1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
7 use IPC::Run qw( run binary );
9 use Text::Tradition::Collation::Reading;
10 use Text::Tradition::Collation::RelationshipStore;
11 use Text::Tradition::Error;
12 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
14 use XML::LibXML::XPathContext;
20 default => sub { Graph->new() },
28 isa => 'Text::Tradition::Collation::RelationshipStore',
30 relationships => 'relationships',
31 related_readings => 'related_readings',
32 get_relationship => 'get_relationship',
33 del_relationship => 'del_relationship',
34 equivalence => 'equivalence',
35 equivalence_graph => 'equivalence_graph',
37 writer => '_set_relations',
42 isa => 'Text::Tradition',
43 writer => '_set_tradition',
48 isa => 'HashRef[Text::Tradition::Collation::Reading]',
52 _add_reading => 'set',
53 del_reading => 'delete',
54 has_reading => 'exists',
57 default => sub { {} },
60 has 'wit_list_separator' => (
69 default => 'base text',
92 isa => 'Text::Tradition::Collation::Reading',
93 writer => '_set_start',
99 isa => 'Text::Tradition::Collation::Reading',
100 writer => '_set_end',
104 has 'cached_svg' => (
107 predicate => 'has_cached_svg',
108 clearer => 'wipe_svg',
111 has 'cached_table' => (
114 predicate => 'has_cached_table',
115 clearer => 'wipe_table',
118 has '_graphcalc_done' => (
126 Text::Tradition::Collation - a software model for a text collation
131 my $t = Text::Tradition->new(
132 'name' => 'this is a text',
134 'file' => '/path/to/tei_parallel_seg_file.xml' );
136 my $c = $t->collation;
137 my @readings = $c->readings;
138 my @paths = $c->paths;
139 my @relationships = $c->relationships;
141 my $svg_variant_graph = $t->collation->as_svg();
145 Text::Tradition is a library for representation and analysis of collated
146 texts, particularly medieval ones. The Collation is the central feature of
147 a Tradition, where the text, its sequence of readings, and its relationships
148 between readings are actually kept.
154 The constructor. Takes a hash or hashref of the following arguments:
158 =item * tradition - The Text::Tradition object to which the collation
161 =item * linear - Whether the collation should be linear; that is, whether
162 transposed readings should be treated as two linked readings rather than one,
163 and therefore whether the collation graph is acyclic. Defaults to true.
165 =item * baselabel - The default label for the path taken by a base text
166 (if any). Defaults to 'base text'.
168 =item * wit_list_separator - The string to join a list of witnesses for
169 purposes of making labels in display graphs. Defaults to ', '.
171 =item * ac_label - The extra label to tack onto a witness sigil when
172 representing another layer of path for the given witness - that is, when
173 a text has more than one possible reading due to scribal corrections or
174 the like. Defaults to ' (a.c.)'.
176 =item * wordsep - The string used to separate words in the original text.
187 =head2 wit_list_separator
195 Simple accessors for collation attributes.
199 The meta-reading at the start of every witness path.
203 The meta-reading at the end of every witness path.
207 Returns all Reading objects in the graph.
209 =head2 reading( $id )
211 Returns the Reading object corresponding to the given ID.
213 =head2 add_reading( $reading_args )
215 Adds a new reading object to the collation.
216 See L<Text::Tradition::Collation::Reading> for the available arguments.
218 =head2 del_reading( $object_or_id )
220 Removes the given reading from the collation, implicitly removing its
221 paths and relationships.
223 =head2 merge_readings( $main, $second, $concatenate, $with_str )
225 Merges the $second reading into the $main one. If $concatenate is true, then
226 the merged node will carry the text of both readings, concatenated with either
227 $with_str (if specified) or a sensible default (the empty string if the
228 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
230 The first two arguments may be either readings or reading IDs.
232 =head2 has_reading( $id )
234 Predicate to see whether a given reading ID is in the graph.
236 =head2 reading_witnesses( $object_or_id )
238 Returns a list of sigils whose witnesses contain the reading.
242 Returns all reading paths within the document - that is, all edges in the
243 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
245 =head2 add_path( $source, $target, $sigil )
247 Links the given readings in the collation in sequence, under the given witness
248 sigil. The readings may be specified by object or ID.
250 =head2 del_path( $source, $target, $sigil )
252 Links the given readings in the collation in sequence, under the given witness
253 sigil. The readings may be specified by object or ID.
255 =head2 has_path( $source, $target );
257 Returns true if the two readings are linked in sequence in any witness.
258 The readings may be specified by object or ID.
262 Returns all Relationship objects in the collation.
264 =head2 add_relationship( $reading, $other_reading, $options )
266 Adds a new relationship of the type given in $options between the two readings,
267 which may be specified by object or ID. Returns a value of ( $status, @vectors)
268 where $status is true on success, and @vectors is a list of relationship edges
269 that were ultimately added.
270 See L<Text::Tradition::Collation::Relationship> for the available options.
276 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
277 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
278 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
281 ### Reading construct/destruct functions
284 my( $self, $reading ) = @_;
285 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
286 my %args = %$reading;
287 if( $self->tradition->has_language && !exists $args{'language'} ) {
288 $args{'language'} = $self->tradition->language;
290 $reading = Text::Tradition::Collation::Reading->new(
291 'collation' => $self,
294 # First check to see if a reading with this ID exists.
295 if( $self->reading( $reading->id ) ) {
296 throw( "Collation already has a reading with id " . $reading->id );
298 $self->_graphcalc_done(0);
299 $self->_add_reading( $reading->id => $reading );
300 # Once the reading has been added, put it in both graphs.
