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 ( grep { $_->sub_class eq 'path' } $pr->incoming ) {
250 if( exists $pr_ids{ $e->from->name } ) {
254 foreach my $e ( grep { $_->sub_class eq 'path' } $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 $graph_name = $self->tradition->name;
317 $graph_name =~ s/[^\w\s]//g;
318 $graph_name = join( '_', split( /\s+/, $graph_name ) );
319 my $dot = sprintf( "digraph %s {\n", $graph_name );
320 $dot .= "\tedge [ arrowhead=open ];\n";
321 $dot .= "\tgraph [ rankdir=LR ];\n";
322 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
323 11, "white", "filled", $self->graph->get_attribute( 'node', 'shape' ) );
325 foreach my $reading ( $self->readings ) {
326 # Need not output nodes without separate labels
327 next if $reading->name eq $reading->label;
328 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->name, $reading->label );
331 my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
332 foreach my $edge ( @edges ) {
333 my %variables = ( 'color' => '#000000',
334 'fontcolor' => '#000000',
335 'label' => $edge->label,
337 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
338 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
339 $edge->from->name, $edge->to->name, $varopts );
347 print $graph->as_graphml( $recalculate )
349 Returns a GraphML representation of the collation graph, with
350 transposition information and position information. Unless
351 $recalculate is passed (and is a true value), the method will return a
352 cached copy of the SVG after the first call to the method.
357 my( $self, $recalc ) = @_;
358 return $self->graphml if $self->has_graphml;
361 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
362 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
363 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
364 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
366 # Create the document and root node
367 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
368 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
369 $graphml->setDocumentElement( $root );
370 $root->setNamespace( $xsi_ns, 'xsi', 0 );
371 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
373 # Add the data keys for the graph
376 my @graph_attributes = qw/ wit_list_separator baselabel linear ac_label /;
377 foreach my $datum ( @graph_attributes ) {
378 $graph_data_keys{$datum} = 'dg'.$gdi++;
379 my $key = $root->addNewChild( $graphml_ns, 'key' );
380 $key->setAttribute( 'attr.name', $datum );
381 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
382 $key->setAttribute( 'for', 'graph' );
383 $key->setAttribute( 'id', $graph_data_keys{$datum} );
386 # Add the data keys for nodes
389 foreach my $datum ( qw/ name reading identical rank class / ) {
390 $node_data_keys{$datum} = 'dn'.$ndi++;
391 my $key = $root->addNewChild( $graphml_ns, 'key' );
392 $key->setAttribute( 'attr.name', $datum );
393 $key->setAttribute( 'attr.type', 'string' );
394 $key->setAttribute( 'for', 'node' );
395 $key->setAttribute( 'id', $node_data_keys{$datum} );
398 # Add the data keys for edges, i.e. witnesses
401 my @string_keys = qw/ class witness relationship /;
402 my @bool_keys = qw/ extra equal_rank non_correctable non_independent /;
403 foreach my $edge_key( @string_keys ) {
404 $edge_data_keys{$edge_key} = 'de'.$edi++;
405 my $key = $root->addNewChild( $graphml_ns, 'key' );
406 $key->setAttribute( 'attr.name', $edge_key );
407 $key->setAttribute( 'attr.type', 'string' );
408 $key->setAttribute( 'for', 'edge' );
409 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
411 foreach my $edge_key( @bool_keys ) {
412 $edge_data_keys{$edge_key} = 'de'.$edi++;
413 my $key = $root->addNewChild( $graphml_ns, 'key' );
414 $key->setAttribute( 'attr.name', $edge_key );
415 $key->setAttribute( 'attr.type', 'boolean' );
416 $key->setAttribute( 'for', 'edge' );
417 $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
420 # Add the graph, its nodes, and its edges
421 my $graph = $root->addNewChild( $graphml_ns, 'graph' );
422 $graph->setAttribute( 'edgedefault', 'directed' );
423 $graph->setAttribute( 'id', $self->tradition->name );
424 $graph->setAttribute( 'parse.edgeids', 'canonical' );
425 $graph->setAttribute( 'parse.edges', scalar($self->paths) );
426 $graph->setAttribute( 'parse.nodeids', 'canonical' );
