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 $self->relations->add_reading( $reading->id );
306 around del_reading => sub {
311 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
314 # Remove the reading from the graphs.
315 $self->_graphcalc_done(0);
316 $self->_clear_cache; # Explicitly clear caches to GC the reading
317 $self->sequence->delete_vertex( $arg );
318 $self->relations->delete_reading( $arg );
321 $self->$orig( $arg );
328 my $cxfile = 't/data/Collatex-16.xml';
329 my $t = Text::Tradition->new(
331 'input' => 'CollateX',
334 my $c = $t->collation;
336 my $rno = scalar $c->readings;
337 # Split n21 for testing purposes
338 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
339 my $old_r = $c->reading( 'n21' );
340 $old_r->alter_text( 'to' );
341 $c->del_path( 'n20', 'n21', 'A' );
342 $c->add_path( 'n20', 'n21p0', 'A' );
343 $c->add_path( 'n21p0', 'n21', 'A' );
345 ok( $c->reading( 'n21p0' ), "New reading exists" );
346 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
348 # Combine n3 and n4 ( with his )
349 $c->merge_readings( 'n3', 'n4', 1 );
350 ok( !$c->reading('n4'), "Reading n4 is gone" );
351 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
353 # Collapse n9 and n10 ( rood / root )
354 $c->merge_readings( 'n9', 'n10' );
355 ok( !$c->reading('n10'), "Reading n10 is gone" );
356 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
358 # Combine n21 and n21p0
359 my $remaining = $c->reading('n21');
360 $remaining ||= $c->reading('n22'); # one of these should still exist
361 $c->merge_readings( 'n21p0', $remaining, 1 );
362 ok( !$c->reading('n21'), "Reading $remaining is gone" );
363 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
372 # We only need the IDs for adding paths to the graph, not the reading
373 # objects themselves.
374 my( $kept, $deleted, $combine, $combine_char ) = $self->_stringify_args( @_ );
375 $self->_graphcalc_done(0);
377 # The kept reading should inherit the paths and the relationships
378 # of the deleted reading.
379 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
380 my @vector = ( $kept );
381 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
382 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
383 next if $vector[0] eq $vector[1]; # Don't add a self loop
384 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
385 $self->sequence->add_edge( @vector );
386 my $fwits = $self->sequence->get_edge_attributes( @vector );
387 @wits{keys %$fwits} = values %$fwits;
388 $self->sequence->set_edge_attributes( @vector, \%wits );
390 $self->relations->merge_readings( $kept, $deleted, $combine_char );
392 # Do the deletion deed.
394 my $kept_obj = $self->reading( $kept );
395 my $del_obj = $self->reading( $deleted );
396 my $joinstr = $combine_char;
397 unless( defined $joinstr ) {
398 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
399 $joinstr = $self->wordsep unless defined $joinstr;
401 $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) );
403 $self->del_reading( $deleted );
407 # Helper function for manipulating the graph.
408 sub _stringify_args {
409 my( $self, $first, $second, @args ) = @_;
411 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
412 $second = $second->id
413 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
414 return( $first, $second, @args );
417 # Helper function for manipulating the graph.
418 sub _objectify_args {
419 my( $self, $first, $second, $arg ) = @_;
420 $first = $self->reading( $first )
421 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
422 $second = $self->reading( $second )
423 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
424 return( $first, $second, $arg );
431 # We only need the IDs for adding paths to the graph, not the reading
432 # objects themselves.
433 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
435 $self->_graphcalc_done(0);
436 # Connect the readings
437 unless( $self->sequence->has_edge( $source, $target ) ) {
438 $self->sequence->add_edge( $source, $target );
439 $self->relations->add_equivalence_edge( $source, $target );
441 # Note the witness in question
442 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
448 if( ref( $_[0] ) eq 'ARRAY' ) {
455 # We only need the IDs for adding paths to the graph, not the reading
456 # objects themselves.
457 my( $source, $target, $wit ) = $self->_stringify_args( @args );
459 $self->_graphcalc_done(0);
460 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
461 $self->sequence->delete_edge_attribute( $source, $target, $wit );
463 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
464 $self->sequence->delete_edge( $source, $target );
465 $self->relations->delete_equivalence_edge( $source, $target );
470 # Extra graph-alike utility
473 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
474 return undef unless $self->sequence->has_edge( $source, $target );
475 return $self->sequence->has_edge_attribute( $source, $target, $wit );
478 =head2 clear_witness( @sigil_list )
480 Clear the given witnesses out of the collation entirely, removing references
481 to them in paths, and removing readings that belong only to them. Should only
482 be called via $tradition->del_witness.
