1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
6 use IPC::Run qw( run binary );
8 use Text::Tradition::Collation::Path;
9 use Text::Tradition::Collation::Reading;
10 use Text::Tradition::Collation::Relationship;
11 use Text::Tradition::Collation::Segment;
19 add_reading => 'add_node',
20 add_lacuna => 'add_node',
21 del_reading => 'del_node',
22 del_segment => 'del_node',
23 add_path => 'add_edge',
24 del_path => 'del_edge',
30 relationships => 'edges',
32 default => sub { Graph::Easy->new( undirected => 0 ) },
36 has 'tradition' => ( # TODO should this not be ro?
38 isa => 'Text::Tradition',
44 writer => '_save_svg',
45 predicate => 'has_svg',
51 writer => '_save_graphml',
52 predicate => 'has_graphml',
58 writer => '_save_csv',
59 predicate => 'has_csv',
62 # Keeps track of the lemmas within the collation. At most one lemma
63 # per position in the graph.
66 isa => 'HashRef[Maybe[Str]]',
67 default => sub { {} },
70 has 'wit_list_separator' => (
79 default => 'base text',
100 # The collation can be created two ways:
101 # 1. Collate a set of witnesses (with CollateX I guess) and process
102 # the results as in 2.
103 # 2. Read a pre-prepared collation in one of a variety of formats,
104 # and make the graph from that.
106 # The graph itself will (for now) be immutable, and the positions
107 # within the graph will also be immutable. We need to calculate those
108 # positions upon graph construction. The equivalences between graph
109 # nodes will be mutable, entirely determined by the user (or possibly
110 # by some semantic pre-processing provided by the user.) So the
111 # constructor should just make an empty equivalences object. The
112 # constructor will also need to make the witness objects, if we didn't
113 # come through option 1.
116 my( $self, $args ) = @_;
117 $self->graph->use_class('node', 'Text::Tradition::Collation::Reading');
118 $self->graph->use_class('edge', 'Text::Tradition::Collation::Path');
120 # Pass through any graph-specific options.
121 my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse';
122 $self->graph->set_attribute( 'node', 'shape', $shape );
125 around add_lacuna => sub {
129 my $l = $self->$orig( '#LACUNA_' . $id . '#' );
134 # Wrapper around add_path
136 around add_path => sub {
140 # Make sure there are three arguments
142 warn "Call add_path with args source, target, witness";
145 # Make sure the proposed path does not yet exist
146 # NOTE 'reading' will currently return readings and segments
147 my( $source, $target, $wit ) = @_;
148 $source = $self->reading( $source )
149 unless ref( $source ) eq 'Text::Tradition::Collation::Reading';
150 $target = $self->reading( $target )
151 unless ref( $target ) eq 'Text::Tradition::Collation::Reading';
152 foreach my $path ( $source->edges_to( $target ) ) {
153 if( $path->label eq $wit && $path->class eq 'edge.path' ) {
161 # Wrapper around paths
162 around paths => sub {
166 my @result = grep { $_->sub_class eq 'path' } $self->$orig( @_ );
170 around relationships => sub {
173 my @result = grep { $_->sub_class eq 'relationship' } $self->$orig( @_ );
177 around readings => sub {
180 my @result = grep { $_->sub_class ne 'segment' } $self->$orig( @_ );
184 around segments => sub {
187 my @result = grep { $_->sub_class eq 'segment' } $self->$orig( @_ );
191 # Wrapper around merge_nodes
195 my $first_node = shift;
196 my $second_node = shift;
197 $first_node->merge_from( $second_node );
198 unshift( @_, $first_node, $second_node );
199 return $self->graph->merge_nodes( @_ );
202 # Extra graph-alike utility
204 my( $self, $source, $target, $label ) = @_;
205 my @paths = $source->edges_to( $target );
206 my @relevant = grep { $_->label eq $label } @paths;
207 return scalar @relevant;
210 ## Dealing with groups of readings, i.e. segments.
213 my( $self, @items ) = @_;
214 my $segment = Text::Tradition::Collation::Segment->new( 'members' => \@items );
218 ## Dealing with relationships between readings. This is a different
219 ## sort of graph edge. Return a success/failure value and a list of
220 ## node pairs that have been linked.
