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, @del_rdgs ) = $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( scalar @del_rdgs, 1, "Deleted reading was returned by API call" );
489 is( $sc->get_relationship( 'n130', 'n135' ), undef, "Bad transposition relationship is gone" );
491 # The collation should not be fixed
492 my @pairs = $sc->identical_readings();
493 is( scalar @pairs, 0, "Not re-collated yet" );
495 ok( $sc->merge_readings( 'n124', 'n131_0' ), "Collated the readings correctly" );
496 @pairs = $sc->identical_readings( start => 'n124', end => $csucc->id );
497 is( scalar @pairs, 3, "Found three more identical readings" );
498 is( $sc->end->rank, 11, "The ranks shifted appropriately" );
499 $sc->flatten_ranks();
500 is( scalar( $sc->readings ), $numr - 3, "Now we are collated correctly" );
506 sub duplicate_reading {
507 my( $self, $r, @wits ) = @_;
508 # Add the new reading, duplicating $r.
509 unless( ref( $r ) eq 'Text::Tradition::Collation::Reading' ) {
510 $r = $self->reading( $r );
512 throw( "Cannot duplicate a meta-reading" )
515 # Get all the reading attributes and duplicate them.
516 my $rmeta = Text::Tradition::Collation::Reading->meta;
518 foreach my $attr( $rmeta->get_all_attributes ) {
519 next if $attr->name =~ /^_/;
520 my $acc = $attr->get_read_method;
521 if( !$acc && $attr->has_applied_traits ) {
522 my $tr = $attr->applied_traits;
523 if( $tr->[0] =~ /::(Array|Hash)$/ ) {
525 my %methods = reverse %{$attr->handles};
526 $acc = $methods{elements};
527 $args{$attr->name} = $which eq 'Array'
528 ? [ $r->$acc ] : { $r->$acc };
531 $args{$attr->name} = $r->$acc if $acc;
534 # By definition the new reading will no longer be common.
535 $args{is_common} = 0;
536 # The new reading also needs its own ID.
537 $args{id} = $self->_generate_dup_id( $r->id );
539 # Try to make the new reading.
540 my $newr = $self->add_reading( \%args );
541 # The old reading is also no longer common.
544 # For each of the witnesses, dissociate from the old reading and
545 # associate with the new.
546 foreach my $wit ( @wits ) {
547 my $prior = $self->prior_reading( $r, $wit );
548 my $next = $self->next_reading( $r, $wit );
549 $self->del_path( $prior, $r, $wit );
550 $self->add_path( $prior, $newr, $wit );
551 $self->del_path( $r, $next, $wit );
552 $self->add_path( $newr, $next, $wit );
555 # If the graph is ranked, we need to look for relationships that are now
556 # invalid (i.e. 'non-colocation' types that might now be colocated) and
557 # remove them. If not, we can skip it.
560 my @deleted_relations;
561 if( $self->end->has_rank ) {
562 # Find the point where we can stop checking
563 $succ = $self->common_successor( $r, $newr );
565 # Hash the existing ranks
566 foreach my $rdg ( $self->readings ) {
567 $rrk{$rdg->id} = $rdg->rank;
569 # Calculate the new ranks
570 $self->calculate_ranks();
572 # Check for invalid non-colocated relationships among changed-rank readings
573 # from where the ranks start changing up to $succ
574 my $lastrank = $succ->rank;
575 foreach my $rdg ( $self->readings ) {
576 next if $rdg->rank > $lastrank;
577 next if $rdg->rank == $rrk{$rdg->id};
578 my @noncolo = $rdg->related_readings( sub { !$_[0]->colocated } );
579 next unless @noncolo;
580 foreach my $nc ( @noncolo ) {
581 unless( $self->relations->verify_or_delete( $rdg, $nc ) ) {
582 push( @deleted_relations, [ $rdg->id, $nc->id ] );
587 return ( $newr, @deleted_relations );
590 sub _generate_dup_id {
591 my( $self, $rid ) = @_;
596 if( $self->has_reading( $newid ) ) {
609 # We only need the IDs for adding paths to the graph, not the reading
610 # objects themselves.
611 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
613 $self->_graphcalc_done(0);
614 # Connect the readings
615 unless( $self->sequence->has_edge( $source, $target ) ) {
616 $self->sequence->add_edge( $source, $target );
617 $self->relations->add_equivalence_edge( $source, $target );
619 # Note the witness in question
620 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
626 if( ref( $_[0] ) eq 'ARRAY' ) {
633 # We only need the IDs for removing paths from the graph, not the reading
634 # objects themselves.
635 my( $source, $target, $wit ) = $self->_stringify_args( @args );
637 $self->_graphcalc_done(0);
638 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
639 $self->sequence->delete_edge_attribute( $source, $target, $wit );
641 unless( $self->sequence->has_edge_attributes( $source, $target ) ) {
642 $self->sequence->delete_edge( $source, $target );
643 $self->relations->delete_equivalence_edge( $source, $target );
648 # Extra graph-alike utility
651 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
652 return undef unless $self->sequence->has_edge( $source, $target );
653 return $self->sequence->has_edge_attribute( $source, $target, $wit );
656 =head2 clear_witness( @sigil_list )
658 Clear the given witnesses out of the collation entirely, removing references
659 to them in paths, and removing readings that belong only to them. Should only
660 be called via $tradition->del_witness.
