self->start->id unless $use_readings{$e->[0]}; |
1162 | $e->[1] = $self->end->id unless $use_readings{$e->[1]}; |
1163 | # Skip any path from start to end; that witness is not in the subgraph. |
1164 | next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id ); |
a30ca502 |
1165 | foreach my $wit ( @edge_wits ) { |
3a2ebbf4 |
1166 | my( $id, $from, $to ) = ( 'e'.$edge_ctr++, |
1167 | $node_hash{ $e->[0] }, |
1168 | $node_hash{ $e->[1] } ); |
2c669bca |
1169 | my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' ); |
3a2ebbf4 |
1170 | $edge_el->setAttribute( 'source', $from ); |
1171 | $edge_el->setAttribute( 'target', $to ); |
1172 | $edge_el->setAttribute( 'id', $id ); |
3a2ebbf4 |
1173 | |
1174 | # It's a witness path, so add the witness |
1175 | my $base = $wit; |
1176 | my $key = $edge_data_keys{'witness'}; |
1177 | # Is this an ante-corr witness? |
1178 | my $aclabel = $self->ac_label; |
1179 | if( $wit =~ /^(.*)\Q$aclabel\E$/ ) { |
1180 | # Keep the base witness |
1181 | $base = $1; |
1182 | # ...and record that this is an 'extra' reading path |
1183 | _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel ); |
1184 | } |
1185 | _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base ); |
1186 | } |
1187 | } |
1188 | |
cc31ebaa |
1189 | # Report the actual number of nodes and edges that went in |
1190 | $sgraph->setAttribute( 'parse.edges', $edge_ctr ); |
1191 | $sgraph->setAttribute( 'parse.nodes', $node_ctr ); |
1192 | |
22222af9 |
1193 | # Add the relationship graph to the XML |
bbd064a9 |
1194 | map { delete $edge_data_keys{$_} } @path_attributes; |
826d8773 |
1195 | $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash, |
1196 | $node_data_keys{'id'}, \%edge_data_keys ); |
8e1394aa |
1197 | |
94c00c71 |
1198 | # Save and return the thing |
1199 | my $result = decode_utf8( $graphml->toString(1) ); |
94c00c71 |
1200 | return $result; |
df6d9812 |
1201 | } |
1202 | |
b15511bf |
1203 | sub _add_graphml_data { |
1204 | my( $el, $key, $value ) = @_; |
b15511bf |
1205 | return unless defined $value; |
c9bf3dbf |
1206 | my $data_el = $el->addNewChild( $el->namespaceURI, 'data' ); |
b15511bf |
1207 | $data_el->setAttribute( 'key', $key ); |
1208 | $data_el->appendText( $value ); |
8e1394aa |
1209 | } |
1210 | |
4e5a7b2c |
1211 | =head2 as_csv |
910a0a6d |
1212 | |
1213 | Returns a CSV alignment table representation of the collation graph, one |
2c669bca |
1214 | row per witness (or witness uncorrected.) |
910a0a6d |
1215 | |
1216 | =cut |
1217 | |
1218 | sub as_csv { |
3a2ebbf4 |
1219 | my( $self ) = @_; |
1dd07bda |
1220 | my $table = $self->alignment_table; |
82fa4d57 |
1221 | my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } ); |
910a0a6d |
1222 | my @result; |
2c669bca |
1223 | # Make the header row |
1224 | $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} ); |
1225 | push( @result, decode_utf8( $csv->string ) ); |
1226 | # Make the rest of the rows |
1227 | foreach my $idx ( 0 .. $table->{'length'} - 1 ) { |
566f4595 |
1228 | my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}}; |
1dd07bda |
1229 | my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs; |
2c669bca |
1230 | $csv->combine( @row ); |
910a0a6d |
1231 | push( @result, decode_utf8( $csv->string ) ); |
1232 | } |
3a2ebbf4 |
1233 | return join( "\n", @result ); |
910a0a6d |
1234 | } |
1235 | |
1dd07bda |
1236 | =head2 alignment_table( $use_refs, $include_witnesses ) |
2c669bca |
1237 | |
566f4595 |
1238 | Return a reference to an alignment table, in a slightly enhanced CollateX |
1239 | format which looks like this: |
1240 | |
1241 | $table = { alignment => [ { witness => "SIGIL", |
4e5a7b2c |