222 sub add_relationship {
223 my( $self, $source, $target, $options ) = @_;
225 # Make sure there is not another relationship between these two
226 # readings or segments already
227 $source = $self->reading( $source )
228 unless ref( $source ) && $source->isa( 'Graph::Easy::Node' );
229 $target = $self->reading( $target )
230 unless ref( $target ) && $target->isa( 'Graph::Easy::Node' );
231 foreach my $rel ( $source->edges_to( $target ), $target->edges_to( $source ) ) {
232 if( $rel->class eq 'edge.relationship' ) {
233 return ( undef, "Relationship already exists between these readings" );
236 if( $options->{'equal_rank'} && !relationship_valid( $source, $target ) ) {
237 return ( undef, 'Relationship creates witness loop' );
240 # TODO Think about positional hilarity if relationships are added after positions
243 my @joined = ( [ $source->name, $target->name ] ); # Keep track of the nodes we join.
245 $options->{'this_relation'} = [ $source, $target ];
247 eval { $rel = Text::Tradition::Collation::Relationship->new( %$options ) };
249 return ( undef, $@ );
251 $self->graph->add_edge( $source, $target, $rel );
253 # TODO Handle global relationship setting
255 return( 1, @joined );
258 sub relationship_valid {
259 my( $source, $target ) = @_;
260 # Check that linking the source and target in a relationship won't lead
261 # to a path loop for any witness.
262 my @proposed_related = ( $source, $target );
263 push( @proposed_related, $source->related_readings );
264 push( @proposed_related, $target->related_readings );
266 map { $pr_ids{ $_->name } = 1 } @proposed_related;
267 # The lists of 'in' and 'out' should not have any element that appears
268 # in 'proposed_related'.
269 foreach my $pr ( @proposed_related ) {
270 foreach my $e ( $pr->incoming ) {
271 if( exists $pr_ids{ $e->from->name } ) {
275 foreach my $e ( $pr->outgoing ) {
276 if( exists $pr_ids{ $e->to->name } ) {
284 =head2 Output method(s)
290 print $graph->as_svg( $recalculate );
292 Returns an SVG string that represents the graph. Uses GraphViz to do
293 this, because Graph::Easy doesn\'t cope well with long graphs. Unless
294 $recalculate is passed (and is a true value), the method will return a
295 cached copy of the SVG after the first call to the method.
300 my( $self, $recalc ) = @_;
301 return $self->svg if $self->has_svg;
303 $self->collapse_graph_paths();
305 my @cmd = qw/dot -Tsvg/;
307 my $dotfile = File::Temp->new();
309 # $dotfile->unlink_on_destroy(0);
310 binmode $dotfile, ':utf8';
311 print $dotfile $self->as_dot();
312 push( @cmd, $dotfile->filename );
313 run( \@cmd, ">", binary(), \$svg );
314 $svg = decode_utf8( $svg );
315 $self->_save_svg( $svg );
316 $self->expand_graph_paths();
322 print $graph->as_dot( $view, $recalculate );
324 Returns a string that is the collation graph expressed in dot
325 (i.e. GraphViz) format. The 'view' argument determines what kind of
327 * 'path': a graph of witness paths through the collation (DEFAULT)
328 * 'relationship': a graph of how collation readings relate to
334 my( $self, $view ) = @_;
335 $view = 'path' unless $view;
336 # TODO consider making some of these things configurable
337 my $dot = sprintf( "digraph %s {\n", $self->tradition->name );
338 $dot .= "\tedge [ arrowhead=open ];\n";
339 $dot .= "\tgraph [ rankdir=LR ];\n";
340 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
341 11, "white", "filled", $self->graph->get_attribute( 'node', 'shape' ) );
343 foreach my $reading ( $self->readings ) {
344 # Need not output nodes without separate labels
345 next if $reading->name eq $reading->label;
346 # TODO output readings or segments, but not both
347 next if $reading->class eq 'node.segment';
348 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->name, $reading->label );
351 my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
352 foreach my $edge ( @edges ) {
353 my %variables = ( 'color' => '#000000',
354 'fontcolor' => '#000000',
355 'label' => $edge->label,
357 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
358 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
359 $edge->from->name, $edge->to->name, $varopts );
367 print $graph->as_graphml( $recalculate )
369 Returns a GraphML representation of the collation graph, with
370 transposition information and position information. Unless
371 $recalculate is passed (and is a true value), the method will return a
372 cached copy of the SVG after the first call to the method.