665 my( $self, @sigils ) = @_;
667 $self->_graphcalc_done(0);
668 # Clear the witness(es) out of the paths
669 foreach my $e ( $self->paths ) {
670 foreach my $sig ( @sigils ) {
671 $self->del_path( $e, $sig );
675 # Clear out the newly unused readings
676 foreach my $r ( $self->readings ) {
677 unless( $self->reading_witnesses( $r ) ) {
678 $self->del_reading( $r );
683 sub add_relationship {
685 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
686 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
687 foreach my $v ( @vectors ) {
688 next unless $self->get_relationship( $v )->colocated;
689 if( $self->reading( $v->[0] )->has_rank && $self->reading( $v->[1] )->has_rank
690 && $self->reading( $v->[0] )->rank ne $self->reading( $v->[1] )->rank ) {
691 $self->_graphcalc_done(0);
699 around qw/ get_relationship del_relationship / => sub {
703 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
706 my @stringargs = $self->_stringify_args( @args );
707 $self->$orig( @stringargs );
710 =head2 reading_witnesses( $reading )
712 Return a list of sigils corresponding to the witnesses in which the reading appears.
716 sub reading_witnesses {
717 my( $self, $reading ) = @_;
718 # We need only check either the incoming or the outgoing edges; I have
719 # arbitrarily chosen "incoming". Thus, special-case the start node.
720 if( $reading eq $self->start ) {
721 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
724 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
725 my $wits = $self->sequence->get_edge_attributes( @$e );
726 @all_witnesses{ keys %$wits } = 1;
728 my $acstr = $self->ac_label;
729 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
730 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
732 return keys %all_witnesses;
735 =head1 OUTPUT METHODS
737 =head2 as_svg( \%options )
739 Returns an SVG string that represents the graph, via as_dot and graphviz.
740 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
745 my( $self, $opts ) = @_;
746 throw( "Need GraphViz installed to output SVG" )
747 unless File::Which::which( 'dot' );
748 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
749 $self->calculate_ranks()
750 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
751 my @cmd = qw/dot -Tsvg/;
753 my $dotfile = File::Temp->new();
755 # $dotfile->unlink_on_destroy(0);
756 binmode $dotfile, ':utf8';
757 print $dotfile $self->as_dot( $opts );
758 push( @cmd, $dotfile->filename );
759 run( \@cmd, ">", binary(), \$svg );
760 $svg = decode_utf8( $svg );
765 =head2 as_dot( \%options )
767 Returns a string that is the collation graph expressed in dot
768 (i.e. GraphViz) format. Options include:
783 my( $self, $opts ) = @_;
784 my $startrank = $opts->{'from'} if $opts;
785 my $endrank = $opts->{'to'} if $opts;
786 my $color_common = $opts->{'color_common'} if $opts;
787 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
788 && $self->end->rank > 100;
789 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
791 # Check the arguments
793 return if $endrank && $startrank > $endrank;
794 return if $startrank > $self->end->rank;
796 if( defined $endrank ) {
797 return if $endrank < 0;
798 $endrank = undef if $endrank == $self->end->rank;
801 my $graph_name = $self->tradition->name;
802 $graph_name =~ s/[^\w\s]//g;
803 $graph_name = join( '_', split( /\s+/, $graph_name ) );
811 'fillcolor' => 'white',
816 'arrowhead' => 'open',
817 'color' => '#000000',
818 'fontcolor' => '#000000',
821 my $dot = sprintf( "digraph %s {\n", $graph_name );
822 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
823 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
825 # Output substitute start/end readings if necessary
827 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
830 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
832 if( $STRAIGHTENHACK ) {
834 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
835 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
836 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
838 my %used; # Keep track of the readings that actually appear in the graph
839 # Sort the readings by rank if we have ranks; this speeds layout.
840 my @all_readings = $self->end->has_rank
841 ? sort { $a->rank <=> $b->rank } $self->readings
843 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
844 foreach my $reading ( @all_readings ) {
845 # Only output readings within our rank range.
846 next if $startrank && $reading->rank < $startrank;
847 next if $endrank && $reading->rank > $endrank;
848 $used{$reading->id} = 1;
849 # Need not output nodes without separate labels
850 next if $reading->id eq $reading->text;
852 my $label = $reading->text;
853 $label .= '-' if $reading->join_next;
854 $label = "-$label" if $reading->join_prior;
855 $label =~ s/\"/\\\"/g;
856 $rattrs->{'label'} = $label;
857 $rattrs->{'id'} = $reading->id;
858 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
859 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
862 # Add the real edges. Need to weight one edge per rank jump, in a
864 # my $weighted = $self->_add_edge_weights;
865 my @edges = $self->paths;
866 my( %substart, %subend );
867 foreach my $edge ( @edges ) {
868 # Do we need to output this edge?