301 $self->sequence->add_vertex( $reading->id );
302 # All meta readings save 'start' and 'end' get disregarded for relationships.
303 unless( $reading->is_nonrel ) {
304 $self->relations->add_reading( $reading->id );
309 around del_reading => sub {
314 unless( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
315 $arg = $self->reading( $arg )
317 my $argid = $arg->id;
318 # Remove the reading from the graphs.
319 $self->_graphcalc_done(0);
320 $self->_clear_cache; # Explicitly clear caches to GC the reading
321 $self->sequence->delete_vertex( $argid );
322 $self->relations->delete_reading( $argid )
323 unless $arg->is_nonrel;
326 $self->$orig( $argid );
333 my $cxfile = 't/data/Collatex-16.xml';
334 my $t = Text::Tradition->new(
336 'input' => 'CollateX',
339 my $c = $t->collation;
341 my $rno = scalar $c->readings;
342 # Split n21 for testing purposes
343 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
344 my $old_r = $c->reading( 'n21' );
345 $old_r->alter_text( 'to' );
346 $c->del_path( 'n20', 'n21', 'A' );
347 $c->add_path( 'n20', 'n21p0', 'A' );
348 $c->add_path( 'n21p0', 'n21', 'A' );
350 ok( $c->reading( 'n21p0' ), "New reading exists" );
351 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
353 # Combine n3 and n4 ( with his )
354 $c->merge_readings( 'n3', 'n4', 1 );
355 ok( !$c->reading('n4'), "Reading n4 is gone" );
356 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
358 # Collapse n9 and n10 ( rood / root )
359 $c->merge_readings( 'n9', 'n10' );
360 ok( !$c->reading('n10'), "Reading n10 is gone" );
361 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
363 # Combine n21 and n21p0
364 my $remaining = $c->reading('n21');
365 $remaining ||= $c->reading('n22'); # one of these should still exist
366 $c->merge_readings( 'n21p0', $remaining, 1 );
367 ok( !$c->reading('n21'), "Reading $remaining is gone" );
368 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
378 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
379 my $mergemeta = $kept_obj->is_meta;
380 throw( "Cannot merge meta and non-meta reading" )
381 unless ( $mergemeta && $del_obj->is_meta )
382 || ( !$mergemeta && !$del_obj->is_meta );
384 throw( "Cannot merge with start or end node" )
385 if( $kept_obj eq $self->start || $kept_obj eq $self->end
386 || $del_obj eq $self->start || $del_obj eq $self->end );
388 # We only need the IDs for adding paths to the graph, not the reading
389 # objects themselves.
390 my $kept = $kept_obj->id;
391 my $deleted = $del_obj->id;
392 $self->_graphcalc_done(0);
394 # The kept reading should inherit the paths and the relationships
395 # of the deleted reading.
396 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
397 my @vector = ( $kept );
398 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
399 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
400 next if $vector[0] eq $vector[1]; # Don't add a self loop
401 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
402 $self->sequence->add_edge( @vector );
403 my $fwits = $self->sequence->get_edge_attributes( @vector );
404 @wits{keys %$fwits} = values %$fwits;
405 $self->sequence->set_edge_attributes( @vector, \%wits );
407 $self->relations->merge_readings( $kept, $deleted, $combine )
410 # Do the deletion deed.
412 my $joinstr = $combine_char;
413 unless( defined $joinstr ) {
414 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
415 $joinstr = $self->wordsep unless defined $joinstr;
417 $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) );
419 $self->del_reading( $deleted );
423 # Helper function for manipulating the graph.
424 sub _stringify_args {
425 my( $self, $first, $second, @args ) = @_;
427 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
428 $second = $second->id
429 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
430 return( $first, $second, @args );
433 # Helper function for manipulating the graph.
434 sub _objectify_args {
435 my( $self, $first, $second, $arg ) = @_;
436 $first = $self->reading( $first )
437 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
438 $second = $self->reading( $second )
439 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
440 return( $first, $second, $arg );
447 # We only need the IDs for adding paths to the graph, not the reading
448 # objects themselves.
449 my( $source, $target, $wit ) = $self->_objectify_args( @_ );
451 $self->_graphcalc_done(0);
452 # Connect the readings
453 unless( $self->sequence->has_edge( $source, $target ) ) {
454 $self->sequence->add_edge( $source, $target );
455 $self->relations->add_equivalence_edge( $source, $target );
457 # Note the witness in question
458 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
464 if( ref( $_[0] ) eq 'ARRAY' ) {
471 # We only need the IDs for adding paths to the graph, not the reading
472 # objects themselves.
473 my( $source, $target, $wit ) = $self->_stringify_args( @args );
475 $self->_graphcalc_done(0);
476 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
477 $self->sequence->delete_edge_attribute( $source, $target, $wit );
479 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
480 $self->sequence->delete_edge( $source, $target );
481 $self->relations->delete_equivalence_edge( $source, $target );
486 # Extra graph-alike utility
489 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
490 return undef unless $self->sequence->has_edge( $source, $target );
491 return $self->sequence->has_edge_attribute( $source, $target, $wit );
494 =head2 clear_witness( @sigil_list )
496 Clear the given witnesses out of the collation entirely, removing references
497 to them in paths, and removing readings that belong only to them. Should only
498 be called via $tradition->del_witness.
