1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Data;
11 use Text::Tradition::Collation::Reading;
12 use Text::Tradition::Collation::RelationshipStore;
13 use Text::Tradition::Error;
14 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
16 use XML::LibXML::XPathContext;
20 isa => 'Text::Tradition::Collation::Data',
58 isa => 'Text::Tradition',
59 writer => '_set_tradition',
65 Text::Tradition::Collation - a software model for a text collation
70 my $t = Text::Tradition->new(
71 'name' => 'this is a text',
73 'file' => '/path/to/tei_parallel_seg_file.xml' );
75 my $c = $t->collation;
76 my @readings = $c->readings;
77 my @paths = $c->paths;
78 my @relationships = $c->relationships;
80 my $svg_variant_graph = $t->collation->as_svg();
84 Text::Tradition is a library for representation and analysis of collated
85 texts, particularly medieval ones. The Collation is the central feature of
86 a Tradition, where the text, its sequence of readings, and its relationships
87 between readings are actually kept.
93 The constructor. Takes a hash or hashref of the following arguments:
97 =item * tradition - The Text::Tradition object to which the collation
100 =item * linear - Whether the collation should be linear; that is, whether
101 transposed readings should be treated as two linked readings rather than one,
102 and therefore whether the collation graph is acyclic. Defaults to true.
104 =item * baselabel - The default label for the path taken by a base text
105 (if any). Defaults to 'base text'.
107 =item * wit_list_separator - The string to join a list of witnesses for
108 purposes of making labels in display graphs. Defaults to ', '.
110 =item * ac_label - The extra label to tack onto a witness sigil when
111 representing another layer of path for the given witness - that is, when
112 a text has more than one possible reading due to scribal corrections or
113 the like. Defaults to ' (a.c.)'.
115 =item * wordsep - The string used to separate words in the original text.
126 =head2 wit_list_separator
134 Simple accessors for collation attributes.
138 The meta-reading at the start of every witness path.
142 The meta-reading at the end of every witness path.
146 Returns all Reading objects in the graph.
148 =head2 reading( $id )
150 Returns the Reading object corresponding to the given ID.
152 =head2 add_reading( $reading_args )
154 Adds a new reading object to the collation.
155 See L<Text::Tradition::Collation::Reading> for the available arguments.
157 =head2 del_reading( $object_or_id )
159 Removes the given reading from the collation, implicitly removing its
160 paths and relationships.
162 =head2 merge_readings( $main, $second, $concatenate, $with_str )
164 Merges the $second reading into the $main one. If $concatenate is true, then
165 the merged node will carry the text of both readings, concatenated with either
166 $with_str (if specified) or a sensible default (the empty string if the
167 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
169 The first two arguments may be either readings or reading IDs.
171 =head2 has_reading( $id )
173 Predicate to see whether a given reading ID is in the graph.
175 =head2 reading_witnesses( $object_or_id )
177 Returns a list of sigils whose witnesses contain the reading.
181 Returns all reading paths within the document - that is, all edges in the
182 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
184 =head2 add_path( $source, $target, $sigil )
186 Links the given readings in the collation in sequence, under the given witness
187 sigil. The readings may be specified by object or ID.
189 =head2 del_path( $source, $target, $sigil )
191 Links the given readings in the collation in sequence, under the given witness
192 sigil. The readings may be specified by object or ID.
194 =head2 has_path( $source, $target );
196 Returns true if the two readings are linked in sequence in any witness.
197 The readings may be specified by object or ID.
201 Returns all Relationship objects in the collation.
203 =head2 add_relationship( $reading, $other_reading, $options )
205 Adds a new relationship of the type given in $options between the two readings,
206 which may be specified by object or ID. Returns a value of ( $status, @vectors)
207 where $status is true on success, and @vectors is a list of relationship edges
208 that were ultimately added.
209 See L<Text::Tradition::Collation::Relationship> for the available options.
214 my ( $class, @args ) = @_;
215 my %args = @args == 1 ? %{ $args[0] } : @args;
216 # TODO determine these from the Moose::Meta object
217 my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel
218 linear wordsep start end cached_table _graphcalc_done);
220 for my $attr (@delegate_attrs) {
221 $data_args{$attr} = delete $args{$attr} if exists $args{$attr};
223 $args{_data} = Text::Tradition::Collation::Data->new(%data_args);
229 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
230 $self->_set_start( $self->add_reading(
231 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
232 $self->_set_end( $self->add_reading(
233 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
236 ### Reading construct/destruct functions
239 my( $self, $reading ) = @_;
240 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
241 my %args = %$reading;
242 if( $args{'init'} ) {
243 # If we are initializing an empty collation, don't assume that we
244 # have set a tradition.
245 delete $args{'init'};
246 } elsif( $self->tradition->can('language') && $self->tradition->has_language
247 && !exists $args{'language'} ) {
248 $args{'language'} = $self->tradition->language;
250 $reading = Text::Tradition::Collation::Reading->new(
251 'collation' => $self,
254 # First check to see if a reading with this ID exists.