487 my( $self, @sigils ) = @_;
489 $self->_graphcalc_done(0);
490 # Clear the witness(es) out of the paths
491 foreach my $e ( $self->paths ) {
492 foreach my $sig ( @sigils ) {
493 $self->del_path( $e, $sig );
497 # Clear out the newly unused readings
498 foreach my $r ( $self->readings ) {
499 unless( $self->reading_witnesses( $r ) ) {
500 $self->del_reading( $r );
505 sub add_relationship {
507 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
508 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
509 $self->_graphcalc_done(0);
513 around qw/ get_relationship del_relationship / => sub {
517 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
520 my( $source, $target ) = $self->_stringify_args( @args );
521 $self->$orig( $source, $target );
524 =head2 reading_witnesses( $reading )
526 Return a list of sigils corresponding to the witnesses in which the reading appears.
530 sub reading_witnesses {
531 my( $self, $reading ) = @_;
532 # We need only check either the incoming or the outgoing edges; I have
533 # arbitrarily chosen "incoming". Thus, special-case the start node.
534 if( $reading eq $self->start ) {
535 return map { $_->sigil } $self->tradition->witnesses;
538 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
539 my $wits = $self->sequence->get_edge_attributes( @$e );
540 @all_witnesses{ keys %$wits } = 1;
542 my $acstr = $self->ac_label;
543 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
544 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
546 return keys %all_witnesses;
549 =head1 OUTPUT METHODS
551 =head2 as_svg( \%options )
553 Returns an SVG string that represents the graph, via as_dot and graphviz.
554 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
559 my( $self, $opts ) = @_;
560 throw( "Need GraphViz installed to output SVG" )
561 unless File::Which::which( 'dot' );
562 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
563 $self->calculate_ranks() unless( $self->_graphcalc_done || $opts->{'nocalc'} );
564 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
565 my @cmd = qw/dot -Tsvg/;
567 my $dotfile = File::Temp->new();
569 # $dotfile->unlink_on_destroy(0);
570 binmode $dotfile, ':utf8';
571 print $dotfile $self->as_dot( $opts );
572 push( @cmd, $dotfile->filename );
573 run( \@cmd, ">", binary(), \$svg );
574 $svg = decode_utf8( $svg );
575 $self->cached_svg( $svg ) unless $want_subgraph;
578 return $self->cached_svg;
583 =head2 as_dot( \%options )
585 Returns a string that is the collation graph expressed in dot
586 (i.e. GraphViz) format. Options include:
601 my( $self, $opts ) = @_;
602 my $startrank = $opts->{'from'} if $opts;
603 my $endrank = $opts->{'to'} if $opts;
604 my $color_common = $opts->{'color_common'} if $opts;
605 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
606 && $self->end->rank > 100;
607 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
609 # Check the arguments
611 return if $endrank && $startrank > $endrank;
612 return if $startrank > $self->end->rank;
614 if( defined $endrank ) {
615 return if $endrank < 0;
616 $endrank = undef if $endrank == $self->end->rank;
619 my $graph_name = $self->tradition->name;
620 $graph_name =~ s/[^\w\s]//g;
621 $graph_name = join( '_', split( /\s+/, $graph_name ) );
629 'fillcolor' => 'white',
634 'arrowhead' => 'open',
635 'color' => '#000000',
636 'fontcolor' => '#000000',
639 my $dot = sprintf( "digraph %s {\n", $graph_name );
640 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
641 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
643 # Output substitute start/end readings if necessary
645 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
648 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
650 if( $STRAIGHTENHACK ) {
652 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
653 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
654 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
656 my %used; # Keep track of the readings that actually appear in the graph
657 # Sort the readings by rank if we have ranks; this speeds layout.
658 my @all_readings = $self->end->has_rank
659 ? sort { $a->rank <=> $b->rank } $self->readings
661 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
662 foreach my $reading ( @all_readings ) {
663 # Only output readings within our rank range.
664 next if $startrank && $reading->rank < $startrank;
665 next if $endrank && $reading->rank > $endrank;
666 $used{$reading->id} = 1;
667 # Need not output nodes without separate labels
668 next if $reading->id eq $reading->text;
670 my $label = $reading->text;
671 $label .= '-' if $reading->join_next;
672 $label = "-$label" if $reading->join_prior;
673 $label =~ s/\"/\\\"/g;
674 $rattrs->{'label'} = $label;
675 $rattrs->{'id'} = $reading->id;
676 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
677 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
680 # Add the real edges. Need to weight one edge per rank jump, in a
682 # my $weighted = $self->_add_edge_weights;
683 my @edges = $self->paths;
684 my( %substart, %subend );
685 foreach my $edge ( @edges ) {
686 # Do we need to output this edge?