427 $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
428 $graph->setAttribute( 'parse.order', 'nodesfirst' );
430 # Collation attribute data
431 foreach my $datum ( @graph_attributes ) {
432 _add_graphml_data( $graph, $graph_data_keys{$datum}, $self->$datum );
437 # Add our readings to the graph
438 foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
439 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
440 my $node_xmlid = 'n' . $node_ctr++;
441 $node_hash{ $n->name } = $node_xmlid;
442 $node_el->setAttribute( 'id', $node_xmlid );
443 _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
444 _add_graphml_data( $node_el, $node_data_keys{'reading'}, $n->label );
445 _add_graphml_data( $node_el, $node_data_keys{'rank'}, $n->rank )
447 _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
448 _add_graphml_data( $node_el, $node_data_keys{'identical'}, $n->primary->name )
449 if $n->has_primary && $n->primary ne $n;
452 # Add the path and relationship edges
454 foreach my $e ( sort { $a->from->name cmp $b->from->name } $self->graph->edges() ) {
455 my( $name, $from, $to ) = ( 'e'.$edge_ctr++,
456 $node_hash{ $e->from->name() },
457 $node_hash{ $e->to->name() } );
458 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
459 $edge_el->setAttribute( 'source', $from );
460 $edge_el->setAttribute( 'target', $to );
461 $edge_el->setAttribute( 'id', $name );
463 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, $e->sub_class );
465 # For some classes we have extra information to save.
466 if( $e->sub_class eq 'path' ) {
467 # It's a witness path, so add the witness
468 my $base = $e->label;
469 my $key = $edge_data_keys{'witness_main'};
470 # Is this an ante-corr witness?
471 my $aclabel = $self->ac_label;
472 if( $e->label =~ /^(.*)\Q$aclabel\E$/ ) {
473 # Keep the base witness
475 # ...and record that this is an 'extra' reading path
476 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, 'true' );
478 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
479 } elsif( $e->sub_class eq 'relationship' ) {
480 # It's a relationship, so save the relationship data
481 _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label );
482 _add_graphml_data( $edge_el, $edge_data_keys{'equal_rank'}, $e->equal_rank );
483 _add_graphml_data( $edge_el, $edge_data_keys{'non_correctable'}, $e->non_correctable );
484 _add_graphml_data( $edge_el, $edge_data_keys{'non_independent'}, $e->non_independent );
488 # Save and return the thing
489 my $result = decode_utf8( $graphml->toString(1) );
490 $self->_save_graphml( $result );
494 sub _add_graphml_data {
495 my( $el, $key, $value ) = @_;
496 return unless defined $value;
497 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
498 $data_el->setAttribute( 'key', $key );
499 $data_el->appendText( $value );
504 print $graph->as_csv( $recalculate )
506 Returns a CSV alignment table representation of the collation graph, one
507 row per witness (or witness uncorrected.) Unless $recalculate is passed
508 (and is a true value), the method will return a cached copy of the CSV
509 after the first call to the method.
514 my( $self, $recalc ) = @_;
515 return $self->csv if $self->has_csv;
516 my $table = $self->make_alignment_table;
517 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
519 foreach my $row ( @$table ) {
520 $csv->combine( @$row );
521 push( @result, decode_utf8( $csv->string ) );
523 $self->_save_csv( join( "\n", @result ) );
527 # Make an alignment table - $noderefs controls whether the objects
528 # in the table are the nodes or simply their readings.
530 sub make_alignment_table {
531 my( $self, $noderefs, $include ) = @_;
532 unless( $self->linear ) {
533 warn "Need a linear graph in order to make an alignment table";
537 my @all_pos = sort { $a <=> $b } $self->possible_positions;
538 foreach my $wit ( $self->tradition->witnesses ) {
539 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
540 my @row = _make_witness_row( $wit->path, \@all_pos, $noderefs );
541 unshift( @row, $wit->sigil );
542 push( @$table, \@row );
543 if( $wit->has_ante_corr ) {
544 my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos, $noderefs );
545 unshift( @ac_row, $wit->sigil . $self->ac_label );
546 push( @$table, \@ac_row );
552 # Winnow out the rows for any witness not included.