255 if( $self->reading( $reading->id ) ) {
256 throw( "Collation already has a reading with id " . $reading->id );
258 $self->_graphcalc_done(0);
259 $self->_add_reading( $reading->id => $reading );
260 # Once the reading has been added, put it in both graphs.
261 $self->sequence->add_vertex( $reading->id );
262 $self->relations->add_reading( $reading->id );
266 around del_reading => sub {
271 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
274 # Remove the reading from the graphs.
275 $self->_graphcalc_done(0);
276 $self->_clear_cache; # Explicitly clear caches to GC the reading
277 $self->sequence->delete_vertex( $arg );
278 $self->relations->delete_reading( $arg );
281 $self->$orig( $arg );
288 my $cxfile = 't/data/Collatex-16.xml';
289 my $t = Text::Tradition->new(
291 'input' => 'CollateX',
294 my $c = $t->collation;
296 my $rno = scalar $c->readings;
297 # Split n21 for testing purposes
298 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
299 my $old_r = $c->reading( 'n21' );
300 $old_r->alter_text( 'to' );
301 $c->del_path( 'n20', 'n21', 'A' );
302 $c->add_path( 'n20', 'n21p0', 'A' );
303 $c->add_path( 'n21p0', 'n21', 'A' );
305 ok( $c->reading( 'n21p0' ), "New reading exists" );
306 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
308 # Combine n3 and n4 ( with his )
309 $c->merge_readings( 'n3', 'n4', 1 );
310 ok( !$c->reading('n4'), "Reading n4 is gone" );
311 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
313 # Collapse n9 and n10 ( rood / root )
314 $c->merge_readings( 'n9', 'n10' );
315 ok( !$c->reading('n10'), "Reading n10 is gone" );
316 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
318 # Combine n21 and n21p0
319 my $remaining = $c->reading('n21');
320 $remaining ||= $c->reading('n22'); # one of these should still exist
321 $c->merge_readings( 'n21p0', $remaining, 1 );
322 ok( !$c->reading('n21'), "Reading $remaining is gone" );
323 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
333 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
334 my $mergemeta = $kept_obj->is_meta;
335 throw( "Cannot merge meta and non-meta reading" )
336 unless ( $mergemeta && $del_obj->is_meta )
337 || ( !$mergemeta && !$del_obj->is_meta );
339 throw( "Cannot merge with start or end node" )
340 if( $kept_obj eq $self->start || $kept_obj eq $self->end
341 || $del_obj eq $self->start || $del_obj eq $self->end );
342 throw( "Cannot combine text of meta readings" ) if $combine;
344 # We only need the IDs for adding paths to the graph, not the reading
345 # objects themselves.
346 my $kept = $kept_obj->id;
347 my $deleted = $del_obj->id;
348 $self->_graphcalc_done(0);
350 # The kept reading should inherit the paths and the relationships
351 # of the deleted reading.
352 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
353 my @vector = ( $kept );
354 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
355 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
356 next if $vector[0] eq $vector[1]; # Don't add a self loop
357 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
358 $self->sequence->add_edge( @vector );
359 my $fwits = $self->sequence->get_edge_attributes( @vector );
360 @wits{keys %$fwits} = values %$fwits;
361 $self->sequence->set_edge_attributes( @vector, \%wits );
363 $self->relations->merge_readings( $kept, $deleted, $combine );
365 # Do the deletion deed.
367 # Combine the text of the readings
368 my $joinstr = $combine_char;
369 unless( defined $joinstr ) {
370 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
371 $joinstr = $self->wordsep unless defined $joinstr;
373 $kept_obj->_combine( $del_obj, $joinstr );
375 $self->del_reading( $deleted );
378 =head2 compress_readings
380 Where possible in the graph, compresses plain sequences of readings into a
381 single reading. The sequences must consist of readings with no
382 relationships to other readings, with only a single witness path between
383 them and no other witness paths from either that would skip the other. The
384 readings must also not be marked as nonsense or bad grammar.
386 WARNING: This operation cannot be undone.
390 sub compress_readings {
392 # Anywhere in the graph that there is a reading that joins only to a single
393 # successor, and neither of these have any relationships, just join the two
395 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
396 # Now look for readings that can be joined to their successors.
397 next unless $rdg->is_combinable;
399 while( $self->sequence->successors( $rdg ) == 1 ) {
400 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
401 throw( "Infinite loop" ) if $seen{$next->id};
402 $seen{$next->id} = 1;
403 last if $self->sequence->predecessors( $next ) > 1;
404 last unless $next->is_combinable;
405 say "Joining readings $rdg and $next";
406 $self->merge_readings( $rdg, $next, 1 );
409 # Make sure we haven't screwed anything up
410 foreach my $wit ( $self->tradition->witnesses ) {
411 my $pathtext = $self->path_text( $wit->sigil );
412 my $origtext = join( ' ', @{$wit->text} );
413 throw( "Text differs for witness " . $wit->sigil )
414 unless $pathtext eq $origtext;
415 if( $wit->is_layered ) {
416 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
417 $origtext = join( ' ', @{$wit->layertext} );
418 throw( "Ante-corr text differs for witness " . $wit->sigil )
419 unless $pathtext eq $origtext;
423 $self->relations->rebuild_equivalence();
424 $self->calculate_ranks();
427 # Helper function for manipulating the graph.
