1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Data;
11 use Text::Tradition::Collation::Reading;
12 use Text::Tradition::Collation::RelationshipStore;
13 use Text::Tradition::Error;
14 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
16 use XML::LibXML::XPathContext;
20 isa => 'Text::Tradition::Collation::Data',
58 isa => 'Text::Tradition',
59 writer => '_set_tradition',
65 Text::Tradition::Collation - a software model for a text collation
70 my $t = Text::Tradition->new(
71 'name' => 'this is a text',
73 'file' => '/path/to/tei_parallel_seg_file.xml' );
75 my $c = $t->collation;
76 my @readings = $c->readings;
77 my @paths = $c->paths;
78 my @relationships = $c->relationships;
80 my $svg_variant_graph = $t->collation->as_svg();
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.
93 The constructor. Takes a hash or hashref of the following arguments:
97 =item * tradition - The Text::Tradition object to which the collation
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.
104 =item * baselabel - The default label for the path taken by a base text
105 (if any). Defaults to 'base text'.
107 =item * wit_list_separator - The string to join a list of witnesses for
108 purposes of making labels in display graphs. Defaults to ', '.
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.)'.
115 =item * wordsep - The string used to separate words in the original text.
126 =head2 wit_list_separator
134 Simple accessors for collation attributes.
138 The meta-reading at the start of every witness path.
142 The meta-reading at the end of every witness path.
146 Returns all Reading objects in the graph.
148 =head2 reading( $id )
150 Returns the Reading object corresponding to the given ID.
152 =head2 add_reading( $reading_args )
154 Adds a new reading object to the collation.
155 See L<Text::Tradition::Collation::Reading> for the available arguments.
157 =head2 del_reading( $object_or_id )
159 Removes the given reading from the collation, implicitly removing its
160 paths and relationships.
162 =head2 has_reading( $id )
164 Predicate to see whether a given reading ID is in the graph.
166 =head2 reading_witnesses( $object_or_id )
168 Returns a list of sigils whose witnesses contain the reading.
172 Returns all reading paths within the document - that is, all edges in the
173 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
175 =head2 add_path( $source, $target, $sigil )
177 Links the given readings in the collation in sequence, under the given witness
178 sigil. The readings may be specified by object or ID.
180 =head2 del_path( $source, $target, $sigil )
182 Links the given readings in the collation in sequence, under the given witness
183 sigil. The readings may be specified by object or ID.
185 =head2 has_path( $source, $target );
187 Returns true if the two readings are linked in sequence in any witness.
188 The readings may be specified by object or ID.
192 Returns all Relationship objects in the collation.
194 =head2 add_relationship( $reading, $other_reading, $options )
196 Adds a new relationship of the type given in $options between the two readings,
197 which may be specified by object or ID. Returns a value of ( $status, @vectors)
198 where $status is true on success, and @vectors is a list of relationship edges
199 that were ultimately added.
200 See L<Text::Tradition::Collation::Relationship> for the available options.
205 my ( $class, @args ) = @_;
206 my %args = @args == 1 ? %{ $args[0] } : @args;
207 # TODO determine these from the Moose::Meta object
208 my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel
209 linear wordsep start end cached_table _graphcalc_done);
211 for my $attr (@delegate_attrs) {
212 $data_args{$attr} = delete $args{$attr} if exists $args{$attr};
214 $args{_data} = Text::Tradition::Collation::Data->new(%data_args);
220 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
221 $self->_set_start( $self->add_reading(
222 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
223 $self->_set_end( $self->add_reading(
224 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
227 sub register_relationship_type {
229 my %args = @_ == 1 ? %{$_[0]} : @_;
230 if( $self->relations->has_type( $args{name} ) ) {
231 throw( 'Relationship type ' . $args{name} . ' already registered' );
233 $self->relations->add_type( %args );
236 sub get_relationship_type {
237 my( $self, $name ) = @_;
238 return $self->relations->has_type( $name )
239 ? $self->relations->type( $name ) : undef;
242 ### Reading construct/destruct functions
245 my( $self, $reading ) = @_;
246 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
247 my %args = %$reading;
248 if( $args{'init'} ) {
249 # If we are initializing an empty collation, don't assume that we
250 # have set a tradition.
251 delete $args{'init'};
252 } elsif( $self->tradition->can('language') && $self->tradition->has_language
253 && !exists $args{'language'} ) {
254 $args{'language'} = $self->tradition->language;
256 $reading = Text::Tradition::Collation::Reading->new(
257 'collation' => $self,
260 # First check to see if a reading with this ID exists.
261 if( $self->reading( $reading->id ) ) {
262 throw( "Collation already has a reading with id " . $reading->id );
264 $self->_graphcalc_done(0);
265 $self->_add_reading( $reading->id => $reading );
266 # Once the reading has been added, put it in both graphs.
267 $self->sequence->add_vertex( $reading->id );
268 $self->relations->add_reading( $reading->id );
272 around del_reading => sub {
277 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
280 # Remove the reading from the graphs.
281 $self->_graphcalc_done(0);
282 $self->_clear_cache; # Explicitly clear caches to GC the reading
283 $self->sequence->delete_vertex( $arg );
284 $self->relations->delete_reading( $arg );
287 $self->$orig( $arg );
290 =head2 merge_readings( $main, $second, $concatenate, $with_str )
292 Merges the $second reading into the $main one. If $concatenate is true, then
293 the merged node will carry the text of both readings, concatenated with either
294 $with_str (if specified) or a sensible default (the empty string if the
295 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
297 The first two arguments may be either readings or reading IDs.
303 my $cxfile = 't/data/Collatex-16.xml';
304 my $t = Text::Tradition->new(
306 'input' => 'CollateX',
309 my $c = $t->collation;
311 my $rno = scalar $c->readings;
312 # Split n21 ('unto') for testing purposes
313 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
314 my $old_r = $c->reading( 'n21' );
315 $old_r->alter_text( 'to' );
316 $c->del_path( 'n20', 'n21', 'A' );
317 $c->add_path( 'n20', 'n21p0', 'A' );
318 $c->add_path( 'n21p0', 'n21', 'A' );
319 $c->add_relationship( 'n21', 'n22', { type => 'collated', scope => 'local' } );
321 ok( $c->reading( 'n21p0' ), "New reading exists" );
322 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
324 # Combine n3 and n4 ( with his )
325 $c->merge_readings( 'n3', 'n4', 1 );
326 ok( !$c->reading('n4'), "Reading n4 is gone" );
327 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
329 # Collapse n9 and n10 ( rood / root )
330 $c->merge_readings( 'n9', 'n10' );
331 ok( !$c->reading('n10'), "Reading n10 is gone" );
332 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
334 # Combine n21 and n21p0
335 my $remaining = $c->reading('n21');
336 $remaining ||= $c->reading('n22'); # one of these should still exist
337 $c->merge_readings( 'n21p0', $remaining, 1 );
338 ok( !$c->reading('n21'), "Reading $remaining is gone" );
339 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
349 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
350 my $mergemeta = $kept_obj->is_meta;
351 throw( "Cannot merge meta and non-meta reading" )
352 unless ( $mergemeta && $del_obj->is_meta )
353 || ( !$mergemeta && !$del_obj->is_meta );
355 throw( "Cannot merge with start or end node" )
356 if( $kept_obj eq $self->start || $kept_obj eq $self->end
357 || $del_obj eq $self->start || $del_obj eq $self->end );
358 throw( "Cannot combine text of meta readings" ) if $combine;
360 # We can only merge readings in a linear graph if:
361 # - they are contiguous with only one edge between them, OR
362 # - they are at equivalent ranks in the graph.
363 if( $self->linear ) {
364 my @delpred = $del_obj->predecessors;
365 my @keptsuc = $kept_obj->successors;
366 unless ( @delpred == 1 && $delpred[0] eq $kept_obj
367 && @keptsuc == 1 && $keptsuc[0] eq $del_obj ) {
368 my( $is_ok, $msg ) = $self->relations->relationship_valid(
369 $kept_obj, $del_obj, 'collated' );
371 throw( "Readings $kept_obj and $del_obj can be neither concatenated nor collated" );
376 # We only need the IDs for adding paths to the graph, not the reading
377 # objects themselves.