687 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
688 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
689 my $variables = { %edge_attrs, 'label' => $label };
691 # Account for the rank gap if necessary
692 my $rank0 = $self->reading( $edge->[0] )->rank
693 if $self->reading( $edge->[0] )->has_rank;
694 my $rank1 = $self->reading( $edge->[1] )->rank
695 if $self->reading( $edge->[1] )->has_rank;
696 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
697 $variables->{'minlen'} = $rank1 - $rank0;
700 # Add the calculated edge weights
701 # if( exists $weighted->{$edge->[0]}
702 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
703 # # $variables->{'color'} = 'red';
704 # $variables->{'weight'} = 3.0;
707 # EXPERIMENTAL: make edge width reflect no. of witnesses
708 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
709 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
711 my $varopts = _dot_attr_string( $variables );
712 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
713 $edge->[0], $edge->[1], $varopts );
714 } elsif( $used{$edge->[0]} ) {
715 $subend{$edge->[0]} = 1;
716 } elsif( $used{$edge->[1]} ) {
717 $substart{$edge->[1]} = 1;
720 # Add substitute start and end edges if necessary
721 foreach my $node ( keys %substart ) {
722 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
723 my $variables = { %edge_attrs, 'label' => $witstr };
724 my $varopts = _dot_attr_string( $variables );
725 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;";
727 foreach my $node ( keys %subend ) {
728 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
729 my $variables = { %edge_attrs, 'label' => $witstr };
730 my $varopts = _dot_attr_string( $variables );
731 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;";
734 if( $STRAIGHTENHACK ) {
735 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
736 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
743 sub _dot_attr_string {
746 foreach my $k ( sort keys %$hash ) {
748 push( @attrs, $k.'="'.$v.'"' );
750 return( '[ ' . join( ', ', @attrs ) . ' ]' );
753 sub _add_edge_weights {
755 # Walk the graph from START to END, choosing the successor node with
756 # the largest number of witness paths each time.
758 my $curr = $self->start->id;
759 my $ranked = $self->end->has_rank;
760 while( $curr ne $self->end->id ) {
761 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
762 my @succ = sort { $self->path_witnesses( $curr, $a )
763 <=> $self->path_witnesses( $curr, $b ) }
764 $self->sequence->successors( $curr );
765 my $next = pop @succ;
766 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
767 # Try to avoid lacunae in the weighted path.
769 ( $self->reading( $next )->is_lacuna ||
770 $nextrank - $rank > 1 ) ){
773 $weighted->{$curr} = $next;
779 =head2 path_witnesses( $edge )
781 Returns the list of sigils whose witnesses are associated with the given edge.
782 The edge can be passed as either an array or an arrayref of ( $source, $target ).
787 my( $self, @edge ) = @_;
788 # If edge is an arrayref, cope.
789 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
793 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
797 # Helper function. Make a display label for the given witnesses, showing a.c.
798 # witnesses only where the main witness is not also in the list.
799 sub _path_display_label {
802 map { $wits{$_} = 1 } @_;
804 # If an a.c. wit is listed, remove it if the main wit is also listed.
805 # Otherwise keep it for explicit listing.
806 my $aclabel = $self->ac_label;
808 foreach my $w ( sort keys %wits ) {
809 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
810 if( exists $wits{$1} ) {
813 push( @disp_ac, $w );
818 # See if we are in a majority situation.
819 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
820 if( scalar keys %wits > $maj ) {
821 unshift( @disp_ac, 'majority' );
822 return join( ', ', @disp_ac );
824 return join( ', ', sort keys %wits );
828 =head2 readings_at_rank( $rank )
830 Returns a list of readings at a given rank, taken from the alignment table.
834 sub readings_at_rank {
835 my( $self, $rank ) = @_;
836 my $table = $self->alignment_table;
837 # Table rank is real rank - 1.
838 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
840 foreach my $e ( @elements ) {
841 next unless ref( $e ) eq 'HASH';
842 next unless exists $e->{'t'};
843 $readings{$e->{'t'}->id} = $e->{'t'};
845 return values %readings;
850 Returns a GraphML representation of the collation. The GraphML will contain
851 two graphs. The first expresses the attributes of the readings and the witness
852 paths that link them; the second expresses the relationships that link the
853 readings. This is the native transfer format for a tradition.
862 my $datafile = 't/data/florilegium_tei_ps.xml';
863 my $tradition = Text::Tradition->new( 'input' => 'TEI',
868 ok( $tradition, "Got a tradition object" );
869 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
870 ok( $tradition->collation, "Tradition has a collation" );
872 my $c = $tradition->collation;
873 is( scalar $c->readings, $READINGS, "Collation has all readings" );
874 is( scalar $c->paths, $PATHS, "Collation has all paths" );
875 is( scalar $c->relationships, 0, "Collation has all relationships" );
877 # Add a few relationships
878 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
879 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
880 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
882 # Now write it to GraphML and parse it again.
884 my $graphml = $c->as_graphml;
885 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
886 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
887 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
888 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
890 # Now add a stemma, write to GraphML, and parse again.
891 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
892 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
893 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
894 $graphml = $c->as_graphml;
895 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
902 my( $self, $options ) = @_;
903 $self->calculate_ranks unless $self->_graphcalc_done;
905 my $start = $options->{'from'}
906 ? $self->reading( $options->{'from'} ) : $self->start;
907 my $end = $options->{'to'}
908 ? $self->reading( $options->{'to'} ) : $self->end;
909 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
910 throw( 'Start node must be before end node' );
912 # The readings need to be ranked for this to work.
