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 ### Reading construct/destruct functions
239 my( $self, $reading ) = @_;
240 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
241 my %args = %$reading;
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'};
246 } elsif( $self->tradition->can('language') && $self->tradition->has_language
247 && !exists $args{'language'} ) {
248 $args{'language'} = $self->tradition->language;
250 $reading = Text::Tradition::Collation::Reading->new(
251 'collation' => $self,
254 # First check to see if a reading with this ID exists.
255 if( $self->reading( $reading->id ) ) {
256 throw( "Collation already has a reading with id " . $reading->id );
258 $self->_graphcalc_done(0);
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 );
262 $self->relations->add_reading( $reading->id );
266 around del_reading => sub {
271 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
274 # Remove the reading from the graphs.
275 $self->_graphcalc_done(0);
276 $self->_clear_cache; # Explicitly clear caches to GC the reading
277 $self->sequence->delete_vertex( $arg );
278 $self->relations->delete_reading( $arg );
281 $self->$orig( $arg );
284 =head2 merge_readings( $main, $second, $concatenate, $with_str )
286 Merges the $second reading into the $main one. If $concatenate is true, then
287 the merged node will carry the text of both readings, concatenated with either
288 $with_str (if specified) or a sensible default (the empty string if the
289 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
291 The first two arguments may be either readings or reading IDs.
297 my $cxfile = 't/data/Collatex-16.xml';
298 my $t = Text::Tradition->new(
300 'input' => 'CollateX',
303 my $c = $t->collation;
305 my $rno = scalar $c->readings;
306 # Split n21 ('unto') for testing purposes
307 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
308 my $old_r = $c->reading( 'n21' );
309 $old_r->alter_text( 'to' );
310 $c->del_path( 'n20', 'n21', 'A' );
311 $c->add_path( 'n20', 'n21p0', 'A' );
312 $c->add_path( 'n21p0', 'n21', 'A' );
313 $c->add_relationship( 'n21', 'n22', { type => 'collated', scope => 'local' } );
315 ok( $c->reading( 'n21p0' ), "New reading exists" );
316 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
318 # Combine n3 and n4 ( with his )
319 $c->merge_readings( 'n3', 'n4', 1 );
320 ok( !$c->reading('n4'), "Reading n4 is gone" );
321 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
323 # Collapse n9 and n10 ( rood / root )
324 $c->merge_readings( 'n9', 'n10' );
325 ok( !$c->reading('n10'), "Reading n10 is gone" );
326 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
328 # Combine n21 and n21p0
329 my $remaining = $c->reading('n21');
330 $remaining ||= $c->reading('n22'); # one of these should still exist
331 $c->merge_readings( 'n21p0', $remaining, 1 );
332 ok( !$c->reading('n21'), "Reading $remaining is gone" );
333 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
343 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
344 my $mergemeta = $kept_obj->is_meta;
345 throw( "Cannot merge meta and non-meta reading" )
346 unless ( $mergemeta && $del_obj->is_meta )
347 || ( !$mergemeta && !$del_obj->is_meta );
349 throw( "Cannot merge with start or end node" )
350 if( $kept_obj eq $self->start || $kept_obj eq $self->end
351 || $del_obj eq $self->start || $del_obj eq $self->end );
352 throw( "Cannot combine text of meta readings" ) if $combine;
354 # We only need the IDs for adding paths to the graph, not the reading
355 # objects themselves.
356 my $kept = $kept_obj->id;
357 my $deleted = $del_obj->id;
358 $self->_graphcalc_done(0);
360 # The kept reading should inherit the paths and the relationships
361 # of the deleted reading.
362 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
363 my @vector = ( $kept );
364 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
365 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
366 next if $vector[0] eq $vector[1]; # Don't add a self loop
367 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
368 $self->sequence->add_edge( @vector );
369 my $fwits = $self->sequence->get_edge_attributes( @vector );
370 @wits{keys %$fwits} = values %$fwits;
371 $self->sequence->set_edge_attributes( @vector, \%wits );
373 $self->relations->merge_readings( $kept, $deleted, $combine );
375 # Do the deletion deed.
377 # Combine the text of the readings
378 my $joinstr = $combine_char;
379 unless( defined $joinstr ) {
380 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
381 $joinstr = $self->wordsep unless defined $joinstr;
383 $kept_obj->_combine( $del_obj, $joinstr );
385 $self->del_reading( $deleted );
388 =head2 compress_readings
390 Where possible in the graph, compresses plain sequences of readings into a
391 single reading. The sequences must consist of readings with no
392 relationships to other readings, with only a single witness path between
393 them and no other witness paths from either that would skip the other. The
394 readings must also not be marked as nonsense or bad grammar.
396 WARNING: This operation cannot be undone.
400 sub compress_readings {
402 # Anywhere in the graph that there is a reading that joins only to a single
403 # successor, and neither of these have any relationships, just join the two
405 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
406 # Now look for readings that can be joined to their successors.
407 next unless $rdg->is_combinable;
409 while( $self->sequence->successors( $rdg ) == 1 ) {
410 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
411 throw( "Infinite loop" ) if $seen{$next->id};
412 $seen{$next->id} = 1;
413 last if $self->sequence->predecessors( $next ) > 1;
414 last unless $next->is_combinable;
415 say "Joining readings $rdg and $next";
416 $self->merge_readings( $rdg, $next, 1 );
419 # Make sure we haven't screwed anything up
420 foreach my $wit ( $self->tradition->witnesses ) {
421 my $pathtext = $self->path_text( $wit->sigil );
422 my $origtext = join( ' ', @{$wit->text} );
423 throw( "Text differs for witness " . $wit->sigil )
424 unless $pathtext eq $origtext;
425 if( $wit->is_layered ) {
426 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
427 $origtext = join( ' ', @{$wit->layertext} );
428 throw( "Ante-corr text differs for witness " . $wit->sigil )
429 unless $pathtext eq $origtext;
433 $self->relations->rebuild_equivalence();
434 $self->calculate_ranks();
437 # Helper function for manipulating the graph.
438 sub _stringify_args {
439 my( $self, $first, $second, @args ) = @_;
441 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
442 $second = $second->id
443 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
444 return( $first, $second, @args );
447 # Helper function for manipulating the graph.