553 foreach my $row ( @$table ) {
554 next unless $include->{$row->[0]};
555 push( @$winnowed, $row );
560 # Return a table where the witnesses read in columns rather than rows.
561 my $turned = _turn_table( $table );
562 # TODO We should really go through and delete empty rows.
566 sub _make_witness_row {
567 my( $path, $positions, $noderefs ) = @_;
569 map { $char_hash{$_} = undef } @$positions;
570 foreach my $rdg ( @$path ) {
571 my $rtext = $rdg->text;
572 $rtext = '#LACUNA#' if $rdg->is_lacuna;
573 # print STDERR "No rank for " . $rdg->name . "\n" unless defined $rdg->rank;
574 $char_hash{$rdg->rank} = $noderefs ? $rdg : $rtext;
576 my @row = map { $char_hash{$_} } @$positions;
577 # Fill in lacuna markers for undef spots in the row
578 my $last_el = shift @row;
579 my @filled_row = ( $last_el );
580 foreach my $el ( @row ) {
581 # If we are using node reference, make the lacuna node appear many times
582 # in the table. If not, use the lacuna tag.
583 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
584 $el = $noderefs ? $last_el : '#LACUNA#';
586 push( @filled_row, $el );
592 # Tiny utility function to say if a table element is a lacuna
595 return 1 if $el eq '#LACUNA#';
596 return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading'
601 # Helper to turn the witnesses along columns rather than rows. Assumes
606 return $result unless scalar @$table;
607 my $nrows = scalar @{$table->[0]};
608 foreach my $idx ( 0 .. $nrows - 1 ) {
609 foreach my $wit ( 0 .. $#{$table} ) {
610 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
617 sub collapse_graph_paths {
619 # Our collation graph has an path per witness. This is great for
620 # calculation purposes, but terrible for display. Thus we want to
621 # display only one path between any two nodes.
623 return if $self->collapsed;
625 print STDERR "Collapsing witness paths in graph...\n";
627 # Don't list out every witness if we have more than half to list.
628 my $majority = int( scalar( $self->tradition->witnesses ) / 2 ) + 1;
629 # But don't compress if there are only a few witnesses.
630 $majority = 4 if $majority < 4;
631 foreach my $node ( $self->readings ) {
633 # We will visit each node, so we only look ahead.
634 foreach my $edge ( $node->outgoing() ) {
635 next unless $edge->class eq 'edge.path';
636 add_hash_entry( $newlabels, $edge->to->name, $edge->name );
637 $self->del_path( $edge );
640 foreach my $newdest ( keys %$newlabels ) {
642 my @compressed_wits = @{$newlabels->{$newdest}};
643 if( @compressed_wits < $majority ) {
644 $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
646 ## TODO FIX THIS HACK
648 foreach my $wit ( @compressed_wits ) {
649 push( @aclabels, $wit ) if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ );
651 $label = join( ', ', 'majority', sort( @aclabels ) );
654 my $newpath = $self->add_path( $node, $self->reading( $newdest ), $label );
655 $newpath->hidden_witnesses( \@compressed_wits );
659 $self->collapsed( 1 );
662 sub expand_graph_paths {
664 # Our collation graph has only one path between any two nodes.
665 # This is great for display, but not so great for analysis.
666 # Expand this so that each witness has its own path between any
668 return unless $self->collapsed;
670 print STDERR "Expanding witness paths in graph...\n";
671 foreach my $path( $self->paths ) {
672 my $from = $path->from;
674 warn sprintf( "No hidden witnesses on %s -> %s ?", $from->name, $to->name )
675 unless $path->has_hidden_witnesses;
676 my @wits = @{$path->hidden_witnesses};
677 $self->del_path( $path );
679 $self->add_path( $from, $to, $_ );
682 $self->collapsed( 0 );
687 =head2 Navigation methods
693 my $beginning = $collation->start();
695 Returns the beginning of the collation, a meta-reading with label '#START#'.
700 # Return the beginning reading of the graph.
701 my( $self, $new_start ) = @_;
702 my $start = $self->reading( '#START#' );
703 if( ref( $new_start ) eq 'Text::Tradition::Collation::Reading' ) {
704 # Replace the existing start node.