913 $start = $self->start unless $start->has_rank;
914 $end = $self->end unless $end->has_rank;
916 unless( $start eq $self->start ) {
917 $rankoffset = $start->rank - 1;
922 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
923 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
924 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
925 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
927 # Create the document and root node
928 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
929 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
930 $graphml->setDocumentElement( $root );
931 $root->setNamespace( $xsi_ns, 'xsi', 0 );
932 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
934 # List of attribute types to save on our objects and their corresponding
940 'ReadingID' => 'string',
941 'RelationshipType' => 'string',
942 'RelationshipScope' => 'string',
945 # List of attribute names *not* to save on our objects.
946 # We will also not save any attribute beginning with _.
948 map { $skipsave{$_} = 1 } qw/ cached_svg /;
950 # Add the data keys for the graph. Include an extra key 'version' for the
951 # GraphML output version.
954 my %graph_attributes = ( 'version' => 'string' );
955 # Graph attributes include those of Tradition and those of Collation.
957 my $tmeta = $self->tradition->meta;
958 my $cmeta = $self->meta;
959 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
960 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
961 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
962 next if $attr->name =~ /^_/;
963 next if $skipsave{$attr->name};
964 next unless $save_types{$attr->type_constraint->name};
965 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
967 # Extra custom key for the tradition stemma(ta)
968 $graph_attributes{'stemmata'} = 'string';
970 foreach my $datum ( sort keys %graph_attributes ) {
971 $graph_data_keys{$datum} = 'dg'.$gdi++;
972 my $key = $root->addNewChild( $graphml_ns, 'key' );
973 $key->setAttribute( 'attr.name', $datum );
974 $key->setAttribute( 'attr.type', $graph_attributes{$datum} );
975 $key->setAttribute( 'for', 'graph' );
976 $key->setAttribute( 'id', $graph_data_keys{$datum} );
979 # Add the data keys for reading nodes
980 my %reading_attributes;
981 my $rmeta = Text::Tradition::Collation::Reading->meta;
982 foreach my $attr( $rmeta->get_all_attributes ) {
983 next if $attr->name =~ /^_/;
984 next if $skipsave{$attr->name};
985 next unless $save_types{$attr->type_constraint->name};
986 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
988 # Extra custom key for the reading morphology
989 $reading_attributes{'lexemes'} = 'string';
993 foreach my $datum ( sort keys %reading_attributes ) {
994 $node_data_keys{$datum} = 'dn'.$ndi++;
995 my $key = $root->addNewChild( $graphml_ns, 'key' );
996 $key->setAttribute( 'attr.name', $datum );
997 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
998 $key->setAttribute( 'for', 'node' );
999 $key->setAttribute( 'id', $node_data_keys{$datum} );
1002 # Add the data keys for edges, that is, paths and relationships. Path
1003 # data does not come from a Moose class so is here manually.
1006 my %edge_attributes = (
1007 witness => 'string', # ID/label for a path
1008 extra => 'boolean', # Path key
1010 my @path_attributes = keys %edge_attributes; # track our manual additions
1011 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1012 foreach my $attr( $pmeta->get_all_attributes ) {
1013 next if $attr->name =~ /^_/;
1014 next if $skipsave{$attr->name};
1015 next unless $save_types{$attr->type_constraint->name};
1016 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1018 foreach my $datum ( sort keys %edge_attributes ) {
1019 $edge_data_keys{$datum} = 'de'.$edi++;
1020 my $key = $root->addNewChild( $graphml_ns, 'key' );
1021 $key->setAttribute( 'attr.name', $datum );
1022 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1023 $key->setAttribute( 'for', 'edge' );
1024 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1027 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1028 my $xmlidname = $self->tradition->name;
1029 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1030 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1031 $xmlidname = '_'.$xmlidname;
1033 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1034 $sgraph->setAttribute( 'edgedefault', 'directed' );
1035 $sgraph->setAttribute( 'id', $xmlidname );
1036 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1037 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1038 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1039 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1040 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1042 # Tradition/collation attribute data
1043 foreach my $datum ( keys %graph_attributes ) {
1045 if( $datum eq 'version' ) {
1047 } elsif( $datum eq 'stemmata' ) {
1049 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1050 $self->tradition->stemmata;
1051 $value = join( "\n", @stemstrs );
1052 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1053 $value = $self->tradition->$datum;
1055 $value = $self->$datum;
1057 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1062 # Add our readings to the graph
1063 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1064 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1065 ( $n->rank < $start->rank || $n->rank > $end->rank );
1066 $use_readings{$n->id} = 1;
1067 # Add to the main graph
1068 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1069 my $node_xmlid = 'n' . $node_ctr++;
1070 $node_hash{ $n->id } = $node_xmlid;
1071 $node_el->setAttribute( 'id', $node_xmlid );
1072 foreach my $d ( keys %reading_attributes ) {
1074 # Custom serialization
1075 if( $d eq 'lexemes' ) {
1076 # If nval is a true value, we have lexemes so we need to
1077 # serialize them. Otherwise set nval to undef so that the
1078 # key is excluded from this reading.