448 sub _objectify_args {
449 my( $self, $first, $second, $arg ) = @_;
450 $first = $self->reading( $first )
451 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
452 $second = $self->reading( $second )
453 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
454 return( $first, $second, $arg );
457 =head2 duplicate_reading( $reading, @witlist )
459 Split the given reading into two, so that the new reading is in the path for
460 the witnesses given in @witlist. If the result is that certain non-colocated
461 relationships (e.g. transpositions) are no longer valid, these will be removed.
462 Returns the newly-created reading.
468 my $st = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/collatecorr.xml' );
469 is( ref( $st ), 'Text::Tradition', "Got a tradition from test file" );
470 ok( $st->has_witness('Ba96'), "Tradition has the affected witness" );
472 my $sc = $st->collation;
474 ok( $sc->reading('n131'), "Tradition has the affected reading" );
475 is( scalar( $sc->readings ), $numr, "There are $numr readings in the graph" );
476 is( $sc->end->rank, 14, "There are fourteen ranks in the graph" );
478 # Detach the erroneously collated reading
479 my $newr = $sc->duplicate_reading( 'n131', 'Ba96' );
480 ok( $newr, "New reading was created" );
481 ok( $sc->reading('n131_0'), "Detached the bad collation with a new reading" );
482 is( scalar( $sc->readings ), $numr + 1, "A reading was added to the graph" );
483 is( $sc->end->rank, 10, "There are now only ten ranks in the graph" );
484 my $csucc = $sc->common_successor( 'n131', 'n131_0' );
485 is( $csucc->id, 'n136', "Found correct common successor to duped reading" );
487 # Check that the bad transposition is gone
488 is( $sc->get_relationship( 'n130', 'n135' ), undef, "Bad transposition relationship is gone" );
491 ok( $sc->merge_readings( 'n124', 'n131_0' ), "Collated the readings correctly" );
492 my @pairs = $sc->identical_readings( start => 'n124', end => $csucc->id );
493 is( $sc->end->rank, 11, "The ranks shifted appropriately" );
494 is( scalar @pairs, 3, "Found three more identical readings" );
495 $sc->flatten_ranks();
496 is( scalar( $sc->readings ), $numr - 3, "Now we are collated correctly" );
502 sub duplicate_reading {
503 my( $self, $r, @wits ) = @_;
504 # Add the new reading, duplicating $r.
505 unless( ref( $r ) eq 'Text::Tradition::Collation::Reading' ) {
506 $r = $self->reading( $r );
508 throw( "Cannot duplicate a meta-reading" )
511 # Get all the reading attributes and duplicate them.
512 my $rmeta = Text::Tradition::Collation::Reading->meta;
514 foreach my $attr( $rmeta->get_all_attributes ) {
515 next if $attr->name =~ /^_/;
516 my $acc = $attr->get_read_method;
517 if( !$acc && $attr->has_applied_traits ) {
518 my $tr = $attr->applied_traits;
519 if( $tr->[0] =~ /::(Array|Hash)$/ ) {
521 my %methods = reverse %{$attr->handles};
522 $acc = $methods{elements};
523 $args{$attr->name} = $which eq 'Array'
524 ? [ $r->$acc ] : { $r->$acc };
527 $args{$attr->name} = $r->$acc if $acc;
530 # By definition the new reading will no longer be common.
531 $args{is_common} = 0;
532 # The new reading also needs its own ID.
533 $args{id} = $self->_generate_dup_id( $r->id );
535 # Try to make the new reading.
536 my $newr = $self->add_reading( \%args );
537 # The old reading is also no longer common.
540 # For each of the witnesses, dissociate from the old reading and
541 # associate with the new.
542 foreach my $wit ( @wits ) {
543 my $prior = $self->prior_reading( $r, $wit );
544 my $next = $self->next_reading( $r, $wit );
545 $self->del_path( $prior, $r, $wit );
546 $self->add_path( $prior, $newr, $wit );
547 $self->del_path( $r, $next, $wit );
548 $self->add_path( $newr, $next, $wit );
551 # If the graph is ranked, we need to look for relationships that are now
552 # invalid (i.e. 'non-colocation' types that might now be colocated) and
553 # remove them. If not, we can skip it.
556 if( $self->end->has_rank ) {
557 # Find the point where we can stop checking
558 $succ = $self->common_successor( $r, $newr );
560 # Hash the existing ranks
561 foreach my $rdg ( $self->readings ) {
562 $rrk{$rdg->id} = $rdg->rank;
564 # Calculate the new ranks
565 $self->calculate_ranks();
567 # Check for invalid non-colocated relationships among changed-rank readings
568 # from where the ranks start changing up to $succ
569 my $lastrank = $succ->rank;
570 foreach my $rdg ( $self->readings ) {
571 next if $rdg->rank > $lastrank;
572 next if $rdg->rank == $rrk{$rdg->id};
573 my @noncolo = $rdg->related_readings( sub { !$_[0]->colocated } );
574 next unless @noncolo;
575 foreach my $nc ( @noncolo ) {
576 $self->relations->verify_or_delete( $rdg, $nc );
583 sub _generate_dup_id {
584 my( $self, $rid ) = @_;
589 if( $self->has_reading( $newid ) ) {
602 # We only need the IDs for adding paths to the graph, not the reading
603 # objects themselves.
604 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
606 $self->_graphcalc_done(0);
607 # Connect the readings
608 unless( $self->sequence->has_edge( $source, $target ) ) {
609 $self->sequence->add_edge( $source, $target );
610 $self->relations->add_equivalence_edge( $source, $target );
612 # Note the witness in question
613 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
619 if( ref( $_[0] ) eq 'ARRAY' ) {
626 # We only need the IDs for removing paths from the graph, not the reading
627 # objects themselves.