378 my $kept = $kept_obj->id;
379 my $deleted = $del_obj->id;
380 $self->_graphcalc_done(0);
382 # The kept reading should inherit the paths and the relationships
383 # of the deleted reading.
384 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
385 my @vector = ( $kept );
386 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
387 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
388 next if $vector[0] eq $vector[1]; # Don't add a self loop
389 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
390 $self->sequence->add_edge( @vector );
391 my $fwits = $self->sequence->get_edge_attributes( @vector );
392 @wits{keys %$fwits} = values %$fwits;
393 $self->sequence->set_edge_attributes( @vector, \%wits );
395 $self->relations->merge_readings( $kept, $deleted, $combine );
397 # Do the deletion deed.
399 # Combine the text of the readings
400 my $joinstr = $combine_char;
401 unless( defined $joinstr ) {
402 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
403 $joinstr = $self->wordsep unless defined $joinstr;
405 $kept_obj->_combine( $del_obj, $joinstr );
407 $self->del_reading( $deleted );
410 =head2 merge_related( @relationship_types )
412 Merge all readings linked with the relationship types given. If any of the selected type(s) is not a colocation, the graph will no longer be linear. The majority/plurality reading in each case will be the one kept.
414 WARNING: This operation cannot be undone.
423 my $t = Text::Tradition->new(
426 'file' => 't/data/legendfrag.xml',
428 my $c = $t->collation;
431 map { $rdg_ids{$_} = 1 } $c->readings;
432 $c->merge_related( 'orthographic' );
433 is( scalar( $c->readings ), keys( %rdg_ids ) - 8,
434 "Successfully collapsed orthographic variation" );
435 map { $rdg_ids{$_} = undef } qw/ r13.3 r11.4 r8.5 r8.2 r7.7 r7.5 r7.4 r7.1 /;
436 foreach my $rid ( keys %rdg_ids ) {
437 my $exp = $rdg_ids{$rid};
438 is( !$c->reading( $rid ), !$exp, "Reading $rid correctly " .
439 ( $exp ? "retained" : "removed" ) );
441 ok( $c->linear, "Graph is still linear" );
443 $c->calculate_ranks; # This should succeed
444 ok( 1, "Can still calculate ranks on the new graph" );
446 ok( 0, "Rank calculation on merged graph failed: $@" );
449 # Now add some transpositions
450 $c->add_relationship( 'r8.4', 'r10.4', { type => 'transposition' } );
451 $c->merge_related( 'transposition' );
452 is( scalar( $c->readings ), keys( %rdg_ids ) - 9,
453 "Transposed relationship is merged away" );
454 ok( !$c->reading('r8.4'), "Correct transposed reading removed" );
455 ok( !$c->linear, "Graph is no longer linear" );
457 $c->calculate_ranks; # This should fail
458 ok( 0, "Rank calculation happened on nonlinear graph?!" );
459 } catch ( Text::Tradition::Error $e ) {
460 is( $e->message, 'Cannot calculate ranks on a non-linear graph',
461 "Rank calculation on merged graph threw an error" );
470 # TODO: there should be a way to display merged without affecting the underlying data!
475 map { $reltypehash{$_} = 1 } @_;
477 # Set up the filter for finding related readings
479 exists $reltypehash{$_[0]->type};
483 # Go through all readings looking for related ones
484 foreach my $r ( $self->readings ) {
485 next unless $self->reading( "$r" ); # might have been deleted meanwhile
486 my @related = $self->related_readings( $r, $filter );
488 push( @related, $r );
490 scalar $b->witnesses <=> scalar $a->witnesses
492 my $keep = shift @related;
493 foreach my $delr ( @related ) {
495 unless( $self->get_relationship( $keep, $delr )->colocated );
496 $self->merge_readings( $keep, $delr );
500 $self->linear( $linear );
503 =head2 compress_readings
505 Where possible in the graph, compresses plain sequences of readings into a
506 single reading. The sequences must consist of readings with no
507 relationships to other readings, with only a single witness path between
508 them and no other witness paths from either that would skip the other. The
509 readings must also not be marked as nonsense or bad grammar.
511 WARNING: This operation cannot be undone.
515 sub compress_readings {
517 # Anywhere in the graph that there is a reading that joins only to a single
518 # successor, and neither of these have any relationships, just join the two
520 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
521 # Now look for readings that can be joined to their successors.
522 next unless $rdg->is_combinable;
524 while( $self->sequence->successors( $rdg ) == 1 ) {
525 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
526 throw( "Infinite loop" ) if $seen{$next->id};
527 $seen{$next->id} = 1;
528 last if $self->sequence->predecessors( $next ) > 1;
529 last unless $next->is_combinable;
530 say "Joining readings $rdg and $next";
531 $self->merge_readings( $rdg, $next, 1 );
534 # Make sure we haven't screwed anything up
535 foreach my $wit ( $self->tradition->witnesses ) {
536 my $pathtext = $self->path_text( $wit->sigil );
537 my $origtext = join( ' ', @{$wit->text} );
538 throw( "Text differs for witness " . $wit->sigil )
539 unless $pathtext eq $origtext;
540 if( $wit->is_layered ) {
541 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
542 $origtext = join( ' ', @{$wit->layertext} );
543 throw( "Ante-corr text differs for witness " . $wit->sigil )
544 unless $pathtext eq $origtext;
548 $self->relations->rebuild_equivalence();
549 $self->calculate_ranks();
552 # Helper function for manipulating the graph.
553 sub _stringify_args {
554 my( $self, $first, $second, @args ) = @_;
556 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
557 $second = $second->id
558 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
559 return( $first, $second, @args );
562 # Helper function for manipulating the graph.
563 sub _objectify_args {
564 my( $self, $first, $second, $arg ) = @_;
565 $first = $self->reading( $first )
566 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
567 $second = $self->reading( $second )
568 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
569 return( $first, $second, $arg );
572 =head2 duplicate_reading( $reading, @witlist )
574 Split the given reading into two, so that the new reading is in the path for
575 the witnesses given in @witlist. If the result is that certain non-colocated
576 relationships (e.g. transpositions) are no longer valid, these will be removed.