428 sub _stringify_args {
429 my( $self, $first, $second, @args ) = @_;
431 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
432 $second = $second->id
433 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
434 return( $first, $second, @args );
437 # Helper function for manipulating the graph.
438 sub _objectify_args {
439 my( $self, $first, $second, $arg ) = @_;
440 $first = $self->reading( $first )
441 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
442 $second = $self->reading( $second )
443 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
444 return( $first, $second, $arg );
451 # We only need the IDs for adding paths to the graph, not the reading
452 # objects themselves.
453 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
455 $self->_graphcalc_done(0);
456 # Connect the readings
457 unless( $self->sequence->has_edge( $source, $target ) ) {
458 $self->sequence->add_edge( $source, $target );
459 $self->relations->add_equivalence_edge( $source, $target );
461 # Note the witness in question
462 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
468 if( ref( $_[0] ) eq 'ARRAY' ) {
475 # We only need the IDs for adding paths to the graph, not the reading
476 # objects themselves.
477 my( $source, $target, $wit ) = $self->_stringify_args( @args );
479 $self->_graphcalc_done(0);
480 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
481 $self->sequence->delete_edge_attribute( $source, $target, $wit );
483 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
484 $self->sequence->delete_edge( $source, $target );
485 $self->relations->delete_equivalence_edge( $source, $target );
490 # Extra graph-alike utility
493 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
494 return undef unless $self->sequence->has_edge( $source, $target );
495 return $self->sequence->has_edge_attribute( $source, $target, $wit );
498 =head2 clear_witness( @sigil_list )
500 Clear the given witnesses out of the collation entirely, removing references
501 to them in paths, and removing readings that belong only to them. Should only
502 be called via $tradition->del_witness.
507 my( $self, @sigils ) = @_;
509 $self->_graphcalc_done(0);
510 # Clear the witness(es) out of the paths
511 foreach my $e ( $self->paths ) {
512 foreach my $sig ( @sigils ) {
513 $self->del_path( $e, $sig );
517 # Clear out the newly unused readings
518 foreach my $r ( $self->readings ) {
519 unless( $self->reading_witnesses( $r ) ) {
520 $self->del_reading( $r );
525 sub add_relationship {
527 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
528 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
529 $self->_graphcalc_done(0);
533 around qw/ get_relationship del_relationship / => sub {
537 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
540 my( $source, $target ) = $self->_stringify_args( @args );
541 $self->$orig( $source, $target );
544 =head2 reading_witnesses( $reading )
546 Return a list of sigils corresponding to the witnesses in which the reading appears.
550 sub reading_witnesses {
551 my( $self, $reading ) = @_;
552 # We need only check either the incoming or the outgoing edges; I have
553 # arbitrarily chosen "incoming". Thus, special-case the start node.
554 if( $reading eq $self->start ) {
555 return map { $_->sigil } $self->tradition->witnesses;
558 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
559 my $wits = $self->sequence->get_edge_attributes( @$e );
560 @all_witnesses{ keys %$wits } = 1;
562 my $acstr = $self->ac_label;
563 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
564 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
566 return keys %all_witnesses;
569 =head1 OUTPUT METHODS
571 =head2 as_svg( \%options )
573 Returns an SVG string that represents the graph, via as_dot and graphviz.
574 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
579 my( $self, $opts ) = @_;
580 throw( "Need GraphViz installed to output SVG" )
581 unless File::Which::which( 'dot' );
582 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
583 $self->calculate_ranks()
584 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
585 my @cmd = qw/dot -Tsvg/;
587 my $dotfile = File::Temp->new();
589 # $dotfile->unlink_on_destroy(0);
590 binmode $dotfile, ':utf8';
591 print $dotfile $self->as_dot( $opts );
592 push( @cmd, $dotfile->filename );
593 run( \@cmd, ">", binary(), \$svg );
594 $svg = decode_utf8( $svg );
599 =head2 as_dot( \%options )
601 Returns a string that is the collation graph expressed in dot
602 (i.e. GraphViz) format. Options include:
617 my( $self, $opts ) = @_;
618 my $startrank = $opts->{'from'} if $opts;
619 my $endrank = $opts->{'to'} if $opts;
620 my $color_common = $opts->{'color_common'} if $opts;
621 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
622 && $self->end->rank > 100;
623 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
625 # Check the arguments
627 return if $endrank && $startrank > $endrank;
628 return if $startrank > $self->end->rank;
630 if( defined $endrank ) {
631 return if $endrank < 0;
632 $endrank = undef if $endrank == $self->end->rank;
635 my $graph_name = $self->tradition->name;
636 $graph_name =~ s/[^\w\s]//g;
637 $graph_name = join( '_', split( /\s+/, $graph_name ) );
645 'fillcolor' => 'white',
650 'arrowhead' => 'open',
651 'color' => '#000000',
652 'fontcolor' => '#000000',
655 my $dot = sprintf( "digraph %s {\n", $graph_name );
656 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
657 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
659 # Output substitute start/end readings if necessary
661 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
664 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
666 if( $STRAIGHTENHACK ) {
668 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
669 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
670 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
672 my %used; # Keep track of the readings that actually appear in the graph
673 # Sort the readings by rank if we have ranks; this speeds layout.