869 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
870 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
871 my $variables = { %edge_attrs, 'label' => $label };
873 # Account for the rank gap if necessary
874 my $rank0 = $self->reading( $edge->[0] )->rank
875 if $self->reading( $edge->[0] )->has_rank;
876 my $rank1 = $self->reading( $edge->[1] )->rank
877 if $self->reading( $edge->[1] )->has_rank;
878 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
879 $variables->{'minlen'} = $rank1 - $rank0;
882 # Add the calculated edge weights
883 # if( exists $weighted->{$edge->[0]}
884 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
885 # # $variables->{'color'} = 'red';
886 # $variables->{'weight'} = 3.0;
889 # EXPERIMENTAL: make edge width reflect no. of witnesses
890 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
891 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
893 my $varopts = _dot_attr_string( $variables );
894 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
895 $edge->[0], $edge->[1], $varopts );
896 } elsif( $used{$edge->[0]} ) {
897 $subend{$edge->[0]} = $edge->[1];
898 } elsif( $used{$edge->[1]} ) {
899 $substart{$edge->[1]} = $edge->[0];
903 # If we are asked to, add relationship links
904 if( exists $opts->{show_relations} ) {
905 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
906 if( $filter eq 'transposition' ) {
907 $filter =~ qr/^transposition$/;
909 foreach my $redge ( $self->relationships ) {
910 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
911 if( $filter ne 'all' ) {
912 my $rel = $self->get_relationship( $redge );
913 next unless $rel->type =~ /$filter/;
917 constraint => 'false',
918 label => uc( substr( $rel->type, 0, 4 ) ),
921 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
922 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
928 # Add substitute start and end edges if necessary
929 foreach my $node ( keys %substart ) {
930 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
931 my $variables = { %edge_attrs, 'label' => $witstr };
932 my $nrdg = $self->reading( $node );
933 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
934 # Substart is actually one lower than $startrank
935 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
937 my $varopts = _dot_attr_string( $variables );
938 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
940 foreach my $node ( keys %subend ) {
941 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
942 my $variables = { %edge_attrs, 'label' => $witstr };
943 my $varopts = _dot_attr_string( $variables );
944 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
947 if( $STRAIGHTENHACK ) {
948 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
949 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
956 sub _dot_attr_string {
959 foreach my $k ( sort keys %$hash ) {
961 push( @attrs, $k.'="'.$v.'"' );
963 return( '[ ' . join( ', ', @attrs ) . ' ]' );
966 sub _add_edge_weights {
968 # Walk the graph from START to END, choosing the successor node with
969 # the largest number of witness paths each time.
971 my $curr = $self->start->id;
972 my $ranked = $self->end->has_rank;
973 while( $curr ne $self->end->id ) {
974 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
975 my @succ = sort { $self->path_witnesses( $curr, $a )
976 <=> $self->path_witnesses( $curr, $b ) }
977 $self->sequence->successors( $curr );
978 my $next = pop @succ;
979 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
980 # Try to avoid lacunae in the weighted path.
982 ( $self->reading( $next )->is_lacuna ||
983 $nextrank - $rank > 1 ) ){
986 $weighted->{$curr} = $next;
992 =head2 path_witnesses( $edge )
994 Returns the list of sigils whose witnesses are associated with the given edge.
995 The edge can be passed as either an array or an arrayref of ( $source, $target ).
1000 my( $self, @edge ) = @_;
1001 # If edge is an arrayref, cope.
1002 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
1003 my $e = shift @edge;
1006 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
1010 # Helper function. Make a display label for the given witnesses, showing a.c.
1011 # witnesses only where the main witness is not also in the list.
1012 sub _path_display_label {
1015 map { $wits{$_} = 1 } @_;
1017 # If an a.c. wit is listed, remove it if the main wit is also listed.
1018 # Otherwise keep it for explicit listing.
1019 my $aclabel = $self->ac_label;
1021 foreach my $w ( sort keys %wits ) {
1022 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
1023 if( exists $wits{$1} ) {
1026 push( @disp_ac, $w );
1031 # See if we are in a majority situation.
1032 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
1033 $maj = $maj > 5 ? $maj : 5;
1034 if( scalar keys %wits > $maj ) {
1035 unshift( @disp_ac, 'majority' );
1036 return join( ', ', @disp_ac );
1038 return join( ', ', sort keys %wits );
1042 =head2 readings_at_rank( $rank )
1044 Returns a list of readings at a given rank, taken from the alignment table.
1048 sub readings_at_rank {
1049 my( $self, $rank ) = @_;
1050 my $table = $self->alignment_table;
1051 # Table rank is real rank - 1.
1052 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
1054 foreach my $e ( @elements ) {
1055 next unless ref( $e ) eq 'HASH';
1056 next unless exists $e->{'t'};
1057 $readings{$e->{'t'}->id} = $e->{'t'};
1059 return values %readings;
1064 Returns a GraphML representation of the collation. The GraphML will contain
1065 two graphs. The first expresses the attributes of the readings and the witness
1066 paths that link them; the second expresses the relationships that link the
1067 readings. This is the native transfer format for a tradition.
1071 use Text::Tradition;
1077 my $datafile = 't/data/florilegium_tei_ps.xml';
1078 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1080 'file' => $datafile,
1083 ok( $tradition, "Got a tradition object" );
1084 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
1085 ok( $tradition->collation, "Tradition has a collation" );
1087 my $c = $tradition->collation;
1088 is( scalar $c->readings, $READINGS, "Collation has all readings" );
1089 is( scalar $c->paths, $PATHS, "Collation has all paths" );
1090 is( scalar $c->relationships, 0, "Collation has all relationships" );
1092 # Add a few relationships
1093 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
1094 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
1095 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
1097 # Now write it to GraphML and parse it again.