377 my( $self, $recalc ) = @_;
378 return $self->graphml if $self->has_graphml;
381 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
382 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
383 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
384 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
386 # Create the document and root node
387 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
388 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
389 $graphml->setDocumentElement( $root );
390 $root->setNamespace( $xsi_ns, 'xsi', 0 );
391 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
393 # TODO Add some global graph data
395 # Add the data keys for nodes
398 foreach my $datum ( qw/ name reading identical rank class / ) {
399 $node_data_keys{$datum} = 'dn'.$ndi++;
400 my $key = $root->addNewChild( $graphml_ns, 'key' );
401 $key->setAttribute( 'attr.name', $datum );
402 $key->setAttribute( 'attr.type', 'string' );
403 $key->setAttribute( 'for', 'node' );
404 $key->setAttribute( 'id', $node_data_keys{$datum} );
407 # Add the data keys for edges, i.e. witnesses
410 foreach my $edge_key( qw/ witness_main witness_ante_corr relationship class / ) {
411 $edge_data_keys{$edge_key} = 'de'.$edi++;
412 my $key = $root->addNewChild( $graphml_ns, 'key' );
413 $key->setAttribute( 'attr.name', $edge_key );
414 $key->setAttribute( 'attr.type', 'string' );
415 $key->setAttribute( 'for', 'edge' );
416 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
419 # Add the graph, its nodes, and its edges
420 my $graph = $root->addNewChild( $graphml_ns, 'graph' );
421 $graph->setAttribute( 'edgedefault', 'directed' );
422 $graph->setAttribute( 'id', 'g0' ); # TODO make this meaningful
423 $graph->setAttribute( 'parse.edgeids', 'canonical' );
424 $graph->setAttribute( 'parse.edges', scalar($self->paths) );
425 $graph->setAttribute( 'parse.nodeids', 'canonical' );
426 $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
427 $graph->setAttribute( 'parse.order', 'nodesfirst' );
431 # Add our readings to the graph
432 foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
433 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
434 my $node_xmlid = 'n' . $node_ctr++;
435 $node_hash{ $n->name } = $node_xmlid;
436 $node_el->setAttribute( 'id', $node_xmlid );
437 _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
438 _add_graphml_data( $node_el, $node_data_keys{'reading'}, $n->label );
439 _add_graphml_data( $node_el, $node_data_keys{'rank'}, $n->rank )
441 _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
442 _add_graphml_data( $node_el, $node_data_keys{'identical'}, $n->primary->name )
446 # Add any segments we have
447 foreach my $n ( sort { $a->name cmp $b->name } $self->segments ) {
448 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
449 my $node_xmlid = 'n' . $node_ctr++;
450 $node_hash{ $n->name } = $node_xmlid;
451 $node_el->setAttribute( 'id', $node_xmlid );
452 _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
453 _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
456 # Add the path, relationship, and segment edges
458 foreach my $e ( sort { $a->from->name cmp $b->from->name } $self->graph->edges() ) {
459 my( $name, $from, $to ) = ( 'e'.$edge_ctr++,
460 $node_hash{ $e->from->name() },
461 $node_hash{ $e->to->name() } );
462 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
463 $edge_el->setAttribute( 'source', $from );
464 $edge_el->setAttribute( 'target', $to );
465 $edge_el->setAttribute( 'id', $name );
467 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, $e->sub_class );
468 if( $e->sub_class eq 'path' ) {
469 # It's a witness path, so add the witness
470 my $base = $e->label;
471 my $key = $edge_data_keys{'witness_main'};
473 if( $e->label =~ /^(.*?)\s+(\(a\.c\.\))$/ ) {
475 $key = $edge_data_keys{'witness_ante_corr'};
477 _add_graphml_data( $edge_el, $key, $base );
478 } elsif( $e->sub_class eq 'relationship' ) {
479 # It's a relationship
480 _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label );
481 } # else a segment, nothing to record but source, target, class
485 $self->_save_graphml( $graphml->toString(1) );
486 return $graphml->toString(1);
489 sub _add_graphml_data {
490 my( $el, $key, $value ) = @_;
491 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
492 return unless defined $value;
493 $data_el->setAttribute( 'key', $key );
494 $data_el->appendText( $value );
499 print $graph->as_csv( $recalculate )
501 Returns a CSV alignment table representation of the collation graph, one
502 row per witness (or witness uncorrected.) Unless $recalculate is passed
503 (and is a true value), the method will return a cached copy of the CSV
504 after the first call to the method.