674 my @all_readings = $self->end->has_rank
675 ? sort { $a->rank <=> $b->rank } $self->readings
677 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
678 foreach my $reading ( @all_readings ) {
679 # Only output readings within our rank range.
680 next if $startrank && $reading->rank < $startrank;
681 next if $endrank && $reading->rank > $endrank;
682 $used{$reading->id} = 1;
683 # Need not output nodes without separate labels
684 next if $reading->id eq $reading->text;
686 my $label = $reading->text;
687 $label .= '-' if $reading->join_next;
688 $label = "-$label" if $reading->join_prior;
689 $label =~ s/\"/\\\"/g;
690 $rattrs->{'label'} = $label;
691 $rattrs->{'id'} = $reading->id;
692 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
693 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
696 # Add the real edges. Need to weight one edge per rank jump, in a
698 # my $weighted = $self->_add_edge_weights;
699 my @edges = $self->paths;
700 my( %substart, %subend );
701 foreach my $edge ( @edges ) {
702 # Do we need to output this edge?
703 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
704 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
705 my $variables = { %edge_attrs, 'label' => $label };
707 # Account for the rank gap if necessary
708 my $rank0 = $self->reading( $edge->[0] )->rank
709 if $self->reading( $edge->[0] )->has_rank;
710 my $rank1 = $self->reading( $edge->[1] )->rank
711 if $self->reading( $edge->[1] )->has_rank;
712 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
713 $variables->{'minlen'} = $rank1 - $rank0;
716 # Add the calculated edge weights
717 # if( exists $weighted->{$edge->[0]}
718 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
719 # # $variables->{'color'} = 'red';
720 # $variables->{'weight'} = 3.0;
723 # EXPERIMENTAL: make edge width reflect no. of witnesses
724 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
725 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
727 my $varopts = _dot_attr_string( $variables );
728 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
729 $edge->[0], $edge->[1], $varopts );
730 } elsif( $used{$edge->[0]} ) {
731 $subend{$edge->[0]} = $edge->[1];
732 } elsif( $used{$edge->[1]} ) {
733 $substart{$edge->[1]} = $edge->[0];
736 # Add substitute start and end edges if necessary
737 foreach my $node ( keys %substart ) {
738 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
739 my $variables = { %edge_attrs, 'label' => $witstr };
740 my $nrdg = $self->reading( $node );
741 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
742 # Substart is actually one lower than $startrank
743 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
745 my $varopts = _dot_attr_string( $variables );
746 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
748 foreach my $node ( keys %subend ) {
749 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
750 my $variables = { %edge_attrs, 'label' => $witstr };
751 my $varopts = _dot_attr_string( $variables );
752 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
755 if( $STRAIGHTENHACK ) {
756 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
757 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
764 sub _dot_attr_string {
767 foreach my $k ( sort keys %$hash ) {
769 push( @attrs, $k.'="'.$v.'"' );
771 return( '[ ' . join( ', ', @attrs ) . ' ]' );
774 sub _add_edge_weights {
776 # Walk the graph from START to END, choosing the successor node with
777 # the largest number of witness paths each time.
779 my $curr = $self->start->id;
780 my $ranked = $self->end->has_rank;
781 while( $curr ne $self->end->id ) {
782 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
783 my @succ = sort { $self->path_witnesses( $curr, $a )
784 <=> $self->path_witnesses( $curr, $b ) }
785 $self->sequence->successors( $curr );
786 my $next = pop @succ;
787 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
788 # Try to avoid lacunae in the weighted path.
790 ( $self->reading( $next )->is_lacuna ||
791 $nextrank - $rank > 1 ) ){
794 $weighted->{$curr} = $next;
800 =head2 path_witnesses( $edge )
802 Returns the list of sigils whose witnesses are associated with the given edge.
803 The edge can be passed as either an array or an arrayref of ( $source, $target ).
808 my( $self, @edge ) = @_;
809 # If edge is an arrayref, cope.
810 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
814 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
818 # Helper function. Make a display label for the given witnesses, showing a.c.
819 # witnesses only where the main witness is not also in the list.
820 sub _path_display_label {
823 map { $wits{$_} = 1 } @_;
825 # If an a.c. wit is listed, remove it if the main wit is also listed.
826 # Otherwise keep it for explicit listing.
827 my $aclabel = $self->ac_label;
829 foreach my $w ( sort keys %wits ) {
830 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
831 if( exists $wits{$1} ) {
834 push( @disp_ac, $w );
839 # See if we are in a majority situation.
840 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
841 $maj = $maj > 5 ? $maj : 5;
842 if( scalar keys %wits > $maj ) {
843 unshift( @disp_ac, 'majority' );
844 return join( ', ', @disp_ac );
846 return join( ', ', sort keys %wits );
850 =head2 readings_at_rank( $rank )
852 Returns a list of readings at a given rank, taken from the alignment table.
856 sub readings_at_rank {
857 my( $self, $rank ) = @_;
858 my $table = $self->alignment_table;
859 # Table rank is real rank - 1.