503 my( $self, @sigils ) = @_;
505 $self->_graphcalc_done(0);
506 # Clear the witness(es) out of the paths
507 foreach my $e ( $self->paths ) {
508 foreach my $sig ( @sigils ) {
509 $self->del_path( $e, $sig );
513 # Clear out the newly unused readings
514 foreach my $r ( $self->readings ) {
515 unless( $self->reading_witnesses( $r ) ) {
516 $self->del_reading( $r );
521 sub add_relationship {
523 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
524 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
525 $self->_graphcalc_done(0);
529 around qw/ get_relationship del_relationship / => sub {
533 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
536 my( $source, $target ) = $self->_stringify_args( @args );
537 $self->$orig( $source, $target );
540 =head2 reading_witnesses( $reading )
542 Return a list of sigils corresponding to the witnesses in which the reading appears.
546 sub reading_witnesses {
547 my( $self, $reading ) = @_;
548 # We need only check either the incoming or the outgoing edges; I have
549 # arbitrarily chosen "incoming". Thus, special-case the start node.
550 if( $reading eq $self->start ) {
551 return map { $_->sigil } $self->tradition->witnesses;
554 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
555 my $wits = $self->sequence->get_edge_attributes( @$e );
556 @all_witnesses{ keys %$wits } = 1;
558 my $acstr = $self->ac_label;
559 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
560 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
562 return keys %all_witnesses;
565 =head1 OUTPUT METHODS
567 =head2 as_svg( \%options )
569 Returns an SVG string that represents the graph, via as_dot and graphviz.
570 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
575 my( $self, $opts ) = @_;
576 throw( "Need GraphViz installed to output SVG" )
577 unless File::Which::which( 'dot' );
578 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
579 $self->calculate_ranks()
580 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
581 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
582 my @cmd = qw/dot -Tsvg/;
584 my $dotfile = File::Temp->new();
586 # $dotfile->unlink_on_destroy(0);
587 binmode $dotfile, ':utf8';
588 print $dotfile $self->as_dot( $opts );
589 push( @cmd, $dotfile->filename );
590 run( \@cmd, ">", binary(), \$svg );
591 $svg = decode_utf8( $svg );
592 $self->cached_svg( $svg ) unless $want_subgraph;
595 return $self->cached_svg;
600 =head2 as_dot( \%options )
602 Returns a string that is the collation graph expressed in dot
603 (i.e. GraphViz) format. Options include:
618 my( $self, $opts ) = @_;
619 my $startrank = $opts->{'from'} if $opts;
620 my $endrank = $opts->{'to'} if $opts;
621 my $color_common = $opts->{'color_common'} if $opts;
622 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
623 && $self->end->rank > 100;
624 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
626 # Check the arguments
628 return if $endrank && $startrank > $endrank;
629 return if $startrank > $self->end->rank;
631 if( defined $endrank ) {
632 return if $endrank < 0;
633 $endrank = undef if $endrank == $self->end->rank;
636 my $graph_name = $self->tradition->name;
637 $graph_name =~ s/[^\w\s]//g;
638 $graph_name = join( '_', split( /\s+/, $graph_name ) );
646 'fillcolor' => 'white',
651 'arrowhead' => 'open',
652 'color' => '#000000',
653 'fontcolor' => '#000000',
656 my $dot = sprintf( "digraph %s {\n", $graph_name );
657 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
658 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
660 # Output substitute start/end readings if necessary
662 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
665 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
667 if( $STRAIGHTENHACK ) {
669 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
670 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
671 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
673 my %used; # Keep track of the readings that actually appear in the graph
674 # Sort the readings by rank if we have ranks; this speeds layout.
675 my @all_readings = $self->end->has_rank
676 ? sort { $a->rank <=> $b->rank } $self->readings
678 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
679 foreach my $reading ( @all_readings ) {
680 # Only output readings within our rank range.
681 next if $startrank && $reading->rank < $startrank;
682 next if $endrank && $reading->rank > $endrank;
683 $used{$reading->id} = 1;
684 # Need not output nodes without separate labels
685 next if $reading->id eq $reading->text;
687 my $label = $reading->text;
688 $label .= '-' if $reading->join_next;
689 $label = "-$label" if $reading->join_prior;
690 $label =~ s/\"/\\\"/g;
691 $rattrs->{'label'} = $label;
692 $rattrs->{'id'} = $reading->id;
693 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
694 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
697 # Add the real edges. Need to weight one edge per rank jump, in a
699 # my $weighted = $self->_add_edge_weights;
700 my @edges = $self->paths;
701 my( %substart, %subend );
702 foreach my $edge ( @edges ) {
703 # Do we need to output this edge?
704 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
705 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
706 my $variables = { %edge_attrs, 'label' => $label };
708 # Account for the rank gap if necessary
709 my $rank0 = $self->reading( $edge->[0] )->rank
710 if $self->reading( $edge->[0] )->has_rank;
711 my $rank1 = $self->reading( $edge->[1] )->rank
712 if $self->reading( $edge->[1] )->has_rank;
713 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
714 $variables->{'minlen'} = $rank1 - $rank0;
717 # Add the calculated edge weights
718 # if( exists $weighted->{$edge->[0]}
719 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
720 # # $variables->{'color'} = 'red';
721 # $variables->{'weight'} = 3.0;
724 # EXPERIMENTAL: make edge width reflect no. of witnesses
725 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
726 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
728 my $varopts = _dot_attr_string( $variables );
729 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
730 $edge->[0], $edge->[1], $varopts );
731 } elsif( $used{$edge->[0]} ) {
732 $subend{$edge->[0]} = 1;
733 } elsif( $used{$edge->[1]} ) {
734 $substart{$edge->[1]} = 1;
737 # Add substitute start and end edges if necessary
738 foreach my $node ( keys %substart ) {
739 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
740 my $variables = { %edge_attrs, 'label' => $witstr };
741 my $varopts = _dot_attr_string( $variables );
742 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;";
744 foreach my $node ( keys %subend ) {
745 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
746 my $variables = { %edge_attrs, 'label' => $witstr };
747 my $varopts = _dot_attr_string( $variables );
748 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;";
751 if( $STRAIGHTENHACK ) {
752 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
753 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
760 sub _dot_attr_string {
763 foreach my $k ( sort keys %$hash ) {
765 push( @attrs, $k.'="'.$v.'"' );
767 return( '[ ' . join( ', ', @attrs ) . ' ]' );
770 sub _add_edge_weights {
772 # Walk the graph from START to END, choosing the successor node with
773 # the largest number of witness paths each time.
