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',
42 writer => '_save_svg',
43 predicate => 'has_svg',
49 writer => '_save_graphml',
50 predicate => 'has_graphml',
56 writer => '_save_csv',
57 predicate => 'has_csv',
60 # Keeps track of the lemmas within the collation. At most one lemma
61 # per position in the graph.
64 isa => 'HashRef[Maybe[Str]]',
65 default => sub { {} },
68 has 'wit_list_separator' => (
77 default => 'base text',
98 # The collation can be created two ways:
99 # 1. Collate a set of witnesses (with CollateX I guess) and process
100 # the results as in 2.
101 # 2. Read a pre-prepared collation in one of a variety of formats,
102 # and make the graph from that.
104 # The graph itself will (for now) be immutable, and the positions
105 # within the graph will also be immutable. We need to calculate those
106 # positions upon graph construction. The equivalences between graph
107 # nodes will be mutable, entirely determined by the user (or possibly
108 # by some semantic pre-processing provided by the user.) So the
109 # constructor should just make an empty equivalences object. The
110 # constructor will also need to make the witness objects, if we didn't
111 # come through option 1.
114 my( $self, $args ) = @_;
115 $self->graph->use_class('node', 'Text::Tradition::Collation::Reading');
116 $self->graph->use_class('edge', 'Text::Tradition::Collation::Path');
118 # Pass through any graph-specific options.
119 my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse';
120 $self->graph->set_attribute( 'node', 'shape', $shape );
122 # Start and end points for all texts
123 $self->start( 'INIT' );
124 $self->end( 'INIT' );
127 around add_lacuna => sub {
131 my $l = $self->$orig( '#LACUNA_' . $id . '#' );
136 # Wrapper around add_path
138 around add_path => sub {
142 # Make sure there are three arguments
144 warn "Call add_path with args source, target, witness";
147 # Make sure the proposed path does not yet exist
148 # NOTE 'reading' will currently return readings and segments
149 my( $source, $target, $wit ) = @_;
150 $source = $self->reading( $source )
151 unless ref( $source ) eq 'Text::Tradition::Collation::Reading';
152 $target = $self->reading( $target )
153 unless ref( $target ) eq 'Text::Tradition::Collation::Reading';
154 foreach my $path ( $source->edges_to( $target ) ) {
155 if( $path->label eq $wit && $path->class eq 'edge.path' ) {
163 # Wrapper around paths
164 around paths => sub {
168 my @result = grep { $_->sub_class eq 'path' } $self->$orig( @_ );
172 around relationships => sub {
175 my @result = grep { $_->sub_class eq 'relationship' } $self->$orig( @_ );
179 # Wrapper around merge_nodes
182 my $first_node = shift;
183 my $second_node = shift;
184 $first_node->merge_from( $second_node );
185 unshift( @_, $first_node, $second_node );
186 return $self->graph->merge_nodes( @_ );
189 # Extra graph-alike utility
191 my( $self, $source, $target, $label ) = @_;
192 my @paths = $source->edges_to( $target );
193 my @relevant = grep { $_->label eq $label } @paths;
194 return scalar @relevant;
197 ## Dealing with relationships between readings. This is a different
198 ## sort of graph edge. Return a success/failure value and a list of
199 ## node pairs that have been linked.
201 sub add_relationship {
202 my( $self, $source, $target, $options ) = @_;
204 # Make sure there is not another relationship between these two
206 $source = $self->reading( $source )
207 unless ref( $source ) && $source->isa( 'Graph::Easy::Node' );
208 $target = $self->reading( $target )
209 unless ref( $target ) && $target->isa( 'Graph::Easy::Node' );
210 foreach my $rel ( $source->edges_to( $target ), $target->edges_to( $source ) ) {
211 if( $rel->class eq 'edge.relationship' ) {
212 return ( undef, "Relationship already exists between these readings" );
215 if( $options->{'equal_rank'} && !relationship_valid( $source, $target ) ) {
216 return ( undef, 'Relationship creates witness loop' );
219 # TODO Think about positional hilarity if relationships are added after positions
222 my @joined = ( [ $source->name, $target->name ] ); # Keep track of the nodes we join.