860 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
862 foreach my $e ( @elements ) {
863 next unless ref( $e ) eq 'HASH';
864 next unless exists $e->{'t'};
865 $readings{$e->{'t'}->id} = $e->{'t'};
867 return values %readings;
872 Returns a GraphML representation of the collation. The GraphML will contain
873 two graphs. The first expresses the attributes of the readings and the witness
874 paths that link them; the second expresses the relationships that link the
875 readings. This is the native transfer format for a tradition.
885 my $datafile = 't/data/florilegium_tei_ps.xml';
886 my $tradition = Text::Tradition->new( 'input' => 'TEI',
891 ok( $tradition, "Got a tradition object" );
892 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
893 ok( $tradition->collation, "Tradition has a collation" );
895 my $c = $tradition->collation;
896 is( scalar $c->readings, $READINGS, "Collation has all readings" );
897 is( scalar $c->paths, $PATHS, "Collation has all paths" );
898 is( scalar $c->relationships, 0, "Collation has all relationships" );
900 # Add a few relationships
901 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
902 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
903 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
905 # Now write it to GraphML and parse it again.
907 my $graphml = $c->as_graphml;
908 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
909 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
910 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
911 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
913 # Now add a stemma, write to GraphML, and look at the output.
915 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
916 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
917 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
918 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
919 $graphml = $c->as_graphml;
920 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
927 ## TODO MOVE this to Tradition.pm and modularize it better
929 my( $self, $options ) = @_;
930 $self->calculate_ranks unless $self->_graphcalc_done;
932 my $start = $options->{'from'}
933 ? $self->reading( $options->{'from'} ) : $self->start;
934 my $end = $options->{'to'}
935 ? $self->reading( $options->{'to'} ) : $self->end;
936 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
937 throw( 'Start node must be before end node' );
939 # The readings need to be ranked for this to work.
940 $start = $self->start unless $start->has_rank;
941 $end = $self->end unless $end->has_rank;
943 unless( $start eq $self->start ) {
944 $rankoffset = $start->rank - 1;
949 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
950 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
951 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
952 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
954 # Create the document and root node
956 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
957 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
958 $graphml->setDocumentElement( $root );
959 $root->setNamespace( $xsi_ns, 'xsi', 0 );
960 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
962 # List of attribute types to save on our objects and their corresponding
968 'ReadingID' => 'string',
969 'RelationshipType' => 'string',
970 'RelationshipScope' => 'string',
973 # Add the data keys for the graph. Include an extra key 'version' for the
974 # GraphML output version.
977 my %graph_attributes = ( 'version' => 'string' );
978 # Graph attributes include those of Tradition and those of Collation.
980 my $tmeta = $self->tradition->meta;
981 my $cmeta = $self->meta;
982 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
983 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
984 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
985 next if $attr->name =~ /^_/;
986 next unless $save_types{$attr->type_constraint->name};
987 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
989 # Extra custom keys for complex objects that should be saved in some form.
990 # The subroutine should return a string, or undef/empty.
991 if( $tmeta->has_method('stemmata') ) {
992 $graph_attributes{'stemmata'} = sub {
994 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
995 $self->tradition->stemmata;
996 join( "\n", @stemstrs );
1000 if( $tmeta->has_method('user') ) {
1001 $graph_attributes{'user'} = sub {
1002 $self->tradition->user ? $self->tradition->user->id : undef
1006 foreach my $datum ( sort keys %graph_attributes ) {
1007 $graph_data_keys{$datum} = 'dg'.$gdi++;
1008 my $key = $root->addNewChild( $graphml_ns, 'key' );
1009 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1010 : $graph_attributes{$datum};
1011 $key->setAttribute( 'attr.name', $datum );
1012 $key->setAttribute( 'attr.type', $dtype );
1013 $key->setAttribute( 'for', 'graph' );
1014 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1017 # Add the data keys for reading nodes
1018 my %reading_attributes;
1019 my $rmeta = Text::Tradition::Collation::Reading->meta;
1020 foreach my $attr( $rmeta->get_all_attributes ) {
1021 next if $attr->name =~ /^_/;
1022 next unless $save_types{$attr->type_constraint->name};
1023 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1025 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1026 # Extra custom key for the reading morphology
1027 $reading_attributes{'lexemes'} = 'string';
1032 foreach my $datum ( sort keys %reading_attributes ) {
1033 $node_data_keys{$datum} = 'dn'.$ndi++;
1034 my $key = $root->addNewChild( $graphml_ns, 'key' );
1035 $key->setAttribute( 'attr.name', $datum );
1036 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1037 $key->setAttribute( 'for', 'node' );
1038 $key->setAttribute( 'id', $node_data_keys{$datum} );
1041 # Add the data keys for edges, that is, paths and relationships. Path
1042 # data does not come from a Moose class so is here manually.