509 my( $self, $recalc ) = @_;
510 return $self->csv if $self->has_csv;
511 my $table = $self->make_alignment_table;
512 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
514 foreach my $row ( @$table ) {
515 $csv->combine( @$row );
516 push( @result, decode_utf8( $csv->string ) );
518 $self->_save_csv( join( "\n", @result ) );
522 # Make an alignment table - $noderefs controls whether the objects
523 # in the table are the nodes or simply their readings.
525 sub make_alignment_table {
526 my( $self, $noderefs ) = @_;
527 unless( $self->linear ) {
528 warn "Need a linear graph in order to make an alignment table";
532 my @all_pos = sort { $a <=> $b } $self->possible_positions;
533 foreach my $wit ( $self->tradition->witnesses ) {
534 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
535 my @row = _make_witness_row( $wit->path, \@all_pos, $noderefs );
536 unshift( @row, $wit->sigil );
537 push( @$table, \@row );
538 if( $wit->has_ante_corr ) {
539 my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos, $noderefs );
540 unshift( @ac_row, $wit->sigil . $self->ac_label );
541 push( @$table, \@ac_row );
545 # Return a table where the witnesses read in columns rather than rows.
546 my $turned = _turn_table( $table );
550 sub _make_witness_row {
551 my( $path, $positions, $noderefs ) = @_;
553 map { $char_hash{$_} = undef } @$positions;
554 foreach my $rdg ( @$path ) {
555 my $rtext = $rdg->text;
556 $rtext = '#LACUNA#' if $rdg->is_lacuna;
557 $char_hash{$rdg->rank} = $noderefs ? $rdg : $rtext;
559 my @row = map { $char_hash{$_} } @$positions;
560 # Fill in lacuna markers for undef spots in the row
561 my $last_el = shift @row;
562 my @filled_row = ( $last_el );
563 foreach my $el ( @row ) {
564 # If we are using node reference, make the lacuna node appear many times
565 # in the table. If not, use the lacuna tag.
566 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
567 $el = $noderefs ? $last_el : '#LACUNA#';
569 push( @filled_row, $el );
575 # Tiny utility function to say if a table element is a lacuna
578 return 1 if $el eq '#LACUNA#';
579 return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading'
584 # Helper to turn the witnesses along columns rather than rows. Assumes
589 return $result unless scalar @$table;
590 my $nrows = scalar @{$table->[0]};
591 foreach my $idx ( 0 .. $nrows - 1 ) {
592 foreach my $wit ( 0 .. $#{$table} ) {
593 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
600 sub collapse_graph_paths {
602 # Our collation graph has an path per witness. This is great for
603 # calculation purposes, but terrible for display. Thus we want to
604 # display only one path between any two nodes.
606 return if $self->collapsed;
608 print STDERR "Collapsing witness paths in graph...\n";
610 # Don't list out every witness if we have more than half to list.
611 my $majority = int( scalar( $self->tradition->witnesses ) / 2 ) + 1;
612 # But don't compress if there are only a few witnesses.
613 $majority = 4 if $majority < 4;
614 foreach my $node ( $self->readings ) {
616 # We will visit each node, so we only look ahead.
617 foreach my $edge ( $node->outgoing() ) {
618 next unless $edge->class eq 'edge.path';
619 add_hash_entry( $newlabels, $edge->to->name, $edge->name );
620 $self->del_path( $edge );
623 foreach my $newdest ( keys %$newlabels ) {
625 my @compressed_wits = @{$newlabels->{$newdest}};
626 if( @compressed_wits < $majority ) {
627 $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
629 ## TODO FIX THIS HACK
631 foreach my $wit ( @compressed_wits ) {
632 push( @aclabels, $wit ) if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ );
634 $label = join( ', ', 'majority', sort( @aclabels ) );
637 my $newpath = $self->add_path( $node, $self->reading( $newdest ), $label );
638 $newpath->hidden_witnesses( \@compressed_wits );
642 $self->collapsed( 1 );
645 sub expand_graph_paths {
647 # Our collation graph has only one path between any two nodes.