224 $options->{'this_relation'} = [ $source, $target ];
226 eval { $rel = Text::Tradition::Collation::Relationship->new( %$options ) };
228 return ( undef, $@ );
230 $self->graph->add_edge( $source, $target, $rel );
232 # TODO Handle global relationship setting
234 return( 1, @joined );
237 sub relationship_valid {
238 my( $source, $target ) = @_;
239 # Check that linking the source and target in a relationship won't lead
240 # to a path loop for any witness.
241 my @proposed_related = ( $source, $target );
242 push( @proposed_related, $source->related_readings );
243 push( @proposed_related, $target->related_readings );
245 map { $pr_ids{ $_->name } = 1 } @proposed_related;
246 # The lists of 'in' and 'out' should not have any element that appears
247 # in 'proposed_related'.
248 foreach my $pr ( @proposed_related ) {
249 foreach my $e ( $pr->incoming ) {
250 if( exists $pr_ids{ $e->from->name } ) {
254 foreach my $e ( $pr->outgoing ) {
255 if( exists $pr_ids{ $e->to->name } ) {
263 =head2 Output method(s)
269 print $graph->as_svg( $recalculate );
271 Returns an SVG string that represents the graph. Uses GraphViz to do
272 this, because Graph::Easy doesn\'t cope well with long graphs. Unless
273 $recalculate is passed (and is a true value), the method will return a
274 cached copy of the SVG after the first call to the method.
279 my( $self, $recalc ) = @_;
280 return $self->svg if $self->has_svg;
282 $self->collapse_graph_paths();
284 my @cmd = qw/dot -Tsvg/;
286 my $dotfile = File::Temp->new();
288 # $dotfile->unlink_on_destroy(0);
289 binmode $dotfile, ':utf8';
290 print $dotfile $self->as_dot();
291 push( @cmd, $dotfile->filename );
292 run( \@cmd, ">", binary(), \$svg );
293 $svg = decode_utf8( $svg );
294 $self->_save_svg( $svg );
295 $self->expand_graph_paths();
301 print $graph->as_dot( $view, $recalculate );
303 Returns a string that is the collation graph expressed in dot
304 (i.e. GraphViz) format. The 'view' argument determines what kind of
306 * 'path': a graph of witness paths through the collation (DEFAULT)
307 * 'relationship': a graph of how collation readings relate to
313 my( $self, $view ) = @_;
314 $view = 'path' unless $view;
315 # TODO consider making some of these things configurable
316 my $dot = sprintf( "digraph %s {\n", $self->tradition->name );
317 $dot .= "\tedge [ arrowhead=open ];\n";
318 $dot .= "\tgraph [ rankdir=LR ];\n";
319 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
320 11, "white", "filled", $self->graph->get_attribute( 'node', 'shape' ) );
322 foreach my $reading ( $self->readings ) {
323 # Need not output nodes without separate labels
324 next if $reading->name eq $reading->label;
325 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->name, $reading->label );
328 my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
329 foreach my $edge ( @edges ) {
330 my %variables = ( 'color' => '#000000',
331 'fontcolor' => '#000000',
332 'label' => $edge->label,
334 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
335 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
336 $edge->from->name, $edge->to->name, $varopts );
344 print $graph->as_graphml( $recalculate )
346 Returns a GraphML representation of the collation graph, with
347 transposition information and position information. Unless
348 $recalculate is passed (and is a true value), the method will return a
349 cached copy of the SVG after the first call to the method.