577 Returns the newly-created reading.
581 use Test::More::UTF8;
585 my $st = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/collatecorr.xml' );
586 is( ref( $st ), 'Text::Tradition', "Got a tradition from test file" );
587 ok( $st->has_witness('Ba96'), "Tradition has the affected witness" );
589 my $sc = $st->collation;
591 ok( $sc->reading('n131'), "Tradition has the affected reading" );
592 is( scalar( $sc->readings ), $numr, "There are $numr readings in the graph" );
593 is( $sc->end->rank, 14, "There are fourteen ranks in the graph" );
595 # Detach the erroneously collated reading
596 my( $newr, @del_rdgs ) = $sc->duplicate_reading( 'n131', 'Ba96' );
597 ok( $newr, "New reading was created" );
598 ok( $sc->reading('n131_0'), "Detached the bad collation with a new reading" );
599 is( scalar( $sc->readings ), $numr + 1, "A reading was added to the graph" );
600 is( $sc->end->rank, 10, "There are now only ten ranks in the graph" );
601 my $csucc = $sc->common_successor( 'n131', 'n131_0' );
602 is( $csucc->id, 'n136', "Found correct common successor to duped reading" );
604 # Check that the bad transposition is gone
605 is( scalar @del_rdgs, 1, "Deleted reading was returned by API call" );
606 is( $sc->get_relationship( 'n130', 'n135' ), undef, "Bad transposition relationship is gone" );
608 # The collation should not be fixed
609 my @pairs = $sc->identical_readings();
610 is( scalar @pairs, 0, "Not re-collated yet" );
612 ok( $sc->merge_readings( 'n124', 'n131_0' ), "Collated the readings correctly" );
613 @pairs = $sc->identical_readings( start => 'n124', end => $csucc->id );
614 is( scalar @pairs, 3, "Found three more identical readings" );
615 is( $sc->end->rank, 11, "The ranks shifted appropriately" );
616 $sc->flatten_ranks();
617 is( scalar( $sc->readings ), $numr - 3, "Now we are collated correctly" );
619 # Check that we can't "duplicate" a reading with no wits or with all wits
621 my( $badr, @del_rdgs ) = $sc->duplicate_reading( 'n124' );
622 ok( 0, "Reading duplication without witnesses throws an error" );
623 } catch( Text::Tradition::Error $e ) {
624 like( $e->message, qr/Must specify one or more witnesses/,
625 "Reading duplication without witnesses throws the expected error" );
627 ok( 0, "Reading duplication without witnesses threw the wrong error" );
631 my( $badr, @del_rdgs ) = $sc->duplicate_reading( 'n124', 'Ba96', 'Mü11475' );
632 ok( 0, "Reading duplication with all witnesses throws an error" );
633 } catch( Text::Tradition::Error $e ) {
634 like( $e->message, qr/Cannot join all witnesses/,
635 "Reading duplication with all witnesses throws the expected error" );
637 ok( 0, "Reading duplication with all witnesses threw the wrong error" );
644 sub duplicate_reading {
645 my( $self, $r, @wits ) = @_;
646 # Check that we are not doing anything unwise.
647 throw( "Must specify one or more witnesses for the duplicated reading" )
649 unless( ref( $r ) eq 'Text::Tradition::Collation::Reading' ) {
650 $r = $self->reading( $r );
652 throw( "Cannot duplicate a meta-reading" )
654 throw( "Cannot join all witnesses to the new reading" )
655 if scalar( @wits ) == scalar( $r->witnesses );
657 # Get all the reading attributes and duplicate them.
658 my $rmeta = Text::Tradition::Collation::Reading->meta;
660 foreach my $attr( $rmeta->get_all_attributes ) {
661 next if $attr->name =~ /^_/;
662 my $acc = $attr->get_read_method;
663 if( !$acc && $attr->has_applied_traits ) {
664 my $tr = $attr->applied_traits;
665 if( $tr->[0] =~ /::(Array|Hash)$/ ) {
667 my %methods = reverse %{$attr->handles};
668 $acc = $methods{elements};
669 $args{$attr->name} = $which eq 'Array'
670 ? [ $r->$acc ] : { $r->$acc };
673 $args{$attr->name} = $r->$acc if $acc;
676 # By definition the new reading will no longer be common.
677 $args{is_common} = 0;
678 # The new reading also needs its own ID.
679 $args{id} = $self->_generate_dup_id( $r->id );
681 # Try to make the new reading.
682 my $newr = $self->add_reading( \%args );
683 # The old reading is also no longer common.
686 # For each of the witnesses, dissociate from the old reading and
687 # associate with the new.
688 foreach my $wit ( @wits ) {
689 my $prior = $self->prior_reading( $r, $wit );
690 my $next = $self->next_reading( $r, $wit );
691 $self->del_path( $prior, $r, $wit );
692 $self->add_path( $prior, $newr, $wit );
693 $self->del_path( $r, $next, $wit );
694 $self->add_path( $newr, $next, $wit );
697 # If the graph is ranked, we need to look for relationships that are now
698 # invalid (i.e. 'non-colocation' types that might now be colocated) and
699 # remove them. If not, we can skip it.
702 my @deleted_relations;
703 if( $self->end->has_rank ) {
704 # Find the point where we can stop checking
705 $succ = $self->common_successor( $r, $newr );
707 # Hash the existing ranks
708 foreach my $rdg ( $self->readings ) {
709 $rrk{$rdg->id} = $rdg->rank;
711 # Calculate the new ranks
712 $self->calculate_ranks();
714 # Check for invalid non-colocated relationships among changed-rank readings
715 # from where the ranks start changing up to $succ
716 my $lastrank = $succ->rank;
717 foreach my $rdg ( $self->readings ) {
718 next if $rdg->rank > $lastrank;
719 next if $rdg->rank == $rrk{$rdg->id};
720 my @noncolo = $rdg->related_readings( sub { !$_[0]->colocated } );
721 next unless @noncolo;
722 foreach my $nc ( @noncolo ) {
723 unless( $self->relations->verify_or_delete( $rdg, $nc ) ) {
724 push( @deleted_relations, [ $rdg->id, $nc->id ] );
729 return ( $newr, @deleted_relations );
732 sub _generate_dup_id {
733 my( $self, $rid ) = @_;
738 if( $self->has_reading( $newid ) ) {
751 # We only need the IDs for adding paths to the graph, not the reading
752 # objects themselves.
753 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
755 $self->_graphcalc_done(0);
756 # Connect the readings
757 unless( $self->sequence->has_edge( $source, $target ) ) {
758 $self->sequence->add_edge( $source, $target );
759 $self->relations->add_equivalence_edge( $source, $target );
761 # Note the witness in question
762 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
768 if( ref( $_[0] ) eq 'ARRAY' ) {
775 # We only need the IDs for removing paths from the graph, not the reading
776 # objects themselves.
777 my( $source, $target, $wit ) = $self->_stringify_args( @args );
779 $self->_graphcalc_done(0);
780 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
781 $self->sequence->delete_edge_attribute( $source, $target, $wit );
783 unless( $self->sequence->has_edge_attributes( $source, $target ) ) {
784 $self->sequence->delete_edge( $source, $target );
785 $self->relations->delete_equivalence_edge( $source, $target );
790 # Extra graph-alike utility
793 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
794 return undef unless $self->sequence->has_edge( $source, $target );
795 return $self->sequence->has_edge_attribute( $source, $target, $wit );
798 =head2 clear_witness( @sigil_list )
800 Clear the given witnesses out of the collation entirely, removing references
801 to them in paths, and removing readings that belong only to them. Should only
802 be called via $tradition->del_witness.
807 my( $self, @sigils ) = @_;
809 $self->_graphcalc_done(0);
810 # Clear the witness(es) out of the paths
811 foreach my $e ( $self->paths ) {
812 foreach my $sig ( @sigils ) {
813 $self->del_path( $e, $sig );
817 # Clear out the newly unused readings
818 foreach my $r ( $self->readings ) {
819 unless( $self->reading_witnesses( $r ) ) {
820 $self->del_reading( $r );
825 sub add_relationship {
827 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
828 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
829 foreach my $v ( @vectors ) {
830 next unless $self->get_relationship( $v )->colocated;
831 if( $self->reading( $v->[0] )->has_rank && $self->reading( $v->[1] )->has_rank
832 && $self->reading( $v->[0] )->rank ne $self->reading( $v->[1] )->rank ) {
833 $self->_graphcalc_done(0);
841 around qw/ get_relationship del_relationship / => sub {
845 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
848 my @stringargs = $self->_stringify_args( @args );
849 $self->$orig( @stringargs );
852 =head2 reading_witnesses( $reading )
854 Return a list of sigils corresponding to the witnesses in which the reading appears.
858 sub reading_witnesses {
859 my( $self, $reading ) = @_;
860 # We need only check either the incoming or the outgoing edges; I have
861 # arbitrarily chosen "incoming". Thus, special-case the start node.