705 $self->del_reading( '#START#' );
706 $self->graph->rename_node( $new_start, '#START#' );
708 } elsif ( $new_start && $new_start eq 'INIT' ) {
709 # Make a new start node.
710 $start = $self->add_reading( '#START#' );
712 # Make sure the start node is a meta node
713 $start->is_meta( 1 );
714 # Make sure the start node has a start position.
715 unless( $start->has_rank ) {
723 my $end = $collation->end();
725 Returns the end of the collation, a meta-reading with label '#END#'.
732 my $end = $self->reading( '#END#' );
733 if( ref( $new_end ) eq 'Text::Tradition::Collation::Reading' ) {
734 $self->del_reading( '#END#' );
735 $self->graph->rename_node( $new_end, '#END#' );
737 } elsif ( $new_end && $new_end eq 'INIT' ) {
738 # Make a new start node.
739 $end = $self->add_reading( '#END#' );
741 # Make sure the start node is a meta node
746 =item B<reading_sequence>
748 my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );
750 Returns the ordered list of readings, starting with $first and ending
751 with $last, along the given witness path. If no path is specified,
752 assume that the path is that of the base text (if any.)
756 # TODO Think about returning some lazy-eval iterator.
758 sub reading_sequence {
759 my( $self, $start, $end, $witness, $backup ) = @_;
761 $witness = $self->baselabel unless $witness;
762 my @readings = ( $start );
765 while( $n && $n ne $end ) {
766 if( exists( $seen{$n->name()} ) ) {
767 warn "Detected loop at " . $n->name();
770 $seen{$n->name()} = 1;
772 my $next = $self->next_reading( $n, $witness, $backup );
773 warn "Did not find any path for $witness from reading " . $n->name
775 push( @readings, $next );
778 # Check that the last reading is our end reading.
779 my $last = $readings[$#readings];
780 warn "Last reading found from " . $start->label() .
781 " for witness $witness is not the end!"
782 unless $last eq $end;
787 =item B<next_reading>
789 my $next_reading = $graph->next_reading( $reading, $witpath );
791 Returns the reading that follows the given reading along the given witness
797 # Return the successor via the corresponding path.
799 return $self->_find_linked_reading( 'next', @_ );
802 =item B<prior_reading>
804 my $prior_reading = $graph->prior_reading( $reading, $witpath );
806 Returns the reading that precedes the given reading along the given witness
812 # Return the predecessor via the corresponding path.
814 return $self->_find_linked_reading( 'prior', @_ );
817 sub _find_linked_reading {
818 my( $self, $direction, $node, $path, $alt_path ) = @_;
819 my @linked_paths = $direction eq 'next'
820 ? $node->outgoing() : $node->incoming();
821 return undef unless scalar( @linked_paths );
823 # We have to find the linked path that contains all of the
824 # witnesses supplied in $path.
825 my( @path_wits, @alt_path_wits );
826 @path_wits = $self->witnesses_of_label( $path ) if $path;
827 @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path;
830 foreach my $le ( @linked_paths ) {
831 if( $le->name eq $self->baselabel ) {
834 my @le_wits = $self->witnesses_of_label( $le->name );
835 if( _is_within( \@path_wits, \@le_wits ) ) {
836 # This is the right path.
837 return $direction eq 'next' ? $le->to() : $le->from();
838 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
843 # Got this far? Return the alternate path if it exists.
844 return $direction eq 'next' ? $alt_le->to() : $alt_le->from()
847 # Got this far? Return the base path if it exists.
848 return $direction eq 'next' ? $base_le->to() : $base_le->from()
851 # Got this far? We have no appropriate path.
852 warn "Could not find $direction node from " . $node->label
853 . " along path $path";
859 my( $set1, $set2 ) = @_;
860 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
861 foreach my $el ( @$set1 ) {
862 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
868 ## INITIALIZATION METHODS - for use by parsers
869 # Walk the paths for each witness in the graph, and return the nodes
870 # that the graph has in common. If $using_base is true, some
871 # different logic is needed.
872 # NOTE This does not create paths; it merely finds common readings.
874 sub walk_witness_paths {
876 # For each witness, walk the path through the graph.
877 # Then we need to find the common nodes.
878 # TODO This method is going to fall down if we have a very gappy
879 # text in the collation.