354 my( $self, $recalc ) = @_;
355 return $self->graphml if $self->has_graphml;
358 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
359 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
360 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
361 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
363 # Create the document and root node
364 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
365 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
366 $graphml->setDocumentElement( $root );
367 $root->setNamespace( $xsi_ns, 'xsi', 0 );
368 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
370 # Add the data keys for the graph
373 my @graph_attributes = qw/ wit_list_separator baselabel linear ac_label /;
374 foreach my $datum ( @graph_attributes ) {
375 $graph_data_keys{$datum} = 'dg'.$gdi++;
376 my $key = $root->addNewChild( $graphml_ns, 'key' );
377 $key->setAttribute( 'attr.name', $datum );
378 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
379 $key->setAttribute( 'for', 'graph' );
380 $key->setAttribute( 'id', $graph_data_keys{$datum} );
383 # Add the data keys for nodes
386 foreach my $datum ( qw/ name reading identical rank class / ) {
387 $node_data_keys{$datum} = 'dn'.$ndi++;
388 my $key = $root->addNewChild( $graphml_ns, 'key' );
389 $key->setAttribute( 'attr.name', $datum );
390 $key->setAttribute( 'attr.type', 'string' );
391 $key->setAttribute( 'for', 'node' );
392 $key->setAttribute( 'id', $node_data_keys{$datum} );
395 # Add the data keys for edges, i.e. witnesses
398 my @string_keys = qw/ class witness relationship /;
399 my @bool_keys = qw/ extra equal_rank non_correctable non_independent /;
400 foreach my $edge_key( @string_keys ) {
401 $edge_data_keys{$edge_key} = 'de'.$edi++;
402 my $key = $root->addNewChild( $graphml_ns, 'key' );
403 $key->setAttribute( 'attr.name', $edge_key );
404 $key->setAttribute( 'attr.type', 'string' );
405 $key->setAttribute( 'for', 'edge' );
406 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
408 foreach my $edge_key( @bool_keys ) {
409 $edge_data_keys{$edge_key} = 'de'.$edi++;
410 my $key = $root->addNewChild( $graphml_ns, 'key' );
411 $key->setAttribute( 'attr.name', $edge_key );
412 $key->setAttribute( 'attr.type', 'boolean' );
413 $key->setAttribute( 'for', 'edge' );
414 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
417 # Add the graph, its nodes, and its edges
418 my $graph = $root->addNewChild( $graphml_ns, 'graph' );
419 $graph->setAttribute( 'edgedefault', 'directed' );
420 $graph->setAttribute( 'id', $self->tradition->name );
421 $graph->setAttribute( 'parse.edgeids', 'canonical' );
422 $graph->setAttribute( 'parse.edges', scalar($self->paths) );
423 $graph->setAttribute( 'parse.nodeids', 'canonical' );
424 $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
425 $graph->setAttribute( 'parse.order', 'nodesfirst' );
427 # Collation attribute data
428 foreach my $datum ( @graph_attributes ) {
429 _add_graphml_data( $graph, $graph_data_keys{$datum}, $self->$datum );
434 # Add our readings to the graph
435 foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
436 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
437 my $node_xmlid = 'n' . $node_ctr++;
438 $node_hash{ $n->name } = $node_xmlid;
439 $node_el->setAttribute( 'id', $node_xmlid );
440 _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
441 _add_graphml_data( $node_el, $node_data_keys{'reading'}, $n->label );
442 _add_graphml_data( $node_el, $node_data_keys{'rank'}, $n->rank )
444 _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
445 _add_graphml_data( $node_el, $node_data_keys{'identical'}, $n->primary->name )
446 if $n->has_primary && $n->primary ne $n;
449 # Add the path and relationship edges
451 foreach my $e ( sort { $a->from->name cmp $b->from->name } $self->graph->edges() ) {
452 my( $name, $from, $to ) = ( 'e'.$edge_ctr++,
453 $node_hash{ $e->from->name() },
454 $node_hash{ $e->to->name() } );
455 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
456 $edge_el->setAttribute( 'source', $from );
457 $edge_el->setAttribute( 'target', $to );
458 $edge_el->setAttribute( 'id', $name );
460 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, $e->sub_class );
462 # For some classes we have extra information to save.
463 if( $e->sub_class eq 'path' ) {
464 # It's a witness path, so add the witness
465 my $base = $e->label;
466 my $key = $edge_data_keys{'witness_main'};
467 # Is this an ante-corr witness?
468 my $aclabel = $self->ac_label;
469 if( $e->label =~ /^(.*)\Q$aclabel\E$/ ) {
470 # Keep the base witness
472 # ...and record that this is an 'extra' reading path
473 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, 'true' );
475 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
476 } elsif( $e->sub_class eq 'relationship' ) {
477 # It's a relationship, so save the relationship data
478 _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label );
479 _add_graphml_data( $edge_el, $edge_data_keys{'equal_rank'}, $e->equal_rank );
480 _add_graphml_data( $edge_el, $edge_data_keys{'non_correctable'}, $e->non_correctable );
481 _add_graphml_data( $edge_el, $edge_data_keys{'non_independent'}, $e->non_independent );
485 # Save and return the thing
486 my $result = decode_utf8( $graphml->toString(1) );
487 $self->_save_graphml( $result );
491 sub _add_graphml_data {
492 my( $el, $key, $value ) = @_;
493 return unless defined $value;
494 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
495 $data_el->setAttribute( 'key', $key );
496 $data_el->appendText( $value );
501 print $graph->as_csv( $recalculate )
503 Returns a CSV alignment table representation of the collation graph, one
504 row per witness (or witness uncorrected.) Unless $recalculate is passed
505 (and is a true value), the method will return a cached copy of the CSV
506 after the first call to the method.