1099 my $graphml = $c->as_graphml;
1100 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
1101 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
1102 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
1103 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
1105 # Now add a stemma, write to GraphML, and look at the output.
1107 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
1108 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
1109 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
1110 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
1111 $graphml = $c->as_graphml;
1112 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
1119 ## TODO MOVE this to Tradition.pm and modularize it better
1121 my( $self, $options ) = @_;
1122 $self->calculate_ranks unless $self->_graphcalc_done;
1124 my $start = $options->{'from'}
1125 ? $self->reading( $options->{'from'} ) : $self->start;
1126 my $end = $options->{'to'}
1127 ? $self->reading( $options->{'to'} ) : $self->end;
1128 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1129 throw( 'Start node must be before end node' );
1131 # The readings need to be ranked for this to work.
1132 $start = $self->start unless $start->has_rank;
1133 $end = $self->end unless $end->has_rank;
1135 unless( $start eq $self->start ) {
1136 $rankoffset = $start->rank - 1;
1141 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1142 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1143 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1144 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1146 # Create the document and root node
1147 require XML::LibXML;
1148 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1149 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1150 $graphml->setDocumentElement( $root );
1151 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1152 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1154 # List of attribute types to save on our objects and their corresponding
1159 'Bool' => 'boolean',
1160 'ReadingID' => 'string',
1161 'RelationshipType' => 'string',
1162 'RelationshipScope' => 'string',
1165 # Add the data keys for the graph. Include an extra key 'version' for the
1166 # GraphML output version.
1167 my %graph_data_keys;
1169 my %graph_attributes = ( 'version' => 'string' );
1170 # Graph attributes include those of Tradition and those of Collation.
1172 # TODO Use meta introspection method from duplicate_reading to do this
1173 # instead of naming custom keys.
1174 my $tmeta = $self->tradition->meta;
1175 my $cmeta = $self->meta;
1176 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1177 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1178 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1179 next if $attr->name =~ /^_/;
1180 next unless $save_types{$attr->type_constraint->name};
1181 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1183 # Extra custom keys for complex objects that should be saved in some form.
1184 # The subroutine should return a string, or undef/empty.
1185 if( $tmeta->has_method('stemmata') ) {
1186 $graph_attributes{'stemmata'} = sub {
1188 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1189 $self->tradition->stemmata;
1190 join( "\n", @stemstrs );
1194 if( $tmeta->has_method('user') ) {
1195 $graph_attributes{'user'} = sub {
1196 $self->tradition->user ? $self->tradition->user->id : undef
1200 foreach my $datum ( sort keys %graph_attributes ) {
1201 $graph_data_keys{$datum} = 'dg'.$gdi++;
1202 my $key = $root->addNewChild( $graphml_ns, 'key' );
1203 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1204 : $graph_attributes{$datum};
1205 $key->setAttribute( 'attr.name', $datum );
1206 $key->setAttribute( 'attr.type', $dtype );
1207 $key->setAttribute( 'for', 'graph' );
1208 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1211 # Add the data keys for reading nodes
1212 my %reading_attributes;
1213 my $rmeta = Text::Tradition::Collation::Reading->meta;
1214 foreach my $attr( $rmeta->get_all_attributes ) {
1215 next if $attr->name =~ /^_/;
1216 next unless $save_types{$attr->type_constraint->name};
1217 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1219 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1220 # Extra custom key for the reading morphology
1221 $reading_attributes{'lexemes'} = 'string';
1226 foreach my $datum ( sort keys %reading_attributes ) {
1227 $node_data_keys{$datum} = 'dn'.$ndi++;
1228 my $key = $root->addNewChild( $graphml_ns, 'key' );
1229 $key->setAttribute( 'attr.name', $datum );
1230 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1231 $key->setAttribute( 'for', 'node' );
1232 $key->setAttribute( 'id', $node_data_keys{$datum} );
1235 # Add the data keys for edges, that is, paths and relationships. Path
1236 # data does not come from a Moose class so is here manually.
1239 my %edge_attributes = (
1240 witness => 'string', # ID/label for a path
1241 extra => 'boolean', # Path key
1243 my @path_attributes = keys %edge_attributes; # track our manual additions
1244 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1245 foreach my $attr( $pmeta->get_all_attributes ) {
1246 next if $attr->name =~ /^_/;
1247 next unless $save_types{$attr->type_constraint->name};
1248 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1250 foreach my $datum ( sort keys %edge_attributes ) {
1251 $edge_data_keys{$datum} = 'de'.$edi++;
1252 my $key = $root->addNewChild( $graphml_ns, 'key' );
1253 $key->setAttribute( 'attr.name', $datum );
1254 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1255 $key->setAttribute( 'for', 'edge' );
1256 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1259 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1260 my $xmlidname = $self->tradition->name;
1261 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1262 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1263 $xmlidname = '_'.$xmlidname;
1265 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1266 $sgraph->setAttribute( 'edgedefault', 'directed' );
1267 $sgraph->setAttribute( 'id', $xmlidname );
1268 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1269 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1270 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1271 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1272 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1274 # Tradition/collation attribute data
1275 foreach my $datum ( keys %graph_attributes ) {
1277 if( $datum eq 'version' ) {
1279 } elsif( ref( $graph_attributes{$datum} ) ) {
1280 my $sub = $graph_attributes{$datum};
1282 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1283 $value = $self->tradition->$datum;
1285 $value = $self->$datum;
1287 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1292 # Add our readings to the graph
1293 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1294 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1295 ( $n->rank < $start->rank || $n->rank > $end->rank );
1296 $use_readings{$n->id} = 1;
1297 # Add to the main graph
1298 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1299 my $node_xmlid = 'n' . $node_ctr++;
1300 $node_hash{ $n->id } = $node_xmlid;
1301 $node_el->setAttribute( 'id', $node_xmlid );
1302 foreach my $d ( keys %reading_attributes ) {
1304 # Custom serialization
1305 if( $d eq 'lexemes' ) {
1306 # If nval is a true value, we have lexemes so we need to
1307 # serialize them. Otherwise set nval to undef so that the
1308 # key is excluded from this reading.
