1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
7 use IPC::Run qw( run binary );
9 use Text::Tradition::Collation::Path;
10 use Text::Tradition::Collation::Reading;
11 use Text::Tradition::Collation::Relationship;
19 add_reading => 'add_node',
20 add_lacuna => 'add_node',
21 del_reading => 'del_node',
22 add_path => 'add_edge',
23 del_path => 'del_edge',
28 relationships => 'edges',
30 default => sub { Graph::Easy->new( undirected => 0 ) },
34 has 'tradition' => ( # TODO should this not be ro?
36 isa => 'Text::Tradition',
43 writer => '_save_svg',
44 predicate => 'has_svg',
50 writer => '_save_graphml',
51 predicate => 'has_graphml',
57 writer => '_save_csv',
58 predicate => 'has_csv',
61 has 'wit_list_separator' => (
70 default => 'base text',
91 # The collation can be created two ways:
92 # 1. Collate a set of witnesses (with CollateX I guess) and process
93 # the results as in 2.
94 # 2. Read a pre-prepared collation in one of a variety of formats,
95 # and make the graph from that.
97 # The graph itself will (for now) be immutable, and the positions
98 # within the graph will also be immutable. We need to calculate those
99 # positions upon graph construction. The equivalences between graph
100 # nodes will be mutable, entirely determined by the user (or possibly
101 # by some semantic pre-processing provided by the user.) So the
102 # constructor should just make an empty equivalences object. The
103 # constructor will also need to make the witness objects, if we didn't
104 # come through option 1.
107 my( $self, $args ) = @_;
108 $self->graph->use_class('node', 'Text::Tradition::Collation::Reading');
109 $self->graph->use_class('edge', 'Text::Tradition::Collation::Path');
111 # Pass through any graph-specific options.
112 my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse';
113 $self->graph->set_attribute( 'node', 'shape', $shape );
115 # Start and end points for all texts
116 $self->start( 'INIT' );
117 $self->end( 'INIT' );
120 around add_lacuna => sub {
124 my $l = $self->$orig( '#LACUNA_' . $id . '#' );
129 # Wrapper around add_path
131 around add_path => sub {
135 # Make sure there are three arguments
137 warn "Call add_path with args source, target, witness";
140 # Make sure the proposed path does not yet exist
141 # NOTE 'reading' will currently return readings and segments
142 my( $source, $target, $wit ) = @_;
143 $source = $self->reading( $source )
144 unless ref( $source ) eq 'Text::Tradition::Collation::Reading';
145 $target = $self->reading( $target )
146 unless ref( $target ) eq 'Text::Tradition::Collation::Reading';
147 foreach my $path ( $source->edges_to( $target ) ) {
148 if( $path->label eq $wit && $path->class eq 'edge.path' ) {
156 # Wrapper around paths
157 around paths => sub {
161 my @result = grep { $_->sub_class eq 'path' } $self->$orig( @_ );
165 around relationships => sub {
168 my @result = grep { $_->sub_class eq 'relationship' } $self->$orig( @_ );
172 # Wrapper around merge_nodes
175 my $first_node = shift;
176 my $second_node = shift;
177 $first_node->merge_from( $second_node );
178 unshift( @_, $first_node, $second_node );
179 return $self->graph->merge_nodes( @_ );
182 # Extra graph-alike utility
184 my( $self, $source, $target, $label ) = @_;
185 my @paths = $source->edges_to( $target );
186 my @relevant = grep { $_->label eq $label } @paths;
187 return scalar @relevant;
190 ## Dealing with relationships between readings. This is a different
191 ## sort of graph edge. Return a success/failure value and a list of
192 ## node pairs that have been linked.
194 sub add_relationship {
195 my( $self, $source, $target, $options ) = @_;
197 # Make sure there is not another relationship between these two
199 $source = $self->reading( $source )
200 unless ref( $source ) && $source->isa( 'Graph::Easy::Node' );
201 $target = $self->reading( $target )
202 unless ref( $target ) && $target->isa( 'Graph::Easy::Node' );
203 foreach my $rel ( $source->edges_to( $target ), $target->edges_to( $source ) ) {
204 if( $rel->class eq 'edge.relationship' ) {
205 return ( undef, "Relationship already exists between these readings" );
208 if( $options->{'equal_rank'} && !relationship_valid( $source, $target ) ) {
209 return ( undef, 'Relationship creates witness loop' );
212 # TODO Think about positional hilarity if relationships are added after positions
215 my @joined = ( [ $source->name, $target->name ] ); # Keep track of the nodes we join.