1045 my %edge_attributes = (
1046 witness => 'string', # ID/label for a path
1047 extra => 'boolean', # Path key
1049 my @path_attributes = keys %edge_attributes; # track our manual additions
1050 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1051 foreach my $attr( $pmeta->get_all_attributes ) {
1052 next if $attr->name =~ /^_/;
1053 next unless $save_types{$attr->type_constraint->name};
1054 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1056 foreach my $datum ( sort keys %edge_attributes ) {
1057 $edge_data_keys{$datum} = 'de'.$edi++;
1058 my $key = $root->addNewChild( $graphml_ns, 'key' );
1059 $key->setAttribute( 'attr.name', $datum );
1060 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1061 $key->setAttribute( 'for', 'edge' );
1062 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1065 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1066 my $xmlidname = $self->tradition->name;
1067 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1068 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1069 $xmlidname = '_'.$xmlidname;
1071 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1072 $sgraph->setAttribute( 'edgedefault', 'directed' );
1073 $sgraph->setAttribute( 'id', $xmlidname );
1074 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1075 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1076 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1077 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1078 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1080 # Tradition/collation attribute data
1081 foreach my $datum ( keys %graph_attributes ) {
1083 if( $datum eq 'version' ) {
1085 } elsif( ref( $graph_attributes{$datum} ) ) {
1086 my $sub = $graph_attributes{$datum};
1088 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1089 $value = $self->tradition->$datum;
1091 $value = $self->$datum;
1093 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1098 # Add our readings to the graph
1099 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1100 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1101 ( $n->rank < $start->rank || $n->rank > $end->rank );
1102 $use_readings{$n->id} = 1;
1103 # Add to the main graph
1104 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1105 my $node_xmlid = 'n' . $node_ctr++;
1106 $node_hash{ $n->id } = $node_xmlid;
1107 $node_el->setAttribute( 'id', $node_xmlid );
1108 foreach my $d ( keys %reading_attributes ) {
1110 # Custom serialization
1111 if( $d eq 'lexemes' ) {
1112 # If nval is a true value, we have lexemes so we need to
1113 # serialize them. Otherwise set nval to undef so that the
1114 # key is excluded from this reading.
1115 $nval = $nval ? $n->_serialize_lexemes : undef;
1116 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1119 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1120 # Adjust the ranks within the subgraph.
1121 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1122 : $nval - $rankoffset;
1124 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1129 # Add the path edges to the sequence graph
1131 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1132 # We add an edge in the graphml for every witness in $e.
1133 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1134 my @edge_wits = sort $self->path_witnesses( $e );
1135 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1136 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1137 # Skip any path from start to end; that witness is not in the subgraph.
1138 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1139 foreach my $wit ( @edge_wits ) {
1140 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1141 $node_hash{ $e->[0] },
1142 $node_hash{ $e->[1] } );
1143 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1144 $edge_el->setAttribute( 'source', $from );
1145 $edge_el->setAttribute( 'target', $to );
1146 $edge_el->setAttribute( 'id', $id );
1148 # It's a witness path, so add the witness
1150 my $key = $edge_data_keys{'witness'};
1151 # Is this an ante-corr witness?
1152 my $aclabel = $self->ac_label;
1153 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1154 # Keep the base witness
1156 # ...and record that this is an 'extra' reading path
1157 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1159 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1163 # Report the actual number of nodes and edges that went in
1164 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1165 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1167 # Add the relationship graph to the XML
1168 map { delete $edge_data_keys{$_} } @path_attributes;
1169 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1170 $node_data_keys{'id'}, \%edge_data_keys );
1172 # Save and return the thing
1173 my $result = decode_utf8( $graphml->toString(1) );
1177 sub _add_graphml_data {
1178 my( $el, $key, $value ) = @_;
1179 return unless defined $value;
1180 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1181 $data_el->setAttribute( 'key', $key );
1182 $data_el->appendText( $value );
1187 Returns a CSV alignment table representation of the collation graph, one
1188 row per witness (or witness uncorrected.)
1194 my $table = $self->alignment_table;
1195 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1197 # Make the header row
1198 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1199 push( @result, decode_utf8( $csv->string ) );
1200 # Make the rest of the rows
1201 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1202 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1203 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1204 $csv->combine( @row );
1205 push( @result, decode_utf8( $csv->string ) );
1207 return join( "\n", @result );
1210 =head2 alignment_table( $use_refs, $include_witnesses )
1212 Return a reference to an alignment table, in a slightly enhanced CollateX
1213 format which looks like this:
1215 $table = { alignment => [ { witness => "SIGIL",
1216 tokens => [ { t => "TEXT" }, ... ] },
1217 { witness => "SIG2",
1218 tokens => [ { t => "TEXT" }, ... ] },
1220 length => TEXTLEN };
1222 If $use_refs is set to 1, the reading object is returned in the table
1223 instead of READINGTEXT; if not, the text of the reading is returned.
1225 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1226 keys have a true hash value will be included.