1242 | tokens => [ { t => "TEXT" }, ... ] }, |
566f4595 |
1243 | { witness => "SIG2", |
4e5a7b2c |
1244 | tokens => [ { t => "TEXT" }, ... ] }, |
566f4595 |
1245 | ... ], |
1246 | length => TEXTLEN }; |
1247 | |
1248 | If $use_refs is set to 1, the reading object is returned in the table |
1249 | instead of READINGTEXT; if not, the text of the reading is returned. |
4e5a7b2c |
1250 | |
1251 | If $include_witnesses is set to a hashref, only the witnesses whose sigil |
566f4595 |
1252 | keys have a true hash value will be included. |
2c669bca |
1253 | |
1254 | =cut |
9f3ba6f7 |
1255 | |
1dd07bda |
1256 | sub alignment_table { |
1257 | my( $self ) = @_; |
c1915ab9 |
1258 | $self->calculate_ranks() unless $self->_graphcalc_done; |
1dd07bda |
1259 | return $self->cached_table if $self->has_cached_table; |
1260 | |
0ecb975c |
1261 | # Make sure we can do this |
1262 | throw( "Need a linear graph in order to make an alignment table" ) |
1263 | unless $self->linear; |
1264 | $self->calculate_ranks unless $self->end->has_rank; |
1265 | |
2c669bca |
1266 | my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 }; |
3a2ebbf4 |
1267 | my @all_pos = ( 1 .. $self->end->rank - 1 ); |
68454b71 |
1268 | foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) { |
6771a1b1 |
1269 | # say STDERR "Making witness row(s) for " . $wit->sigil; |
1f7aa795 |
1270 | my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); |
1dd07bda |
1271 | my @row = _make_witness_row( \@wit_path, \@all_pos ); |
bed6ce83 |
1272 | my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row }; |
1273 | $witobj->{'identifier'} = $wit->identifier if $wit->identifier; |
1274 | push( @{$table->{'alignment'}}, $witobj ); |
1f7aa795 |
1275 | if( $wit->is_layered ) { |
1276 | my @wit_ac_path = $self->reading_sequence( $self->start, $self->end, |
861c3e27 |
1277 | $wit->sigil.$self->ac_label ); |
1dd07bda |
1278 | my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos ); |
bed6ce83 |
1279 | my $witacobj = { 'witness' => $wit->sigil.$self->ac_label, |
1280 | 'tokens' => \@ac_row }; |
1281 | $witacobj->{'identifier'} = $wit->identifier if $wit->identifier; |
1282 | push( @{$table->{'alignment'}}, $witacobj ); |
910a0a6d |
1283 | } |
1284 | } |
1dd07bda |
1285 | $self->cached_table( $table ); |
1286 | return $table; |
910a0a6d |
1287 | } |
1288 | |
1289 | sub _make_witness_row { |
1dd07bda |
1290 | my( $path, $positions ) = @_; |
910a0a6d |
1291 | my %char_hash; |
1292 | map { $char_hash{$_} = undef } @$positions; |
2c669bca |
1293 | my $debug = 0; |
910a0a6d |
1294 | foreach my $rdg ( @$path ) { |
6771a1b1 |
1295 | say STDERR "rank " . $rdg->rank if $debug; |
1296 | # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank; |
1dd07bda |
1297 | $char_hash{$rdg->rank} = { 't' => $rdg }; |
910a0a6d |
1298 | } |
1299 | my @row = map { $char_hash{$_} } @$positions; |
eca16057 |
1300 | # Fill in lacuna markers for undef spots in the row |
1301 | my $last_el = shift @row; |
1302 | my @filled_row = ( $last_el ); |
1303 | foreach my $el ( @row ) { |
0e476982 |
1304 | # If we are using node reference, make the lacuna node appear many times |
1305 | # in the table. If not, use the lacuna tag. |
1dd07bda |
1306 | if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) { |
1307 | $el = $last_el; |
eca16057 |
1308 | } |
1309 | push( @filled_row, $el ); |
1310 | $last_el = $el; |
1311 | } |
1312 | return @filled_row; |
910a0a6d |
1313 | } |
1314 | |
4e5a7b2c |
1315 | =head1 NAVIGATION METHODS |
910a0a6d |
1316 | |
4e5a7b2c |
1317 | =head2 reading_sequence( $first, $last, $sigil, $backup ) |
e2902068 |
1318 | |
1319 | Returns the ordered list of readings, starting with $first and ending |