628 my( $source, $target, $wit ) = $self->_stringify_args( @args );
630 $self->_graphcalc_done(0);
631 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
632 $self->sequence->delete_edge_attribute( $source, $target, $wit );
634 unless( $self->sequence->has_edge_attributes( $source, $target ) ) {
635 $self->sequence->delete_edge( $source, $target );
636 $self->relations->delete_equivalence_edge( $source, $target );
641 # Extra graph-alike utility
644 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
645 return undef unless $self->sequence->has_edge( $source, $target );
646 return $self->sequence->has_edge_attribute( $source, $target, $wit );
649 =head2 clear_witness( @sigil_list )
651 Clear the given witnesses out of the collation entirely, removing references
652 to them in paths, and removing readings that belong only to them. Should only
653 be called via $tradition->del_witness.
658 my( $self, @sigils ) = @_;
660 $self->_graphcalc_done(0);
661 # Clear the witness(es) out of the paths
662 foreach my $e ( $self->paths ) {
663 foreach my $sig ( @sigils ) {
664 $self->del_path( $e, $sig );
668 # Clear out the newly unused readings
669 foreach my $r ( $self->readings ) {
670 unless( $self->reading_witnesses( $r ) ) {
671 $self->del_reading( $r );
676 sub add_relationship {
678 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
679 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
680 foreach my $v ( @vectors ) {
681 next unless $self->get_relationship( $v )->colocated;
682 if( $self->reading( $v->[0] )->has_rank && $self->reading( $v->[1] )->has_rank
683 && $self->reading( $v->[0] )->rank ne $self->reading( $v->[1] )->rank ) {
684 $self->_graphcalc_done(0);
692 around qw/ get_relationship del_relationship / => sub {
696 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
699 my( $source, $target ) = $self->_stringify_args( @args );
700 $self->$orig( $source, $target );
703 =head2 reading_witnesses( $reading )
705 Return a list of sigils corresponding to the witnesses in which the reading appears.
709 sub reading_witnesses {
710 my( $self, $reading ) = @_;
711 # We need only check either the incoming or the outgoing edges; I have
712 # arbitrarily chosen "incoming". Thus, special-case the start node.
713 if( $reading eq $self->start ) {
714 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
717 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
718 my $wits = $self->sequence->get_edge_attributes( @$e );
719 @all_witnesses{ keys %$wits } = 1;
721 my $acstr = $self->ac_label;
722 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
723 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
725 return keys %all_witnesses;
728 =head1 OUTPUT METHODS
730 =head2 as_svg( \%options )
732 Returns an SVG string that represents the graph, via as_dot and graphviz.
733 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
738 my( $self, $opts ) = @_;
739 throw( "Need GraphViz installed to output SVG" )
740 unless File::Which::which( 'dot' );
741 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
742 $self->calculate_ranks()
743 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
744 my @cmd = qw/dot -Tsvg/;
746 my $dotfile = File::Temp->new();
748 # $dotfile->unlink_on_destroy(0);
749 binmode $dotfile, ':utf8';
750 print $dotfile $self->as_dot( $opts );
751 push( @cmd, $dotfile->filename );
752 run( \@cmd, ">", binary(), \$svg );
753 $svg = decode_utf8( $svg );
758 =head2 as_dot( \%options )
760 Returns a string that is the collation graph expressed in dot
761 (i.e. GraphViz) format. Options include:
776 my( $self, $opts ) = @_;
777 my $startrank = $opts->{'from'} if $opts;
778 my $endrank = $opts->{'to'} if $opts;
779 my $color_common = $opts->{'color_common'} if $opts;
780 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
781 && $self->end->rank > 100;
782 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
784 # Check the arguments
786 return if $endrank && $startrank > $endrank;
787 return if $startrank > $self->end->rank;
789 if( defined $endrank ) {
790 return if $endrank < 0;
791 $endrank = undef if $endrank == $self->end->rank;
794 my $graph_name = $self->tradition->name;
795 $graph_name =~ s/[^\w\s]//g;
796 $graph_name = join( '_', split( /\s+/, $graph_name ) );
804 'fillcolor' => 'white',
809 'arrowhead' => 'open',
810 'color' => '#000000',
811 'fontcolor' => '#000000',
814 my $dot = sprintf( "digraph %s {\n", $graph_name );
815 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
816 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
818 # Output substitute start/end readings if necessary
820 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
823 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
825 if( $STRAIGHTENHACK ) {
827 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
828 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
829 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
831 my %used; # Keep track of the readings that actually appear in the graph
832 # Sort the readings by rank if we have ranks; this speeds layout.
833 my @all_readings = $self->end->has_rank
834 ? sort { $a->rank <=> $b->rank } $self->readings
836 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
837 foreach my $reading ( @all_readings ) {
838 # Only output readings within our rank range.
839 next if $startrank && $reading->rank < $startrank;
840 next if $endrank && $reading->rank > $endrank;
841 $used{$reading->id} = 1;
842 # Need not output nodes without separate labels
843 next if $reading->id eq $reading->text;
845 my $label = $reading->text;
846 $label .= '-' if $reading->join_next;
847 $label = "-$label" if $reading->join_prior;
848 $label =~ s/\"/\\\"/g;
849 $rattrs->{'label'} = $label;
850 $rattrs->{'id'} = $reading->id;
851 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
852 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
855 # Add the real edges. Need to weight one edge per rank jump, in a
857 # my $weighted = $self->_add_edge_weights;
858 my @edges = $self->paths;
859 my( %substart, %subend );
860 foreach my $edge ( @edges ) {
861 # Do we need to output this edge?