775 my $curr = $self->start->id;
776 my $ranked = $self->end->has_rank;
777 while( $curr ne $self->end->id ) {
778 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
779 my @succ = sort { $self->path_witnesses( $curr, $a )
780 <=> $self->path_witnesses( $curr, $b ) }
781 $self->sequence->successors( $curr );
782 my $next = pop @succ;
783 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
784 # Try to avoid lacunae in the weighted path.
786 ( $self->reading( $next )->is_lacuna ||
787 $nextrank - $rank > 1 ) ){
790 $weighted->{$curr} = $next;
796 =head2 path_witnesses( $edge )
798 Returns the list of sigils whose witnesses are associated with the given edge.
799 The edge can be passed as either an array or an arrayref of ( $source, $target ).
804 my( $self, @edge ) = @_;
805 # If edge is an arrayref, cope.
806 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
810 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
814 # Helper function. Make a display label for the given witnesses, showing a.c.
815 # witnesses only where the main witness is not also in the list.
816 sub _path_display_label {
819 map { $wits{$_} = 1 } @_;
821 # If an a.c. wit is listed, remove it if the main wit is also listed.
822 # Otherwise keep it for explicit listing.
823 my $aclabel = $self->ac_label;
825 foreach my $w ( sort keys %wits ) {
826 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
827 if( exists $wits{$1} ) {
830 push( @disp_ac, $w );
835 # See if we are in a majority situation.
836 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
837 $maj = $maj > 5 ? $maj : 5;
838 if( scalar keys %wits > $maj ) {
839 unshift( @disp_ac, 'majority' );
840 return join( ', ', @disp_ac );
842 return join( ', ', sort keys %wits );
846 =head2 readings_at_rank( $rank )
848 Returns a list of readings at a given rank, taken from the alignment table.
852 sub readings_at_rank {
853 my( $self, $rank ) = @_;
854 my $table = $self->alignment_table;
855 # Table rank is real rank - 1.
856 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
858 foreach my $e ( @elements ) {
859 next unless ref( $e ) eq 'HASH';
860 next unless exists $e->{'t'};
861 $readings{$e->{'t'}->id} = $e->{'t'};
863 return values %readings;
868 Returns a GraphML representation of the collation. The GraphML will contain
869 two graphs. The first expresses the attributes of the readings and the witness
870 paths that link them; the second expresses the relationships that link the
871 readings. This is the native transfer format for a tradition.
880 my $datafile = 't/data/florilegium_tei_ps.xml';
881 my $tradition = Text::Tradition->new( 'input' => 'TEI',
886 ok( $tradition, "Got a tradition object" );
887 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
888 ok( $tradition->collation, "Tradition has a collation" );
890 my $c = $tradition->collation;
891 is( scalar $c->readings, $READINGS, "Collation has all readings" );
892 is( scalar $c->paths, $PATHS, "Collation has all paths" );
893 is( scalar $c->relationships, 0, "Collation has all relationships" );
895 # Add a few relationships
896 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
897 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
898 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
900 # Now write it to GraphML and parse it again.
902 my $graphml = $c->as_graphml;
903 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
904 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
905 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
906 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
908 # Now add a stemma, write to GraphML, and parse again.
909 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
910 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
911 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
912 $graphml = $c->as_graphml;
913 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
920 my( $self, $options ) = @_;
921 $self->calculate_ranks unless $self->_graphcalc_done;
923 my $start = $options->{'from'}
924 ? $self->reading( $options->{'from'} ) : $self->start;
925 my $end = $options->{'to'}
926 ? $self->reading( $options->{'to'} ) : $self->end;
927 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
928 throw( 'Start node must be before end node' );
930 # The readings need to be ranked for this to work.
931 $start = $self->start unless $start->has_rank;
932 $end = $self->end unless $end->has_rank;
934 unless( $start eq $self->start ) {
935 $rankoffset = $start->rank - 1;
940 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
941 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
942 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
943 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
945 # Create the document and root node
946 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
947 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
948 $graphml->setDocumentElement( $root );
949 $root->setNamespace( $xsi_ns, 'xsi', 0 );
950 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
952 # List of attribute types to save on our objects and their corresponding
958 'ReadingID' => 'string',
959 'RelationshipType' => 'string',
960 'RelationshipScope' => 'string',
963 # List of attribute names *not* to save on our objects.
964 # We will also not save any attribute beginning with _.
966 map { $skipsave{$_} = 1 } qw/ cached_svg /;
968 # Add the data keys for the graph. Include an extra key 'version' for the
969 # GraphML output version.
972 my %graph_attributes = ( 'version' => 'string' );
973 # Graph attributes include those of Tradition and those of Collation.