4e5a7b2c |
1320 | with $last, for the witness given in $sigil. If a $backup sigil is |
1321 | specified (e.g. when walking a layered witness), it will be used wherever |
1322 | no $sigil path exists. If there is a base text reading, that will be |
1323 | used wherever no path exists for $sigil or $backup. |
e2902068 |
1324 | |
1325 | =cut |
1326 | |
910a0a6d |
1327 | # TODO Think about returning some lazy-eval iterator. |
b0b4421a |
1328 | # TODO Get rid of backup; we should know from what witness is whether we need it. |
910a0a6d |
1329 | |
e2902068 |
1330 | sub reading_sequence { |
861c3e27 |
1331 | my( $self, $start, $end, $witness ) = @_; |
e2902068 |
1332 | |
930ff666 |
1333 | $witness = $self->baselabel unless $witness; |
e2902068 |
1334 | my @readings = ( $start ); |
1335 | my %seen; |
1336 | my $n = $start; |
3a2ebbf4 |
1337 | while( $n && $n->id ne $end->id ) { |
1338 | if( exists( $seen{$n->id} ) ) { |
63778331 |
1339 | throw( "Detected loop for $witness at " . $n->id ); |
910a0a6d |
1340 | } |
3a2ebbf4 |
1341 | $seen{$n->id} = 1; |
910a0a6d |
1342 | |
861c3e27 |
1343 | my $next = $self->next_reading( $n, $witness ); |
44771cf2 |
1344 | unless( $next ) { |
63778331 |
1345 | throw( "Did not find any path for $witness from reading " . $n->id ); |
44771cf2 |
1346 | } |
910a0a6d |
1347 | push( @readings, $next ); |
1348 | $n = $next; |
e2902068 |
1349 | } |
1350 | # Check that the last reading is our end reading. |
1351 | my $last = $readings[$#readings]; |
63778331 |
1352 | throw( "Last reading found from " . $start->text . |
1353 | " for witness $witness is not the end!" ) # TODO do we get this far? |
3a2ebbf4 |
1354 | unless $last->id eq $end->id; |
e2902068 |
1355 | |
1356 | return @readings; |
1357 | } |
1358 | |
4e5a7b2c |
1359 | =head2 next_reading( $reading, $sigil ); |
8e1394aa |
1360 | |
4a8828f0 |
1361 | Returns the reading that follows the given reading along the given witness |
930ff666 |
1362 | path. |
8e1394aa |
1363 | |
1364 | =cut |
1365 | |
4a8828f0 |
1366 | sub next_reading { |
e2902068 |
1367 | # Return the successor via the corresponding path. |
8e1394aa |
1368 | my $self = shift; |
3a2ebbf4 |
1369 | my $answer = $self->_find_linked_reading( 'next', @_ ); |
2c669bca |
1370 | return undef unless $answer; |
3a2ebbf4 |
1371 | return $self->reading( $answer ); |
8e1394aa |
1372 | } |
1373 | |
4e5a7b2c |
1374 | =head2 prior_reading( $reading, $sigil ) |
8e1394aa |
1375 | |
4a8828f0 |
1376 | Returns the reading that precedes the given reading along the given witness |
930ff666 |
1377 | path. |
8e1394aa |
1378 | |
1379 | =cut |
1380 | |
4a8828f0 |
1381 | sub prior_reading { |
e2902068 |
1382 | # Return the predecessor via the corresponding path. |
8e1394aa |
1383 | my $self = shift; |
3a2ebbf4 |
1384 | my $answer = $self->_find_linked_reading( 'prior', @_ ); |
1385 | return $self->reading( $answer ); |
8e1394aa |
1386 | } |
1387 | |
4a8828f0 |
1388 | sub _find_linked_reading { |
861c3e27 |
1389 | my( $self, $direction, $node, $path ) = @_; |
1390 | |
1391 | # Get a backup if we are dealing with a layered witness |
1392 | my $alt_path; |
1393 | my $aclabel = $self->ac_label; |
1394 | if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) { |
1395 | $alt_path = $1; |
1396 | } |
1397 | |
e2902068 |
1398 | my @linked_paths = $direction eq 'next' |
3a2ebbf4 |
1399 | ? $self->sequence->edges_from( $node ) |
1400 | : $self->sequence->edges_to( $node ); |
e2902068 |
1401 | return undef unless scalar( @linked_paths ); |
8e1394aa |
1402 | |
e2902068 |
1403 | # We have to find the linked path that contains all of the |
1404 | # witnesses supplied in $path. |