511 my( $self, $recalc ) = @_;
512 return $self->csv if $self->has_csv;
513 my $table = $self->make_alignment_table;
514 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
516 foreach my $row ( @$table ) {
517 $csv->combine( @$row );
518 push( @result, decode_utf8( $csv->string ) );
520 $self->_save_csv( join( "\n", @result ) );
524 # Make an alignment table - $noderefs controls whether the objects
525 # in the table are the nodes or simply their readings.
527 sub make_alignment_table {
528 my( $self, $noderefs ) = @_;
529 unless( $self->linear ) {
530 warn "Need a linear graph in order to make an alignment table";
534 my @all_pos = sort { $a <=> $b } $self->possible_positions;
536 foreach my $wit ( $self->tradition->witnesses ) {
537 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
538 my @row = _make_witness_row( $wit->path, \@all_pos, $noderefs );
539 unshift( @row, $wit->sigil );
540 push( @$table, \@row );
541 if( $wit->has_ante_corr ) {
542 my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos, $noderefs );
543 unshift( @ac_row, $wit->sigil . $self->ac_label );
544 push( @$table, \@ac_row );
548 # Return a table where the witnesses read in columns rather than rows.
549 my $turned = _turn_table( $table );
553 sub _make_witness_row {
554 my( $path, $positions, $noderefs ) = @_;
556 map { $char_hash{$_} = undef } @$positions;
557 foreach my $rdg ( @$path ) {
558 my $rtext = $rdg->text;
559 $rtext = '#LACUNA#' if $rdg->is_lacuna;
560 $char_hash{$rdg->rank} = $noderefs ? $rdg : $rtext;
562 my @row = map { $char_hash{$_} } @$positions;
563 # Fill in lacuna markers for undef spots in the row
564 my $last_el = shift @row;
565 my @filled_row = ( $last_el );
566 foreach my $el ( @row ) {
567 # If we are using node reference, make the lacuna node appear many times
568 # in the table. If not, use the lacuna tag.
569 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
570 $el = $noderefs ? $last_el : '#LACUNA#';
572 push( @filled_row, $el );
578 # Tiny utility function to say if a table element is a lacuna
581 return 1 if $el eq '#LACUNA#';
582 return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading'
587 # Helper to turn the witnesses along columns rather than rows. Assumes
592 return $result unless scalar @$table;
593 my $nrows = scalar @{$table->[0]};
594 foreach my $idx ( 0 .. $nrows - 1 ) {
595 foreach my $wit ( 0 .. $#{$table} ) {
596 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
603 sub collapse_graph_paths {
605 # Our collation graph has an path per witness. This is great for
606 # calculation purposes, but terrible for display. Thus we want to
607 # display only one path between any two nodes.
609 return if $self->collapsed;
611 print STDERR "Collapsing witness paths in graph...\n";
613 # Don't list out every witness if we have more than half to list.
614 my $majority = int( scalar( $self->tradition->witnesses ) / 2 ) + 1;
615 # But don't compress if there are only a few witnesses.
616 $majority = 4 if $majority < 4;
617 foreach my $node ( $self->readings ) {
619 # We will visit each node, so we only look ahead.
620 foreach my $edge ( $node->outgoing() ) {
621 next unless $edge->class eq 'edge.path';
622 add_hash_entry( $newlabels, $edge->to->name, $edge->name );
623 $self->del_path( $edge );
626 foreach my $newdest ( keys %$newlabels ) {
628 my @compressed_wits = @{$newlabels->{$newdest}};
629 if( @compressed_wits < $majority ) {
630 $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
632 ## TODO FIX THIS HACK
634 foreach my $wit ( @compressed_wits ) {
635 push( @aclabels, $wit ) if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ );
637 $label = join( ', ', 'majority', sort( @aclabels ) );
640 my $newpath = $self->add_path( $node, $self->reading( $newdest ), $label );
641 $newpath->hidden_witnesses( \@compressed_wits );
645 $self->collapsed( 1 );
648 sub expand_graph_paths {
650 # Our collation graph has only one path between any two nodes.
651 # This is great for display, but not so great for analysis.