862 if( $reading eq $self->start ) {
863 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
866 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
867 my $wits = $self->sequence->get_edge_attributes( @$e );
868 @all_witnesses{ keys %$wits } = 1;
870 my $acstr = $self->ac_label;
871 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
872 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
874 return keys %all_witnesses;
877 =head1 OUTPUT METHODS
879 =head2 as_svg( \%options )
881 Returns an SVG string that represents the graph, via as_dot and graphviz.
882 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
887 my( $self, $opts ) = @_;
888 throw( "Need GraphViz installed to output SVG" )
889 unless File::Which::which( 'dot' );
890 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
891 $self->calculate_ranks()
892 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
893 my @cmd = qw/dot -Tsvg/;
895 my $dotfile = File::Temp->new();
897 # $dotfile->unlink_on_destroy(0);
898 binmode $dotfile, ':utf8';
899 print $dotfile $self->as_dot( $opts );
900 push( @cmd, $dotfile->filename );
901 run( \@cmd, ">", binary(), \$svg );
902 $svg = decode_utf8( $svg );
907 =head2 as_dot( \%options )
909 Returns a string that is the collation graph expressed in dot
910 (i.e. GraphViz) format. Options include:
925 my( $self, $opts ) = @_;
926 my $startrank = $opts->{'from'} if $opts;
927 my $endrank = $opts->{'to'} if $opts;
928 my $color_common = $opts->{'color_common'} if $opts;
929 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
930 && $self->end->rank > 100;
931 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
933 # Check the arguments
935 return if $endrank && $startrank > $endrank;
936 return if $startrank > $self->end->rank;
938 if( defined $endrank ) {
939 return if $endrank < 0;
940 $endrank = undef if $endrank == $self->end->rank;
943 my $graph_name = $self->tradition->name;
944 $graph_name =~ s/[^\w\s]//g;
945 $graph_name = join( '_', split( /\s+/, $graph_name ) );
953 'fillcolor' => 'white',
958 'arrowhead' => 'open',
959 'color' => '#000000',
960 'fontcolor' => '#000000',
963 my $dot = sprintf( "digraph %s {\n", $graph_name );
964 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
965 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
967 # Output substitute start/end readings if necessary
969 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
972 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
974 if( $STRAIGHTENHACK ) {
976 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
977 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
978 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
980 my %used; # Keep track of the readings that actually appear in the graph
981 # Sort the readings by rank if we have ranks; this speeds layout.
982 my @all_readings = $self->end->has_rank
983 ? sort { $a->rank <=> $b->rank } $self->readings
985 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
986 foreach my $reading ( @all_readings ) {
987 # Only output readings within our rank range.
988 next if $startrank && $reading->rank < $startrank;
989 next if $endrank && $reading->rank > $endrank;
990 $used{$reading->id} = 1;
991 # Need not output nodes without separate labels
992 next if $reading->id eq $reading->text;
994 my $label = $reading->text;
995 $label .= '-' if $reading->join_next;
996 $label = "-$label" if $reading->join_prior;
997 $label =~ s/\"/\\\"/g;
998 $rattrs->{'label'} = $label;
999 $rattrs->{'id'} = $reading->id;
1000 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
1001 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
1004 # Add the real edges. Need to weight one edge per rank jump, in a
1006 # my $weighted = $self->_add_edge_weights;
1007 my @edges = $self->paths;
1008 my( %substart, %subend );
1009 foreach my $edge ( @edges ) {
1010 # Do we need to output this edge?
1011 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
1012 my $label = $self->_path_display_label( $opts,
1013 $self->path_witnesses( $edge ) );
1014 my $variables = { %edge_attrs, 'label' => $label };
1016 # Account for the rank gap if necessary
1017 my $rank0 = $self->reading( $edge->[0] )->rank
1018 if $self->reading( $edge->[0] )->has_rank;
1019 my $rank1 = $self->reading( $edge->[1] )->rank
1020 if $self->reading( $edge->[1] )->has_rank;
1021 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
1022 $variables->{'minlen'} = $rank1 - $rank0;
1025 # Add the calculated edge weights
1026 # if( exists $weighted->{$edge->[0]}
1027 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
1028 # # $variables->{'color'} = 'red';
1029 # $variables->{'weight'} = 3.0;
1032 # EXPERIMENTAL: make edge width reflect no. of witnesses
1033 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
1034 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
1036 my $varopts = _dot_attr_string( $variables );
1037 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
1038 $edge->[0], $edge->[1], $varopts );
1039 } elsif( $used{$edge->[0]} ) {
1040 $subend{$edge->[0]} = $edge->[1];
1041 } elsif( $used{$edge->[1]} ) {
1042 $substart{$edge->[1]} = $edge->[0];
1046 # If we are asked to, add relationship links
1047 if( exists $opts->{show_relations} ) {
1048 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
1049 if( $filter eq 'transposition' ) {
1050 $filter =~ qr/^transposition$/;
1053 my @types = sort( map { $_->name } $self->relations->types );
1054 if( exists $opts->{graphcolors} ) {
1055 foreach my $tdx ( 0 .. $#types ) {
1056 $typecolors{$types[$tdx]} = $opts->{graphcolors}->[$tdx];
1059 map { $typecolors{$_} = '#FFA14F' } @types;
1061 foreach my $redge ( $self->relationships ) {
1062 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
1063 my $rel = $self->get_relationship( $redge );
1064 next unless $filter eq 'all' || $rel->type =~ /$filter/;
1066 arrowhead => 'none',
1067 color => $typecolors{$rel->type},
1068 constraint => 'false',
1071 unless( exists $opts->{graphcolors} ) {
1072 $variables->{label} = uc( substr( $rel->type, 0, 4 ) ),
1074 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
1075 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
1080 # Add substitute start and end edges if necessary
1081 foreach my $node ( keys %substart ) {
1082 my $witstr = $self->_path_display_label( $opts,
1083 $self->path_witnesses( $substart{$node}, $node ) );
1084 my $variables = { %edge_attrs, 'label' => $witstr };
1085 my $nrdg = $self->reading( $node );
1086 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
1087 # Substart is actually one lower than $startrank
1088 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
1090 my $varopts = _dot_attr_string( $variables );
1091 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
1093 foreach my $node ( keys %subend ) {
1094 my $witstr = $self->_path_display_label( $opts,
1095 $self->path_witnesses( $node, $subend{$node} ) );
1096 my $variables = { %edge_attrs, 'label' => $witstr };
1097 my $varopts = _dot_attr_string( $variables );
1098 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
1101 if( $STRAIGHTENHACK ) {
1102 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
1103 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
1110 sub _dot_attr_string {
1113 foreach my $k ( sort keys %$hash ) {
1114 my $v = $hash->{$k};
1115 push( @attrs, $k.'="'.$v.'"' );
1117 return( '[ ' . join( ', ', @attrs ) . ' ]' );
1120 sub _add_edge_weights {
1122 # Walk the graph from START to END, choosing the successor node with
1123 # the largest number of witness paths each time.
1125 my $curr = $self->start->id;
1126 my $ranked = $self->end->has_rank;
1127 while( $curr ne $self->end->id ) {
1128 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
1129 my @succ = sort { $self->path_witnesses( $curr, $a )
1130 <=> $self->path_witnesses( $curr, $b ) }
1131 $self->sequence->successors( $curr );
1132 my $next = pop @succ;
1133 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
1134 # Try to avoid lacunae in the weighted path.
1136 ( $self->reading( $next )->is_lacuna ||
1137 $nextrank - $rank > 1 ) ){
1140 $weighted->{$curr} = $next;
1146 =head2 path_witnesses( $edge )
1148 Returns the list of sigils whose witnesses are associated with the given edge.
1149 The edge can be passed as either an array or an arrayref of ( $source, $target ).
1153 sub path_witnesses {
1154 my( $self, @edge ) = @_;
1155 # If edge is an arrayref, cope.
1156 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
1157 my $e = shift @edge;
1160 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
1164 # Helper function. Make a display label for the given witnesses, showing a.c.
1165 # witnesses only where the main witness is not also in the list.
1166 sub _path_display_label {
1170 map { $wits{$_} = 1 } @_;
1172 # If an a.c. wit is listed, remove it if the main wit is also listed.
1173 # Otherwise keep it for explicit listing.