1405 | my( @path_wits, @alt_path_wits ); |
4e5a7b2c |
1406 | @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path; |
1407 | @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path; |
e2902068 |
1408 | my $base_le; |
1409 | my $alt_le; |
1410 | foreach my $le ( @linked_paths ) { |
3a2ebbf4 |
1411 | if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) { |
910a0a6d |
1412 | $base_le = $le; |
910a0a6d |
1413 | } |
508fd430 |
1414 | my @le_wits = sort $self->path_witnesses( $le ); |
3a2ebbf4 |
1415 | if( _is_within( \@path_wits, \@le_wits ) ) { |
1416 | # This is the right path. |
1417 | return $direction eq 'next' ? $le->[1] : $le->[0]; |
1418 | } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) { |
1419 | $alt_le = $le; |
1420 | } |
8e1394aa |
1421 | } |
e2902068 |
1422 | # Got this far? Return the alternate path if it exists. |
3a2ebbf4 |
1423 | return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0] |
910a0a6d |
1424 | if $alt_le; |
e2902068 |
1425 | |
1426 | # Got this far? Return the base path if it exists. |
3a2ebbf4 |
1427 | return $direction eq 'next' ? $base_le->[1] : $base_le->[0] |
910a0a6d |
1428 | if $base_le; |
e2902068 |
1429 | |
1430 | # Got this far? We have no appropriate path. |
2c669bca |
1431 | warn "Could not find $direction node from " . $node->id |
910a0a6d |
1432 | . " along path $path"; |
8e1394aa |
1433 | return undef; |
1434 | } |
1435 | |
4a8828f0 |
1436 | # Some set logic. |
1437 | sub _is_within { |
1438 | my( $set1, $set2 ) = @_; |
7854e12e |
1439 | my $ret = @$set1; # will be 0, i.e. false, if set1 is empty |
4a8828f0 |
1440 | foreach my $el ( @$set1 ) { |
910a0a6d |
1441 | $ret = 0 unless grep { /^\Q$el\E$/ } @$set2; |
4a8828f0 |
1442 | } |
1443 | return $ret; |
1444 | } |
1445 | |
4e5a7b2c |
1446 | # Return the string that joins together a list of witnesses for |
1447 | # display on a single path. |
1448 | sub _witnesses_of_label { |
1449 | my( $self, $label ) = @_; |
1450 | my $regex = $self->wit_list_separator; |
1451 | my @answer = split( /\Q$regex\E/, $label ); |
1452 | return @answer; |
b0b4421a |
1453 | } |
1454 | |
d4b75f44 |
1455 | =head2 common_readings |
1456 | |
1457 | Returns the list of common readings in the graph (i.e. those readings that are |
1458 | shared by all non-lacunose witnesses.) |
1459 | |
1460 | =cut |
1461 | |
1462 | sub common_readings { |
1463 | my $self = shift; |
1464 | my @common = grep { $_->is_common } $self->readings; |
1465 | return @common; |
1466 | } |
1467 | |
fae52efd |
1468 | =head2 path_text( $sigil, [, $start, $end ] ) |
b0b4421a |
1469 | |
1470 | Returns the text of a witness (plus its backup, if we are using a layer) |
1471 | as stored in the collation. The text is returned as a string, where the |
1472 | individual readings are joined with spaces and the meta-readings (e.g. |
1473 | lacunae) are omitted. Optional specification of $start and $end allows |
1474 | the generation of a subset of the witness text. |
4e5a7b2c |
1475 | |
b0b4421a |
1476 | =cut |
1477 | |
1478 | sub path_text { |
861c3e27 |
1479 | my( $self, $wit, $start, $end ) = @_; |
b0b4421a |
1480 | $start = $self->start unless $start; |
1481 | $end = $self->end unless $end; |
861c3e27 |
1482 | my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit ); |
629e27b0 |
1483 | my $pathtext = ''; |
1484 | my $last; |
1485 | foreach my $r ( @path ) { |
6ad2ce78 |
1486 | unless ( $r->join_prior || !$last || $last->join_next ) { |
1487 | $pathtext .= ' '; |
1488 | } |
1489 | $pathtext .= $r->text; |
629e27b0 |
1490 | $last = $r; |
1491 | } |
1492 | return $pathtext; |
b0b4421a |
1493 | } |
4e5a7b2c |
1494 | |
1495 | =head1 INITIALIZATION METHODS |
1496 | |