975 my $tmeta = $self->tradition->meta;
976 my $cmeta = $self->meta;
977 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
978 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
979 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
980 next if $attr->name =~ /^_/;
981 next if $skipsave{$attr->name};
982 next unless $save_types{$attr->type_constraint->name};
983 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
985 # Extra custom key for the tradition stemma(ta)
986 $graph_attributes{'stemmata'} = 'string';
988 foreach my $datum ( sort keys %graph_attributes ) {
989 $graph_data_keys{$datum} = 'dg'.$gdi++;
990 my $key = $root->addNewChild( $graphml_ns, 'key' );
991 $key->setAttribute( 'attr.name', $datum );
992 $key->setAttribute( 'attr.type', $graph_attributes{$datum} );
993 $key->setAttribute( 'for', 'graph' );
994 $key->setAttribute( 'id', $graph_data_keys{$datum} );
997 # Add the data keys for reading nodes
998 my %reading_attributes;
999 my $rmeta = Text::Tradition::Collation::Reading->meta;
1000 foreach my $attr( $rmeta->get_all_attributes ) {
1001 next if $attr->name =~ /^_/;
1002 next if $skipsave{$attr->name};
1003 next unless $save_types{$attr->type_constraint->name};
1004 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1006 # Extra custom key for the reading morphology
1007 $reading_attributes{'lexemes'} = 'string';
1011 foreach my $datum ( sort keys %reading_attributes ) {
1012 $node_data_keys{$datum} = 'dn'.$ndi++;
1013 my $key = $root->addNewChild( $graphml_ns, 'key' );
1014 $key->setAttribute( 'attr.name', $datum );
1015 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1016 $key->setAttribute( 'for', 'node' );
1017 $key->setAttribute( 'id', $node_data_keys{$datum} );
1020 # Add the data keys for edges, that is, paths and relationships. Path
1021 # data does not come from a Moose class so is here manually.
1024 my %edge_attributes = (
1025 witness => 'string', # ID/label for a path
1026 extra => 'boolean', # Path key
1028 my @path_attributes = keys %edge_attributes; # track our manual additions
1029 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1030 foreach my $attr( $pmeta->get_all_attributes ) {
1031 next if $attr->name =~ /^_/;
1032 next if $skipsave{$attr->name};
1033 next unless $save_types{$attr->type_constraint->name};
1034 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1036 foreach my $datum ( sort keys %edge_attributes ) {
1037 $edge_data_keys{$datum} = 'de'.$edi++;
1038 my $key = $root->addNewChild( $graphml_ns, 'key' );
1039 $key->setAttribute( 'attr.name', $datum );
1040 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1041 $key->setAttribute( 'for', 'edge' );
1042 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1045 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1046 my $xmlidname = $self->tradition->name;
1047 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1048 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1049 $xmlidname = '_'.$xmlidname;
1051 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1052 $sgraph->setAttribute( 'edgedefault', 'directed' );
1053 $sgraph->setAttribute( 'id', $xmlidname );
1054 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1055 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1056 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1057 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1058 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1060 # Tradition/collation attribute data
1061 foreach my $datum ( keys %graph_attributes ) {
1063 if( $datum eq 'version' ) {
1065 } elsif( $datum eq 'stemmata' ) {
1067 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1068 $self->tradition->stemmata;
1069 $value = join( "\n", @stemstrs );
1070 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1071 $value = $self->tradition->$datum;
1073 $value = $self->$datum;
1075 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1080 # Add our readings to the graph
1081 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1082 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1083 ( $n->rank < $start->rank || $n->rank > $end->rank );
1084 $use_readings{$n->id} = 1;
1085 # Add to the main graph
1086 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1087 my $node_xmlid = 'n' . $node_ctr++;
1088 $node_hash{ $n->id } = $node_xmlid;
1089 $node_el->setAttribute( 'id', $node_xmlid );
1090 foreach my $d ( keys %reading_attributes ) {
1092 # Custom serialization
1093 if( $d eq 'lexemes' ) {
1094 # If nval is a true value, we have lexemes so we need to
1095 # serialize them. Otherwise set nval to undef so that the
1096 # key is excluded from this reading.
1097 $nval = $nval ? $n->_serialize_lexemes : undef;
1098 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1101 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1102 # Adjust the ranks within the subgraph.
1103 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1104 : $nval - $rankoffset;
1106 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1111 # Add the path edges to the sequence graph
1113 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1114 # We add an edge in the graphml for every witness in $e.
1115 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1116 my @edge_wits = sort $self->path_witnesses( $e );
1117 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1118 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1119 # Skip any path from start to end; that witness is not in the subgraph.
1120 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1121 foreach my $wit ( @edge_wits ) {
1122 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1123 $node_hash{ $e->[0] },
1124 $node_hash{ $e->[1] } );
1125 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1126 $edge_el->setAttribute( 'source', $from );
1127 $edge_el->setAttribute( 'target', $to );
1128 $edge_el->setAttribute( 'id', $id );
1130 # It's a witness path, so add the witness
1132 my $key = $edge_data_keys{'witness'};
1133 # Is this an ante-corr witness?
1134 my $aclabel = $self->ac_label;
1135 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1136 # Keep the base witness
1138 # ...and record that this is an 'extra' reading path
1139 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1141 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1145 # Report the actual number of nodes and edges that went in
1146 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1147 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1149 # Add the relationship graph to the XML
1150 map { delete $edge_data_keys{$_} } @path_attributes;
1151 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1152 $node_data_keys{'id'}, \%edge_data_keys );
1154 # Save and return the thing
1155 my $result = decode_utf8( $graphml->toString(1) );
1159 sub _add_graphml_data {
1160 my( $el, $key, $value ) = @_;
1161 return unless defined $value;
1162 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1163 $data_el->setAttribute( 'key', $key );
1164 $data_el->appendText( $value );
1169 Returns a CSV alignment table representation of the collation graph, one
1170 row per witness (or witness uncorrected.)