1174 my $aclabel = $self->ac_label;
1176 foreach my $w ( sort keys %wits ) {
1177 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
1178 if( exists $wits{$1} ) {
1181 push( @disp_ac, $w );
1186 if( $opts->{'explicit_wits'} ) {
1187 return join( ', ', sort keys %wits );
1189 # See if we are in a majority situation.
1190 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
1191 $maj = $maj > 5 ? $maj : 5;
1192 if( scalar keys %wits > $maj ) {
1193 unshift( @disp_ac, 'majority' );
1194 return join( ', ', @disp_ac );
1196 return join( ', ', sort keys %wits );
1201 =head2 readings_at_rank( $rank )
1203 Returns a list of readings at a given rank, taken from the alignment table.
1207 sub readings_at_rank {
1208 my( $self, $rank ) = @_;
1209 my $table = $self->alignment_table;
1210 # Table rank is real rank - 1.
1211 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
1213 foreach my $e ( @elements ) {
1214 next unless ref( $e ) eq 'HASH';
1215 next unless exists $e->{'t'};
1216 $readings{$e->{'t'}->id} = $e->{'t'};
1218 return values %readings;
1223 Returns a GraphML representation of the collation. The GraphML will contain
1224 two graphs. The first expresses the attributes of the readings and the witness
1225 paths that link them; the second expresses the relationships that link the
1226 readings. This is the native transfer format for a tradition.
1230 use Text::Tradition;
1236 my $datafile = 't/data/florilegium_tei_ps.xml';
1237 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1239 'file' => $datafile,
1242 ok( $tradition, "Got a tradition object" );
1243 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
1244 ok( $tradition->collation, "Tradition has a collation" );
1246 my $c = $tradition->collation;
1247 is( scalar $c->readings, $READINGS, "Collation has all readings" );
1248 is( scalar $c->paths, $PATHS, "Collation has all paths" );
1249 is( scalar $c->relationships, 0, "Collation has all relationships" );
1251 # Add a few relationships
1252 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
1253 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
1254 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
1256 # Now write it to GraphML and parse it again.
1258 my $graphml = $c->as_graphml;
1259 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
1260 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
1261 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
1262 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
1264 # Now add a stemma, write to GraphML, and look at the output.
1266 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
1267 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
1268 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
1269 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
1270 $graphml = $c->as_graphml;
1271 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
1278 ## TODO MOVE this to Tradition.pm and modularize it better
1280 my( $self, $options ) = @_;
1281 $self->calculate_ranks unless $self->_graphcalc_done;
1283 my $start = $options->{'from'}
1284 ? $self->reading( $options->{'from'} ) : $self->start;
1285 my $end = $options->{'to'}
1286 ? $self->reading( $options->{'to'} ) : $self->end;
1287 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1288 throw( 'Start node must be before end node' );
1290 # The readings need to be ranked for this to work.
1291 $start = $self->start unless $start->has_rank;
1292 $end = $self->end unless $end->has_rank;
1294 unless( $start eq $self->start ) {
1295 $rankoffset = $start->rank - 1;
1300 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1301 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1302 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1303 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1305 # Create the document and root node
1306 require XML::LibXML;
1307 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1308 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1309 $graphml->setDocumentElement( $root );
1310 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1311 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1313 # List of attribute types to save on our objects and their corresponding
1318 'Bool' => 'boolean',
1319 'ReadingID' => 'string',
1320 'RelationshipType' => 'string',
1321 'RelationshipScope' => 'string',
1324 # Add the data keys for the graph. Include an extra key 'version' for the
1325 # GraphML output version.
1326 my %graph_data_keys;
1328 my %graph_attributes = ( 'version' => 'string' );
1329 # Graph attributes include those of Tradition and those of Collation.
1331 # TODO Use meta introspection method from duplicate_reading to do this
1332 # instead of naming custom keys.
1333 my $tmeta = $self->tradition->meta;
1334 my $cmeta = $self->meta;
1335 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1336 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1337 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1338 next if $attr->name =~ /^_/;
1339 next unless $save_types{$attr->type_constraint->name};
1340 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1342 # Extra custom keys for complex objects that should be saved in some form.
1343 # The subroutine should return a string, or undef/empty.
1344 if( $tmeta->has_method('stemmata') ) {
1345 $graph_attributes{'stemmata'} = sub {
1347 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1348 $self->tradition->stemmata;
1349 join( "\n", @stemstrs );
1353 if( $tmeta->has_method('user') ) {
1354 $graph_attributes{'user'} = sub {
1355 $self->tradition->user ? $self->tradition->user->id : undef
1359 foreach my $datum ( sort keys %graph_attributes ) {
1360 $graph_data_keys{$datum} = 'dg'.$gdi++;
1361 my $key = $root->addNewChild( $graphml_ns, 'key' );
1362 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1363 : $graph_attributes{$datum};
1364 $key->setAttribute( 'attr.name', $datum );
1365 $key->setAttribute( 'attr.type', $dtype );
1366 $key->setAttribute( 'for', 'graph' );
1367 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1370 # Add the data keys for reading nodes
1371 my %reading_attributes;
1372 my $rmeta = Text::Tradition::Collation::Reading->meta;
1373 foreach my $attr( $rmeta->get_all_attributes ) {
1374 next if $attr->name =~ /^_/;
1375 next unless $save_types{$attr->type_constraint->name};
1376 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1378 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1379 # Extra custom key for the reading morphology
1380 $reading_attributes{'lexemes'} = 'string';
1385 foreach my $datum ( sort keys %reading_attributes ) {
1386 $node_data_keys{$datum} = 'dn'.$ndi++;
1387 my $key = $root->addNewChild( $graphml_ns, 'key' );
1388 $key->setAttribute( 'attr.name', $datum );
1389 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1390 $key->setAttribute( 'for', 'node' );
1391 $key->setAttribute( 'id', $node_data_keys{$datum} );
1394 # Add the data keys for edges, that is, paths and relationships. Path
1395 # data does not come from a Moose class so is here manually.
1398 my %edge_attributes = (
1399 witness => 'string', # ID/label for a path
1400 extra => 'boolean', # Path key
1402 my @path_attributes = keys %edge_attributes; # track our manual additions
1403 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1404 foreach my $attr( $pmeta->get_all_attributes ) {
1405 next if $attr->name =~ /^_/;
1406 next unless $save_types{$attr->type_constraint->name};
1407 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1409 foreach my $datum ( sort keys %edge_attributes ) {
1410 $edge_data_keys{$datum} = 'de'.$edi++;
1411 my $key = $root->addNewChild( $graphml_ns, 'key' );
1412 $key->setAttribute( 'attr.name', $datum );
1413 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1414 $key->setAttribute( 'for', 'edge' );
1415 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1418 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1419 my $xmlidname = $self->tradition->name;
1420 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1421 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1422 $xmlidname = '_'.$xmlidname;
1424 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1425 $sgraph->setAttribute( 'edgedefault', 'directed' );
1426 $sgraph->setAttribute( 'id', $xmlidname );
1427 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1428 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1429 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1430 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1431 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1433 # Tradition/collation attribute data
1434 foreach my $datum ( keys %graph_attributes ) {
1436 if( $datum eq 'version' ) {
1438 } elsif( ref( $graph_attributes{$datum} ) ) {
1439 my $sub = $graph_attributes{$datum};
1441 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1442 $value = $self->tradition->$datum;
1444 $value = $self->$datum;
1446 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1451 # Add our readings to the graph
1452 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1453 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1454 ( $n->rank < $start->rank || $n->rank > $end->rank );
1455 $use_readings{$n->id} = 1;
1456 # Add to the main graph
1457 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1458 my $node_xmlid = 'n' . $node_ctr++;
1459 $node_hash{ $n->id } = $node_xmlid;
1460 $node_el->setAttribute( 'id', $node_xmlid );
1461 foreach my $d ( keys %reading_attributes ) {
1463 # Custom serialization
1464 if( $d eq 'lexemes' ) {
1465 # If nval is a true value, we have lexemes so we need to
1466 # serialize them. Otherwise set nval to undef so that the
1467 # key is excluded from this reading.