217 $options->{'this_relation'} = [ $source, $target ];
219 eval { $rel = Text::Tradition::Collation::Relationship->new( %$options ) };
221 return ( undef, $@ );
223 $self->graph->add_edge( $source, $target, $rel );
225 # TODO Handle global relationship setting
227 return( 1, @joined );
230 sub relationship_valid {
231 my( $source, $target ) = @_;
232 # Check that linking the source and target in a relationship won't lead
233 # to a path loop for any witness.
234 my @proposed_related = ( $source, $target );
235 push( @proposed_related, $source->related_readings );
236 push( @proposed_related, $target->related_readings );
238 map { $pr_ids{ $_->name } = 1 } @proposed_related;
239 # The lists of 'in' and 'out' should not have any element that appears
240 # in 'proposed_related'.
241 foreach my $pr ( @proposed_related ) {
242 foreach my $e ( grep { $_->sub_class eq 'path' } $pr->incoming ) {
243 if( exists $pr_ids{ $e->from->name } ) {
247 foreach my $e ( grep { $_->sub_class eq 'path' } $pr->outgoing ) {
248 if( exists $pr_ids{ $e->to->name } ) {
256 =head2 Output method(s)
262 print $graph->as_svg( $recalculate );
264 Returns an SVG string that represents the graph. Uses GraphViz to do
265 this, because Graph::Easy doesn\'t cope well with long graphs. Unless
266 $recalculate is passed (and is a true value), the method will return a
267 cached copy of the SVG after the first call to the method.
272 my( $self, $recalc ) = @_;
273 return $self->svg if $self->has_svg;
275 $self->collapse_graph_paths();
277 my @cmd = qw/dot -Tsvg/;
279 my $dotfile = File::Temp->new();
281 # $dotfile->unlink_on_destroy(0);
282 binmode $dotfile, ':utf8';
283 print $dotfile $self->as_dot();
284 push( @cmd, $dotfile->filename );
285 run( \@cmd, ">", binary(), \$svg );
286 $svg = decode_utf8( $svg );
287 $self->_save_svg( $svg );
288 $self->expand_graph_paths();
294 print $graph->as_dot( $view, $recalculate );
296 Returns a string that is the collation graph expressed in dot
297 (i.e. GraphViz) format. The 'view' argument determines what kind of
299 * 'path': a graph of witness paths through the collation (DEFAULT)
300 * 'relationship': a graph of how collation readings relate to
306 my( $self, $view ) = @_;
307 $view = 'path' unless $view;
308 # TODO consider making some of these things configurable
309 my $graph_name = $self->tradition->name;
310 $graph_name =~ s/[^\w\s]//g;
311 $graph_name = join( '_', split( /\s+/, $graph_name ) );
312 my $dot = sprintf( "digraph %s {\n", $graph_name );
313 $dot .= "\tedge [ arrowhead=open ];\n";
314 $dot .= "\tgraph [ rankdir=LR ];\n";
315 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
316 11, "white", "filled", $self->graph->get_attribute( 'node', 'shape' ) );
318 foreach my $reading ( $self->readings ) {
319 # Need not output nodes without separate labels
320 next if $reading->name eq $reading->label;
321 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->name, $reading->label );
324 my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
325 foreach my $edge ( @edges ) {
326 my %variables = ( 'color' => '#000000',
327 'fontcolor' => '#000000',
328 'label' => $edge->label,
330 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
331 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
332 $edge->from->name, $edge->to->name, $varopts );
340 print $graph->as_graphml( $recalculate )
342 Returns a GraphML representation of the collation graph, with
343 transposition information and position information. Unless
344 $recalculate is passed (and is a true value), the method will return a
345 cached copy of the SVG after the first call to the method.