652 # Expand this so that each witness has its own path between any
654 return unless $self->collapsed;
656 print STDERR "Expanding witness paths in graph...\n";
657 foreach my $path( $self->paths ) {
658 my $from = $path->from;
660 warn sprintf( "No hidden witnesses on %s -> %s ?", $from->name, $to->name )
661 unless $path->has_hidden_witnesses;
662 my @wits = @{$path->hidden_witnesses};
663 $self->del_path( $path );
665 $self->add_path( $from, $to, $_ );
668 $self->collapsed( 0 );
673 =head2 Navigation methods
679 my $beginning = $collation->start();
681 Returns the beginning of the collation, a meta-reading with label '#START#'.
686 # Return the beginning reading of the graph.
687 my( $self, $new_start ) = @_;
688 my $start = $self->reading( '#START#' );
689 if( ref( $new_start ) eq 'Text::Tradition::Collation::Reading' ) {
690 # Replace the existing start node.
691 $self->del_reading( '#START#' );
692 $self->graph->rename_node( $new_start, '#START#' );
694 } elsif ( $new_start && $new_start eq 'INIT' ) {
695 # Make a new start node.
696 $start = $self->add_reading( '#START#' );
698 # Make sure the start node is a meta node
699 $start->is_meta( 1 );
700 # Make sure the start node has a start position.
701 unless( $start->has_rank ) {
709 my $end = $collation->end();
711 Returns the end of the collation, a meta-reading with label '#END#'.
718 my $end = $self->reading( '#END#' );
719 if( ref( $new_end ) eq 'Text::Tradition::Collation::Reading' ) {
720 $self->del_reading( '#END#' );
721 $self->graph->rename_node( $new_end, '#END#' );
723 } elsif ( $new_end && $new_end eq 'INIT' ) {
724 # Make a new start node.
725 $end = $self->add_reading( '#END#' );
727 # Make sure the start node is a meta node
732 =item B<reading_sequence>
734 my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );
736 Returns the ordered list of readings, starting with $first and ending
737 with $last, along the given witness path. If no path is specified,
738 assume that the path is that of the base text (if any.)
742 # TODO Think about returning some lazy-eval iterator.
744 sub reading_sequence {
745 my( $self, $start, $end, $witness, $backup ) = @_;
747 $witness = $self->baselabel unless $witness;
748 my @readings = ( $start );
751 while( $n && $n ne $end ) {
752 if( exists( $seen{$n->name()} ) ) {
753 warn "Detected loop at " . $n->name();
756 $seen{$n->name()} = 1;
758 my $next = $self->next_reading( $n, $witness, $backup );
759 warn "Did not find any path for $witness from reading " . $n->name
761 push( @readings, $next );
764 # Check that the last reading is our end reading.
765 my $last = $readings[$#readings];
766 warn "Last reading found from " . $start->label() .
767 " for witness $witness is not the end!"
768 unless $last eq $end;
773 =item B<next_reading>
775 my $next_reading = $graph->next_reading( $reading, $witpath );
777 Returns the reading that follows the given reading along the given witness
783 # Return the successor via the corresponding path.
785 return $self->_find_linked_reading( 'next', @_ );
788 =item B<prior_reading>
790 my $prior_reading = $graph->prior_reading( $reading, $witpath );
792 Returns the reading that precedes the given reading along the given witness
798 # Return the predecessor via the corresponding path.
800 return $self->_find_linked_reading( 'prior', @_ );
803 sub _find_linked_reading {
804 my( $self, $direction, $node, $path, $alt_path ) = @_;
805 my @linked_paths = $direction eq 'next'
806 ? $node->outgoing() : $node->incoming();
807 return undef unless scalar( @linked_paths );
809 # We have to find the linked path that contains all of the
810 # witnesses supplied in $path.
811 my( @path_wits, @alt_path_wits );
812 @path_wits = $self->witnesses_of_label( $path ) if $path;
813 @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path;
816 foreach my $le ( @linked_paths ) {
817 if( $le->name eq $self->baselabel ) {
820 my @le_wits = $self->witnesses_of_label( $le->name );
821 if( _is_within( \@path_wits, \@le_wits ) ) {
822 # This is the right path.
823 return $direction eq 'next' ? $le->to() : $le->from();
824 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
829 # Got this far? Return the alternate path if it exists.
830 return $direction eq 'next' ? $alt_le->to() : $alt_le->from()
833 # Got this far? Return the base path if it exists.