1497 | These are mostly for use by parsers. |
1498 | |
1499 | =head2 make_witness_path( $witness ) |
1500 | |
1501 | Link the array of readings contained in $witness->path (and in |
1502 | $witness->uncorrected_path if it exists) into collation paths. |
1503 | Clear out the arrays when finished. |
de51424a |
1504 | |
4e5a7b2c |
1505 | =head2 make_witness_paths |
1506 | |
1507 | Call make_witness_path for all witnesses in the tradition. |
1508 | |
1509 | =cut |
930ff666 |
1510 | |
7e450e44 |
1511 | # For use when a collation is constructed from a base text and an apparatus. |
1512 | # We have the sequences of readings and just need to add path edges. |
1f7aa795 |
1513 | # When we are done, clear out the witness path attributes, as they are no |
1514 | # longer needed. |
1515 | # TODO Find a way to replace the witness path attributes with encapsulated functions? |
e2902068 |
1516 | |
6a222840 |
1517 | sub make_witness_paths { |
1518 | my( $self ) = @_; |
910a0a6d |
1519 | foreach my $wit ( $self->tradition->witnesses ) { |
6771a1b1 |
1520 | # say STDERR "Making path for " . $wit->sigil; |
910a0a6d |
1521 | $self->make_witness_path( $wit ); |
7854e12e |
1522 | } |
7854e12e |
1523 | } |
1524 | |
6a222840 |
1525 | sub make_witness_path { |
7854e12e |
1526 | my( $self, $wit ) = @_; |
1527 | my @chain = @{$wit->path}; |
15d2d3df |
1528 | my $sig = $wit->sigil; |
fae52efd |
1529 | # Add start and end if necessary |
1530 | unshift( @chain, $self->start ) unless $chain[0] eq $self->start; |
1531 | push( @chain, $self->end ) unless $chain[-1] eq $self->end; |
7854e12e |
1532 | foreach my $idx ( 0 .. $#chain-1 ) { |
910a0a6d |
1533 | $self->add_path( $chain[$idx], $chain[$idx+1], $sig ); |
7854e12e |
1534 | } |
1f7aa795 |
1535 | if( $wit->is_layered ) { |
d9e873d0 |
1536 | @chain = @{$wit->uncorrected_path}; |
fae52efd |
1537 | unshift( @chain, $self->start ) unless $chain[0] eq $self->start; |
1538 | push( @chain, $self->end ) unless $chain[-1] eq $self->end; |
d9e873d0 |
1539 | foreach my $idx( 0 .. $#chain-1 ) { |
1540 | my $source = $chain[$idx]; |
1541 | my $target = $chain[$idx+1]; |
1542 | $self->add_path( $source, $target, $sig.$self->ac_label ) |
1543 | unless $self->has_path( $source, $target, $sig ); |
1544 | } |
15d2d3df |
1545 | } |
1f7aa795 |
1546 | $wit->clear_path; |
1547 | $wit->clear_uncorrected_path; |
e2902068 |
1548 | } |
1549 | |
4e5a7b2c |
1550 | =head2 calculate_ranks |
1551 | |
1552 | Calculate the reading ranks (that is, their aligned positions relative |
1553 | to each other) for the graph. This can only be called on linear collations. |
1554 | |
b365fbae |
1555 | =begin testing |
1556 | |
1557 | use Text::Tradition; |
1558 | |
1559 | my $cxfile = 't/data/Collatex-16.xml'; |
1560 | my $t = Text::Tradition->new( |
1561 | 'name' => 'inline', |
1562 | 'input' => 'CollateX', |
1563 | 'file' => $cxfile, |
1564 | ); |
1565 | my $c = $t->collation; |
1566 | |
1567 | # Make an svg |
bfcbcecb |
1568 | my $table = $c->alignment_table; |
1569 | ok( $c->has_cached_table, "Alignment table was cached" ); |
1570 | is( $c->alignment_table, $table, "Cached table returned upon second call" ); |
b365fbae |
1571 | $c->calculate_ranks; |
bfcbcecb |
1572 | is( $c->alignment_table, $table, "Cached table retained with no rank change" ); |
679f17e1 |
1573 | $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } ); |
bfcbcecb |
1574 | isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" ); |
b365fbae |
1575 | |
1576 | =end testing |
1577 | |
4e5a7b2c |
1578 | =cut |
1579 | |
910a0a6d |
1580 | sub calculate_ranks { |
1581 | my $self = shift; |
b365fbae |
1582 | # Save the existing ranks, in case we need to invalidate the cached SVG. |