350 my( $self, $recalc ) = @_;
351 return $self->graphml if $self->has_graphml;
354 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
355 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
356 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
357 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
359 # Create the document and root node
360 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
361 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
362 $graphml->setDocumentElement( $root );
363 $root->setNamespace( $xsi_ns, 'xsi', 0 );
364 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
366 # Add the data keys for the graph
369 my @graph_attributes = qw/ wit_list_separator baselabel linear ac_label /;
370 foreach my $datum ( @graph_attributes ) {
371 $graph_data_keys{$datum} = 'dg'.$gdi++;
372 my $key = $root->addNewChild( $graphml_ns, 'key' );
373 $key->setAttribute( 'attr.name', $datum );
374 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
375 $key->setAttribute( 'for', 'graph' );
376 $key->setAttribute( 'id', $graph_data_keys{$datum} );
379 # Add the data keys for nodes
382 foreach my $datum ( qw/ name reading identical rank class / ) {
383 $node_data_keys{$datum} = 'dn'.$ndi++;
384 my $key = $root->addNewChild( $graphml_ns, 'key' );
385 $key->setAttribute( 'attr.name', $datum );
386 $key->setAttribute( 'attr.type', 'string' );
387 $key->setAttribute( 'for', 'node' );
388 $key->setAttribute( 'id', $node_data_keys{$datum} );
391 # Add the data keys for edges, i.e. witnesses
394 my @string_keys = qw/ class witness relationship /;
395 my @bool_keys = qw/ extra equal_rank non_correctable non_independent /;
396 foreach my $edge_key( @string_keys ) {
397 $edge_data_keys{$edge_key} = 'de'.$edi++;
398 my $key = $root->addNewChild( $graphml_ns, 'key' );
399 $key->setAttribute( 'attr.name', $edge_key );
400 $key->setAttribute( 'attr.type', 'string' );
401 $key->setAttribute( 'for', 'edge' );
402 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
404 foreach my $edge_key( @bool_keys ) {
405 $edge_data_keys{$edge_key} = 'de'.$edi++;
406 my $key = $root->addNewChild( $graphml_ns, 'key' );
407 $key->setAttribute( 'attr.name', $edge_key );
408 $key->setAttribute( 'attr.type', 'boolean' );
409 $key->setAttribute( 'for', 'edge' );
410 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
413 # Add the graph, its nodes, and its edges
414 my $graph = $root->addNewChild( $graphml_ns, 'graph' );
415 $graph->setAttribute( 'edgedefault', 'directed' );
416 $graph->setAttribute( 'id', $self->tradition->name );
417 $graph->setAttribute( 'parse.edgeids', 'canonical' );
418 $graph->setAttribute( 'parse.edges', scalar($self->paths) );
419 $graph->setAttribute( 'parse.nodeids', 'canonical' );
420 $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
421 $graph->setAttribute( 'parse.order', 'nodesfirst' );
423 # Collation attribute data
424 foreach my $datum ( @graph_attributes ) {
425 _add_graphml_data( $graph, $graph_data_keys{$datum}, $self->$datum );
430 # Add our readings to the graph
431 foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
432 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
433 my $node_xmlid = 'n' . $node_ctr++;
434 $node_hash{ $n->name } = $node_xmlid;
435 $node_el->setAttribute( 'id', $node_xmlid );
436 _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
437 _add_graphml_data( $node_el, $node_data_keys{'reading'}, $n->label );
438 _add_graphml_data( $node_el, $node_data_keys{'rank'}, $n->rank )
440 _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
441 _add_graphml_data( $node_el, $node_data_keys{'identical'}, $n->primary->name )
442 if $n->has_primary && $n->primary ne $n;
445 # Add the path and relationship edges
447 foreach my $e ( sort { $a->from->name cmp $b->from->name } $self->graph->edges() ) {
448 my( $name, $from, $to ) = ( 'e'.$edge_ctr++,
449 $node_hash{ $e->from->name() },
450 $node_hash{ $e->to->name() } );
451 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
452 $edge_el->setAttribute( 'source', $from );
453 $edge_el->setAttribute( 'target', $to );
454 $edge_el->setAttribute( 'id', $name );
456 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, $e->sub_class );
458 # For some classes we have extra information to save.
459 if( $e->sub_class eq 'path' ) {
460 # It's a witness path, so add the witness
461 my $base = $e->label;
462 my $key = $edge_data_keys{'witness_main'};
463 # Is this an ante-corr witness?