1468 $nval = $nval ? $n->_serialize_lexemes : undef;
1469 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1472 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1473 # Adjust the ranks within the subgraph.
1474 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1475 : $nval - $rankoffset;
1477 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1482 # Add the path edges to the sequence graph
1484 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1485 # We add an edge in the graphml for every witness in $e.
1486 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1487 my @edge_wits = sort $self->path_witnesses( $e );
1488 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1489 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1490 # Skip any path from start to end; that witness is not in the subgraph.
1491 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1492 foreach my $wit ( @edge_wits ) {
1493 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1494 $node_hash{ $e->[0] },
1495 $node_hash{ $e->[1] } );
1496 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1497 $edge_el->setAttribute( 'source', $from );
1498 $edge_el->setAttribute( 'target', $to );
1499 $edge_el->setAttribute( 'id', $id );
1501 # It's a witness path, so add the witness
1503 my $key = $edge_data_keys{'witness'};
1504 # Is this an ante-corr witness?
1505 my $aclabel = $self->ac_label;
1506 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1507 # Keep the base witness
1509 # ...and record that this is an 'extra' reading path
1510 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1512 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1516 # Report the actual number of nodes and edges that went in
1517 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1518 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1520 # Add the relationship graph to the XML
1521 map { delete $edge_data_keys{$_} } @path_attributes;
1522 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1523 $node_data_keys{'id'}, \%edge_data_keys );
1525 # Save and return the thing
1526 my $result = decode_utf8( $graphml->toString(1) );
1530 sub _add_graphml_data {
1531 my( $el, $key, $value ) = @_;
1532 return unless defined $value;
1533 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1534 $data_el->setAttribute( 'key', $key );
1535 $data_el->appendText( $value );
1540 Returns a CSV alignment table representation of the collation graph, one
1541 row per witness (or witness uncorrected.)
1545 Returns a tab-separated alignment table representation of the collation graph,
1546 one row per witness (or witness uncorrected.)
1550 use Text::Tradition;
1558 my $datafile = 't/data/florilegium_tei_ps.xml';
1559 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1561 'file' => $datafile,
1564 my $c = $tradition->collation;
1565 # Export the thing to CSV
1566 my $csvstr = $c->as_csv();
1568 my $csv = Text::CSV->new({ sep_char => ',', binary => 1 });
1569 my @lines = split(/\n/, $csvstr );
1570 ok( $csv->parse( $lines[0] ), "Successfully parsed first line of CSV" );
1571 is( scalar( $csv->fields ), $WITS + $WITAC, "CSV has correct number of witness columns" );
1572 my @q_ac = grep { $_ eq 'Q'.$c->ac_label } $csv->fields;
1573 ok( @q_ac, "Found a layered witness" );
1575 my $t2 = Text::Tradition->new( input => 'Tabular',
1579 is( scalar $t2->collation->readings, $READINGS, "Reparsed CSV collation has all readings" );
1580 is( scalar $t2->collation->paths, $PATHS, "Reparsed CSV collation has all paths" );
1582 # Now do it with TSV
1583 my $tsvstr = $c->as_tsv();
1584 my $t3 = Text::Tradition->new( input => 'Tabular',
1588 is( scalar $t3->collation->readings, $READINGS, "Reparsed TSV collation has all readings" );
1589 is( scalar $t3->collation->paths, $PATHS, "Reparsed TSV collation has all paths" );
1591 my $table = $c->alignment_table;
1592 my $noaccsv = $c->as_csv({ noac => 1 });
1593 my @noaclines = split(/\n/, $noaccsv );
1594 ok( $csv->parse( $noaclines[0] ), "Successfully parsed first line of no-ac CSV" );
1595 is( scalar( $csv->fields ), $WITS, "CSV has correct number of witness columns" );
1596 is( $c->alignment_table, $table, "Request for CSV did not alter the alignment table" );
1598 my $safecsv = $c->as_csv({ safe_ac => 1});
1599 my @safelines = split(/\n/, $safecsv );
1600 ok( $csv->parse( $safelines[0] ), "Successfully parsed first line of safe CSV" );
1601 is( scalar( $csv->fields ), $WITS + $WITAC, "CSV has correct number of witness columns" );
1602 @q_ac = grep { $_ eq 'Q__L' } $csv->fields;
1603 ok( @q_ac, "Found a sanitized layered witness" );
1604 is( $c->alignment_table, $table, "Request for CSV did not alter the alignment table" );
1611 my( $self, $opts ) = @_;
1612 my $table = $self->alignment_table( $opts );
1613 my $csv_options = { binary => 1, quote_null => 0 };
1614 $csv_options->{'sep_char'} = $opts->{fieldsep};
1615 if( $opts->{fieldsep} eq "\t" ) {
1616 # If it is really tab separated, nothing is an escape char.
1617 $csv_options->{'quote_char'} = undef;
1618 $csv_options->{'escape_char'} = '';
1620 my $csv = Text::CSV->new( $csv_options );
1622 # Make the header row
1623 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1624 push( @result, $csv->string );
1625 # Make the rest of the rows
1626 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1627 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1628 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1629 $csv->combine( @row );
1630 push( @result, $csv->string );
1632 return join( "\n", @result );
1637 my $opts = shift || {};
1638 $opts->{fieldsep} = ',';
1639 return $self->_tabular( $opts );
1644 my $opts = shift || {};
1645 $opts->{fieldsep} = "\t";
1646 return $self->_tabular( $opts );
1649 =head2 alignment_table
1651 Return a reference to an alignment table, in a slightly enhanced CollateX
1652 format which looks like this:
1654 $table = { alignment => [ { witness => "SIGIL",
1655 tokens => [ { t => "TEXT" }, ... ] },
1656 { witness => "SIG2",
1657 tokens => [ { t => "TEXT" }, ... ] },
1659 length => TEXTLEN };
1663 sub alignment_table {
1664 my( $self, $opts ) = @_;
1665 if( $self->has_cached_table ) {
1666 return $self->cached_table
1667 unless $opts->{noac} || $opts->{safe_ac};
1670 # Make sure we can do this
1671 throw( "Need a linear graph in order to make an alignment table" )
1672 unless $self->linear;
1673 $self->calculate_ranks()
1674 unless $self->_graphcalc_done && $self->end->has_rank;
1676 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1677 my @all_pos = ( 1 .. $self->end->rank - 1 );
1678 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1679 # say STDERR "Making witness row(s) for " . $wit->sigil;
1680 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1681 my @row = _make_witness_row( \@wit_path, \@all_pos );
1682 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1683 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1684 push( @{$table->{'alignment'}}, $witobj );
1685 if( $wit->is_layered && !$opts->{noac} ) {
1686 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1687 $wit->sigil.$self->ac_label );
1688 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1689 my $witlabel = $opts->{safe_ac}
1690 ? $wit->sigil . '__L' : $wit->sigil.$self->ac_label;
1691 my $witacobj = { 'witness' => $witlabel,
1692 'tokens' => \@ac_row };
1693 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1694 push( @{$table->{'alignment'}}, $witacobj );
1697 unless( $opts->{noac} || $opts->{safe_ac} ) {
1698 $self->cached_table( $table );
1703 sub _make_witness_row {
1704 my( $path, $positions ) = @_;
1706 map { $char_hash{$_} = undef } @$positions;
1708 foreach my $rdg ( @$path ) {
1709 say STDERR "rank " . $rdg->rank if $debug;
1710 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1711 $char_hash{$rdg->rank} = { 't' => $rdg };
1713 my @row = map { $char_hash{$_} } @$positions;
1714 # Fill in lacuna markers for undef spots in the row
1715 my $last_el = shift @row;
1716 my @filled_row = ( $last_el );
1717 foreach my $el ( @row ) {
1718 # If we are using node reference, make the lacuna node appear many times
1719 # in the table. If not, use the lacuna tag.