834 return $direction eq 'next' ? $base_le->to() : $base_le->from()
837 # Got this far? We have no appropriate path.
838 warn "Could not find $direction node from " . $node->label
839 . " along path $path";
845 my( $set1, $set2 ) = @_;
846 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
847 foreach my $el ( @$set1 ) {
848 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
854 ## INITIALIZATION METHODS - for use by parsers
855 # Walk the paths for each witness in the graph, and return the nodes
856 # that the graph has in common. If $using_base is true, some
857 # different logic is needed.
858 # NOTE This does not create paths; it merely finds common readings.
860 sub walk_witness_paths {
862 # For each witness, walk the path through the graph.
863 # Then we need to find the common nodes.
864 # TODO This method is going to fall down if we have a very gappy
865 # text in the collation.
868 foreach my $wit ( $self->tradition->witnesses ) {
869 my $curr_reading = $self->start;
870 my @wit_path = $self->reading_sequence( $self->start, $self->end,
872 $wit->path( \@wit_path );
874 # Detect the common readings.
875 @common_readings = _find_common( \@common_readings, \@wit_path );
878 # Mark all the nodes as either common or not.
879 foreach my $cn ( @common_readings ) {
880 print STDERR "Setting " . $cn->name . " / " . $cn->label
881 . " as common node\n";
884 foreach my $n ( $self->readings() ) {
885 $n->make_variant unless $n->is_common;
887 # Return an array of the common nodes in order.
888 return @common_readings;
892 my( $common_readings, $new_path ) = @_;
894 if( @$common_readings ) {
895 foreach my $n ( @$new_path ) {
896 push( @cr, $n ) if grep { $_ eq $n } @$common_readings;
899 push( @cr, @$new_path );
905 my( $common_readings, $divergence ) = @_;
908 map { $diverged{$_->name} = 1 } @$divergence;
909 foreach( @$common_readings ) {
910 push( @cr, $_ ) unless $diverged{$_->name};
916 # For use when a collation is constructed from a base text and an apparatus.
917 # We have the sequences of readings and just need to add path edges.
919 sub make_witness_paths {
921 foreach my $wit ( $self->tradition->witnesses ) {
922 print STDERR "Making path for " . $wit->sigil . "\n";
923 $self->make_witness_path( $wit );
927 sub make_witness_path {
928 my( $self, $wit ) = @_;
929 my @chain = @{$wit->path};
930 my $sig = $wit->sigil;
931 foreach my $idx ( 0 .. $#chain-1 ) {
932 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
934 if( $wit->has_ante_corr ) {
935 @chain = @{$wit->uncorrected_path};
936 foreach my $idx( 0 .. $#chain-1 ) {
937 my $source = $chain[$idx];
938 my $target = $chain[$idx+1];
939 $self->add_path( $source, $target, $sig.$self->ac_label )
940 unless $self->has_path( $source, $target, $sig );
945 sub calculate_ranks {
947 # Walk a version of the graph where every node linked by a relationship
948 # edge is fundamentally the same node, and do a topological ranking on
949 # the nodes in this graph.
950 my $topo_graph = Graph->new();
954 foreach my $r ( $self->readings ) {
955 next if exists $rel_containers{$r->name};
956 my @rels = $r->related_readings( 'colocated' );
958 # Make a relationship container.
960 my $rn = 'rel_container_' . $rel_ctr++;
961 $topo_graph->add_vertex( $rn );
963 $rel_containers{$_->name} = $rn;
966 # Add a new node to mirror the old node.
967 $rel_containers{$r->name} = $r->name;
968 $topo_graph->add_vertex( $r->name );
972 # Add the edges. Need only one edge between any pair of nodes.
973 foreach my $r ( $self->readings ) {
974 foreach my $n ( $r->neighbor_readings( 'forward' ) ) {
975 my( $tfrom, $tto ) = ( $rel_containers{$r->name},
976 $rel_containers{$n->name} );
977 $topo_graph->add_edge( $tfrom, $tto )
978 unless $topo_graph->has_edge( $tfrom, $tto );
982 # Now do the rankings, starting with the start node.
983 my $topo_start = $rel_containers{$self->start->name};
984 my $node_ranks = { $topo_start => 0 };
985 my @curr_origin = ( $topo_start );
986 # A little iterative function.
987 while( @curr_origin ) {
988 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
990 # Transfer our rankings from the topological graph to the real one.