648 # This is great for display, but not so great for analysis.
649 # Expand this so that each witness has its own path between any
651 return unless $self->collapsed;
653 print STDERR "Expanding witness paths in graph...\n";
654 foreach my $path( $self->paths ) {
655 my $from = $path->from;
657 warn sprintf( "No hidden witnesses on %s -> %s ?", $from->name, $to->name )
658 unless $path->has_hidden_witnesses;
659 my @wits = @{$path->hidden_witnesses};
660 $self->del_path( $path );
662 $self->add_path( $from, $to, $_ );
665 $self->collapsed( 0 );
670 =head2 Navigation methods
676 my $beginning = $collation->start();
678 Returns the beginning of the collation, a meta-reading with label '#START#'.
683 # Return the beginning reading of the graph.
685 my( $new_start ) = @_;
687 $self->del_reading( '#START#' );
688 $self->graph->rename_node( $new_start, '#START#' );
690 # Make sure the start node has a start position.
691 unless( $self->reading( '#START#' )->has_rank ) {
692 $self->reading( '#START#' )->rank( '0' );
694 return $self->reading('#START#');
699 my $end = $collation->end();
701 Returns the end of the collation, a meta-reading with label '#END#'.
709 $self->del_reading( '#END#' );
710 $self->graph->rename_node( $new_end, '#END#' );
712 return $self->reading('#END#');
715 =item B<reading_sequence>
717 my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );
719 Returns the ordered list of readings, starting with $first and ending
720 with $last, along the given witness path. If no path is specified,
721 assume that the path is that of the base text (if any.)
725 # TODO Think about returning some lazy-eval iterator.
727 sub reading_sequence {
728 my( $self, $start, $end, $witness, $backup ) = @_;
730 $witness = $self->baselabel unless $witness;
731 my @readings = ( $start );
734 while( $n && $n ne $end ) {
735 if( exists( $seen{$n->name()} ) ) {
736 warn "Detected loop at " . $n->name();
739 $seen{$n->name()} = 1;
741 my $next = $self->next_reading( $n, $witness, $backup );
742 warn "Did not find any path for $witness from reading " . $n->name
744 push( @readings, $next );
747 # Check that the last reading is our end reading.
748 my $last = $readings[$#readings];
749 warn "Last reading found from " . $start->label() .
750 " for witness $witness is not the end!"
751 unless $last eq $end;
756 =item B<next_reading>
758 my $next_reading = $graph->next_reading( $reading, $witpath );
760 Returns the reading that follows the given reading along the given witness
766 # Return the successor via the corresponding path.
768 return $self->_find_linked_reading( 'next', @_ );
771 =item B<prior_reading>
773 my $prior_reading = $graph->prior_reading( $reading, $witpath );
775 Returns the reading that precedes the given reading along the given witness
781 # Return the predecessor via the corresponding path.
783 return $self->_find_linked_reading( 'prior', @_ );
786 sub _find_linked_reading {
787 my( $self, $direction, $node, $path, $alt_path ) = @_;
788 my @linked_paths = $direction eq 'next'
789 ? $node->outgoing() : $node->incoming();
790 return undef unless scalar( @linked_paths );
792 # We have to find the linked path that contains all of the
793 # witnesses supplied in $path.
794 my( @path_wits, @alt_path_wits );
795 @path_wits = $self->witnesses_of_label( $path ) if $path;
796 @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path;
799 foreach my $le ( @linked_paths ) {
800 if( $le->name eq $self->baselabel ) {
803 my @le_wits = $self->witnesses_of_label( $le->name );
804 if( _is_within( \@path_wits, \@le_wits ) ) {
805 # This is the right path.
806 return $direction eq 'next' ? $le->to() : $le->from();
807 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
812 # Got this far? Return the alternate path if it exists.
813 return $direction eq 'next' ? $alt_le->to() : $alt_le->from()
816 # Got this far? Return the base path if it exists.
817 return $direction eq 'next' ? $base_le->to() : $base_le->from()
820 # Got this far? We have no appropriate path.