1079 $nval = $nval ? $n->_serialize_lexemes : undef;
1081 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1082 # Adjust the ranks within the subgraph.
1083 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1084 : $nval - $rankoffset;
1086 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1091 # Add the path edges to the sequence graph
1093 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1094 # We add an edge in the graphml for every witness in $e.
1095 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1096 my @edge_wits = sort $self->path_witnesses( $e );
1097 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1098 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1099 # Skip any path from start to end; that witness is not in the subgraph.
1100 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1101 foreach my $wit ( @edge_wits ) {
1102 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1103 $node_hash{ $e->[0] },
1104 $node_hash{ $e->[1] } );
1105 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1106 $edge_el->setAttribute( 'source', $from );
1107 $edge_el->setAttribute( 'target', $to );
1108 $edge_el->setAttribute( 'id', $id );
1110 # It's a witness path, so add the witness
1112 my $key = $edge_data_keys{'witness'};
1113 # Is this an ante-corr witness?
1114 my $aclabel = $self->ac_label;
1115 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1116 # Keep the base witness
1118 # ...and record that this is an 'extra' reading path
1119 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1121 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1125 # Report the actual number of nodes and edges that went in
1126 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1127 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1129 # Add the relationship graph to the XML
1130 map { delete $edge_data_keys{$_} } @path_attributes;
1131 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1132 $node_data_keys{'id'}, \%edge_data_keys );
1134 # Save and return the thing
1135 my $result = decode_utf8( $graphml->toString(1) );
1139 sub _add_graphml_data {
1140 my( $el, $key, $value ) = @_;
1141 return unless defined $value;
1142 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1143 $data_el->setAttribute( 'key', $key );
1144 $data_el->appendText( $value );
1149 Returns a CSV alignment table representation of the collation graph, one
1150 row per witness (or witness uncorrected.)
1156 my $table = $self->alignment_table;
1157 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1159 # Make the header row
1160 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1161 push( @result, decode_utf8( $csv->string ) );
1162 # Make the rest of the rows
1163 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1164 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1165 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1166 $csv->combine( @row );
1167 push( @result, decode_utf8( $csv->string ) );
1169 return join( "\n", @result );
1172 =head2 alignment_table( $use_refs, $include_witnesses )
1174 Return a reference to an alignment table, in a slightly enhanced CollateX
1175 format which looks like this:
1177 $table = { alignment => [ { witness => "SIGIL",
1178 tokens => [ { t => "TEXT" }, ... ] },
1179 { witness => "SIG2",
1180 tokens => [ { t => "TEXT" }, ... ] },
1182 length => TEXTLEN };
1184 If $use_refs is set to 1, the reading object is returned in the table
1185 instead of READINGTEXT; if not, the text of the reading is returned.
1187 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1188 keys have a true hash value will be included.
1192 sub alignment_table {
1194 $self->calculate_ranks() unless $self->_graphcalc_done;
1195 return $self->cached_table if $self->has_cached_table;
1197 # Make sure we can do this
1198 throw( "Need a linear graph in order to make an alignment table" )
1199 unless $self->linear;
1200 $self->calculate_ranks unless $self->end->has_rank;
1202 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1203 my @all_pos = ( 1 .. $self->end->rank - 1 );
1204 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1205 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
1206 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1207 my @row = _make_witness_row( \@wit_path, \@all_pos );
1208 push( @{$table->{'alignment'}},
1209 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1210 if( $wit->is_layered ) {
1211 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1212 $wit->sigil.$self->ac_label );
1213 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1214 push( @{$table->{'alignment'}},
1215 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1218 $self->cached_table( $table );
1222 sub _make_witness_row {
1223 my( $path, $positions ) = @_;
1225 map { $char_hash{$_} = undef } @$positions;
1227 foreach my $rdg ( @$path ) {
1228 my $rtext = $rdg->text;
1229 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1230 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1231 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1232 $char_hash{$rdg->rank} = { 't' => $rdg };
1234 my @row = map { $char_hash{$_} } @$positions;
1235 # Fill in lacuna markers for undef spots in the row
1236 my $last_el = shift @row;
1237 my @filled_row = ( $last_el );
1238 foreach my $el ( @row ) {
1239 # If we are using node reference, make the lacuna node appear many times
1240 # in the table. If not, use the lacuna tag.
1241 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1244 push( @filled_row, $el );
1250 =head1 NAVIGATION METHODS
1252 =head2 reading_sequence( $first, $last, $sigil, $backup )
1254 Returns the ordered list of readings, starting with $first and ending
1255 with $last, for the witness given in $sigil. If a $backup sigil is
1256 specified (e.g. when walking a layered witness), it will be used wherever
1257 no $sigil path exists. If there is a base text reading, that will be
1258 used wherever no path exists for $sigil or $backup.