862 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
863 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
864 my $variables = { %edge_attrs, 'label' => $label };
866 # Account for the rank gap if necessary
867 my $rank0 = $self->reading( $edge->[0] )->rank
868 if $self->reading( $edge->[0] )->has_rank;
869 my $rank1 = $self->reading( $edge->[1] )->rank
870 if $self->reading( $edge->[1] )->has_rank;
871 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
872 $variables->{'minlen'} = $rank1 - $rank0;
875 # Add the calculated edge weights
876 # if( exists $weighted->{$edge->[0]}
877 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
878 # # $variables->{'color'} = 'red';
879 # $variables->{'weight'} = 3.0;
882 # EXPERIMENTAL: make edge width reflect no. of witnesses
883 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
884 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
886 my $varopts = _dot_attr_string( $variables );
887 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
888 $edge->[0], $edge->[1], $varopts );
889 } elsif( $used{$edge->[0]} ) {
890 $subend{$edge->[0]} = $edge->[1];
891 } elsif( $used{$edge->[1]} ) {
892 $substart{$edge->[1]} = $edge->[0];
896 # If we are asked to, add relationship links
897 if( exists $opts->{show_relations} ) {
898 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
899 if( $filter eq 'transposition' ) {
900 $filter =~ qr/^transposition$/;
902 foreach my $redge ( $self->relationships ) {
903 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
904 if( $filter ne 'all' ) {
905 my $rel = $self->get_relationship( $redge );
906 next unless $rel->type =~ /$filter/;
910 constraint => 'false',
911 label => uc( substr( $rel->type, 0, 4 ) ),
914 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
915 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
921 # Add substitute start and end edges if necessary
922 foreach my $node ( keys %substart ) {
923 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
924 my $variables = { %edge_attrs, 'label' => $witstr };
925 my $nrdg = $self->reading( $node );
926 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
927 # Substart is actually one lower than $startrank
928 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
930 my $varopts = _dot_attr_string( $variables );
931 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
933 foreach my $node ( keys %subend ) {
934 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
935 my $variables = { %edge_attrs, 'label' => $witstr };
936 my $varopts = _dot_attr_string( $variables );
937 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
940 if( $STRAIGHTENHACK ) {
941 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
942 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
949 sub _dot_attr_string {
952 foreach my $k ( sort keys %$hash ) {
954 push( @attrs, $k.'="'.$v.'"' );
956 return( '[ ' . join( ', ', @attrs ) . ' ]' );
959 sub _add_edge_weights {
961 # Walk the graph from START to END, choosing the successor node with
962 # the largest number of witness paths each time.
964 my $curr = $self->start->id;
965 my $ranked = $self->end->has_rank;
966 while( $curr ne $self->end->id ) {
967 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
968 my @succ = sort { $self->path_witnesses( $curr, $a )
969 <=> $self->path_witnesses( $curr, $b ) }
970 $self->sequence->successors( $curr );
971 my $next = pop @succ;
972 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
973 # Try to avoid lacunae in the weighted path.
975 ( $self->reading( $next )->is_lacuna ||
976 $nextrank - $rank > 1 ) ){
979 $weighted->{$curr} = $next;
985 =head2 path_witnesses( $edge )
987 Returns the list of sigils whose witnesses are associated with the given edge.
988 The edge can be passed as either an array or an arrayref of ( $source, $target ).
993 my( $self, @edge ) = @_;
994 # If edge is an arrayref, cope.
995 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
999 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
1003 # Helper function. Make a display label for the given witnesses, showing a.c.
1004 # witnesses only where the main witness is not also in the list.
1005 sub _path_display_label {
1008 map { $wits{$_} = 1 } @_;
1010 # If an a.c. wit is listed, remove it if the main wit is also listed.
1011 # Otherwise keep it for explicit listing.
1012 my $aclabel = $self->ac_label;
1014 foreach my $w ( sort keys %wits ) {
1015 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
1016 if( exists $wits{$1} ) {
1019 push( @disp_ac, $w );
1024 # See if we are in a majority situation.
1025 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
1026 $maj = $maj > 5 ? $maj : 5;
1027 if( scalar keys %wits > $maj ) {
1028 unshift( @disp_ac, 'majority' );
1029 return join( ', ', @disp_ac );
1031 return join( ', ', sort keys %wits );
1035 =head2 readings_at_rank( $rank )
1037 Returns a list of readings at a given rank, taken from the alignment table.
1041 sub readings_at_rank {
1042 my( $self, $rank ) = @_;
1043 my $table = $self->alignment_table;
1044 # Table rank is real rank - 1.
1045 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
1047 foreach my $e ( @elements ) {
1048 next unless ref( $e ) eq 'HASH';
1049 next unless exists $e->{'t'};
1050 $readings{$e->{'t'}->id} = $e->{'t'};
1052 return values %readings;
1057 Returns a GraphML representation of the collation. The GraphML will contain
1058 two graphs. The first expresses the attributes of the readings and the witness
1059 paths that link them; the second expresses the relationships that link the
1060 readings. This is the native transfer format for a tradition.
1064 use Text::Tradition;
1070 my $datafile = 't/data/florilegium_tei_ps.xml';
1071 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1073 'file' => $datafile,
1076 ok( $tradition, "Got a tradition object" );
1077 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
1078 ok( $tradition->collation, "Tradition has a collation" );
1080 my $c = $tradition->collation;
1081 is( scalar $c->readings, $READINGS, "Collation has all readings" );
1082 is( scalar $c->paths, $PATHS, "Collation has all paths" );
1083 is( scalar $c->relationships, 0, "Collation has all relationships" );
1085 # Add a few relationships
1086 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
1087 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
1088 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
1090 # Now write it to GraphML and parse it again.
1092 my $graphml = $c->as_graphml;
1093 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
1094 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
1095 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
1096 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
1098 # Now add a stemma, write to GraphML, and look at the output.
1100 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
1101 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
1102 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
1103 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
1104 $graphml = $c->as_graphml;
1105 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
1112 ## TODO MOVE this to Tradition.pm and modularize it better
1114 my( $self, $options ) = @_;
1115 $self->calculate_ranks unless $self->_graphcalc_done;
1117 my $start = $options->{'from'}
1118 ? $self->reading( $options->{'from'} ) : $self->start;
1119 my $end = $options->{'to'}
1120 ? $self->reading( $options->{'to'} ) : $self->end;
1121 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1122 throw( 'Start node must be before end node' );
1124 # The readings need to be ranked for this to work.
1125 $start = $self->start unless $start->has_rank;
1126 $end = $self->end unless $end->has_rank;
1128 unless( $start eq $self->start ) {
1129 $rankoffset = $start->rank - 1;
1134 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1135 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1136 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1137 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1139 # Create the document and root node
1140 require XML::LibXML;
1141 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1142 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1143 $graphml->setDocumentElement( $root );
1144 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1145 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1147 # List of attribute types to save on our objects and their corresponding
1152 'Bool' => 'boolean',
1153 'ReadingID' => 'string',
1154 'RelationshipType' => 'string',
1155 'RelationshipScope' => 'string',
1158 # Add the data keys for the graph. Include an extra key 'version' for the
1159 # GraphML output version.