821 warn "Could not find $direction node from " . $node->label
822 . " along path $path";
828 my( $set1, $set2 ) = @_;
829 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
830 foreach my $el ( @$set1 ) {
831 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
837 ## INITIALIZATION METHODS - for use by parsers
838 # Walk the paths for each witness in the graph, and return the nodes
839 # that the graph has in common. If $using_base is true, some
840 # different logic is needed.
841 # NOTE This does not create paths; it merely finds common readings.
843 sub walk_witness_paths {
845 # For each witness, walk the path through the graph.
846 # Then we need to find the common nodes.
847 # TODO This method is going to fall down if we have a very gappy
848 # text in the collation.
851 foreach my $wit ( $self->tradition->witnesses ) {
852 my $curr_reading = $self->start;
853 my @wit_path = $self->reading_sequence( $self->start, $self->end,
855 $wit->path( \@wit_path );
857 # Detect the common readings.
858 @common_readings = _find_common( \@common_readings, \@wit_path );
861 # Mark all the nodes as either common or not.
862 foreach my $cn ( @common_readings ) {
863 print STDERR "Setting " . $cn->name . " / " . $cn->label
864 . " as common node\n";
867 foreach my $n ( $self->readings() ) {
868 $n->make_variant unless $n->is_common;
870 # Return an array of the common nodes in order.
871 return @common_readings;
875 my( $common_readings, $new_path ) = @_;
877 if( @$common_readings ) {
878 foreach my $n ( @$new_path ) {
879 push( @cr, $n ) if grep { $_ eq $n } @$common_readings;
882 push( @cr, @$new_path );
888 my( $common_readings, $divergence ) = @_;
891 map { $diverged{$_->name} = 1 } @$divergence;
892 foreach( @$common_readings ) {
893 push( @cr, $_ ) unless $diverged{$_->name};
899 # For use when a collation is constructed from a base text and an apparatus.
900 # We have the sequences of readings and just need to add path edges.
902 sub make_witness_paths {
904 foreach my $wit ( $self->tradition->witnesses ) {
905 print STDERR "Making path for " . $wit->sigil . "\n";
906 $self->make_witness_path( $wit );
910 sub make_witness_path {
911 my( $self, $wit ) = @_;
912 my @chain = @{$wit->path};
913 my $sig = $wit->sigil;
914 foreach my $idx ( 0 .. $#chain-1 ) {
915 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
917 if( $wit->has_ante_corr ) {
918 @chain = @{$wit->uncorrected_path};
919 foreach my $idx( 0 .. $#chain-1 ) {
920 my $source = $chain[$idx];
921 my $target = $chain[$idx+1];
922 $self->add_path( $source, $target, $sig.$self->ac_label )
923 unless $self->has_path( $source, $target, $sig );
928 sub calculate_ranks {
930 # Walk a version of the graph where every node linked by a relationship
931 # edge is fundamentally the same node, and do a topological ranking on
932 # the nodes in this graph.
933 my $topo_graph = Graph::Easy->new();
937 foreach my $r ( $self->readings ) {
938 next if exists $rel_containers{$r->name};
939 my @rels = $r->related_readings( 'colocated' );
941 # Make a relationship container.
943 my $rn = $topo_graph->add_node( 'rel_container_' . $rel_ctr++ );
945 $rel_containers{$_->name} = $rn;
948 # Add a new node to mirror the old node.
949 $rel_containers{$r->name} = $topo_graph->add_node( $r->name );
953 # Add the edges. Need only one edge between any pair of nodes.
954 foreach my $r ( $self->readings ) {
955 foreach my $n ( $r->neighbor_readings( 'forward' ) ) {
956 $topo_graph->add_edge_once( $rel_containers{$r->name},
957 $rel_containers{$n->name} );
961 # Now do the rankings, starting with the start node.
962 my $topo_start = $rel_containers{$self->start->name};
963 my $node_ranks = { $topo_start->name => 0 };
964 my @curr_origin = ( $topo_start );
965 # A little iterative function.
966 while( @curr_origin ) {
967 @curr_origin = _assign_rank( $node_ranks, @curr_origin );
969 # Transfer our rankings from the topological graph to the real one.
970 foreach my $r ( $self->readings ) {
971 $r->rank( $node_ranks->{$rel_containers{$r->name}->name} );
976 my( $node_ranks, @current_nodes ) = @_;
977 # Look at each of the children of @current_nodes. If all the child's
978 # parents have a rank, assign it the highest rank + 1 and add it to
979 # @next_nodes. Otherwise skip it.