1230 sub alignment_table {
1232 $self->calculate_ranks() unless $self->_graphcalc_done;
1233 return $self->cached_table if $self->has_cached_table;
1235 # Make sure we can do this
1236 throw( "Need a linear graph in order to make an alignment table" )
1237 unless $self->linear;
1238 $self->calculate_ranks unless $self->end->has_rank;
1240 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1241 my @all_pos = ( 1 .. $self->end->rank - 1 );
1242 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1243 # say STDERR "Making witness row(s) for " . $wit->sigil;
1244 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1245 my @row = _make_witness_row( \@wit_path, \@all_pos );
1246 push( @{$table->{'alignment'}},
1247 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1248 if( $wit->is_layered ) {
1249 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1250 $wit->sigil.$self->ac_label );
1251 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1252 push( @{$table->{'alignment'}},
1253 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1256 $self->cached_table( $table );
1260 sub _make_witness_row {
1261 my( $path, $positions ) = @_;
1263 map { $char_hash{$_} = undef } @$positions;
1265 foreach my $rdg ( @$path ) {
1266 say STDERR "rank " . $rdg->rank if $debug;
1267 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1268 $char_hash{$rdg->rank} = { 't' => $rdg };
1270 my @row = map { $char_hash{$_} } @$positions;
1271 # Fill in lacuna markers for undef spots in the row
1272 my $last_el = shift @row;
1273 my @filled_row = ( $last_el );
1274 foreach my $el ( @row ) {
1275 # If we are using node reference, make the lacuna node appear many times
1276 # in the table. If not, use the lacuna tag.
1277 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1280 push( @filled_row, $el );
1286 =head1 NAVIGATION METHODS
1288 =head2 reading_sequence( $first, $last, $sigil, $backup )
1290 Returns the ordered list of readings, starting with $first and ending
1291 with $last, for the witness given in $sigil. If a $backup sigil is
1292 specified (e.g. when walking a layered witness), it will be used wherever
1293 no $sigil path exists. If there is a base text reading, that will be
1294 used wherever no path exists for $sigil or $backup.
1298 # TODO Think about returning some lazy-eval iterator.
1299 # TODO Get rid of backup; we should know from what witness is whether we need it.
1301 sub reading_sequence {
1302 my( $self, $start, $end, $witness ) = @_;
1304 $witness = $self->baselabel unless $witness;
1305 my @readings = ( $start );
1308 while( $n && $n->id ne $end->id ) {
1309 if( exists( $seen{$n->id} ) ) {
1310 throw( "Detected loop for $witness at " . $n->id );
1314 my $next = $self->next_reading( $n, $witness );
1316 throw( "Did not find any path for $witness from reading " . $n->id );
1318 push( @readings, $next );
1321 # Check that the last reading is our end reading.
1322 my $last = $readings[$#readings];
1323 throw( "Last reading found from " . $start->text .
1324 " for witness $witness is not the end!" ) # TODO do we get this far?
1325 unless $last->id eq $end->id;
1330 =head2 next_reading( $reading, $sigil );
1332 Returns the reading that follows the given reading along the given witness
1338 # Return the successor via the corresponding path.
1340 my $answer = $self->_find_linked_reading( 'next', @_ );
1341 return undef unless $answer;
1342 return $self->reading( $answer );
1345 =head2 prior_reading( $reading, $sigil )
1347 Returns the reading that precedes the given reading along the given witness
1353 # Return the predecessor via the corresponding path.
1355 my $answer = $self->_find_linked_reading( 'prior', @_ );
1356 return $self->reading( $answer );
1359 sub _find_linked_reading {
1360 my( $self, $direction, $node, $path ) = @_;
1362 # Get a backup if we are dealing with a layered witness
1364 my $aclabel = $self->ac_label;
1365 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1369 my @linked_paths = $direction eq 'next'
1370 ? $self->sequence->edges_from( $node )
1371 : $self->sequence->edges_to( $node );
1372 return undef unless scalar( @linked_paths );
1374 # We have to find the linked path that contains all of the
1375 # witnesses supplied in $path.
1376 my( @path_wits, @alt_path_wits );
1377 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1378 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1381 foreach my $le ( @linked_paths ) {
1382 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1385 my @le_wits = sort $self->path_witnesses( $le );
1386 if( _is_within( \@path_wits, \@le_wits ) ) {
1387 # This is the right path.
1388 return $direction eq 'next' ? $le->[1] : $le->[0];
1389 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1393 # Got this far? Return the alternate path if it exists.
1394 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1397 # Got this far? Return the base path if it exists.
1398 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1401 # Got this far? We have no appropriate path.
1402 warn "Could not find $direction node from " . $node->id
1403 . " along path $path";
1409 my( $set1, $set2 ) = @_;
1410 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1411 foreach my $el ( @$set1 ) {
1412 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1417 # Return the string that joins together a list of witnesses for
1418 # display on a single path.
1419 sub _witnesses_of_label {
1420 my( $self, $label ) = @_;
1421 my $regex = $self->wit_list_separator;
1422 my @answer = split( /\Q$regex\E/, $label );
1426 =head2 common_readings
1428 Returns the list of common readings in the graph (i.e. those readings that are
1429 shared by all non-lacunose witnesses.)