1309 $nval = $nval ? $n->_serialize_lexemes : undef;
1310 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1313 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1314 # Adjust the ranks within the subgraph.
1315 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1316 : $nval - $rankoffset;
1318 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1323 # Add the path edges to the sequence graph
1325 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1326 # We add an edge in the graphml for every witness in $e.
1327 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1328 my @edge_wits = sort $self->path_witnesses( $e );
1329 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1330 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1331 # Skip any path from start to end; that witness is not in the subgraph.
1332 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1333 foreach my $wit ( @edge_wits ) {
1334 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1335 $node_hash{ $e->[0] },
1336 $node_hash{ $e->[1] } );
1337 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1338 $edge_el->setAttribute( 'source', $from );
1339 $edge_el->setAttribute( 'target', $to );
1340 $edge_el->setAttribute( 'id', $id );
1342 # It's a witness path, so add the witness
1344 my $key = $edge_data_keys{'witness'};
1345 # Is this an ante-corr witness?
1346 my $aclabel = $self->ac_label;
1347 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1348 # Keep the base witness
1350 # ...and record that this is an 'extra' reading path
1351 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1353 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1357 # Report the actual number of nodes and edges that went in
1358 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1359 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1361 # Add the relationship graph to the XML
1362 map { delete $edge_data_keys{$_} } @path_attributes;
1363 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1364 $node_data_keys{'id'}, \%edge_data_keys );
1366 # Save and return the thing
1367 my $result = decode_utf8( $graphml->toString(1) );
1371 sub _add_graphml_data {
1372 my( $el, $key, $value ) = @_;
1373 return unless defined $value;
1374 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1375 $data_el->setAttribute( 'key', $key );
1376 $data_el->appendText( $value );
1381 Returns a CSV alignment table representation of the collation graph, one
1382 row per witness (or witness uncorrected.)
1388 my $table = $self->alignment_table;
1389 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1391 # Make the header row
1392 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1393 push( @result, decode_utf8( $csv->string ) );
1394 # Make the rest of the rows
1395 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1396 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1397 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1398 $csv->combine( @row );
1399 push( @result, decode_utf8( $csv->string ) );
1401 return join( "\n", @result );
1404 =head2 alignment_table
1406 Return a reference to an alignment table, in a slightly enhanced CollateX
1407 format which looks like this:
1409 $table = { alignment => [ { witness => "SIGIL",
1410 tokens => [ { t => "TEXT" }, ... ] },
1411 { witness => "SIG2",
1412 tokens => [ { t => "TEXT" }, ... ] },
1414 length => TEXTLEN };
1418 sub alignment_table {
1420 return $self->cached_table if $self->has_cached_table;
1422 # Make sure we can do this
1423 throw( "Need a linear graph in order to make an alignment table" )
1424 unless $self->linear;
1425 $self->calculate_ranks()
1426 unless $self->_graphcalc_done && $self->end->has_rank;
1428 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1429 my @all_pos = ( 1 .. $self->end->rank - 1 );
1430 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1431 # say STDERR "Making witness row(s) for " . $wit->sigil;
1432 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1433 my @row = _make_witness_row( \@wit_path, \@all_pos );
1434 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1435 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1436 push( @{$table->{'alignment'}}, $witobj );
1437 if( $wit->is_layered ) {
1438 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1439 $wit->sigil.$self->ac_label );
1440 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1441 my $witacobj = { 'witness' => $wit->sigil.$self->ac_label,
1442 'tokens' => \@ac_row };
1443 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1444 push( @{$table->{'alignment'}}, $witacobj );
1447 $self->cached_table( $table );
1451 sub _make_witness_row {
1452 my( $path, $positions ) = @_;
1454 map { $char_hash{$_} = undef } @$positions;
1456 foreach my $rdg ( @$path ) {
1457 say STDERR "rank " . $rdg->rank if $debug;
1458 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1459 $char_hash{$rdg->rank} = { 't' => $rdg };
1461 my @row = map { $char_hash{$_} } @$positions;
1462 # Fill in lacuna markers for undef spots in the row
1463 my $last_el = shift @row;
1464 my @filled_row = ( $last_el );
1465 foreach my $el ( @row ) {
1466 # If we are using node reference, make the lacuna node appear many times
1467 # in the table. If not, use the lacuna tag.