1160 my %graph_data_keys;
1162 my %graph_attributes = ( 'version' => 'string' );
1163 # Graph attributes include those of Tradition and those of Collation.
1165 # TODO Use meta introspection method from duplicate_reading to do this
1166 # instead of naming custom keys.
1167 my $tmeta = $self->tradition->meta;
1168 my $cmeta = $self->meta;
1169 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1170 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1171 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1172 next if $attr->name =~ /^_/;
1173 next unless $save_types{$attr->type_constraint->name};
1174 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1176 # Extra custom keys for complex objects that should be saved in some form.
1177 # The subroutine should return a string, or undef/empty.
1178 if( $tmeta->has_method('stemmata') ) {
1179 $graph_attributes{'stemmata'} = sub {
1181 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1182 $self->tradition->stemmata;
1183 join( "\n", @stemstrs );
1187 if( $tmeta->has_method('user') ) {
1188 $graph_attributes{'user'} = sub {
1189 $self->tradition->user ? $self->tradition->user->id : undef
1193 foreach my $datum ( sort keys %graph_attributes ) {
1194 $graph_data_keys{$datum} = 'dg'.$gdi++;
1195 my $key = $root->addNewChild( $graphml_ns, 'key' );
1196 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1197 : $graph_attributes{$datum};
1198 $key->setAttribute( 'attr.name', $datum );
1199 $key->setAttribute( 'attr.type', $dtype );
1200 $key->setAttribute( 'for', 'graph' );
1201 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1204 # Add the data keys for reading nodes
1205 my %reading_attributes;
1206 my $rmeta = Text::Tradition::Collation::Reading->meta;
1207 foreach my $attr( $rmeta->get_all_attributes ) {
1208 next if $attr->name =~ /^_/;
1209 next unless $save_types{$attr->type_constraint->name};
1210 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1212 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1213 # Extra custom key for the reading morphology
1214 $reading_attributes{'lexemes'} = 'string';
1219 foreach my $datum ( sort keys %reading_attributes ) {
1220 $node_data_keys{$datum} = 'dn'.$ndi++;
1221 my $key = $root->addNewChild( $graphml_ns, 'key' );
1222 $key->setAttribute( 'attr.name', $datum );
1223 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1224 $key->setAttribute( 'for', 'node' );
1225 $key->setAttribute( 'id', $node_data_keys{$datum} );
1228 # Add the data keys for edges, that is, paths and relationships. Path
1229 # data does not come from a Moose class so is here manually.
1232 my %edge_attributes = (
1233 witness => 'string', # ID/label for a path
1234 extra => 'boolean', # Path key
1236 my @path_attributes = keys %edge_attributes; # track our manual additions
1237 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1238 foreach my $attr( $pmeta->get_all_attributes ) {
1239 next if $attr->name =~ /^_/;
1240 next unless $save_types{$attr->type_constraint->name};
1241 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1243 foreach my $datum ( sort keys %edge_attributes ) {
1244 $edge_data_keys{$datum} = 'de'.$edi++;
1245 my $key = $root->addNewChild( $graphml_ns, 'key' );
1246 $key->setAttribute( 'attr.name', $datum );
1247 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1248 $key->setAttribute( 'for', 'edge' );
1249 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1252 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1253 my $xmlidname = $self->tradition->name;
1254 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1255 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1256 $xmlidname = '_'.$xmlidname;
1258 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1259 $sgraph->setAttribute( 'edgedefault', 'directed' );
1260 $sgraph->setAttribute( 'id', $xmlidname );
1261 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1262 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1263 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1264 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1265 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1267 # Tradition/collation attribute data
1268 foreach my $datum ( keys %graph_attributes ) {
1270 if( $datum eq 'version' ) {
1272 } elsif( ref( $graph_attributes{$datum} ) ) {
1273 my $sub = $graph_attributes{$datum};
1275 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1276 $value = $self->tradition->$datum;
1278 $value = $self->$datum;
1280 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1285 # Add our readings to the graph
1286 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1287 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1288 ( $n->rank < $start->rank || $n->rank > $end->rank );
1289 $use_readings{$n->id} = 1;
1290 # Add to the main graph
1291 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1292 my $node_xmlid = 'n' . $node_ctr++;
1293 $node_hash{ $n->id } = $node_xmlid;
1294 $node_el->setAttribute( 'id', $node_xmlid );
1295 foreach my $d ( keys %reading_attributes ) {
1297 # Custom serialization
1298 if( $d eq 'lexemes' ) {
1299 # If nval is a true value, we have lexemes so we need to
1300 # serialize them. Otherwise set nval to undef so that the
1301 # key is excluded from this reading.
1302 $nval = $nval ? $n->_serialize_lexemes : undef;
1303 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1306 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1307 # Adjust the ranks within the subgraph.
1308 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1309 : $nval - $rankoffset;
1311 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1316 # Add the path edges to the sequence graph
1318 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1319 # We add an edge in the graphml for every witness in $e.
1320 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1321 my @edge_wits = sort $self->path_witnesses( $e );
1322 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1323 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1324 # Skip any path from start to end; that witness is not in the subgraph.
1325 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1326 foreach my $wit ( @edge_wits ) {
1327 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1328 $node_hash{ $e->[0] },
1329 $node_hash{ $e->[1] } );
1330 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1331 $edge_el->setAttribute( 'source', $from );
1332 $edge_el->setAttribute( 'target', $to );
1333 $edge_el->setAttribute( 'id', $id );
1335 # It's a witness path, so add the witness
1337 my $key = $edge_data_keys{'witness'};
1338 # Is this an ante-corr witness?
1339 my $aclabel = $self->ac_label;
1340 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1341 # Keep the base witness
1343 # ...and record that this is an 'extra' reading path
1344 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1346 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1350 # Report the actual number of nodes and edges that went in
1351 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1352 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1354 # Add the relationship graph to the XML
1355 map { delete $edge_data_keys{$_} } @path_attributes;
1356 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1357 $node_data_keys{'id'}, \%edge_data_keys );
1359 # Save and return the thing
1360 my $result = decode_utf8( $graphml->toString(1) );
1364 sub _add_graphml_data {
1365 my( $el, $key, $value ) = @_;
1366 return unless defined $value;
1367 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1368 $data_el->setAttribute( 'key', $key );
1369 $data_el->appendText( $value );
1374 Returns a CSV alignment table representation of the collation graph, one
1375 row per witness (or witness uncorrected.)