1262 # TODO Think about returning some lazy-eval iterator.
1263 # TODO Get rid of backup; we should know from what witness is whether we need it.
1265 sub reading_sequence {
1266 my( $self, $start, $end, $witness ) = @_;
1268 $witness = $self->baselabel unless $witness;
1269 my @readings = ( $start );
1272 while( $n && $n->id ne $end->id ) {
1273 if( exists( $seen{$n->id} ) ) {
1274 throw( "Detected loop for $witness at " . $n->id );
1278 my $next = $self->next_reading( $n, $witness );
1280 throw( "Did not find any path for $witness from reading " . $n->id );
1282 push( @readings, $next );
1285 # Check that the last reading is our end reading.
1286 my $last = $readings[$#readings];
1287 throw( "Last reading found from " . $start->text .
1288 " for witness $witness is not the end!" ) # TODO do we get this far?
1289 unless $last->id eq $end->id;
1294 =head2 next_reading( $reading, $sigil );
1296 Returns the reading that follows the given reading along the given witness
1302 # Return the successor via the corresponding path.
1304 my $answer = $self->_find_linked_reading( 'next', @_ );
1305 return undef unless $answer;
1306 return $self->reading( $answer );
1309 =head2 prior_reading( $reading, $sigil )
1311 Returns the reading that precedes the given reading along the given witness
1317 # Return the predecessor via the corresponding path.
1319 my $answer = $self->_find_linked_reading( 'prior', @_ );
1320 return $self->reading( $answer );
1323 sub _find_linked_reading {
1324 my( $self, $direction, $node, $path ) = @_;
1326 # Get a backup if we are dealing with a layered witness
1328 my $aclabel = $self->ac_label;
1329 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1333 my @linked_paths = $direction eq 'next'
1334 ? $self->sequence->edges_from( $node )
1335 : $self->sequence->edges_to( $node );
1336 return undef unless scalar( @linked_paths );
1338 # We have to find the linked path that contains all of the
1339 # witnesses supplied in $path.
1340 my( @path_wits, @alt_path_wits );
1341 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1342 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1345 foreach my $le ( @linked_paths ) {
1346 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1349 my @le_wits = sort $self->path_witnesses( $le );
1350 if( _is_within( \@path_wits, \@le_wits ) ) {
1351 # This is the right path.
1352 return $direction eq 'next' ? $le->[1] : $le->[0];
1353 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1357 # Got this far? Return the alternate path if it exists.
1358 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1361 # Got this far? Return the base path if it exists.
1362 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1365 # Got this far? We have no appropriate path.
1366 warn "Could not find $direction node from " . $node->id
1367 . " along path $path";
1373 my( $set1, $set2 ) = @_;
1374 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1375 foreach my $el ( @$set1 ) {
1376 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1381 # Return the string that joins together a list of witnesses for
1382 # display on a single path.
1383 sub _witnesses_of_label {
1384 my( $self, $label ) = @_;
1385 my $regex = $self->wit_list_separator;
1386 my @answer = split( /\Q$regex\E/, $label );
1390 =head2 common_readings
1392 Returns the list of common readings in the graph (i.e. those readings that are
1393 shared by all non-lacunose witnesses.)
1397 sub common_readings {
1399 my @common = grep { $_->is_common } $self->readings;
1403 =head2 path_text( $sigil, [, $start, $end ] )
1405 Returns the text of a witness (plus its backup, if we are using a layer)
1406 as stored in the collation. The text is returned as a string, where the
1407 individual readings are joined with spaces and the meta-readings (e.g.
1408 lacunae) are omitted. Optional specification of $start and $end allows
1409 the generation of a subset of the witness text.
1414 my( $self, $wit, $start, $end ) = @_;
1415 $start = $self->start unless $start;
1416 $end = $self->end unless $end;
1417 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1420 foreach my $r ( @path ) {
1421 unless ( $r->join_prior || !$last || $last->join_next ) {
1424 $pathtext .= $r->text;
1430 =head1 INITIALIZATION METHODS
1432 These are mostly for use by parsers.
1434 =head2 make_witness_path( $witness )
1436 Link the array of readings contained in $witness->path (and in
1437 $witness->uncorrected_path if it exists) into collation paths.
1438 Clear out the arrays when finished.
1440 =head2 make_witness_paths
1442 Call make_witness_path for all witnesses in the tradition.
1446 # For use when a collation is constructed from a base text and an apparatus.
1447 # We have the sequences of readings and just need to add path edges.