464 my $aclabel = $self->ac_label;
465 if( $e->label =~ /^(.*)\Q$aclabel\E$/ ) {
466 # Keep the base witness
468 # ...and record that this is an 'extra' reading path
469 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, 'true' );
471 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
472 } elsif( $e->sub_class eq 'relationship' ) {
473 # It's a relationship, so save the relationship data
474 _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label );
475 _add_graphml_data( $edge_el, $edge_data_keys{'equal_rank'}, $e->equal_rank );
476 _add_graphml_data( $edge_el, $edge_data_keys{'non_correctable'}, $e->non_correctable );
477 _add_graphml_data( $edge_el, $edge_data_keys{'non_independent'}, $e->non_independent );
481 # Save and return the thing
482 my $result = decode_utf8( $graphml->toString(1) );
483 $self->_save_graphml( $result );
487 sub _add_graphml_data {
488 my( $el, $key, $value ) = @_;
489 return unless defined $value;
490 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
491 $data_el->setAttribute( 'key', $key );
492 $data_el->appendText( $value );
497 print $graph->as_csv( $recalculate )
499 Returns a CSV alignment table representation of the collation graph, one
500 row per witness (or witness uncorrected.) Unless $recalculate is passed
501 (and is a true value), the method will return a cached copy of the CSV
502 after the first call to the method.
507 my( $self, $recalc ) = @_;
508 return $self->csv if $self->has_csv;
509 my $table = $self->make_alignment_table;
510 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
512 foreach my $row ( @$table ) {
513 $csv->combine( @$row );
514 push( @result, decode_utf8( $csv->string ) );
516 $self->_save_csv( join( "\n", @result ) );
520 # Make an alignment table - $noderefs controls whether the objects
521 # in the table are the nodes or simply their readings.
523 sub make_alignment_table {
524 my( $self, $noderefs, $include ) = @_;
525 unless( $self->linear ) {
526 warn "Need a linear graph in order to make an alignment table";
530 my @all_pos = ( 0 .. $self->end->rank - 1 );
531 foreach my $wit ( $self->tradition->witnesses ) {
532 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
533 my @row = _make_witness_row( $wit->path, \@all_pos, $noderefs );
534 unshift( @row, $wit->sigil );
535 push( @$table, \@row );
536 if( $wit->has_ante_corr ) {
537 my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos, $noderefs );
538 unshift( @ac_row, $wit->sigil . $self->ac_label );
539 push( @$table, \@ac_row );
545 # Winnow out the rows for any witness not included.
546 foreach my $row ( @$table ) {
547 next unless $include->{$row->[0]};
548 push( @$winnowed, $row );
553 # Return a table where the witnesses read in columns rather than rows.
554 my $turned = _turn_table( $table );
555 # TODO We should really go through and delete empty rows.
559 sub _make_witness_row {
560 my( $path, $positions, $noderefs ) = @_;
562 map { $char_hash{$_} = undef } @$positions;
563 foreach my $rdg ( @$path ) {
564 my $rtext = $rdg->text;
565 $rtext = '#LACUNA#' if $rdg->is_lacuna;
566 # print STDERR "No rank for " . $rdg->name . "\n" unless defined $rdg->rank;
567 $char_hash{$rdg->rank} = $noderefs ? $rdg : $rtext;
569 my @row = map { $char_hash{$_} } @$positions;
570 # Fill in lacuna markers for undef spots in the row
571 my $last_el = shift @row;
572 my @filled_row = ( $last_el );
573 foreach my $el ( @row ) {
574 # If we are using node reference, make the lacuna node appear many times
575 # in the table. If not, use the lacuna tag.
576 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
577 $el = $noderefs ? $last_el : '#LACUNA#';
579 push( @filled_row, $el );
585 # Tiny utility function to say if a table element is a lacuna
588 return 1 if $el eq '#LACUNA#';
589 return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading'
594 # Helper to turn the witnesses along columns rather than rows. Assumes
599 return $result unless scalar @$table;
600 my $nrows = scalar @{$table->[0]};
601 foreach my $idx ( 0 .. $nrows - 1 ) {
602 foreach my $wit ( 0 .. $#{$table} ) {
603 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
610 sub collapse_graph_paths {
612 # Our collation graph has an path per witness. This is great for
613 # calculation purposes, but terrible for display. Thus we want to
614 # display only one path between any two nodes.
616 return if $self->collapsed;
618 print STDERR "Collapsing witness paths in graph...\n";
620 # Don't list out every witness if we have more than half to list.