1381 my $table = $self->alignment_table;
1382 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1384 # Make the header row
1385 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1386 push( @result, decode_utf8( $csv->string ) );
1387 # Make the rest of the rows
1388 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1389 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1390 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1391 $csv->combine( @row );
1392 push( @result, decode_utf8( $csv->string ) );
1394 return join( "\n", @result );
1397 =head2 alignment_table
1399 Return a reference to an alignment table, in a slightly enhanced CollateX
1400 format which looks like this:
1402 $table = { alignment => [ { witness => "SIGIL",
1403 tokens => [ { t => "TEXT" }, ... ] },
1404 { witness => "SIG2",
1405 tokens => [ { t => "TEXT" }, ... ] },
1407 length => TEXTLEN };
1411 sub alignment_table {
1413 return $self->cached_table if $self->has_cached_table;
1415 # Make sure we can do this
1416 throw( "Need a linear graph in order to make an alignment table" )
1417 unless $self->linear;
1418 $self->calculate_ranks()
1419 unless $self->_graphcalc_done && $self->end->has_rank;
1421 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1422 my @all_pos = ( 1 .. $self->end->rank - 1 );
1423 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1424 # say STDERR "Making witness row(s) for " . $wit->sigil;
1425 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1426 my @row = _make_witness_row( \@wit_path, \@all_pos );
1427 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1428 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1429 push( @{$table->{'alignment'}}, $witobj );
1430 if( $wit->is_layered ) {
1431 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1432 $wit->sigil.$self->ac_label );
1433 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1434 my $witacobj = { 'witness' => $wit->sigil.$self->ac_label,
1435 'tokens' => \@ac_row };
1436 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1437 push( @{$table->{'alignment'}}, $witacobj );
1440 $self->cached_table( $table );
1444 sub _make_witness_row {
1445 my( $path, $positions ) = @_;
1447 map { $char_hash{$_} = undef } @$positions;
1449 foreach my $rdg ( @$path ) {
1450 say STDERR "rank " . $rdg->rank if $debug;
1451 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1452 $char_hash{$rdg->rank} = { 't' => $rdg };
1454 my @row = map { $char_hash{$_} } @$positions;
1455 # Fill in lacuna markers for undef spots in the row
1456 my $last_el = shift @row;
1457 my @filled_row = ( $last_el );
1458 foreach my $el ( @row ) {
1459 # If we are using node reference, make the lacuna node appear many times
1460 # in the table. If not, use the lacuna tag.
1461 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1464 push( @filled_row, $el );
1471 =head1 NAVIGATION METHODS
1473 =head2 reading_sequence( $first, $last, $sigil, $backup )
1475 Returns the ordered list of readings, starting with $first and ending
1476 with $last, for the witness given in $sigil. If a $backup sigil is
1477 specified (e.g. when walking a layered witness), it will be used wherever
1478 no $sigil path exists. If there is a base text reading, that will be
1479 used wherever no path exists for $sigil or $backup.
1483 # TODO Think about returning some lazy-eval iterator.
1484 # TODO Get rid of backup; we should know from what witness is whether we need it.
1486 sub reading_sequence {
1487 my( $self, $start, $end, $witness ) = @_;
1489 $witness = $self->baselabel unless $witness;
1490 my @readings = ( $start );
1493 while( $n && $n->id ne $end->id ) {
1494 if( exists( $seen{$n->id} ) ) {
1495 throw( "Detected loop for $witness at " . $n->id );
1499 my $next = $self->next_reading( $n, $witness );
1501 throw( "Did not find any path for $witness from reading " . $n->id );
1503 push( @readings, $next );
1506 # Check that the last reading is our end reading.
1507 my $last = $readings[$#readings];
1508 throw( "Last reading found from " . $start->text .
1509 " for witness $witness is not the end!" ) # TODO do we get this far?
1510 unless $last->id eq $end->id;
1515 =head2 next_reading( $reading, $sigil );
1517 Returns the reading that follows the given reading along the given witness
1523 # Return the successor via the corresponding path.
1525 my $answer = $self->_find_linked_reading( 'next', @_ );
1526 return undef unless $answer;
1527 return $self->reading( $answer );
1530 =head2 prior_reading( $reading, $sigil )
1532 Returns the reading that precedes the given reading along the given witness
1538 # Return the predecessor via the corresponding path.
1540 my $answer = $self->_find_linked_reading( 'prior', @_ );
1541 return $self->reading( $answer );
1544 sub _find_linked_reading {
1545 my( $self, $direction, $node, $path ) = @_;
1547 # Get a backup if we are dealing with a layered witness
1549 my $aclabel = $self->ac_label;
1550 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1554 my @linked_paths = $direction eq 'next'
1555 ? $self->sequence->edges_from( $node )
1556 : $self->sequence->edges_to( $node );
1557 return undef unless scalar( @linked_paths );
1559 # We have to find the linked path that contains all of the
1560 # witnesses supplied in $path.
1561 my( @path_wits, @alt_path_wits );
1562 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1563 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1566 foreach my $le ( @linked_paths ) {
1567 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1570 my @le_wits = sort $self->path_witnesses( $le );
1571 if( _is_within( \@path_wits, \@le_wits ) ) {
1572 # This is the right path.
1573 return $direction eq 'next' ? $le->[1] : $le->[0];
1574 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1578 # Got this far? Return the alternate path if it exists.
1579 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1582 # Got this far? Return the base path if it exists.
1583 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1586 # Got this far? We have no appropriate path.
1587 warn "Could not find $direction node from " . $node->id
1588 . " along path $path";
1594 my( $set1, $set2 ) = @_;
1595 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1596 foreach my $el ( @$set1 ) {
1597 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1602 # Return the string that joins together a list of witnesses for
1603 # display on a single path.