882 foreach my $wit ( $self->tradition->witnesses ) {
883 my $curr_reading = $self->start;
884 my @wit_path = $self->reading_sequence( $self->start, $self->end,
886 $wit->path( \@wit_path );
888 # Detect the common readings.
889 @common_readings = _find_common( \@common_readings, \@wit_path );
892 # Mark all the nodes as either common or not.
893 foreach my $cn ( @common_readings ) {
894 print STDERR "Setting " . $cn->name . " / " . $cn->label
895 . " as common node\n";
898 foreach my $n ( $self->readings() ) {
899 $n->make_variant unless $n->is_common;
901 # Return an array of the common nodes in order.
902 return @common_readings;
906 my( $common_readings, $new_path ) = @_;
908 if( @$common_readings ) {
909 foreach my $n ( @$new_path ) {
910 push( @cr, $n ) if grep { $_ eq $n } @$common_readings;
913 push( @cr, @$new_path );
919 my( $common_readings, $divergence ) = @_;
922 map { $diverged{$_->name} = 1 } @$divergence;
923 foreach( @$common_readings ) {
924 push( @cr, $_ ) unless $diverged{$_->name};
930 # For use when a collation is constructed from a base text and an apparatus.
931 # We have the sequences of readings and just need to add path edges.
933 sub make_witness_paths {
935 foreach my $wit ( $self->tradition->witnesses ) {
936 print STDERR "Making path for " . $wit->sigil . "\n";
937 $self->make_witness_path( $wit );
941 sub make_witness_path {
942 my( $self, $wit ) = @_;
943 my @chain = @{$wit->path};
944 my $sig = $wit->sigil;
945 foreach my $idx ( 0 .. $#chain-1 ) {
946 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
948 if( $wit->has_ante_corr ) {
949 @chain = @{$wit->uncorrected_path};
950 foreach my $idx( 0 .. $#chain-1 ) {
951 my $source = $chain[$idx];
952 my $target = $chain[$idx+1];
953 $self->add_path( $source, $target, $sig.$self->ac_label )
954 unless $self->has_path( $source, $target, $sig );
959 sub calculate_ranks {
961 # Walk a version of the graph where every node linked by a relationship
962 # edge is fundamentally the same node, and do a topological ranking on
963 # the nodes in this graph.
964 my $topo_graph = Graph->new();
968 foreach my $r ( $self->readings ) {
969 next if exists $rel_containers{$r->name};
970 my @rels = $r->related_readings( 'colocated' );
972 # Make a relationship container.
974 my $rn = 'rel_container_' . $rel_ctr++;
975 $topo_graph->add_vertex( $rn );
977 $rel_containers{$_->name} = $rn;
980 # Add a new node to mirror the old node.
981 $rel_containers{$r->name} = $r->name;
982 $topo_graph->add_vertex( $r->name );
986 # Add the edges. Need only one edge between any pair of nodes.
987 foreach my $r ( $self->readings ) {
988 foreach my $n ( $r->neighbor_readings( 'forward' ) ) {
989 my( $tfrom, $tto ) = ( $rel_containers{$r->name},
990 $rel_containers{$n->name} );
991 $topo_graph->add_edge( $tfrom, $tto )
992 unless $topo_graph->has_edge( $tfrom, $tto );
996 # Now do the rankings, starting with the start node.
997 my $topo_start = $rel_containers{$self->start->name};
998 my $node_ranks = { $topo_start => 0 };
999 my @curr_origin = ( $topo_start );
1000 # A little iterative function.
1001 while( @curr_origin ) {
1002 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1004 # Transfer our rankings from the topological graph to the real one.
1005 foreach my $r ( $self->readings ) {
1006 if( defined $node_ranks->{$rel_containers{$r->name}} ) {
1007 $r->rank( $node_ranks->{$rel_containers{$r->name}} );
1010 die "No rank calculated for node " . $r->name
1011 . " - do you have a cycle in the graph?";
1017 my( $graph, $node_ranks, @current_nodes ) = @_;
1018 # Look at each of the children of @current_nodes. If all the child's
1019 # parents have a rank, assign it the highest rank + 1 and add it to
1020 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1021 # parent gets a rank.