621 my $majority = int( scalar( $self->tradition->witnesses ) / 2 ) + 1;
622 # But don't compress if there are only a few witnesses.
623 $majority = 4 if $majority < 4;
624 foreach my $node ( $self->readings ) {
626 # We will visit each node, so we only look ahead.
627 foreach my $edge ( $node->outgoing() ) {
628 next unless $edge->class eq 'edge.path';
629 add_hash_entry( $newlabels, $edge->to->name, $edge->name );
630 $self->del_path( $edge );
633 foreach my $newdest ( keys %$newlabels ) {
635 my @compressed_wits = @{$newlabels->{$newdest}};
636 if( @compressed_wits < $majority ) {
637 $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
639 ## TODO FIX THIS HACK
641 foreach my $wit ( @compressed_wits ) {
642 push( @aclabels, $wit ) if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ );
644 $label = join( ', ', 'majority', sort( @aclabels ) );
647 my $newpath = $self->add_path( $node, $self->reading( $newdest ), $label );
648 $newpath->hidden_witnesses( \@compressed_wits );
652 $self->collapsed( 1 );
655 sub expand_graph_paths {
657 # Our collation graph has only one path between any two nodes.
658 # This is great for display, but not so great for analysis.
659 # Expand this so that each witness has its own path between any
661 return unless $self->collapsed;
663 print STDERR "Expanding witness paths in graph...\n";
664 foreach my $path( $self->paths ) {
665 my $from = $path->from;
667 warn sprintf( "No hidden witnesses on %s -> %s ?", $from->name, $to->name )
668 unless $path->has_hidden_witnesses;
669 my @wits = @{$path->hidden_witnesses};
670 $self->del_path( $path );
672 $self->add_path( $from, $to, $_ );
675 $self->collapsed( 0 );
680 =head2 Navigation methods
686 my $beginning = $collation->start();
688 Returns the beginning of the collation, a meta-reading with label '#START#'.
693 # Return the beginning reading of the graph.
694 my( $self, $new_start ) = @_;
695 my $start = $self->reading( '#START#' );
696 if( ref( $new_start ) eq 'Text::Tradition::Collation::Reading' ) {
697 # Replace the existing start node.
698 $self->del_reading( '#START#' );
699 $self->graph->rename_node( $new_start, '#START#' );
701 } elsif ( $new_start && $new_start eq 'INIT' ) {
702 # Make a new start node.
703 $start = $self->add_reading( '#START#' );
705 # Make sure the start node is a meta node
706 $start->is_meta( 1 );
707 # Make sure the start node has a start position.
708 unless( $start->has_rank ) {
716 my $end = $collation->end();
718 Returns the end of the collation, a meta-reading with label '#END#'.
725 my $end = $self->reading( '#END#' );
726 if( ref( $new_end ) eq 'Text::Tradition::Collation::Reading' ) {
727 $self->del_reading( '#END#' );
728 $self->graph->rename_node( $new_end, '#END#' );
730 } elsif ( $new_end && $new_end eq 'INIT' ) {
731 # Make a new start node.
732 $end = $self->add_reading( '#END#' );
734 # Make sure the start node is a meta node
739 =item B<reading_sequence>
741 my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );
743 Returns the ordered list of readings, starting with $first and ending
744 with $last, along the given witness path. If no path is specified,
745 assume that the path is that of the base text (if any.)
749 # TODO Think about returning some lazy-eval iterator.
751 sub reading_sequence {
752 my( $self, $start, $end, $witness, $backup ) = @_;
754 $witness = $self->baselabel unless $witness;
755 my @readings = ( $start );
758 while( $n && $n ne $end ) {
759 if( exists( $seen{$n->name()} ) ) {
760 warn "Detected loop at " . $n->name();
763 $seen{$n->name()} = 1;
765 my $next = $self->next_reading( $n, $witness, $backup );
767 warn "Did not find any path for $witness from reading " . $n->name;
770 push( @readings, $next );
773 # Check that the last reading is our end reading.
774 my $last = $readings[$#readings];
775 warn "Last reading found from " . $start->label() .
776 " for witness $witness is not the end!"