1176 my $table = $self->alignment_table;
1177 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1179 # Make the header row
1180 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1181 push( @result, decode_utf8( $csv->string ) );
1182 # Make the rest of the rows
1183 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1184 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1185 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1186 $csv->combine( @row );
1187 push( @result, decode_utf8( $csv->string ) );
1189 return join( "\n", @result );
1192 =head2 alignment_table( $use_refs, $include_witnesses )
1194 Return a reference to an alignment table, in a slightly enhanced CollateX
1195 format which looks like this:
1197 $table = { alignment => [ { witness => "SIGIL",
1198 tokens => [ { t => "TEXT" }, ... ] },
1199 { witness => "SIG2",
1200 tokens => [ { t => "TEXT" }, ... ] },
1202 length => TEXTLEN };
1204 If $use_refs is set to 1, the reading object is returned in the table
1205 instead of READINGTEXT; if not, the text of the reading is returned.
1207 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1208 keys have a true hash value will be included.
1212 sub alignment_table {
1214 $self->calculate_ranks() unless $self->_graphcalc_done;
1215 return $self->cached_table if $self->has_cached_table;
1217 # Make sure we can do this
1218 throw( "Need a linear graph in order to make an alignment table" )
1219 unless $self->linear;
1220 $self->calculate_ranks unless $self->end->has_rank;
1222 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1223 my @all_pos = ( 1 .. $self->end->rank - 1 );
1224 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1225 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
1226 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1227 my @row = _make_witness_row( \@wit_path, \@all_pos );
1228 push( @{$table->{'alignment'}},
1229 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1230 if( $wit->is_layered ) {
1231 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1232 $wit->sigil.$self->ac_label );
1233 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1234 push( @{$table->{'alignment'}},
1235 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1238 $self->cached_table( $table );
1242 sub _make_witness_row {
1243 my( $path, $positions ) = @_;
1245 map { $char_hash{$_} = undef } @$positions;
1247 foreach my $rdg ( @$path ) {
1248 my $rtext = $rdg->text;
1249 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1250 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1251 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1252 $char_hash{$rdg->rank} = { 't' => $rdg };
1254 my @row = map { $char_hash{$_} } @$positions;
1255 # Fill in lacuna markers for undef spots in the row
1256 my $last_el = shift @row;
1257 my @filled_row = ( $last_el );
1258 foreach my $el ( @row ) {
1259 # If we are using node reference, make the lacuna node appear many times
1260 # in the table. If not, use the lacuna tag.
1261 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1264 push( @filled_row, $el );
1270 =head1 NAVIGATION METHODS
1272 =head2 reading_sequence( $first, $last, $sigil, $backup )
1274 Returns the ordered list of readings, starting with $first and ending
1275 with $last, for the witness given in $sigil. If a $backup sigil is
1276 specified (e.g. when walking a layered witness), it will be used wherever
1277 no $sigil path exists. If there is a base text reading, that will be
1278 used wherever no path exists for $sigil or $backup.
1282 # TODO Think about returning some lazy-eval iterator.
1283 # TODO Get rid of backup; we should know from what witness is whether we need it.
1285 sub reading_sequence {
1286 my( $self, $start, $end, $witness ) = @_;
1288 $witness = $self->baselabel unless $witness;
1289 my @readings = ( $start );
1292 while( $n && $n->id ne $end->id ) {
1293 if( exists( $seen{$n->id} ) ) {
1294 throw( "Detected loop for $witness at " . $n->id );
1298 my $next = $self->next_reading( $n, $witness );
1300 throw( "Did not find any path for $witness from reading " . $n->id );
1302 push( @readings, $next );
1305 # Check that the last reading is our end reading.
1306 my $last = $readings[$#readings];
1307 throw( "Last reading found from " . $start->text .
1308 " for witness $witness is not the end!" ) # TODO do we get this far?
1309 unless $last->id eq $end->id;
1314 =head2 next_reading( $reading, $sigil );
1316 Returns the reading that follows the given reading along the given witness
1322 # Return the successor via the corresponding path.
1324 my $answer = $self->_find_linked_reading( 'next', @_ );
1325 return undef unless $answer;
1326 return $self->reading( $answer );
1329 =head2 prior_reading( $reading, $sigil )
1331 Returns the reading that precedes the given reading along the given witness
1337 # Return the predecessor via the corresponding path.
1339 my $answer = $self->_find_linked_reading( 'prior', @_ );
1340 return $self->reading( $answer );
1343 sub _find_linked_reading {
1344 my( $self, $direction, $node, $path ) = @_;
1346 # Get a backup if we are dealing with a layered witness
1348 my $aclabel = $self->ac_label;
1349 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1353 my @linked_paths = $direction eq 'next'
1354 ? $self->sequence->edges_from( $node )
1355 : $self->sequence->edges_to( $node );
1356 return undef unless scalar( @linked_paths );
1358 # We have to find the linked path that contains all of the
1359 # witnesses supplied in $path.
1360 my( @path_wits, @alt_path_wits );
1361 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1362 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1365 foreach my $le ( @linked_paths ) {
1366 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1369 my @le_wits = sort $self->path_witnesses( $le );
1370 if( _is_within( \@path_wits, \@le_wits ) ) {
1371 # This is the right path.
1372 return $direction eq 'next' ? $le->[1] : $le->[0];
1373 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1377 # Got this far? Return the alternate path if it exists.
1378 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1381 # Got this far? Return the base path if it exists.
1382 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1385 # Got this far? We have no appropriate path.
1386 warn "Could not find $direction node from " . $node->id
1387 . " along path $path";
1393 my( $set1, $set2 ) = @_;
1394 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1395 foreach my $el ( @$set1 ) {
1396 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1401 # Return the string that joins together a list of witnesses for
1402 # display on a single path.