981 foreach my $c ( @current_nodes ) {
982 warn "Current reading " . $c->name . "has no rank!"
983 unless exists $node_ranks->{$c->name};
984 # print STDERR "Looking at child of node " . $c->name . ", rank "
985 # . $node_ranks->{$c->name} . "\n";
986 my @children = map { $_->to } $c->outgoing;
987 foreach my $child ( @children ) {
988 next if exists $node_ranks->{$child->name};
989 my $highest_rank = -1;
991 my @parents = map { $_->from } $child->incoming;
992 foreach my $parent ( @parents ) {
993 if( exists $node_ranks->{$parent->name} ) {
994 $highest_rank = $node_ranks->{$parent->name}
995 if $highest_rank <= $node_ranks->{$parent->name};
1002 # print STDERR "Assigning rank " . ( $highest_rank + 1 ) . " to node " . $child->name . "\n";
1003 $node_ranks->{$child->name} = $highest_rank + 1;
1004 push( @next_nodes, $child );
1010 # Another method to make up for rough collation methods. If the same reading
1011 # appears multiple times at the same rank, collapse the nodes.
1014 my %unique_rank_rdg;
1015 foreach my $rdg ( $self->readings ) {
1016 next unless $rdg->has_rank;
1017 my $key = $rdg->rank . "||" . $rdg->text;
1018 if( exists $unique_rank_rdg{$key} ) {
1020 print STDERR "Combining readings at same rank: $key\n";
1021 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1023 $unique_rank_rdg{$key} = $rdg;
1029 sub possible_positions {
1032 map { $all_pos{ $_->rank } = 1 } $self->readings;
1033 return keys %all_pos;
1036 # TODO think about indexing this.
1037 sub readings_at_position {
1038 my( $self, $position, $strict ) = @_;
1040 foreach my $r ( $self->readings ) {
1041 push( @answer, $r ) if $r->is_at_position( $position, $strict );
1046 ## Lemmatizer functions
1051 foreach my $position ( $self->possible_positions ) {
1052 $self->lemmata->{$position} = undef;
1055 foreach my $cr ( $self->common_readings ) {
1056 $self->lemmata->{$cr->position->maxref} = $cr->name;
1060 sub common_readings {
1062 my @common = grep { $_->is_common } $self->readings();
1063 return sort { $a->rank <=> $b->rank } @common;
1066 =item B<lemma_readings>
1068 my @state = $graph->lemma_readings( @readings_delemmatized );
1070 Takes a list of readings that have just been delemmatized, and returns
1071 a set of tuples of the form ['reading', 'state'] that indicates what
1072 changes need to be made to the graph.
1078 A state of 1 means 'lemmatize this reading'
1082 A state of 0 means 'delemmatize this reading'
1086 A state of undef means 'an ellipsis belongs in the text here because
1087 no decision has been made / an earlier decision was backed out'
1093 sub lemma_readings {
1094 my( $self, @toggled_off_nodes ) = @_;
1096 # First get the positions of those nodes which have been
1098 my $positions_off = {};
1099 map { $positions_off->{ $_->position->reference } = $_->name }
1102 # Now for each position, we have to see if a node is on, and we
1103 # have to see if a node has been turned off. The lemmata hash
1104 # should contain fixed positions, range positions whose node was
1105 # just turned off, and range positions whose node is on.
1107 my %fixed_positions;
1108 # TODO One of these is probably redundant.
1109 map { $fixed_positions{$_} = 0 } keys %{$self->lemmata};
1110 map { $fixed_positions{$_} = 0 } keys %{$positions_off};
1111 map { $fixed_positions{$_} = 1 } $self->possible_positions;
1112 foreach my $pos ( sort { Text::Tradition::Collation::Position::str_cmp( $a, $b ) } keys %fixed_positions ) {
1113 # Find the state of this position. If there is an active node,
1114 # its name will be the state; otherwise the state will be 0
1115 # (nothing at this position) or undef (ellipsis at this position)
1117 $active = $self->lemmata->{$pos} if exists $self->lemmata->{$pos};
1119 # Is there a formerly active node that was toggled off?