1023 foreach my $c ( @current_nodes ) {
1024 warn "Current reading $c has no rank!"
1025 unless exists $node_ranks->{$c};
1026 # print STDERR "Looking at child of node $c, rank "
1027 # . $node_ranks->{$c} . "\n";
1028 foreach my $child ( $graph->successors( $c ) ) {
1029 next if exists $node_ranks->{$child};
1030 my $highest_rank = -1;
1032 foreach my $parent ( $graph->predecessors( $child ) ) {
1033 if( exists $node_ranks->{$parent} ) {
1034 $highest_rank = $node_ranks->{$parent}
1035 if $highest_rank <= $node_ranks->{$parent};
1042 my $c_rank = $highest_rank + 1;
1043 # print STDERR "Assigning rank $c_rank to node $child \n";
1044 $node_ranks->{$child} = $c_rank;
1045 push( @next_nodes, $child );
1051 # Another method to make up for rough collation methods. If the same reading
1052 # appears multiple times at the same rank, collapse the nodes.
1055 my %unique_rank_rdg;
1056 foreach my $rdg ( $self->readings ) {
1057 next unless $rdg->has_rank;
1058 my $key = $rdg->rank . "||" . $rdg->text;
1059 if( exists $unique_rank_rdg{$key} ) {
1061 print STDERR "Combining readings at same rank: $key\n";
1062 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1064 $unique_rank_rdg{$key} = $rdg;
1070 sub possible_positions {
1073 map { $all_pos{ $_->rank } = 1 } $self->readings;
1074 return keys %all_pos;
1077 # TODO think about indexing this.
1078 sub readings_at_position {
1079 my( $self, $position, $strict ) = @_;
1081 foreach my $r ( $self->readings ) {
1082 push( @answer, $r ) if $r->is_at_position( $position, $strict );
1087 ## Lemmatizer functions
1092 foreach my $position ( $self->possible_positions ) {
1093 $self->lemmata->{$position} = undef;
1096 foreach my $cr ( $self->common_readings ) {
1097 $self->lemmata->{$cr->position->maxref} = $cr->name;
1101 sub common_readings {
1103 my @common = grep { $_->is_common } $self->readings();
1104 return sort { $a->rank <=> $b->rank } @common;
1107 =item B<lemma_readings>
1109 my @state = $graph->lemma_readings( @readings_delemmatized );
1111 Takes a list of readings that have just been delemmatized, and returns
1112 a set of tuples of the form ['reading', 'state'] that indicates what
1113 changes need to be made to the graph.
1119 A state of 1 means 'lemmatize this reading'
1123 A state of 0 means 'delemmatize this reading'
1127 A state of undef means 'an ellipsis belongs in the text here because
1128 no decision has been made / an earlier decision was backed out'
1134 sub lemma_readings {
1135 my( $self, @toggled_off_nodes ) = @_;
1137 # First get the positions of those nodes which have been
1139 my $positions_off = {};
1140 map { $positions_off->{ $_->position->reference } = $_->name }
1143 # Now for each position, we have to see if a node is on, and we
1144 # have to see if a node has been turned off. The lemmata hash
1145 # should contain fixed positions, range positions whose node was
1146 # just turned off, and range positions whose node is on.
1148 my %fixed_positions;
1149 # TODO One of these is probably redundant.
1150 map { $fixed_positions{$_} = 0 } keys %{$self->lemmata};
1151 map { $fixed_positions{$_} = 0 } keys %{$positions_off};
1152 map { $fixed_positions{$_} = 1 } $self->possible_positions;
1153 foreach my $pos ( sort { Text::Tradition::Collation::Position::str_cmp( $a, $b ) } keys %fixed_positions ) {
1154 # Find the state of this position. If there is an active node,
1155 # its name will be the state; otherwise the state will be 0
1156 # (nothing at this position) or undef (ellipsis at this position)
1158 $active = $self->lemmata->{$pos} if exists $self->lemmata->{$pos};
1160 # Is there a formerly active node that was toggled off?