1403 sub _witnesses_of_label {
1404 my( $self, $label ) = @_;
1405 my $regex = $self->wit_list_separator;
1406 my @answer = split( /\Q$regex\E/, $label );
1410 =head2 common_readings
1412 Returns the list of common readings in the graph (i.e. those readings that are
1413 shared by all non-lacunose witnesses.)
1417 sub common_readings {
1419 my @common = grep { $_->is_common } $self->readings;
1423 =head2 path_text( $sigil, [, $start, $end ] )
1425 Returns the text of a witness (plus its backup, if we are using a layer)
1426 as stored in the collation. The text is returned as a string, where the
1427 individual readings are joined with spaces and the meta-readings (e.g.
1428 lacunae) are omitted. Optional specification of $start and $end allows
1429 the generation of a subset of the witness text.
1434 my( $self, $wit, $start, $end ) = @_;
1435 $start = $self->start unless $start;
1436 $end = $self->end unless $end;
1437 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1440 foreach my $r ( @path ) {
1441 unless ( $r->join_prior || !$last || $last->join_next ) {
1444 $pathtext .= $r->text;
1450 =head1 INITIALIZATION METHODS
1452 These are mostly for use by parsers.
1454 =head2 make_witness_path( $witness )
1456 Link the array of readings contained in $witness->path (and in
1457 $witness->uncorrected_path if it exists) into collation paths.
1458 Clear out the arrays when finished.
1460 =head2 make_witness_paths
1462 Call make_witness_path for all witnesses in the tradition.
1466 # For use when a collation is constructed from a base text and an apparatus.
1467 # We have the sequences of readings and just need to add path edges.
1468 # When we are done, clear out the witness path attributes, as they are no
1470 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1472 sub make_witness_paths {
1474 foreach my $wit ( $self->tradition->witnesses ) {
1475 # print STDERR "Making path for " . $wit->sigil . "\n";
1476 $self->make_witness_path( $wit );
1480 sub make_witness_path {
1481 my( $self, $wit ) = @_;
1482 my @chain = @{$wit->path};
1483 my $sig = $wit->sigil;
1484 # Add start and end if necessary
1485 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1486 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1487 foreach my $idx ( 0 .. $#chain-1 ) {
1488 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1490 if( $wit->is_layered ) {
1491 @chain = @{$wit->uncorrected_path};
1492 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1493 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1494 foreach my $idx( 0 .. $#chain-1 ) {
1495 my $source = $chain[$idx];
1496 my $target = $chain[$idx+1];
1497 $self->add_path( $source, $target, $sig.$self->ac_label )
1498 unless $self->has_path( $source, $target, $sig );
1502 $wit->clear_uncorrected_path;
1505 =head2 calculate_ranks
1507 Calculate the reading ranks (that is, their aligned positions relative
1508 to each other) for the graph. This can only be called on linear collations.
1512 use Text::Tradition;
1514 my $cxfile = 't/data/Collatex-16.xml';
1515 my $t = Text::Tradition->new(
1517 'input' => 'CollateX',
1520 my $c = $t->collation;
1523 my $table = $c->alignment_table;
1524 ok( $c->has_cached_table, "Alignment table was cached" );
1525 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1526 $c->calculate_ranks;
1527 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1528 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1529 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1535 sub calculate_ranks {
1537 # Save the existing ranks, in case we need to invalidate the cached SVG.
1539 map { $existing_ranks{$_} = $_->rank } $self->readings;
1541 # Do the rankings based on the relationship equivalence graph, starting
1542 # with the start node.
1543 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1545 # Transfer our rankings from the topological graph to the real one.
1546 foreach my $r ( $self->readings ) {
1547 if( $r->is_nonrel ) {
1548 # These are not in the equivalence graph. Grab the rank of the highest
1550 my @preds = $self->sequence->predecessors( $r );
1552 map { my $rk = $node_ranks->{$self->equivalence( $_ )} + 1;
1553 $mrank = $rk > $mrank ? $rk : $mrank; }
1554 $self->sequence->predecessors( $r );
1555 throw( "All predecessors of $r unranked!" ) unless $mrank;
1557 } elsif( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1558 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1560 # Die. Find the last rank we calculated.
1561 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1562 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1564 my $last = pop @all_defined;
1565 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1568 # Do we need to invalidate the cached data?
1569 if( $self->has_cached_svg || $self->has_cached_table ) {
1570 foreach my $r ( $self->readings ) {
1571 next if defined( $existing_ranks{$r} )
1572 && $existing_ranks{$r} == $r->rank;
1573 # Something has changed, so clear the cache
1574 $self->_clear_cache;
1575 # ...and recalculate the common readings.
1576 $self->calculate_common_readings();
1580 # The graph calculation information is now up to date.
1581 $self->_graphcalc_done(1);
1586 $self->wipe_svg if $self->has_cached_svg;
1587 $self->wipe_table if $self->has_cached_table;
1591 =head2 flatten_ranks
1593 A convenience method for parsing collation data. Searches the graph for readings
1594 with the same text at the same rank, and merges any that are found.
1600 my %unique_rank_rdg;
1602 foreach my $rdg ( $self->readings ) {
1603 next unless $rdg->has_rank;
1604 my $key = $rdg->rank . "||" . $rdg->text;
1605 if( exists $unique_rank_rdg{$key} ) {
1606 # Make sure they don't have different grammatical forms
1607 my $ur = $unique_rank_rdg{$key};
1608 if( $rdg->disambiguated && $ur->disambiguated ) {
1609 my $rform = join( '//', map { $_->form->to_string } $rdg->lexemes );
1610 my $uform = join( '//', map { $_->form->to_string } $ur->lexemes );
1611 next unless $rform eq $uform;
1612 } elsif( $rdg->disambiguated xor $ur->disambiguated ) {
1616 print STDERR "Combining readings at same rank: $key\n";
1618 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1619 # TODO see if this now makes a common point.