777 unless $last eq $end;
782 =item B<next_reading>
784 my $next_reading = $graph->next_reading( $reading, $witpath );
786 Returns the reading that follows the given reading along the given witness
792 # Return the successor via the corresponding path.
794 return $self->_find_linked_reading( 'next', @_ );
797 =item B<prior_reading>
799 my $prior_reading = $graph->prior_reading( $reading, $witpath );
801 Returns the reading that precedes the given reading along the given witness
807 # Return the predecessor via the corresponding path.
809 return $self->_find_linked_reading( 'prior', @_ );
812 sub _find_linked_reading {
813 my( $self, $direction, $node, $path, $alt_path ) = @_;
814 my @linked_paths = $direction eq 'next'
815 ? $node->outgoing() : $node->incoming();
816 return undef unless scalar( @linked_paths );
818 # We have to find the linked path that contains all of the
819 # witnesses supplied in $path.
820 my( @path_wits, @alt_path_wits );
821 @path_wits = $self->witnesses_of_label( $path ) if $path;
822 @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path;
825 foreach my $le ( @linked_paths ) {
826 if( $le->name eq $self->baselabel ) {
829 my @le_wits = $self->witnesses_of_label( $le->name );
830 if( _is_within( \@path_wits, \@le_wits ) ) {
831 # This is the right path.
832 return $direction eq 'next' ? $le->to() : $le->from();
833 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
838 # Got this far? Return the alternate path if it exists.
839 return $direction eq 'next' ? $alt_le->to() : $alt_le->from()
842 # Got this far? Return the base path if it exists.
843 return $direction eq 'next' ? $base_le->to() : $base_le->from()
846 # Got this far? We have no appropriate path.
847 warn "Could not find $direction node from " . $node->label
848 . " along path $path";
854 my( $set1, $set2 ) = @_;
855 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
856 foreach my $el ( @$set1 ) {
857 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
863 ## INITIALIZATION METHODS - for use by parsers
864 # Walk the paths for each witness in the graph, and return the nodes
865 # that the graph has in common. If $using_base is true, some
866 # different logic is needed.
867 # NOTE This does not create paths; it merely finds common readings.
869 sub walk_witness_paths {
871 # For each witness, walk the path through the graph.
872 # Then we need to find the common nodes.
873 # TODO This method is going to fall down if we have a very gappy
874 # text in the collation.
877 foreach my $wit ( $self->tradition->witnesses ) {
878 my $curr_reading = $self->start;
879 my @wit_path = $self->reading_sequence( $self->start, $self->end,
881 $wit->path( \@wit_path );
883 # Detect the common readings.
884 @common_readings = _find_common( \@common_readings, \@wit_path );
887 # Mark all the nodes as either common or not.
888 foreach my $cn ( @common_readings ) {
889 print STDERR "Setting " . $cn->name . " / " . $cn->label
890 . " as common node\n";
893 foreach my $n ( $self->readings() ) {
894 $n->make_variant unless $n->is_common;
896 # Return an array of the common nodes in order.
897 return @common_readings;
901 my( $common_readings, $new_path ) = @_;
903 if( @$common_readings ) {
904 foreach my $n ( @$new_path ) {
905 push( @cr, $n ) if grep { $_ eq $n } @$common_readings;
908 push( @cr, @$new_path );
914 my( $common_readings, $divergence ) = @_;
917 map { $diverged{$_->name} = 1 } @$divergence;
918 foreach( @$common_readings ) {
919 push( @cr, $_ ) unless $diverged{$_->name};
925 # For use when a collation is constructed from a base text and an apparatus.
926 # We have the sequences of readings and just need to add path edges.
928 sub make_witness_paths {
930 foreach my $wit ( $self->tradition->witnesses ) {
931 print STDERR "Making path for " . $wit->sigil . "\n";
932 $self->make_witness_path( $wit );
936 sub make_witness_path {
937 my( $self, $wit ) = @_;
938 my @chain = @{$wit->path};
939 my $sig = $wit->sigil;
940 foreach my $idx ( 0 .. $#chain-1 ) {
941 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
943 if( $wit->has_ante_corr ) {
944 @chain = @{$wit->uncorrected_path};
945 foreach my $idx( 0 .. $#chain-1 ) {
946 my $source = $chain[$idx];
947 my $target = $chain[$idx+1];
948 $self->add_path( $source, $target, $sig.$self->ac_label )
949 unless $self->has_path( $source, $target, $sig );
954 sub calculate_ranks {
956 # Walk a version of the graph where every node linked by a relationship
957 # edge is fundamentally the same node, and do a topological ranking on
958 # the nodes in this graph.