1161 if( exists( $positions_off->{$pos} ) ) {
1162 my $off_node = $positions_off->{$pos};
1163 if( $active && $active ne $off_node) {
1164 push( @answer, [ $off_node, 0 ], [ $active, 1 ] );
1166 unless( $fixed_positions{$pos} ) {
1168 delete $self->lemmata->{$pos};
1170 push( @answer, [ $off_node, $active ] );
1173 # No formerly active node, so we just see if there is a currently
1175 } elsif( $active ) {
1176 # Push the active node, whatever it is.
1177 push( @answer, [ $active, 1 ] );
1179 # Push the state that is there. Arbitrarily use the first node
1181 my @pos_nodes = $self->readings_at_position( $pos );
1182 push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] );
1183 delete $self->lemmata->{$pos} unless $fixed_positions{$pos};
1190 =item B<toggle_reading>
1192 my @readings_delemmatized = $graph->toggle_reading( $reading_name );
1194 Takes a reading node name, and either lemmatizes or de-lemmatizes
1195 it. Returns a list of all readings that are de-lemmatized as a result
1200 sub toggle_reading {
1201 my( $self, $rname ) = @_;
1203 return unless $rname;
1204 my $reading = $self->reading( $rname );
1205 if( !$reading || $reading->is_common() ) {
1206 # Do nothing, it's a common node.
1210 my $pos = $reading->position;
1211 my $fixed = $reading->position->fixed;
1212 my $old_state = $self->lemmata->{$pos->reference};
1215 if( $old_state && $old_state eq $rname ) {
1216 # Turn off the node. We turn on no others by default.
1217 push( @readings_off, $reading );
1220 $self->lemmata->{$pos->reference} = $rname;
1221 # Any other 'on' readings in the same position should be off
1222 # if we have a fixed position.
1223 push( @readings_off, $self->same_position_as( $reading, 1 ) )
1225 # Any node that is an identical transposed one should be off.
1226 push( @readings_off, $reading->identical_readings );
1228 @readings_off = unique_list( @readings_off );
1230 # Turn off the readings that need to be turned off.
1231 my @readings_delemmatized;
1232 foreach my $n ( @readings_off ) {
1233 my $npos = $n->position;
1235 $state = $self->lemmata->{$npos->reference}
1236 if defined $self->lemmata->{$npos->reference};
1237 if( $state && $state eq $n->name ) {
1238 # this reading is still on, so turn it off
1239 push( @readings_delemmatized, $n );
1240 my $new_state = undef;
1241 if( $npos->fixed && $n eq $reading ) {
1242 # This is the reading that was clicked, so if there are no
1243 # other readings there and this is a fixed position, turn off
1244 # the position. In all other cases, restore the ellipsis.
1245 my @other_n = $self->same_position_as( $n ); # TODO do we need strict?
1246 $new_state = 0 unless @other_n;
1248 $self->lemmata->{$npos->reference} = $new_state;
1249 } elsif( $old_state && $old_state eq $n->name ) {
1250 # another reading has already been turned on here
1251 push( @readings_delemmatized, $n );
1252 } # else some other reading was on anyway, so pass.
1254 return @readings_delemmatized;
1257 sub same_position_as {
1258 my( $self, $reading, $strict ) = @_;
1259 my $pos = $reading->position;
1260 my %onpath = ( $reading->name => 1 );
1261 # TODO This might not always be sufficient. We really want to
1262 # exclude all readings on this one's path between its two
1264 map { $onpath{$_->name} = 1 } $reading->neighbor_readings;
1265 my @same = grep { !$onpath{$_->name} }
1266 $self->readings_at_position( $reading->position, $strict );
1270 # Return the string that joins together a list of witnesses for
1271 # display on a single path.
1274 return join( $self->wit_list_separator, @_ );
1277 sub witnesses_of_label {
1278 my( $self, $label ) = @_;
1279 my $regex = $self->wit_list_separator;
1280 my @answer = split( /\Q$regex\E/, $label );
1287 map { $h{$_->name} = $_ } @list;
1288 return values( %h );
1291 sub add_hash_entry {
1292 my( $hash, $key, $entry ) = @_;
1293 if( exists $hash->{$key} ) {
1294 push( @{$hash->{$key}}, $entry );
1296 $hash->{$key} = [ $entry ];
1301 __PACKAGE__->meta->make_immutable;
1307 =item * Rationalize edge classes
1309 =item * Port the internal graph from Graph::Easy to Graph