1720 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1723 push( @filled_row, $el );
1730 =head1 NAVIGATION METHODS
1732 =head2 reading_sequence( $first, $last, $sigil, $backup )
1734 Returns the ordered list of readings, starting with $first and ending
1735 with $last, for the witness given in $sigil. If a $backup sigil is
1736 specified (e.g. when walking a layered witness), it will be used wherever
1737 no $sigil path exists. If there is a base text reading, that will be
1738 used wherever no path exists for $sigil or $backup.
1742 # TODO Think about returning some lazy-eval iterator.
1743 # TODO Get rid of backup; we should know from what witness is whether we need it.
1745 sub reading_sequence {
1746 my( $self, $start, $end, $witness ) = @_;
1748 $witness = $self->baselabel unless $witness;
1749 my @readings = ( $start );
1752 while( $n && $n->id ne $end->id ) {
1753 if( exists( $seen{$n->id} ) ) {
1754 throw( "Detected loop for $witness at " . $n->id );
1758 my $next = $self->next_reading( $n, $witness );
1760 throw( "Did not find any path for $witness from reading " . $n->id );
1762 push( @readings, $next );
1765 # Check that the last reading is our end reading.
1766 my $last = $readings[$#readings];
1767 throw( "Last reading found from " . $start->text .
1768 " for witness $witness is not the end!" ) # TODO do we get this far?
1769 unless $last->id eq $end->id;
1774 =head2 next_reading( $reading, $sigil );
1776 Returns the reading that follows the given reading along the given witness
1782 # Return the successor via the corresponding path.
1784 my $answer = $self->_find_linked_reading( 'next', @_ );
1785 return undef unless $answer;
1786 return $self->reading( $answer );
1789 =head2 prior_reading( $reading, $sigil )
1791 Returns the reading that precedes the given reading along the given witness
1797 # Return the predecessor via the corresponding path.
1799 my $answer = $self->_find_linked_reading( 'prior', @_ );
1800 return $self->reading( $answer );
1803 sub _find_linked_reading {
1804 my( $self, $direction, $node, $path ) = @_;
1806 # Get a backup if we are dealing with a layered witness
1808 my $aclabel = $self->ac_label;
1809 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1813 my @linked_paths = $direction eq 'next'
1814 ? $self->sequence->edges_from( $node )
1815 : $self->sequence->edges_to( $node );
1816 return undef unless scalar( @linked_paths );
1818 # We have to find the linked path that contains all of the
1819 # witnesses supplied in $path.
1820 my( @path_wits, @alt_path_wits );
1821 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1822 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1825 foreach my $le ( @linked_paths ) {
1826 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1829 my @le_wits = sort $self->path_witnesses( $le );
1830 if( _is_within( \@path_wits, \@le_wits ) ) {
1831 # This is the right path.
1832 return $direction eq 'next' ? $le->[1] : $le->[0];
1833 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1837 # Got this far? Return the alternate path if it exists.
1838 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1841 # Got this far? Return the base path if it exists.
1842 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1845 # Got this far? We have no appropriate path.
1846 warn "Could not find $direction node from " . $node->id
1847 . " along path $path";
1853 my( $set1, $set2 ) = @_;
1854 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1855 foreach my $el ( @$set1 ) {
1856 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1861 # Return the string that joins together a list of witnesses for
1862 # display on a single path.
1863 sub _witnesses_of_label {
1864 my( $self, $label ) = @_;
1865 my $regex = $self->wit_list_separator;
1866 my @answer = split( /\Q$regex\E/, $label );
1870 =head2 common_readings
1872 Returns the list of common readings in the graph (i.e. those readings that are
1873 shared by all non-lacunose witnesses.)
1877 sub common_readings {
1879 my @common = grep { $_->is_common } $self->readings;
1883 =head2 path_text( $sigil, [, $start, $end ] )
1885 Returns the text of a witness (plus its backup, if we are using a layer)
1886 as stored in the collation. The text is returned as a string, where the
1887 individual readings are joined with spaces and the meta-readings (e.g.
1888 lacunae) are omitted. Optional specification of $start and $end allows
1889 the generation of a subset of the witness text.
1894 my( $self, $wit, $start, $end ) = @_;
1895 $start = $self->start unless $start;
1896 $end = $self->end unless $end;
1897 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1900 foreach my $r ( @path ) {
1901 unless ( $r->join_prior || !$last || $last->join_next ) {
1904 $pathtext .= $r->text;
1910 =head1 INITIALIZATION METHODS
1912 These are mostly for use by parsers.
1914 =head2 make_witness_path( $witness )
1916 Link the array of readings contained in $witness->path (and in
1917 $witness->uncorrected_path if it exists) into collation paths.
1918 Clear out the arrays when finished.
1920 =head2 make_witness_paths
1922 Call make_witness_path for all witnesses in the tradition.
1926 # For use when a collation is constructed from a base text and an apparatus.
1927 # We have the sequences of readings and just need to add path edges.
1928 # When we are done, clear out the witness path attributes, as they are no
1930 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1932 sub make_witness_paths {
1934 foreach my $wit ( $self->tradition->witnesses ) {
1935 # say STDERR "Making path for " . $wit->sigil;
1936 $self->make_witness_path( $wit );
1940 sub make_witness_path {
1941 my( $self, $wit ) = @_;
1942 my @chain = @{$wit->path};
1943 my $sig = $wit->sigil;
1944 # Add start and end if necessary
1945 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1946 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1947 foreach my $idx ( 0 .. $#chain-1 ) {
1948 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1950 if( $wit->is_layered ) {
1951 @chain = @{$wit->uncorrected_path};
1952 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1953 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1954 foreach my $idx( 0 .. $#chain-1 ) {
1955 my $source = $chain[$idx];
1956 my $target = $chain[$idx+1];
1957 $self->add_path( $source, $target, $sig.$self->ac_label )
1958 unless $self->has_path( $source, $target, $sig );
1962 $wit->clear_uncorrected_path;
1965 =head2 calculate_ranks
1967 Calculate the reading ranks (that is, their aligned positions relative
1968 to each other) for the graph. This can only be called on linear collations.
1972 use Text::Tradition;
1974 my $cxfile = 't/data/Collatex-16.xml';
1975 my $t = Text::Tradition->new(
1977 'input' => 'CollateX',
1980 my $c = $t->collation;
1983 my $table = $c->alignment_table;
1984 ok( $c->has_cached_table, "Alignment table was cached" );
1985 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1986 $c->calculate_ranks;
1987 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1988 $c->add_relationship( 'n13', 'n23', { type => 'repetition' } );
1989 is( $c->alignment_table, $table, "Alignment table unchanged after non-colo relationship add" );
1990 $c->add_relationship( 'n24', 'n23', { type => 'spelling' } );
1991 isnt( $c->alignment_table, $table, "Alignment table changed after colo relationship add" );
1997 sub calculate_ranks {
1999 # Save the existing ranks, in case we need to invalidate the cached SVG.
2000 throw( "Cannot calculate ranks on a non-linear graph" )
2001 unless $self->linear;
2003 map { $existing_ranks{$_} = $_->rank } $self->readings;
2005 # Do the rankings based on the relationship equivalence graph, starting
2006 # with the start node.
2007 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
2009 # Transfer our rankings from the topological graph to the real one.
2010 foreach my $r ( $self->readings ) {
2011 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
2012 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
2014 # Die. Find the last rank we calculated.
2015 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
2016 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
2018 my $last = pop @all_defined;
2019 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
2022 # Do we need to invalidate the cached data?
2023 if( $self->has_cached_table ) {
2024 foreach my $r ( $self->readings ) {
2025 next if defined( $existing_ranks{$r} )
2026 && $existing_ranks{$r} == $r->rank;
2027 # Something has changed, so clear the cache
2028 $self->_clear_cache;
2029 # ...and recalculate the common readings.
2030 $self->calculate_common_readings();
2034 # The graph calculation information is now up to date.