959 my $topo_graph = Graph->new();
963 foreach my $r ( $self->readings ) {
964 next if exists $rel_containers{$r->name};
965 my @rels = $r->related_readings( 'colocated' );
967 # Make a relationship container.
969 my $rn = 'rel_container_' . $rel_ctr++;
970 $topo_graph->add_vertex( $rn );
972 $rel_containers{$_->name} = $rn;
975 # Add a new node to mirror the old node.
976 $rel_containers{$r->name} = $r->name;
977 $topo_graph->add_vertex( $r->name );
981 # Add the edges. Need only one edge between any pair of nodes.
982 foreach my $r ( $self->readings ) {
983 foreach my $n ( $r->neighbor_readings( 'forward' ) ) {
984 my( $tfrom, $tto ) = ( $rel_containers{$r->name},
985 $rel_containers{$n->name} );
986 $topo_graph->add_edge( $tfrom, $tto )
987 unless $topo_graph->has_edge( $tfrom, $tto );
991 # Now do the rankings, starting with the start node.
992 my $topo_start = $rel_containers{$self->start->name};
993 my $node_ranks = { $topo_start => 0 };
994 my @curr_origin = ( $topo_start );
995 # A little iterative function.
996 while( @curr_origin ) {
997 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
999 # Transfer our rankings from the topological graph to the real one.
1000 foreach my $r ( $self->readings ) {
1001 if( defined $node_ranks->{$rel_containers{$r->name}} ) {
1002 $r->rank( $node_ranks->{$rel_containers{$r->name}} );
1005 die "No rank calculated for node " . $r->name
1006 . " - do you have a cycle in the graph?";
1012 my( $graph, $node_ranks, @current_nodes ) = @_;
1013 # Look at each of the children of @current_nodes. If all the child's
1014 # parents have a rank, assign it the highest rank + 1 and add it to
1015 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1016 # parent gets a rank.
1018 foreach my $c ( @current_nodes ) {
1019 warn "Current reading $c has no rank!"
1020 unless exists $node_ranks->{$c};
1021 # print STDERR "Looking at child of node $c, rank "
1022 # . $node_ranks->{$c} . "\n";
1023 foreach my $child ( $graph->successors( $c ) ) {
1024 next if exists $node_ranks->{$child};
1025 my $highest_rank = -1;
1027 foreach my $parent ( $graph->predecessors( $child ) ) {
1028 if( exists $node_ranks->{$parent} ) {
1029 $highest_rank = $node_ranks->{$parent}
1030 if $highest_rank <= $node_ranks->{$parent};
1037 my $c_rank = $highest_rank + 1;
1038 # print STDERR "Assigning rank $c_rank to node $child \n";
1039 $node_ranks->{$child} = $c_rank;
1040 push( @next_nodes, $child );
1046 # Another method to make up for rough collation methods. If the same reading
1047 # appears multiple times at the same rank, collapse the nodes.
1050 my %unique_rank_rdg;
1051 foreach my $rdg ( $self->readings ) {
1052 next unless $rdg->has_rank;
1053 my $key = $rdg->rank . "||" . $rdg->text;
1054 if( exists $unique_rank_rdg{$key} ) {
1056 print STDERR "Combining readings at same rank: $key\n";
1057 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1059 $unique_rank_rdg{$key} = $rdg;
1065 ## Utility functions
1067 # Return the string that joins together a list of witnesses for
1068 # display on a single path.
1071 return join( $self->wit_list_separator, @_ );
1074 sub witnesses_of_label {
1075 my( $self, $label ) = @_;
1076 my $regex = $self->wit_list_separator;
1077 my @answer = split( /\Q$regex\E/, $label );
1081 sub add_hash_entry {
1082 my( $hash, $key, $entry ) = @_;
1083 if( exists $hash->{$key} ) {
1084 push( @{$hash->{$key}}, $entry );
1086 $hash->{$key} = [ $entry ];
1091 __PACKAGE__->meta->make_immutable;
1097 =item * Rationalize edge classes
1099 =item * Port the internal graph from Graph::Easy to Graph