1468 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1471 push( @filled_row, $el );
1478 =head1 NAVIGATION METHODS
1480 =head2 reading_sequence( $first, $last, $sigil, $backup )
1482 Returns the ordered list of readings, starting with $first and ending
1483 with $last, for the witness given in $sigil. If a $backup sigil is
1484 specified (e.g. when walking a layered witness), it will be used wherever
1485 no $sigil path exists. If there is a base text reading, that will be
1486 used wherever no path exists for $sigil or $backup.
1490 # TODO Think about returning some lazy-eval iterator.
1491 # TODO Get rid of backup; we should know from what witness is whether we need it.
1493 sub reading_sequence {
1494 my( $self, $start, $end, $witness ) = @_;
1496 $witness = $self->baselabel unless $witness;
1497 my @readings = ( $start );
1500 while( $n && $n->id ne $end->id ) {
1501 if( exists( $seen{$n->id} ) ) {
1502 throw( "Detected loop for $witness at " . $n->id );
1506 my $next = $self->next_reading( $n, $witness );
1508 throw( "Did not find any path for $witness from reading " . $n->id );
1510 push( @readings, $next );
1513 # Check that the last reading is our end reading.
1514 my $last = $readings[$#readings];
1515 throw( "Last reading found from " . $start->text .
1516 " for witness $witness is not the end!" ) # TODO do we get this far?
1517 unless $last->id eq $end->id;
1522 =head2 next_reading( $reading, $sigil );
1524 Returns the reading that follows the given reading along the given witness
1530 # Return the successor via the corresponding path.
1532 my $answer = $self->_find_linked_reading( 'next', @_ );
1533 return undef unless $answer;
1534 return $self->reading( $answer );
1537 =head2 prior_reading( $reading, $sigil )
1539 Returns the reading that precedes the given reading along the given witness
1545 # Return the predecessor via the corresponding path.
1547 my $answer = $self->_find_linked_reading( 'prior', @_ );
1548 return $self->reading( $answer );
1551 sub _find_linked_reading {
1552 my( $self, $direction, $node, $path ) = @_;
1554 # Get a backup if we are dealing with a layered witness
1556 my $aclabel = $self->ac_label;
1557 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1561 my @linked_paths = $direction eq 'next'
1562 ? $self->sequence->edges_from( $node )
1563 : $self->sequence->edges_to( $node );
1564 return undef unless scalar( @linked_paths );
1566 # We have to find the linked path that contains all of the
1567 # witnesses supplied in $path.
1568 my( @path_wits, @alt_path_wits );
1569 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1570 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1573 foreach my $le ( @linked_paths ) {
1574 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1577 my @le_wits = sort $self->path_witnesses( $le );
1578 if( _is_within( \@path_wits, \@le_wits ) ) {
1579 # This is the right path.
1580 return $direction eq 'next' ? $le->[1] : $le->[0];
1581 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1585 # Got this far? Return the alternate path if it exists.
1586 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1589 # Got this far? Return the base path if it exists.
1590 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1593 # Got this far? We have no appropriate path.
1594 warn "Could not find $direction node from " . $node->id
1595 . " along path $path";
1601 my( $set1, $set2 ) = @_;
1602 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1603 foreach my $el ( @$set1 ) {
1604 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1609 # Return the string that joins together a list of witnesses for
1610 # display on a single path.
1611 sub _witnesses_of_label {
1612 my( $self, $label ) = @_;
1613 my $regex = $self->wit_list_separator;
1614 my @answer = split( /\Q$regex\E/, $label );
1618 =head2 common_readings
1620 Returns the list of common readings in the graph (i.e. those readings that are
1621 shared by all non-lacunose witnesses.)
1625 sub common_readings {
1627 my @common = grep { $_->is_common } $self->readings;
1631 =head2 path_text( $sigil, [, $start, $end ] )
1633 Returns the text of a witness (plus its backup, if we are using a layer)
1634 as stored in the collation. The text is returned as a string, where the
1635 individual readings are joined with spaces and the meta-readings (e.g.
1636 lacunae) are omitted. Optional specification of $start and $end allows
1637 the generation of a subset of the witness text.
1642 my( $self, $wit, $start, $end ) = @_;
1643 $start = $self->start unless $start;
1644 $end = $self->end unless $end;
1645 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1648 foreach my $r ( @path ) {
1649 unless ( $r->join_prior || !$last || $last->join_next ) {
1652 $pathtext .= $r->text;
1658 =head1 INITIALIZATION METHODS
1660 These are mostly for use by parsers.
1662 =head2 make_witness_path( $witness )
1664 Link the array of readings contained in $witness->path (and in
1665 $witness->uncorrected_path if it exists) into collation paths.
1666 Clear out the arrays when finished.
1668 =head2 make_witness_paths
1670 Call make_witness_path for all witnesses in the tradition.
1674 # For use when a collation is constructed from a base text and an apparatus.
1675 # We have the sequences of readings and just need to add path edges.