1433 sub common_readings {
1435 my @common = grep { $_->is_common } $self->readings;
1439 =head2 path_text( $sigil, [, $start, $end ] )
1441 Returns the text of a witness (plus its backup, if we are using a layer)
1442 as stored in the collation. The text is returned as a string, where the
1443 individual readings are joined with spaces and the meta-readings (e.g.
1444 lacunae) are omitted. Optional specification of $start and $end allows
1445 the generation of a subset of the witness text.
1450 my( $self, $wit, $start, $end ) = @_;
1451 $start = $self->start unless $start;
1452 $end = $self->end unless $end;
1453 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1456 foreach my $r ( @path ) {
1457 unless ( $r->join_prior || !$last || $last->join_next ) {
1460 $pathtext .= $r->text;
1466 =head1 INITIALIZATION METHODS
1468 These are mostly for use by parsers.
1470 =head2 make_witness_path( $witness )
1472 Link the array of readings contained in $witness->path (and in
1473 $witness->uncorrected_path if it exists) into collation paths.
1474 Clear out the arrays when finished.
1476 =head2 make_witness_paths
1478 Call make_witness_path for all witnesses in the tradition.
1482 # For use when a collation is constructed from a base text and an apparatus.
1483 # We have the sequences of readings and just need to add path edges.
1484 # When we are done, clear out the witness path attributes, as they are no
1486 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1488 sub make_witness_paths {
1490 foreach my $wit ( $self->tradition->witnesses ) {
1491 # say STDERR "Making path for " . $wit->sigil;
1492 $self->make_witness_path( $wit );
1496 sub make_witness_path {
1497 my( $self, $wit ) = @_;
1498 my @chain = @{$wit->path};
1499 my $sig = $wit->sigil;
1500 # Add start and end if necessary
1501 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1502 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1503 foreach my $idx ( 0 .. $#chain-1 ) {
1504 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1506 if( $wit->is_layered ) {
1507 @chain = @{$wit->uncorrected_path};
1508 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1509 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1510 foreach my $idx( 0 .. $#chain-1 ) {
1511 my $source = $chain[$idx];
1512 my $target = $chain[$idx+1];
1513 $self->add_path( $source, $target, $sig.$self->ac_label )
1514 unless $self->has_path( $source, $target, $sig );
1518 $wit->clear_uncorrected_path;
1521 =head2 calculate_ranks
1523 Calculate the reading ranks (that is, their aligned positions relative
1524 to each other) for the graph. This can only be called on linear collations.
1528 use Text::Tradition;
1530 my $cxfile = 't/data/Collatex-16.xml';
1531 my $t = Text::Tradition->new(
1533 'input' => 'CollateX',
1536 my $c = $t->collation;
1539 my $table = $c->alignment_table;
1540 ok( $c->has_cached_table, "Alignment table was cached" );
1541 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1542 $c->calculate_ranks;
1543 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1544 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1545 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1551 sub calculate_ranks {
1553 # Save the existing ranks, in case we need to invalidate the cached SVG.
1555 map { $existing_ranks{$_} = $_->rank } $self->readings;
1557 # Do the rankings based on the relationship equivalence graph, starting
1558 # with the start node.
1559 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1561 # Transfer our rankings from the topological graph to the real one.
1562 foreach my $r ( $self->readings ) {
1563 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1564 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1566 # Die. Find the last rank we calculated.
1567 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1568 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1570 my $last = pop @all_defined;
1571 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1574 # Do we need to invalidate the cached data?
1575 if( $self->has_cached_table ) {
1576 foreach my $r ( $self->readings ) {
1577 next if defined( $existing_ranks{$r} )
1578 && $existing_ranks{$r} == $r->rank;
1579 # Something has changed, so clear the cache
1580 $self->_clear_cache;
1581 # ...and recalculate the common readings.
1582 $self->calculate_common_readings();
1586 # The graph calculation information is now up to date.
1587 $self->_graphcalc_done(1);
1592 $self->wipe_table if $self->has_cached_table;
1596 =head2 flatten_ranks
1598 A convenience method for parsing collation data. Searches the graph for readings
1599 with the same text at the same rank, and merges any that are found.
1605 my %unique_rank_rdg;
1607 foreach my $rdg ( $self->readings ) {
1608 next unless $rdg->has_rank;
1609 my $key = $rdg->rank . "||" . $rdg->text;
1610 if( exists $unique_rank_rdg{$key} ) {
1611 # Make sure they don't have different grammatical forms
1612 my $ur = $unique_rank_rdg{$key};
1613 if( $rdg->is_identical( $ur ) ) {
1615 #say STDERR "Combining readings at same rank: $key";
1617 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1618 # 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 # say STDERR "Finding common $dir for $r1, $r2";
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 # say STDERR "Path candidate $p from $lc";
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 # say STDERR "Path candidate $p from $lc";
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;
1826 =item * Rework XML serialization in a more modular way
1832 This package is free software and is provided "as is" without express
1833 or implied warranty. You can redistribute it and/or modify it under
1834 the same terms as Perl itself.
1838 Tara L Andrews E<lt>aurum@cpan.orgE<gt>