1604 sub _witnesses_of_label {
1605 my( $self, $label ) = @_;
1606 my $regex = $self->wit_list_separator;
1607 my @answer = split( /\Q$regex\E/, $label );
1611 =head2 common_readings
1613 Returns the list of common readings in the graph (i.e. those readings that are
1614 shared by all non-lacunose witnesses.)
1618 sub common_readings {
1620 my @common = grep { $_->is_common } $self->readings;
1624 =head2 path_text( $sigil, [, $start, $end ] )
1626 Returns the text of a witness (plus its backup, if we are using a layer)
1627 as stored in the collation. The text is returned as a string, where the
1628 individual readings are joined with spaces and the meta-readings (e.g.
1629 lacunae) are omitted. Optional specification of $start and $end allows
1630 the generation of a subset of the witness text.
1635 my( $self, $wit, $start, $end ) = @_;
1636 $start = $self->start unless $start;
1637 $end = $self->end unless $end;
1638 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1641 foreach my $r ( @path ) {
1642 unless ( $r->join_prior || !$last || $last->join_next ) {
1645 $pathtext .= $r->text;
1651 =head1 INITIALIZATION METHODS
1653 These are mostly for use by parsers.
1655 =head2 make_witness_path( $witness )
1657 Link the array of readings contained in $witness->path (and in
1658 $witness->uncorrected_path if it exists) into collation paths.
1659 Clear out the arrays when finished.
1661 =head2 make_witness_paths
1663 Call make_witness_path for all witnesses in the tradition.
1667 # For use when a collation is constructed from a base text and an apparatus.
1668 # We have the sequences of readings and just need to add path edges.
1669 # When we are done, clear out the witness path attributes, as they are no
1671 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1673 sub make_witness_paths {
1675 foreach my $wit ( $self->tradition->witnesses ) {
1676 # say STDERR "Making path for " . $wit->sigil;
1677 $self->make_witness_path( $wit );
1681 sub make_witness_path {
1682 my( $self, $wit ) = @_;
1683 my @chain = @{$wit->path};
1684 my $sig = $wit->sigil;
1685 # Add start and end if necessary
1686 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1687 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1688 foreach my $idx ( 0 .. $#chain-1 ) {
1689 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1691 if( $wit->is_layered ) {
1692 @chain = @{$wit->uncorrected_path};
1693 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1694 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1695 foreach my $idx( 0 .. $#chain-1 ) {
1696 my $source = $chain[$idx];
1697 my $target = $chain[$idx+1];
1698 $self->add_path( $source, $target, $sig.$self->ac_label )
1699 unless $self->has_path( $source, $target, $sig );
1703 $wit->clear_uncorrected_path;
1706 =head2 calculate_ranks
1708 Calculate the reading ranks (that is, their aligned positions relative
1709 to each other) for the graph. This can only be called on linear collations.
1713 use Text::Tradition;
1715 my $cxfile = 't/data/Collatex-16.xml';
1716 my $t = Text::Tradition->new(
1718 'input' => 'CollateX',
1721 my $c = $t->collation;
1724 my $table = $c->alignment_table;
1725 ok( $c->has_cached_table, "Alignment table was cached" );
1726 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1727 $c->calculate_ranks;
1728 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1729 $c->add_relationship( 'n13', 'n23', { type => 'repetition' } );
1730 is( $c->alignment_table, $table, "Alignment table unchanged after non-colo relationship add" );
1731 $c->add_relationship( 'n24', 'n23', { type => 'spelling' } );
1732 isnt( $c->alignment_table, $table, "Alignment table changed after colo relationship add" );
1738 sub calculate_ranks {
1740 # Save the existing ranks, in case we need to invalidate the cached SVG.
1742 map { $existing_ranks{$_} = $_->rank } $self->readings;
1744 # Do the rankings based on the relationship equivalence graph, starting
1745 # with the start node.
1746 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1748 # Transfer our rankings from the topological graph to the real one.
1749 foreach my $r ( $self->readings ) {
1750 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1751 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1753 # Die. Find the last rank we calculated.
1754 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1755 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1757 my $last = pop @all_defined;
1758 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1761 # Do we need to invalidate the cached data?
1762 if( $self->has_cached_table ) {
1763 foreach my $r ( $self->readings ) {
1764 next if defined( $existing_ranks{$r} )
1765 && $existing_ranks{$r} == $r->rank;
1766 # Something has changed, so clear the cache
1767 $self->_clear_cache;
1768 # ...and recalculate the common readings.
1769 $self->calculate_common_readings();
1773 # The graph calculation information is now up to date.
1774 $self->_graphcalc_done(1);
1779 $self->wipe_table if $self->has_cached_table;
1783 =head2 flatten_ranks
1785 A convenience method for parsing collation data. Searches the graph for readings
1786 with the same text at the same rank, and merges any that are found.
1791 my ( $self, %args ) = shift;
1792 my %unique_rank_rdg;
1794 foreach my $p ( $self->identical_readings( %args ) ) {
1795 # say STDERR "Combining readings at same rank: @$p";
1797 $self->merge_readings( @$p );
1798 # TODO see if this now makes a common point.
1800 # If we merged readings, the ranks are still fine but the alignment
1801 # table is wrong. Wipe it.
1802 $self->wipe_table() if $changed;
1805 =head2 identical_readings
1806 =head2 identical_readings( start => $startnode, end => $endnode )
1807 =head2 identical_readings( startrank => $startrank, endrank => $endrank )
1809 Goes through the graph identifying all pairs of readings that appear to be
1810 identical, and therefore able to be merged into a single reading. Returns the
1811 relevant identical pairs. Can be restricted to run over only a part of the
1812 graph, specified either by node or by rank.
1816 sub identical_readings {
1817 my ( $self, %args ) = @_;
1818 # Find where we should start and end.
1820 my $startrank = $args{startrank} || 0;
1821 if( $args{start} ) {
1822 throw( "Starting reading has no rank" ) unless $self->reading( $args{start} )
1823 && $self->reading( $args{start} )->has_rank;
1824 $startrank = $self->reading( $args{start} )->rank;
1826 my $endrank = $args{endrank} || $self->end->rank;
1828 throw( "Ending reading has no rank" ) unless $self->reading( $args{end} )
1829 && $self->reading( $args{end} )->has_rank;
1830 $endrank = $self->reading( $args{end} )->rank;
1833 # Make sure the ranks are correct.