2035 $self->_graphcalc_done(1);
2040 $self->wipe_table if $self->has_cached_table;
2044 =head2 flatten_ranks
2046 A convenience method for parsing collation data. Searches the graph for readings
2047 with the same text at the same rank, and merges any that are found.
2052 my ( $self, %args ) = shift;
2053 my %unique_rank_rdg;
2055 foreach my $p ( $self->identical_readings( %args ) ) {
2056 # say STDERR "Combining readings at same rank: @$p";
2058 $self->merge_readings( @$p );
2059 # TODO see if this now makes a common point.
2061 # If we merged readings, the ranks are still fine but the alignment
2062 # table is wrong. Wipe it.
2063 $self->wipe_table() if $changed;
2066 =head2 identical_readings
2067 =head2 identical_readings( start => $startnode, end => $endnode )
2068 =head2 identical_readings( startrank => $startrank, endrank => $endrank )
2070 Goes through the graph identifying all pairs of readings that appear to be
2071 identical, and therefore able to be merged into a single reading. Returns the
2072 relevant identical pairs. Can be restricted to run over only a part of the
2073 graph, specified either by node or by rank.
2077 sub identical_readings {
2078 my ( $self, %args ) = @_;
2079 # Find where we should start and end.
2080 my $startrank = $args{startrank} || 0;
2081 if( $args{start} ) {
2082 throw( "Starting reading has no rank" ) unless $self->reading( $args{start} )
2083 && $self->reading( $args{start} )->has_rank;
2084 $startrank = $self->reading( $args{start} )->rank;
2086 my $endrank = $args{endrank} || $self->end->rank;
2088 throw( "Ending reading has no rank" ) unless $self->reading( $args{end} )
2089 && $self->reading( $args{end} )->has_rank;
2090 $endrank = $self->reading( $args{end} )->rank;
2093 # Make sure the ranks are correct.
2094 unless( $self->_graphcalc_done ) {
2095 $self->calculate_ranks;
2097 # Go through the readings looking for duplicates.
2098 my %unique_rank_rdg;
2100 foreach my $rdg ( $self->readings ) {
2101 next unless $rdg->has_rank;
2102 my $rk = $rdg->rank;
2103 next if $rk > $endrank || $rk < $startrank;
2104 my $key = $rk . "||" . $rdg->text;
2105 if( exists $unique_rank_rdg{$key} ) {
2106 # Make sure they don't have different grammatical forms
2107 my $ur = $unique_rank_rdg{$key};
2108 if( $rdg->is_identical( $ur ) ) {
2109 push( @pairs, [ $ur, $rdg ] );
2112 $unique_rank_rdg{$key} = $rdg;
2120 =head2 calculate_common_readings
2122 Goes through the graph identifying the readings that appear in every witness
2123 (apart from those with lacunae at that spot.) Marks them as common and returns
2128 use Text::Tradition;
2130 my $cxfile = 't/data/Collatex-16.xml';
2131 my $t = Text::Tradition->new(
2133 'input' => 'CollateX',
2136 my $c = $t->collation;
2138 my @common = $c->calculate_common_readings();
2139 is( scalar @common, 8, "Found correct number of common readings" );
2140 my @marked = sort $c->common_readings();
2141 is( scalar @common, 8, "All common readings got marked as such" );
2142 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
2143 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
2149 sub calculate_common_readings {
2152 map { $_->is_common( 0 ) } $self->readings;
2153 # Implicitly calls calculate_ranks
2154 my $table = $self->alignment_table;
2155 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
2156 my @row = map { $_->{'tokens'}->[$idx]
2157 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
2158 @{$table->{'alignment'}};
2160 foreach my $r ( @row ) {
2162 $hash{$r->id} = $r unless $r->is_meta;
2164 $hash{'UNDEF'} = $r;
2167 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
2168 my( $r ) = values %hash;
2170 push( @common, $r );
2176 =head2 text_from_paths
2178 Calculate the text array for all witnesses from the path, for later consistency
2179 checking. Only to be used if there is no non-graph-based way to know the
2184 sub text_from_paths {
2186 foreach my $wit ( $self->tradition->witnesses ) {
2187 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
2189 foreach my $r ( @readings ) {
2190 next if $r->is_meta;
2191 push( @text, $r->text );
2193 $wit->text( \@text );
2194 if( $wit->is_layered ) {
2195 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
2196 $wit->sigil.$self->ac_label );
2198 foreach my $r ( @ucrdgs ) {
2199 next if $r->is_meta;
2200 push( @uctext, $r->text );
2202 $wit->layertext( \@uctext );
2207 =head1 UTILITY FUNCTIONS
2209 =head2 common_predecessor( $reading_a, $reading_b )
2211 Find the last reading that occurs in sequence before both the given readings.
2212 At the very least this should be $self->start.
2214 =head2 common_successor( $reading_a, $reading_b )
2216 Find the first reading that occurs in sequence after both the given readings.
2217 At the very least this should be $self->end.
2221 use Text::Tradition;
2223 my $cxfile = 't/data/Collatex-16.xml';
2224 my $t = Text::Tradition->new(
2226 'input' => 'CollateX',
2229 my $c = $t->collation;
2231 is( $c->common_predecessor( 'n24', 'n23' )->id,
2232 'n20', "Found correct common predecessor" );
2233 is( $c->common_successor( 'n24', 'n23' )->id,
2234 '__END__', "Found correct common successor" );
2236 is( $c->common_predecessor( 'n19', 'n17' )->id,
2237 'n16', "Found correct common predecessor for readings on same path" );
2238 is( $c->common_successor( 'n21', 'n10' )->id,
2239 '__END__', "Found correct common successor for readings on same path" );
2245 ## Return the closest reading that is a predecessor of both the given readings.
2246 sub common_predecessor {
2248 my( $r1, $r2 ) = $self->_objectify_args( @_ );
2249 return $self->_common_in_path( $r1, $r2, 'predecessors' );
2252 sub common_successor {
2254 my( $r1, $r2 ) = $self->_objectify_args( @_ );
2255 return $self->_common_in_path( $r1, $r2, 'successors' );
2259 # TODO think about how to do this without ranks...
2260 sub _common_in_path {
2261 my( $self, $r1, $r2, $dir ) = @_;
2262 my $iter = $self->end->rank;
2264 my @last_r1 = ( $r1 );
2265 my @last_r2 = ( $r2 );
2266 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
2268 # say STDERR "Finding common $dir for $r1, $r2";
2269 while( !@candidates ) {
2270 last unless $iter--; # Avoid looping infinitely
2271 # Iterate separately down the graph from r1 and r2
2272 my( @new_lc1, @new_lc2 );
2273 foreach my $lc ( @last_r1 ) {
2274 foreach my $p ( $lc->$dir ) {
2275 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
2276 # say STDERR "Path candidate $p from $lc";
2277 push( @candidates, $p );
2278 } elsif( !$all_seen{$p->id} ) {
2279 $all_seen{$p->id} = 'r1';
2280 push( @new_lc1, $p );
2284 foreach my $lc ( @last_r2 ) {
2285 foreach my $p ( $lc->$dir ) {
2286 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
2287 # say STDERR "Path candidate $p from $lc";
2288 push( @candidates, $p );
2289 } elsif( !$all_seen{$p->id} ) {
2290 $all_seen{$p->id} = 'r2';
2291 push( @new_lc2, $p );
2295 @last_r1 = @new_lc1;
2296 @last_r2 = @new_lc2;
2298 my @answer = sort { $a->rank <=> $b->rank } @candidates;
2299 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
2303 Text::Tradition::Error->throw(
2304 'ident' => 'Collation error',
2310 __PACKAGE__->meta->make_immutable;
2316 =item * Rework XML serialization in a more modular way
2322 This package is free software and is provided "as is" without express
2323 or implied warranty. You can redistribute it and/or modify it under
2324 the same terms as Perl itself.
2328 Tara L Andrews E<lt>aurum@cpan.orgE<gt>