1621 $unique_rank_rdg{$key} = $rdg;
1624 # If we merged readings, the ranks are still fine but the alignment
1625 # table is wrong. Wipe it.
1626 $self->wipe_table() if $changed;
1630 =head2 calculate_common_readings
1632 Goes through the graph identifying the readings that appear in every witness
1633 (apart from those with lacunae at that spot.) Marks them as common and returns
1638 use Text::Tradition;
1640 my $cxfile = 't/data/Collatex-16.xml';
1641 my $t = Text::Tradition->new(
1643 'input' => 'CollateX',
1646 my $c = $t->collation;
1648 my @common = $c->calculate_common_readings();
1649 is( scalar @common, 8, "Found correct number of common readings" );
1650 my @marked = sort $c->common_readings();
1651 is( scalar @common, 8, "All common readings got marked as such" );
1652 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1653 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1659 sub calculate_common_readings {
1662 map { $_->is_common( 0 ) } $self->readings;
1663 # Implicitly calls calculate_ranks
1664 my $table = $self->alignment_table;
1665 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1666 my @row = map { $_->{'tokens'}->[$idx]
1667 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1668 @{$table->{'alignment'}};
1670 foreach my $r ( @row ) {
1672 $hash{$r->id} = $r unless $r->is_meta;
1674 $hash{'UNDEF'} = $r;
1677 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1678 my( $r ) = values %hash;
1680 push( @common, $r );
1686 =head2 text_from_paths
1688 Calculate the text array for all witnesses from the path, for later consistency
1689 checking. Only to be used if there is no non-graph-based way to know the
1694 sub text_from_paths {
1696 foreach my $wit ( $self->tradition->witnesses ) {
1697 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1699 foreach my $r ( @readings ) {
1700 next if $r->is_meta;
1701 push( @text, $r->text );
1703 $wit->text( \@text );
1704 if( $wit->is_layered ) {
1705 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1706 $wit->sigil.$self->ac_label );
1708 foreach my $r ( @ucrdgs ) {
1709 next if $r->is_meta;
1710 push( @uctext, $r->text );
1712 $wit->layertext( \@uctext );
1717 =head1 UTILITY FUNCTIONS
1719 =head2 common_predecessor( $reading_a, $reading_b )
1721 Find the last reading that occurs in sequence before both the given readings.
1722 At the very least this should be $self->start.
1724 =head2 common_successor( $reading_a, $reading_b )
1726 Find the first reading that occurs in sequence after both the given readings.
1727 At the very least this should be $self->end.
1731 use Text::Tradition;
1733 my $cxfile = 't/data/Collatex-16.xml';
1734 my $t = Text::Tradition->new(
1736 'input' => 'CollateX',
1739 my $c = $t->collation;
1741 is( $c->common_predecessor( 'n24', 'n23' )->id,
1742 'n20', "Found correct common predecessor" );
1743 is( $c->common_successor( 'n24', 'n23' )->id,
1744 '__END__', "Found correct common successor" );
1746 is( $c->common_predecessor( 'n19', 'n17' )->id,
1747 'n16', "Found correct common predecessor for readings on same path" );
1748 is( $c->common_successor( 'n21', 'n10' )->id,
1749 '__END__', "Found correct common successor for readings on same path" );
1755 ## Return the closest reading that is a predecessor of both the given readings.
1756 sub common_predecessor {
1758 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1759 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1762 sub common_successor {
1764 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1765 return $self->_common_in_path( $r1, $r2, 'successors' );
1769 # TODO think about how to do this without ranks...
1770 sub _common_in_path {
1771 my( $self, $r1, $r2, $dir ) = @_;
1772 my $iter = $self->end->rank;
1774 my @last_r1 = ( $r1 );
1775 my @last_r2 = ( $r2 );
1776 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1778 # print STDERR "Finding common $dir for $r1, $r2\n";
1779 while( !@candidates ) {
1780 last unless $iter--; # Avoid looping infinitely
1781 # Iterate separately down the graph from r1 and r2
1782 my( @new_lc1, @new_lc2 );
1783 foreach my $lc ( @last_r1 ) {
1784 foreach my $p ( $lc->$dir ) {
1785 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1786 # print STDERR "Path candidate $p from $lc\n";
1787 push( @candidates, $p );
1788 } elsif( !$all_seen{$p->id} ) {
1789 $all_seen{$p->id} = 'r1';
1790 push( @new_lc1, $p );
1794 foreach my $lc ( @last_r2 ) {
1795 foreach my $p ( $lc->$dir ) {
1796 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1797 # print STDERR "Path candidate $p from $lc\n";
1798 push( @candidates, $p );
1799 } elsif( !$all_seen{$p->id} ) {
1800 $all_seen{$p->id} = 'r2';
1801 push( @new_lc2, $p );
1805 @last_r1 = @new_lc1;
1806 @last_r2 = @new_lc2;
1808 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1809 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1813 Text::Tradition::Error->throw(
1814 'ident' => 'Collation error',
1820 __PACKAGE__->meta->make_immutable;
1824 This package is free software and is provided "as is" without express
1825 or implied warranty. You can redistribute it and/or modify it under
1826 the same terms as Perl itself.
1830 Tara L Andrews E<lt>aurum@cpan.orgE<gt>