1834 unless( $self->_graphcalc_done ) {
1835 $self->calculate_ranks;
1837 # Go through the readings looking for duplicates.
1838 my %unique_rank_rdg;
1840 foreach my $rdg ( $self->readings ) {
1841 next unless $rdg->has_rank;
1842 my $rk = $rdg->rank;
1843 next if $rk > $endrank || $rk < $startrank;
1844 my $key = $rk . "||" . $rdg->text;
1845 if( exists $unique_rank_rdg{$key} ) {
1846 # Make sure they don't have different grammatical forms
1847 my $ur = $unique_rank_rdg{$key};
1848 if( $rdg->is_identical( $ur ) ) {
1849 push( @pairs, [ $ur, $rdg ] );
1852 $unique_rank_rdg{$key} = $rdg;
1860 =head2 calculate_common_readings
1862 Goes through the graph identifying the readings that appear in every witness
1863 (apart from those with lacunae at that spot.) Marks them as common and returns
1868 use Text::Tradition;
1870 my $cxfile = 't/data/Collatex-16.xml';
1871 my $t = Text::Tradition->new(
1873 'input' => 'CollateX',
1876 my $c = $t->collation;
1878 my @common = $c->calculate_common_readings();
1879 is( scalar @common, 8, "Found correct number of common readings" );
1880 my @marked = sort $c->common_readings();
1881 is( scalar @common, 8, "All common readings got marked as such" );
1882 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1883 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1889 sub calculate_common_readings {
1892 map { $_->is_common( 0 ) } $self->readings;
1893 # Implicitly calls calculate_ranks
1894 my $table = $self->alignment_table;
1895 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1896 my @row = map { $_->{'tokens'}->[$idx]
1897 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1898 @{$table->{'alignment'}};
1900 foreach my $r ( @row ) {
1902 $hash{$r->id} = $r unless $r->is_meta;
1904 $hash{'UNDEF'} = $r;
1907 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1908 my( $r ) = values %hash;
1910 push( @common, $r );
1916 =head2 text_from_paths
1918 Calculate the text array for all witnesses from the path, for later consistency
1919 checking. Only to be used if there is no non-graph-based way to know the
1924 sub text_from_paths {
1926 foreach my $wit ( $self->tradition->witnesses ) {
1927 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1929 foreach my $r ( @readings ) {
1930 next if $r->is_meta;
1931 push( @text, $r->text );
1933 $wit->text( \@text );
1934 if( $wit->is_layered ) {
1935 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1936 $wit->sigil.$self->ac_label );
1938 foreach my $r ( @ucrdgs ) {
1939 next if $r->is_meta;
1940 push( @uctext, $r->text );
1942 $wit->layertext( \@uctext );
1947 =head1 UTILITY FUNCTIONS
1949 =head2 common_predecessor( $reading_a, $reading_b )
1951 Find the last reading that occurs in sequence before both the given readings.
1952 At the very least this should be $self->start.
1954 =head2 common_successor( $reading_a, $reading_b )
1956 Find the first reading that occurs in sequence after both the given readings.
1957 At the very least this should be $self->end.
1961 use Text::Tradition;
1963 my $cxfile = 't/data/Collatex-16.xml';
1964 my $t = Text::Tradition->new(
1966 'input' => 'CollateX',
1969 my $c = $t->collation;
1971 is( $c->common_predecessor( 'n24', 'n23' )->id,
1972 'n20', "Found correct common predecessor" );
1973 is( $c->common_successor( 'n24', 'n23' )->id,
1974 '__END__', "Found correct common successor" );
1976 is( $c->common_predecessor( 'n19', 'n17' )->id,
1977 'n16', "Found correct common predecessor for readings on same path" );
1978 is( $c->common_successor( 'n21', 'n10' )->id,
1979 '__END__', "Found correct common successor for readings on same path" );
1985 ## Return the closest reading that is a predecessor of both the given readings.
1986 sub common_predecessor {
1988 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1989 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1992 sub common_successor {
1994 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1995 return $self->_common_in_path( $r1, $r2, 'successors' );
1999 # TODO think about how to do this without ranks...
2000 sub _common_in_path {
2001 my( $self, $r1, $r2, $dir ) = @_;
2002 my $iter = $self->end->rank;
2004 my @last_r1 = ( $r1 );
2005 my @last_r2 = ( $r2 );
2006 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
2008 # say STDERR "Finding common $dir for $r1, $r2";
2009 while( !@candidates ) {
2010 last unless $iter--; # Avoid looping infinitely
2011 # Iterate separately down the graph from r1 and r2
2012 my( @new_lc1, @new_lc2 );
2013 foreach my $lc ( @last_r1 ) {
2014 foreach my $p ( $lc->$dir ) {
2015 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
2016 # say STDERR "Path candidate $p from $lc";
2017 push( @candidates, $p );
2018 } elsif( !$all_seen{$p->id} ) {
2019 $all_seen{$p->id} = 'r1';
2020 push( @new_lc1, $p );
2024 foreach my $lc ( @last_r2 ) {
2025 foreach my $p ( $lc->$dir ) {
2026 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
2027 # say STDERR "Path candidate $p from $lc";
2028 push( @candidates, $p );
2029 } elsif( !$all_seen{$p->id} ) {
2030 $all_seen{$p->id} = 'r2';
2031 push( @new_lc2, $p );
2035 @last_r1 = @new_lc1;
2036 @last_r2 = @new_lc2;
2038 my @answer = sort { $a->rank <=> $b->rank } @candidates;
2039 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
2043 Text::Tradition::Error->throw(
2044 'ident' => 'Collation error',
2050 __PACKAGE__->meta->make_immutable;
2056 =item * Rework XML serialization in a more modular way
2062 This package is free software and is provided "as is" without express
2063 or implied warranty. You can redistribute it and/or modify it under
2064 the same terms as Perl itself.
2068 Tara L Andrews E<lt>aurum@cpan.orgE<gt>