1583 | my %existing_ranks; |
ac4d7ac2 |
1584 | map { $existing_ranks{$_} = $_->rank } $self->readings; |
359944f7 |
1585 | |
1586 | # Do the rankings based on the relationship equivalence graph, starting |
1587 | # with the start node. |
56772e8c |
1588 | my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks(); |
1589 | |
910a0a6d |
1590 | # Transfer our rankings from the topological graph to the real one. |
1591 | foreach my $r ( $self->readings ) { |
cecbe56d |
1592 | if( defined $node_ranks->{$self->equivalence( $r->id )} ) { |
359944f7 |
1593 | $r->rank( $node_ranks->{$self->equivalence( $r->id )} ); |
67da8d6c |
1594 | } else { |
63778331 |
1595 | # Die. Find the last rank we calculated. |
359944f7 |
1596 | my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 ) |
1597 | <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) } |
63778331 |
1598 | $self->readings; |
1599 | my $last = pop @all_defined; |
1600 | throw( "Ranks not calculated after $last - do you have a cycle in the graph?" ); |
67da8d6c |
1601 | } |
de51424a |
1602 | } |
bfcbcecb |
1603 | # Do we need to invalidate the cached data? |
be3af600 |
1604 | if( $self->has_cached_table ) { |
b365fbae |
1605 | foreach my $r ( $self->readings ) { |
7c293912 |
1606 | next if defined( $existing_ranks{$r} ) |
1607 | && $existing_ranks{$r} == $r->rank; |
c1915ab9 |
1608 | # Something has changed, so clear the cache |
bfcbcecb |
1609 | $self->_clear_cache; |
c1915ab9 |
1610 | # ...and recalculate the common readings. |
1611 | $self->calculate_common_readings(); |
b365fbae |
1612 | last; |
1613 | } |
1614 | } |
c1915ab9 |
1615 | # The graph calculation information is now up to date. |
1616 | $self->_graphcalc_done(1); |
8e1394aa |
1617 | } |
3a1f2523 |
1618 | |
c1915ab9 |
1619 | sub _clear_cache { |
1620 | my $self = shift; |
c1915ab9 |
1621 | $self->wipe_table if $self->has_cached_table; |
1622 | } |
1623 | |
1624 | |
4e5a7b2c |
1625 | =head2 flatten_ranks |
1626 | |
1627 | A convenience method for parsing collation data. Searches the graph for readings |
1628 | with the same text at the same rank, and merges any that are found. |
1629 | |
1630 | =cut |
1631 | |
0e476982 |
1632 | sub flatten_ranks { |
1633 | my $self = shift; |
1634 | my %unique_rank_rdg; |
bf6e338d |
1635 | my $changed; |
0e476982 |
1636 | foreach my $rdg ( $self->readings ) { |
1637 | next unless $rdg->has_rank; |
1638 | my $key = $rdg->rank . "||" . $rdg->text; |
1639 | if( exists $unique_rank_rdg{$key} ) { |
07e6765f |
1640 | # Make sure they don't have different grammatical forms |
1641 | my $ur = $unique_rank_rdg{$key}; |
a445ce40 |
1642 | if( $rdg->is_identical( $ur ) ) { |
1643 | # Combine! |
1644 | #say STDERR "Combining readings at same rank: $key"; |
1645 | $changed = 1; |
1646 | $self->merge_readings( $unique_rank_rdg{$key}, $rdg ); |
1647 | # TODO see if this now makes a common point. |
07e6765f |
1648 | } |
0e476982 |
1649 | } else { |
1650 | $unique_rank_rdg{$key} = $rdg; |
1651 | } |
1652 | } |
bf6e338d |
1653 | # If we merged readings, the ranks are still fine but the alignment |
1654 | # table is wrong. Wipe it. |
1655 | $self->wipe_table() if $changed; |
0e476982 |
1656 | } |
4633f9e4 |
1657 | |
1658 | |
d4b75f44 |
1659 | =head2 calculate_common_readings |
1660 | |
1661 | Goes through the graph identifying the readings that appear in every witness |
1662 | (apart from those with lacunae at that spot.) Marks them as common and returns |
1663 | the list. |
1664 | |
1665 | =begin testing |
1666 | |
1667 | use Text::Tradition; |
1668 | |
1669 | my $cxfile = 't/data/Collatex-16.xml'; |
1670 | my $t = Text::Tradition->new( |