1448 # When we are done, clear out the witness path attributes, as they are no
1450 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1452 sub make_witness_paths {
1454 foreach my $wit ( $self->tradition->witnesses ) {
1455 # print STDERR "Making path for " . $wit->sigil . "\n";
1456 $self->make_witness_path( $wit );
1460 sub make_witness_path {
1461 my( $self, $wit ) = @_;
1462 my @chain = @{$wit->path};
1463 my $sig = $wit->sigil;
1464 # Add start and end if necessary
1465 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1466 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1467 foreach my $idx ( 0 .. $#chain-1 ) {
1468 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1470 if( $wit->is_layered ) {
1471 @chain = @{$wit->uncorrected_path};
1472 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1473 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1474 foreach my $idx( 0 .. $#chain-1 ) {
1475 my $source = $chain[$idx];
1476 my $target = $chain[$idx+1];
1477 $self->add_path( $source, $target, $sig.$self->ac_label )
1478 unless $self->has_path( $source, $target, $sig );
1482 $wit->clear_uncorrected_path;
1485 =head2 calculate_ranks
1487 Calculate the reading ranks (that is, their aligned positions relative
1488 to each other) for the graph. This can only be called on linear collations.
1492 use Text::Tradition;
1494 my $cxfile = 't/data/Collatex-16.xml';
1495 my $t = Text::Tradition->new(
1497 'input' => 'CollateX',
1500 my $c = $t->collation;
1503 my $table = $c->alignment_table;
1504 ok( $c->has_cached_table, "Alignment table was cached" );
1505 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1506 $c->calculate_ranks;
1507 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1508 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1509 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1515 sub calculate_ranks {
1517 # Save the existing ranks, in case we need to invalidate the cached SVG.
1519 map { $existing_ranks{$_} = $_->rank } $self->readings;
1521 # Do the rankings based on the relationship equivalence graph, starting
1522 # with the start node.
1523 my $topo_start = $self->equivalence( $self->start->id );
1524 my $node_ranks = { $topo_start => 0 };
1525 my @curr_origin = ( $topo_start );
1526 # A little iterative function.
1527 while( @curr_origin ) {
1528 @curr_origin = _assign_rank( $self->equivalence_graph,
1529 $node_ranks, @curr_origin );
1531 # Transfer our rankings from the topological graph to the real one.
1532 foreach my $r ( $self->readings ) {
1533 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1534 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1536 # Die. Find the last rank we calculated.
1537 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1538 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1540 my $last = pop @all_defined;
1541 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1544 # Do we need to invalidate the cached data?
1545 if( $self->has_cached_svg || $self->has_cached_table ) {
1546 foreach my $r ( $self->readings ) {
1547 next if defined( $existing_ranks{$r} )
1548 && $existing_ranks{$r} == $r->rank;
1549 # Something has changed, so clear the cache
1550 $self->_clear_cache;
1551 # ...and recalculate the common readings.
1552 $self->calculate_common_readings();
1556 # The graph calculation information is now up to date.
1557 $self->_graphcalc_done(1);
1561 my( $graph, $node_ranks, @current_nodes ) = @_;
1562 # Look at each of the children of @current_nodes. If all the child's
1563 # parents have a rank, assign it the highest rank + 1 and add it to
1564 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1565 # parent gets a rank.
1567 foreach my $c ( @current_nodes ) {
1568 warn "Current reading $c has no rank!"
1569 unless exists $node_ranks->{$c};
1570 # print STDERR "Looking at child of node $c, rank "
1571 # . $node_ranks->{$c} . "\n";
1572 foreach my $child ( $graph->successors( $c ) ) {
1573 next if exists $node_ranks->{$child};
1574 my $highest_rank = -1;
1576 foreach my $parent ( $graph->predecessors( $child ) ) {
1577 if( exists $node_ranks->{$parent} ) {
1578 $highest_rank = $node_ranks->{$parent}
1579 if $highest_rank <= $node_ranks->{$parent};
1586 my $c_rank = $highest_rank + 1;
1587 # print STDERR "Assigning rank $c_rank to node $child \n";
1588 $node_ranks->{$child} = $c_rank;
1589 push( @next_nodes, $child );
1597 $self->wipe_svg if $self->has_cached_svg;
1598 $self->wipe_table if $self->has_cached_table;
1602 =head2 flatten_ranks
1604 A convenience method for parsing collation data. Searches the graph for readings
1605 with the same text at the same rank, and merges any that are found.
1611 my %unique_rank_rdg;
1613 foreach my $rdg ( $self->readings ) {
1614 next unless $rdg->has_rank;
1615 my $key = $rdg->rank . "||" . $rdg->text;
1616 if( exists $unique_rank_rdg{$key} ) {
1618 # print STDERR "Combining readings at same rank: $key\n";
1620 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1621 # TODO see if this now makes a common point.
1623 $unique_rank_rdg{$key} = $rdg;
1626 # If we merged readings, the ranks are still fine but the alignment
1627 # table is wrong. Wipe it.