1676 # When we are done, clear out the witness path attributes, as they are no
1678 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1680 sub make_witness_paths {
1682 foreach my $wit ( $self->tradition->witnesses ) {
1683 # say STDERR "Making path for " . $wit->sigil;
1684 $self->make_witness_path( $wit );
1688 sub make_witness_path {
1689 my( $self, $wit ) = @_;
1690 my @chain = @{$wit->path};
1691 my $sig = $wit->sigil;
1692 # Add start and end if necessary
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 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1698 if( $wit->is_layered ) {
1699 @chain = @{$wit->uncorrected_path};
1700 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1701 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1702 foreach my $idx( 0 .. $#chain-1 ) {
1703 my $source = $chain[$idx];
1704 my $target = $chain[$idx+1];
1705 $self->add_path( $source, $target, $sig.$self->ac_label )
1706 unless $self->has_path( $source, $target, $sig );
1710 $wit->clear_uncorrected_path;
1713 =head2 calculate_ranks
1715 Calculate the reading ranks (that is, their aligned positions relative
1716 to each other) for the graph. This can only be called on linear collations.
1720 use Text::Tradition;
1722 my $cxfile = 't/data/Collatex-16.xml';
1723 my $t = Text::Tradition->new(
1725 'input' => 'CollateX',
1728 my $c = $t->collation;
1731 my $table = $c->alignment_table;
1732 ok( $c->has_cached_table, "Alignment table was cached" );
1733 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1734 $c->calculate_ranks;
1735 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1736 $c->add_relationship( 'n13', 'n23', { type => 'repetition' } );
1737 is( $c->alignment_table, $table, "Alignment table unchanged after non-colo relationship add" );
1738 $c->add_relationship( 'n24', 'n23', { type => 'spelling' } );
1739 isnt( $c->alignment_table, $table, "Alignment table changed after colo relationship add" );
1745 sub calculate_ranks {
1747 # Save the existing ranks, in case we need to invalidate the cached SVG.
1749 map { $existing_ranks{$_} = $_->rank } $self->readings;
1751 # Do the rankings based on the relationship equivalence graph, starting
1752 # with the start node.
1753 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1755 # Transfer our rankings from the topological graph to the real one.
1756 foreach my $r ( $self->readings ) {
1757 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1758 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1760 # Die. Find the last rank we calculated.
1761 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1762 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1764 my $last = pop @all_defined;
1765 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1768 # Do we need to invalidate the cached data?
1769 if( $self->has_cached_table ) {
1770 foreach my $r ( $self->readings ) {
1771 next if defined( $existing_ranks{$r} )
1772 && $existing_ranks{$r} == $r->rank;
1773 # Something has changed, so clear the cache
1774 $self->_clear_cache;
1775 # ...and recalculate the common readings.
1776 $self->calculate_common_readings();
1780 # The graph calculation information is now up to date.
1781 $self->_graphcalc_done(1);
1786 $self->wipe_table if $self->has_cached_table;
1790 =head2 flatten_ranks
1792 A convenience method for parsing collation data. Searches the graph for readings
1793 with the same text at the same rank, and merges any that are found.
1798 my ( $self, %args ) = shift;
1799 my %unique_rank_rdg;
1801 foreach my $p ( $self->identical_readings( %args ) ) {
1802 # say STDERR "Combining readings at same rank: @$p";
1804 $self->merge_readings( @$p );
1805 # TODO see if this now makes a common point.
1807 # If we merged readings, the ranks are still fine but the alignment
1808 # table is wrong. Wipe it.
1809 $self->wipe_table() if $changed;
1812 =head2 identical_readings
1813 =head2 identical_readings( start => $startnode, end => $endnode )
1814 =head2 identical_readings( startrank => $startrank, endrank => $endrank )
1816 Goes through the graph identifying all pairs of readings that appear to be
1817 identical, and therefore able to be merged into a single reading. Returns the
1818 relevant identical pairs. Can be restricted to run over only a part of the
1819 graph, specified either by node or by rank.
1823 sub identical_readings {
1824 my ( $self, %args ) = @_;
1825 # Find where we should start and end.
1826 my $startrank = $args{startrank} || 0;
1827 if( $args{start} ) {
1828 throw( "Starting reading has no rank" ) unless $self->reading( $args{start} )
1829 && $self->reading( $args{start} )->has_rank;
1830 $startrank = $self->reading( $args{start} )->rank;
1832 my $endrank = $args{endrank} || $self->end->rank;
1834 throw( "Ending reading has no rank" ) unless $self->reading( $args{end} )
1835 && $self->reading( $args{end} )->has_rank;
1836 $endrank = $self->reading( $args{end} )->rank;
1839 # Make sure the ranks are correct.
1840 unless( $self->_graphcalc_done ) {
1841 $self->calculate_ranks;
1843 # Go through the readings looking for duplicates.