1671 | 'name' => 'inline', |
1672 | 'input' => 'CollateX', |
1673 | 'file' => $cxfile, |
1674 | ); |
1675 | my $c = $t->collation; |
1676 | |
1677 | my @common = $c->calculate_common_readings(); |
1678 | is( scalar @common, 8, "Found correct number of common readings" ); |
1679 | my @marked = sort $c->common_readings(); |
1680 | is( scalar @common, 8, "All common readings got marked as such" ); |
679f17e1 |
1681 | my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /; |
d4b75f44 |
1682 | is_deeply( \@marked, \@expected, "Found correct list of common readings" ); |
1683 | |
1684 | =end testing |
1685 | |
1686 | =cut |
1687 | |
1688 | sub calculate_common_readings { |
1689 | my $self = shift; |
1690 | my @common; |
c1915ab9 |
1691 | map { $_->is_common( 0 ) } $self->readings; |
1692 | # Implicitly calls calculate_ranks |
1dd07bda |
1693 | my $table = $self->alignment_table; |
d4b75f44 |
1694 | foreach my $idx ( 0 .. $table->{'length'} - 1 ) { |
7f52eac8 |
1695 | my @row = map { $_->{'tokens'}->[$idx] |
1696 | ? $_->{'tokens'}->[$idx]->{'t'} : '' } |
1697 | @{$table->{'alignment'}}; |
d4b75f44 |
1698 | my %hash; |
1699 | foreach my $r ( @row ) { |
1700 | if( $r ) { |
1701 | $hash{$r->id} = $r unless $r->is_meta; |
1702 | } else { |
1703 | $hash{'UNDEF'} = $r; |
1704 | } |
1705 | } |
1706 | if( keys %hash == 1 && !exists $hash{'UNDEF'} ) { |
1707 | my( $r ) = values %hash; |
1708 | $r->is_common( 1 ); |
1709 | push( @common, $r ); |
1710 | } |
1711 | } |
1712 | return @common; |
1713 | } |
1714 | |
861c3e27 |
1715 | =head2 text_from_paths |
1716 | |
1717 | Calculate the text array for all witnesses from the path, for later consistency |
1718 | checking. Only to be used if there is no non-graph-based way to know the |
1719 | original texts. |
1720 | |
1721 | =cut |
1722 | |
1723 | sub text_from_paths { |
1724 | my $self = shift; |
1725 | foreach my $wit ( $self->tradition->witnesses ) { |
5164a6f0 |
1726 | my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); |
1727 | my @text; |
1728 | foreach my $r ( @readings ) { |
1729 | next if $r->is_meta; |
1730 | push( @text, $r->text ); |
1731 | } |
861c3e27 |
1732 | $wit->text( \@text ); |
1733 | if( $wit->is_layered ) { |
5164a6f0 |
1734 | my @ucrdgs = $self->reading_sequence( $self->start, $self->end, |
1735 | $wit->sigil.$self->ac_label ); |
1736 | my @uctext; |
1737 | foreach my $r ( @ucrdgs ) { |
1738 | next if $r->is_meta; |
1739 | push( @uctext, $r->text ); |
1740 | } |
1741 | $wit->layertext( \@uctext ); |
861c3e27 |
1742 | } |
1743 | } |
1744 | } |
0e476982 |
1745 | |
4e5a7b2c |
1746 | =head1 UTILITY FUNCTIONS |
1747 | |
1748 | =head2 common_predecessor( $reading_a, $reading_b ) |
8e1394aa |
1749 | |
4e5a7b2c |
1750 | Find the last reading that occurs in sequence before both the given readings. |
414cc046 |
1751 | At the very least this should be $self->start. |
4e5a7b2c |
1752 | |
1753 | =head2 common_successor( $reading_a, $reading_b ) |
1754 | |
1755 | Find the first reading that occurs in sequence after both the given readings. |
414cc046 |
1756 | At the very least this should be $self->end. |
4e5a7b2c |
1757 | |
22222af9 |
1758 | =begin testing |
1759 | |
1760 | use Text::Tradition; |
1761 | |
1762 | my $cxfile = 't/data/Collatex-16.xml'; |
1763 | my $t = Text::Tradition->new( |
1764 | 'name' => 'inline', |
1765 | 'input' => 'CollateX', |
1766 | 'file' => $cxfile, |
1767 | ); |
1768 | my $c = $t->collation; |
1769 | |
679f17e1 |
1770 | is( $c->common_predecessor( 'n24', 'n23' )->id, |
22222af9 |
1771 | 'n20', "Found correct common predecessor" ); |
679f17e1 |
1772 | is( $c->common_successor( 'n24', 'n23' )->id, |
10e4b1ac |