1628 $self->wipe_table() if $changed;
1632 =head2 calculate_common_readings
1634 Goes through the graph identifying the readings that appear in every witness
1635 (apart from those with lacunae at that spot.) Marks them as common and returns
1640 use Text::Tradition;
1642 my $cxfile = 't/data/Collatex-16.xml';
1643 my $t = Text::Tradition->new(
1645 'input' => 'CollateX',
1648 my $c = $t->collation;
1650 my @common = $c->calculate_common_readings();
1651 is( scalar @common, 8, "Found correct number of common readings" );
1652 my @marked = sort $c->common_readings();
1653 is( scalar @common, 8, "All common readings got marked as such" );
1654 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1655 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1661 sub calculate_common_readings {
1664 map { $_->is_common( 0 ) } $self->readings;
1665 # Implicitly calls calculate_ranks
1666 my $table = $self->alignment_table;
1667 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1668 my @row = map { $_->{'tokens'}->[$idx]
1669 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1670 @{$table->{'alignment'}};
1672 foreach my $r ( @row ) {
1674 $hash{$r->id} = $r unless $r->is_meta;
1676 $hash{'UNDEF'} = $r;
1679 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1680 my( $r ) = values %hash;
1682 push( @common, $r );
1688 =head2 text_from_paths
1690 Calculate the text array for all witnesses from the path, for later consistency
1691 checking. Only to be used if there is no non-graph-based way to know the
1696 sub text_from_paths {
1698 foreach my $wit ( $self->tradition->witnesses ) {
1699 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1701 foreach my $r ( @readings ) {
1702 next if $r->is_meta;
1703 push( @text, $r->text );
1705 $wit->text( \@text );
1706 if( $wit->is_layered ) {
1707 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1708 $wit->sigil.$self->ac_label );
1710 foreach my $r ( @ucrdgs ) {
1711 next if $r->is_meta;
1712 push( @uctext, $r->text );
1714 $wit->layertext( \@uctext );
1719 =head1 UTILITY FUNCTIONS
1721 =head2 common_predecessor( $reading_a, $reading_b )
1723 Find the last reading that occurs in sequence before both the given readings.
1724 At the very least this should be $self->start.
1726 =head2 common_successor( $reading_a, $reading_b )
1728 Find the first reading that occurs in sequence after both the given readings.
1729 At the very least this should be $self->end.
1733 use Text::Tradition;
1735 my $cxfile = 't/data/Collatex-16.xml';
1736 my $t = Text::Tradition->new(
1738 'input' => 'CollateX',
1741 my $c = $t->collation;
1743 is( $c->common_predecessor( 'n24', 'n23' )->id,
1744 'n20', "Found correct common predecessor" );
1745 is( $c->common_successor( 'n24', 'n23' )->id,
1746 '__END__', "Found correct common successor" );
1748 is( $c->common_predecessor( 'n19', 'n17' )->id,
1749 'n16', "Found correct common predecessor for readings on same path" );
1750 is( $c->common_successor( 'n21', 'n10' )->id,
1751 '__END__', "Found correct common successor for readings on same path" );
1757 ## Return the closest reading that is a predecessor of both the given readings.
1758 sub common_predecessor {
1760 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1761 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1764 sub common_successor {
1766 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1767 return $self->_common_in_path( $r1, $r2, 'successors' );
1771 # TODO think about how to do this without ranks...
1772 sub _common_in_path {
1773 my( $self, $r1, $r2, $dir ) = @_;
1774 my $iter = $self->end->rank;
1776 my @last_r1 = ( $r1 );
1777 my @last_r2 = ( $r2 );
1778 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1780 # print STDERR "Finding common $dir for $r1, $r2\n";
1781 while( !@candidates ) {
1782 last unless $iter--; # Avoid looping infinitely
1783 # Iterate separately down the graph from r1 and r2
1784 my( @new_lc1, @new_lc2 );
1785 foreach my $lc ( @last_r1 ) {
1786 foreach my $p ( $lc->$dir ) {
1787 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1788 # print STDERR "Path candidate $p from $lc\n";
1789 push( @candidates, $p );
1790 } elsif( !$all_seen{$p->id} ) {
1791 $all_seen{$p->id} = 'r1';
1792 push( @new_lc1, $p );
1796 foreach my $lc ( @last_r2 ) {
1797 foreach my $p ( $lc->$dir ) {
1798 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1799 # print STDERR "Path candidate $p from $lc\n";
1800 push( @candidates, $p );
1801 } elsif( !$all_seen{$p->id} ) {
1802 $all_seen{$p->id} = 'r2';
1803 push( @new_lc2, $p );
1807 @last_r1 = @new_lc1;
1808 @last_r2 = @new_lc2;
1810 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1811 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1815 Text::Tradition::Error->throw(
1816 'ident' => 'Collation error',
1822 __PACKAGE__->meta->make_immutable;
1826 This package is free software and is provided "as is" without express
1827 or implied warranty. You can redistribute it and/or modify it under
1828 the same terms as Perl itself.
1832 Tara L Andrews E<lt>aurum@cpan.orgE<gt>