1120 if( exists( $positions_off->{$pos} ) ) {
1121 my $off_node = $positions_off->{$pos};
1122 if( $active && $active ne $off_node) {
1123 push( @answer, [ $off_node, 0 ], [ $active, 1 ] );
1125 unless( $fixed_positions{$pos} ) {
1127 delete $self->lemmata->{$pos};
1129 push( @answer, [ $off_node, $active ] );
1132 # No formerly active node, so we just see if there is a currently
1134 } elsif( $active ) {
1135 # Push the active node, whatever it is.
1136 push( @answer, [ $active, 1 ] );
1138 # Push the state that is there. Arbitrarily use the first node
1140 my @pos_nodes = $self->readings_at_position( $pos );
1141 push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] );
1142 delete $self->lemmata->{$pos} unless $fixed_positions{$pos};
1149 =item B<toggle_reading>
1151 my @readings_delemmatized = $graph->toggle_reading( $reading_name );
1153 Takes a reading node name, and either lemmatizes or de-lemmatizes
1154 it. Returns a list of all readings that are de-lemmatized as a result
1159 sub toggle_reading {
1160 my( $self, $rname ) = @_;
1162 return unless $rname;
1163 my $reading = $self->reading( $rname );
1164 if( !$reading || $reading->is_common() ) {
1165 # Do nothing, it's a common node.
1169 my $pos = $reading->position;
1170 my $fixed = $reading->position->fixed;
1171 my $old_state = $self->lemmata->{$pos->reference};
1174 if( $old_state && $old_state eq $rname ) {
1175 # Turn off the node. We turn on no others by default.
1176 push( @readings_off, $reading );
1179 $self->lemmata->{$pos->reference} = $rname;
1180 # Any other 'on' readings in the same position should be off
1181 # if we have a fixed position.
1182 push( @readings_off, $self->same_position_as( $reading, 1 ) )
1184 # Any node that is an identical transposed one should be off.
1185 push( @readings_off, $reading->identical_readings );
1187 @readings_off = unique_list( @readings_off );
1189 # Turn off the readings that need to be turned off.
1190 my @readings_delemmatized;
1191 foreach my $n ( @readings_off ) {
1192 my $npos = $n->position;
1194 $state = $self->lemmata->{$npos->reference}
1195 if defined $self->lemmata->{$npos->reference};
1196 if( $state && $state eq $n->name ) {
1197 # this reading is still on, so turn it off
1198 push( @readings_delemmatized, $n );
1199 my $new_state = undef;
1200 if( $npos->fixed && $n eq $reading ) {
1201 # This is the reading that was clicked, so if there are no
1202 # other readings there and this is a fixed position, turn off
1203 # the position. In all other cases, restore the ellipsis.
1204 my @other_n = $self->same_position_as( $n ); # TODO do we need strict?
1205 $new_state = 0 unless @other_n;
1207 $self->lemmata->{$npos->reference} = $new_state;
1208 } elsif( $old_state && $old_state eq $n->name ) {
1209 # another reading has already been turned on here
1210 push( @readings_delemmatized, $n );
1211 } # else some other reading was on anyway, so pass.
1213 return @readings_delemmatized;
1216 sub same_position_as {
1217 my( $self, $reading, $strict ) = @_;
1218 my $pos = $reading->position;
1219 my %onpath = ( $reading->name => 1 );
1220 # TODO This might not always be sufficient. We really want to
1221 # exclude all readings on this one's path between its two
1223 map { $onpath{$_->name} = 1 } $reading->neighbor_readings;
1224 my @same = grep { !$onpath{$_->name} }
1225 $self->readings_at_position( $reading->position, $strict );
1229 # Return the string that joins together a list of witnesses for
1230 # display on a single path.
1233 return join( $self->wit_list_separator, @_ );
1236 sub witnesses_of_label {
1237 my( $self, $label ) = @_;
1238 my $regex = $self->wit_list_separator;
1239 my @answer = split( /\Q$regex\E/, $label );
1246 map { $h{$_->name} = $_ } @list;
1247 return values( %h );
1250 sub add_hash_entry {
1251 my( $hash, $key, $entry ) = @_;
1252 if( exists $hash->{$key} ) {
1253 push( @{$hash->{$key}}, $entry );
1255 $hash->{$key} = [ $entry ];
1260 __PACKAGE__->meta->make_immutable;