1773 | '__END__', "Found correct common successor" ); |
22222af9 |
1774 | |
4e5a7b2c |
1775 | is( $c->common_predecessor( 'n19', 'n17' )->id, |
22222af9 |
1776 | 'n16', "Found correct common predecessor for readings on same path" ); |
679f17e1 |
1777 | is( $c->common_successor( 'n21', 'n10' )->id, |
10e4b1ac |
1778 | '__END__', "Found correct common successor for readings on same path" ); |
22222af9 |
1779 | |
1780 | =end testing |
1781 | |
1782 | =cut |
1783 | |
1784 | ## Return the closest reading that is a predecessor of both the given readings. |
1785 | sub common_predecessor { |
1786 | my $self = shift; |
4e5a7b2c |
1787 | my( $r1, $r2 ) = $self->_objectify_args( @_ ); |
027d819c |
1788 | return $self->_common_in_path( $r1, $r2, 'predecessors' ); |
22222af9 |
1789 | } |
1790 | |
1791 | sub common_successor { |
1792 | my $self = shift; |
4e5a7b2c |
1793 | my( $r1, $r2 ) = $self->_objectify_args( @_ ); |
027d819c |
1794 | return $self->_common_in_path( $r1, $r2, 'successors' ); |
22222af9 |
1795 | } |
1796 | |
414cc046 |
1797 | |
1798 | # TODO think about how to do this without ranks... |
027d819c |
1799 | sub _common_in_path { |
22222af9 |
1800 | my( $self, $r1, $r2, $dir ) = @_; |
414cc046 |
1801 | my $iter = $self->end->rank; |
22222af9 |
1802 | my @candidates; |
414cc046 |
1803 | my @last_r1 = ( $r1 ); |
1804 | my @last_r2 = ( $r2 ); |
1805 | # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' ); |
22222af9 |
1806 | my %all_seen; |
6771a1b1 |
1807 | # say STDERR "Finding common $dir for $r1, $r2"; |
22222af9 |
1808 | while( !@candidates ) { |
414cc046 |
1809 | last unless $iter--; # Avoid looping infinitely |
1810 | # Iterate separately down the graph from r1 and r2 |
1811 | my( @new_lc1, @new_lc2 ); |
1812 | foreach my $lc ( @last_r1 ) { |
1813 | foreach my $p ( $lc->$dir ) { |
1814 | if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) { |
6771a1b1 |
1815 | # say STDERR "Path candidate $p from $lc"; |
414cc046 |
1816 | push( @candidates, $p ); |
002e3600 |
1817 | } elsif( !$all_seen{$p->id} ) { |
414cc046 |
1818 | $all_seen{$p->id} = 'r1'; |
1819 | push( @new_lc1, $p ); |
1820 | } |
1821 | } |
1822 | } |
1823 | foreach my $lc ( @last_r2 ) { |
22222af9 |
1824 | foreach my $p ( $lc->$dir ) { |
414cc046 |
1825 | if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) { |
6771a1b1 |
1826 | # say STDERR "Path candidate $p from $lc"; |
22222af9 |
1827 | push( @candidates, $p ); |
002e3600 |
1828 | } elsif( !$all_seen{$p->id} ) { |
414cc046 |
1829 | $all_seen{$p->id} = 'r2'; |
1830 | push( @new_lc2, $p ); |
22222af9 |
1831 | } |
1832 | } |
1833 | } |
414cc046 |
1834 | @last_r1 = @new_lc1; |
1835 | @last_r2 = @new_lc2; |
22222af9 |
1836 | } |
1837 | my @answer = sort { $a->rank <=> $b->rank } @candidates; |
1838 | return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer ); |
1839 | } |
1840 | |
63778331 |
1841 | sub throw { |
1842 | Text::Tradition::Error->throw( |
1843 | 'ident' => 'Collation error', |
1844 | 'message' => $_[0], |
1845 | ); |
1846 | } |
1847 | |
dd3b58b0 |
1848 | no Moose; |
1849 | __PACKAGE__->meta->make_immutable; |
e867486f |
1850 | |
a445ce40 |
1851 | =head1 BUGS/TODO |
1852 | |
1853 | =over |
1854 | |
1855 | =item * Rework XML serialization in a more modular way |
1856 | |
1857 | =back |
1858 | |
027d819c |
1859 | =head1 LICENSE |
e867486f |
1860 | |
027d819c |
1861 | This package is free software and is provided "as is" without express |
1862 | or implied warranty. You can redistribute it and/or modify it under |
1863 | the same terms as Perl itself. |
e867486f |
1864 | |
027d819c |
1865 | =head1 AUTHOR |
e867486f |
1866 | |
027d819c |
1867 | Tara L Andrews E<lt>aurum@cpan.orgE<gt> |