991 foreach my $r ( $self->readings ) {
992 $r->rank( $node_ranks->{$rel_containers{$r->name}} );
997 my( $graph, $node_ranks, @current_nodes ) = @_;
998 # Look at each of the children of @current_nodes. If all the child's
999 # parents have a rank, assign it the highest rank + 1 and add it to
1000 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1001 # parent gets a rank.
1003 foreach my $c ( @current_nodes ) {
1004 warn "Current reading $c has no rank!"
1005 unless exists $node_ranks->{$c};
1006 # print STDERR "Looking at child of node $c, rank "
1007 # . $node_ranks->{$c} . "\n";
1008 foreach my $child ( $graph->successors( $c ) ) {
1009 next if exists $node_ranks->{$child};
1010 my $highest_rank = -1;
1012 foreach my $parent ( $graph->predecessors( $child ) ) {
1013 if( exists $node_ranks->{$parent} ) {
1014 $highest_rank = $node_ranks->{$parent}
1015 if $highest_rank <= $node_ranks->{$parent};
1022 my $c_rank = $highest_rank + 1;
1023 # print STDERR "Assigning rank $c_rank to node $child \n";
1024 $node_ranks->{$child} = $c_rank;
1025 push( @next_nodes, $child );
1031 # Another method to make up for rough collation methods. If the same reading
1032 # appears multiple times at the same rank, collapse the nodes.
1035 my %unique_rank_rdg;
1036 foreach my $rdg ( $self->readings ) {
1037 next unless $rdg->has_rank;
1038 my $key = $rdg->rank . "||" . $rdg->text;
1039 if( exists $unique_rank_rdg{$key} ) {
1041 print STDERR "Combining readings at same rank: $key\n";
1042 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1044 $unique_rank_rdg{$key} = $rdg;
1050 sub possible_positions {
1053 map { $all_pos{ $_->rank } = 1 } $self->readings;
1054 return keys %all_pos;
1057 # TODO think about indexing this.
1058 sub readings_at_position {
1059 my( $self, $position, $strict ) = @_;
1061 foreach my $r ( $self->readings ) {
1062 push( @answer, $r ) if $r->is_at_position( $position, $strict );
1067 ## Lemmatizer functions
1072 foreach my $position ( $self->possible_positions ) {
1073 $self->lemmata->{$position} = undef;
1076 foreach my $cr ( $self->common_readings ) {
1077 $self->lemmata->{$cr->position->maxref} = $cr->name;
1081 sub common_readings {
1083 my @common = grep { $_->is_common } $self->readings();
1084 return sort { $a->rank <=> $b->rank } @common;
1087 =item B<lemma_readings>
1089 my @state = $graph->lemma_readings( @readings_delemmatized );
1091 Takes a list of readings that have just been delemmatized, and returns
1092 a set of tuples of the form ['reading', 'state'] that indicates what
1093 changes need to be made to the graph.
1099 A state of 1 means 'lemmatize this reading'
1103 A state of 0 means 'delemmatize this reading'
1107 A state of undef means 'an ellipsis belongs in the text here because
1108 no decision has been made / an earlier decision was backed out'
1114 sub lemma_readings {
1115 my( $self, @toggled_off_nodes ) = @_;
1117 # First get the positions of those nodes which have been
1119 my $positions_off = {};
1120 map { $positions_off->{ $_->position->reference } = $_->name }
1123 # Now for each position, we have to see if a node is on, and we
1124 # have to see if a node has been turned off. The lemmata hash
1125 # should contain fixed positions, range positions whose node was
1126 # just turned off, and range positions whose node is on.
1128 my %fixed_positions;
1129 # TODO One of these is probably redundant.
1130 map { $fixed_positions{$_} = 0 } keys %{$self->lemmata};
1131 map { $fixed_positions{$_} = 0 } keys %{$positions_off};
1132 map { $fixed_positions{$_} = 1 } $self->possible_positions;
1133 foreach my $pos ( sort { Text::Tradition::Collation::Position::str_cmp( $a, $b ) } keys %fixed_positions ) {
1134 # Find the state of this position. If there is an active node,
1135 # its name will be the state; otherwise the state will be 0
1136 # (nothing at this position) or undef (ellipsis at this position)
1138 $active = $self->lemmata->{$pos} if exists $self->lemmata->{$pos};
1140 # Is there a formerly active node that was toggled off?