1844 my %unique_rank_rdg;
1846 foreach my $rdg ( $self->readings ) {
1847 next unless $rdg->has_rank;
1848 my $rk = $rdg->rank;
1849 next if $rk > $endrank || $rk < $startrank;
1850 my $key = $rk . "||" . $rdg->text;
1851 if( exists $unique_rank_rdg{$key} ) {
1852 # Make sure they don't have different grammatical forms
1853 my $ur = $unique_rank_rdg{$key};
1854 if( $rdg->is_identical( $ur ) ) {
1855 push( @pairs, [ $ur, $rdg ] );
1858 $unique_rank_rdg{$key} = $rdg;
1866 =head2 calculate_common_readings
1868 Goes through the graph identifying the readings that appear in every witness
1869 (apart from those with lacunae at that spot.) Marks them as common and returns
1874 use Text::Tradition;
1876 my $cxfile = 't/data/Collatex-16.xml';
1877 my $t = Text::Tradition->new(
1879 'input' => 'CollateX',
1882 my $c = $t->collation;
1884 my @common = $c->calculate_common_readings();
1885 is( scalar @common, 8, "Found correct number of common readings" );
1886 my @marked = sort $c->common_readings();
1887 is( scalar @common, 8, "All common readings got marked as such" );
1888 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1889 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1895 sub calculate_common_readings {
1898 map { $_->is_common( 0 ) } $self->readings;
1899 # Implicitly calls calculate_ranks
1900 my $table = $self->alignment_table;
1901 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1902 my @row = map { $_->{'tokens'}->[$idx]
1903 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1904 @{$table->{'alignment'}};
1906 foreach my $r ( @row ) {
1908 $hash{$r->id} = $r unless $r->is_meta;
1910 $hash{'UNDEF'} = $r;
1913 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1914 my( $r ) = values %hash;
1916 push( @common, $r );
1922 =head2 text_from_paths
1924 Calculate the text array for all witnesses from the path, for later consistency
1925 checking. Only to be used if there is no non-graph-based way to know the
1930 sub text_from_paths {
1932 foreach my $wit ( $self->tradition->witnesses ) {
1933 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1935 foreach my $r ( @readings ) {
1936 next if $r->is_meta;
1937 push( @text, $r->text );
1939 $wit->text( \@text );
1940 if( $wit->is_layered ) {
1941 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1942 $wit->sigil.$self->ac_label );
1944 foreach my $r ( @ucrdgs ) {
1945 next if $r->is_meta;
1946 push( @uctext, $r->text );
1948 $wit->layertext( \@uctext );
1953 =head1 UTILITY FUNCTIONS
1955 =head2 common_predecessor( $reading_a, $reading_b )
1957 Find the last reading that occurs in sequence before both the given readings.
1958 At the very least this should be $self->start.
1960 =head2 common_successor( $reading_a, $reading_b )
1962 Find the first reading that occurs in sequence after both the given readings.
1963 At the very least this should be $self->end.
1967 use Text::Tradition;
1969 my $cxfile = 't/data/Collatex-16.xml';
1970 my $t = Text::Tradition->new(
1972 'input' => 'CollateX',
1975 my $c = $t->collation;
1977 is( $c->common_predecessor( 'n24', 'n23' )->id,
1978 'n20', "Found correct common predecessor" );
1979 is( $c->common_successor( 'n24', 'n23' )->id,
1980 '__END__', "Found correct common successor" );
1982 is( $c->common_predecessor( 'n19', 'n17' )->id,
1983 'n16', "Found correct common predecessor for readings on same path" );
1984 is( $c->common_successor( 'n21', 'n10' )->id,
1985 '__END__', "Found correct common successor for readings on same path" );
1991 ## Return the closest reading that is a predecessor of both the given readings.
1992 sub common_predecessor {
1994 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1995 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1998 sub common_successor {
2000 my( $r1, $r2 ) = $self->_objectify_args( @_ );
2001 return $self->_common_in_path( $r1, $r2, 'successors' );
2005 # TODO think about how to do this without ranks...
2006 sub _common_in_path {
2007 my( $self, $r1, $r2, $dir ) = @_;
2008 my $iter = $self->end->rank;
2010 my @last_r1 = ( $r1 );
2011 my @last_r2 = ( $r2 );
2012 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
2014 # say STDERR "Finding common $dir for $r1, $r2";
2015 while( !@candidates ) {
2016 last unless $iter--; # Avoid looping infinitely
2017 # Iterate separately down the graph from r1 and r2
2018 my( @new_lc1, @new_lc2 );
2019 foreach my $lc ( @last_r1 ) {
2020 foreach my $p ( $lc->$dir ) {
2021 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
2022 # say STDERR "Path candidate $p from $lc";
2023 push( @candidates, $p );
2024 } elsif( !$all_seen{$p->id} ) {
2025 $all_seen{$p->id} = 'r1';
2026 push( @new_lc1, $p );
2030 foreach my $lc ( @last_r2 ) {
2031 foreach my $p ( $lc->$dir ) {
2032 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
2033 # say STDERR "Path candidate $p from $lc";
2034 push( @candidates, $p );
2035 } elsif( !$all_seen{$p->id} ) {
2036 $all_seen{$p->id} = 'r2';
2037 push( @new_lc2, $p );
2041 @last_r1 = @new_lc1;
2042 @last_r2 = @new_lc2;
2044 my @answer = sort { $a->rank <=> $b->rank } @candidates;
2045 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
2049 Text::Tradition::Error->throw(
2050 'ident' => 'Collation error',
2056 __PACKAGE__->meta->make_immutable;
2062 =item * Rework XML serialization in a more modular way
2068 This package is free software and is provided "as is" without express
2069 or implied warranty. You can redistribute it and/or modify it under
2070 the same terms as Perl itself.
2074 Tara L Andrews E<lt>aurum@cpan.orgE<gt>