1141 if( exists( $positions_off->{$pos} ) ) {
1142 my $off_node = $positions_off->{$pos};
1143 if( $active && $active ne $off_node) {
1144 push( @answer, [ $off_node, 0 ], [ $active, 1 ] );
1146 unless( $fixed_positions{$pos} ) {
1148 delete $self->lemmata->{$pos};
1150 push( @answer, [ $off_node, $active ] );
1153 # No formerly active node, so we just see if there is a currently
1155 } elsif( $active ) {
1156 # Push the active node, whatever it is.
1157 push( @answer, [ $active, 1 ] );
1159 # Push the state that is there. Arbitrarily use the first node
1161 my @pos_nodes = $self->readings_at_position( $pos );
1162 push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] );
1163 delete $self->lemmata->{$pos} unless $fixed_positions{$pos};
1170 =item B<toggle_reading>
1172 my @readings_delemmatized = $graph->toggle_reading( $reading_name );
1174 Takes a reading node name, and either lemmatizes or de-lemmatizes
1175 it. Returns a list of all readings that are de-lemmatized as a result
1180 sub toggle_reading {
1181 my( $self, $rname ) = @_;
1183 return unless $rname;
1184 my $reading = $self->reading( $rname );
1185 if( !$reading || $reading->is_common() ) {
1186 # Do nothing, it's a common node.
1190 my $pos = $reading->position;
1191 my $fixed = $reading->position->fixed;
1192 my $old_state = $self->lemmata->{$pos->reference};
1195 if( $old_state && $old_state eq $rname ) {
1196 # Turn off the node. We turn on no others by default.
1197 push( @readings_off, $reading );
1200 $self->lemmata->{$pos->reference} = $rname;
1201 # Any other 'on' readings in the same position should be off
1202 # if we have a fixed position.
1203 push( @readings_off, $self->same_position_as( $reading, 1 ) )
1205 # Any node that is an identical transposed one should be off.
1206 push( @readings_off, $reading->identical_readings );
1208 @readings_off = unique_list( @readings_off );
1210 # Turn off the readings that need to be turned off.
1211 my @readings_delemmatized;
1212 foreach my $n ( @readings_off ) {
1213 my $npos = $n->position;
1215 $state = $self->lemmata->{$npos->reference}
1216 if defined $self->lemmata->{$npos->reference};
1217 if( $state && $state eq $n->name ) {
1218 # this reading is still on, so turn it off
1219 push( @readings_delemmatized, $n );
1220 my $new_state = undef;
1221 if( $npos->fixed && $n eq $reading ) {
1222 # This is the reading that was clicked, so if there are no
1223 # other readings there and this is a fixed position, turn off
1224 # the position. In all other cases, restore the ellipsis.
1225 my @other_n = $self->same_position_as( $n ); # TODO do we need strict?
1226 $new_state = 0 unless @other_n;
1228 $self->lemmata->{$npos->reference} = $new_state;
1229 } elsif( $old_state && $old_state eq $n->name ) {
1230 # another reading has already been turned on here
1231 push( @readings_delemmatized, $n );
1232 } # else some other reading was on anyway, so pass.
1234 return @readings_delemmatized;
1237 sub same_position_as {
1238 my( $self, $reading, $strict ) = @_;
1239 my $pos = $reading->position;
1240 my %onpath = ( $reading->name => 1 );
1241 # TODO This might not always be sufficient. We really want to
1242 # exclude all readings on this one's path between its two
1244 map { $onpath{$_->name} = 1 } $reading->neighbor_readings;
1245 my @same = grep { !$onpath{$_->name} }
1246 $self->readings_at_position( $reading->position, $strict );
1250 # Return the string that joins together a list of witnesses for
1251 # display on a single path.
1254 return join( $self->wit_list_separator, @_ );
1257 sub witnesses_of_label {
1258 my( $self, $label ) = @_;
1259 my $regex = $self->wit_list_separator;
1260 my @answer = split( /\Q$regex\E/, $label );
1267 map { $h{$_->name} = $_ } @list;
1268 return values( %h );
1271 sub add_hash_entry {
1272 my( $hash, $key, $entry ) = @_;
1273 if( exists $hash->{$key} ) {
1274 push( @{$hash->{$key}}, $entry );
1276 $hash->{$key} = [ $entry ];
1281 __PACKAGE__->meta->make_immutable;