1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
6 use IPC::Run qw( run binary );
8 use Text::Tradition::Collation::Reading;
15 default => sub { Graph->new() },
24 default => sub { Graph->new( undirected => 1 ) },
26 relationships => 'edges',
32 isa => 'Text::Tradition',
37 isa => 'HashRef[Text::Tradition::Collation::Reading]',
41 _add_reading => 'set',
42 del_reading => 'delete',
43 has_reading => 'exists',
46 default => sub { {} },
49 has 'wit_list_separator' => (
58 default => 'base text',
75 isa => 'Text::Tradition::Collation::Reading',
76 writer => '_set_start',
82 isa => 'Text::Tradition::Collation::Reading',
87 # The collation can be created two ways:
88 # 1. Collate a set of witnesses (with CollateX I guess) and process
89 # the results as in 2.
90 # 2. Read a pre-prepared collation in one of a variety of formats,
91 # and make the graph from that.
93 # The graph itself will (for now) be immutable, and the positions
94 # within the graph will also be immutable. We need to calculate those
95 # positions upon graph construction. The equivalences between graph
96 # nodes will be mutable, entirely determined by the user (or possibly
97 # by some semantic pre-processing provided by the user.) So the
98 # constructor should just make an empty equivalences object. The
99 # constructor will also need to make the witness objects, if we didn't
100 # come through option 1.
104 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
105 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
108 ### Reading construct/destruct functions
111 my( $self, $reading ) = @_;
112 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
113 my %args = %$reading;
114 $reading = Text::Tradition::Collation::Reading->new(
115 'collation' => $self,
118 # First check to see if a reading with this ID exists.
119 if( $self->reading( $reading->id ) ) {
120 warn "Collation already has a reading with id " . $reading->id;
123 $self->_add_reading( $reading->id => $reading );
124 # Once the reading has been added, put it in both graphs.
125 $self->sequence->add_vertex( $reading->id );
126 $self->relations->add_vertex( $reading->id );
130 around del_reading => sub {
135 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
138 # Remove the reading from the graphs.
139 $self->sequence->delete_vertex( $arg );
140 $self->relations->delete_vertex( $arg );
143 $self->$orig( $arg );
146 # merge_readings( $main, $to_be_deleted );
151 # We only need the IDs for adding paths to the graph, not the reading
152 # objects themselves.
153 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
155 # The kept reading should inherit the paths and the relationships
156 # of the deleted reading.
157 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
158 my @vector = ( $kept );
159 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
160 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
161 next if $vector[0] eq $vector[1]; # Don't add a self loop
162 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
163 $self->sequence->add_edge( @vector );
164 my $fwits = $self->sequence->get_edge_attributes( @vector );
165 @wits{keys %$fwits} = values %$fwits;
166 $self->sequence->set_edge_attributes( @vector, \%wits );
168 foreach my $rel ( $self->relations->edges_at( $deleted ) ) {
169 my @vector = ( $kept );
170 push( @vector, $rel->[0] eq $deleted ? $rel->[1] : $rel->[0] );
171 next if $vector[0] eq $vector[1]; # Don't add a self loop
172 # Is there a relationship here already? If so, keep it.
173 # TODO Warn about conflicting relationships
174 next if $self->relations->has_edge( @vector );
175 # If not, adopt the relationship that would be deleted.
176 $self->relations->add_edge( @vector );
177 my $attr = $self->relations->get_edge_attributes( @$rel );
178 $self->relations->set_edge_attributes( @vector, $attr );
181 # Do the deletion deed.
182 if( $combine_char ) {
183 my $kept_obj = $self->reading( $kept );
184 my $new_text = join( $combine_char, $kept_obj->text,
185 $self->reading( $deleted )->text );
186 $kept_obj->alter_text( $new_text );
188 $self->del_reading( $deleted );
192 # Helper function for manipulating the graph.
193 sub _stringify_args {
194 my( $self, $first, $second, $arg ) = @_;
196 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
197 $second = $second->id
198 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
199 return( $first, $second, $arg );
207 # We only need the IDs for adding paths to the graph, not the reading
208 # objects themselves.
209 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
211 # Connect the readings
212 $self->sequence->add_edge( $source, $target );
213 # Note the witness in question
214 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
220 if( ref( $_[0] ) eq 'ARRAY' ) {
227 # We only need the IDs for adding paths to the graph, not the reading
228 # objects themselves.
229 my( $source, $target, $wit ) = $self->_stringify_args( @args );
231 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
232 $self->sequence->delete_edge_attribute( $source, $target, $wit );
234 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
235 $self->sequence->delete_edge( $source, $target );
240 # Extra graph-alike utility
243 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
244 return undef unless $self->sequence->has_edge( $source, $target );
245 return $self->sequence->has_edge_attribute( $source, $target, $wit );
248 ### Relationship logic
250 =head2 add_relationship( $reading1, $reading2, $definition )
252 Adds the specified relationship between the two readings. A relationship
253 is transitive (i.e. undirected), and must have the following attributes
254 specified in the hashref $definition:
258 =item * type - Can be one of spelling, orthographic, grammatical, meaning, repetition, transposition. The first three are only valid relationships between readings that occur at the same point in the text.
260 =item * non_correctable - (Optional) True if the reading would not have been corrected independently.
262 =item * non_independent - (Optional) True if the variant is unlikely to have occurred independently in unrelated witnesses.
264 =item * global - (Optional) A meta-attribute, to set the same relationship between readings with the same text whenever they occur in the same place.
270 # Wouldn't it be lovely if edges could be objects, and all this type checking
271 # and attribute management could be done via Moose?
273 sub add_relationship {
275 my( $source, $target, $options ) = $self->_stringify_args( @_ );
278 if( !defined $options->{'type'} ||
279 $options->{'type'} !~ /^(spelling|orthographic|grammatical|meaning|lexical|repetition|transposition)$/i ) {
280 my $t = $options->{'type'} ? $options->{'type'} : '';
281 return( undef, "Invalid or missing type " . $options->{'type'} );
283 unless ( $options->{'type'} =~ /^(repetition|transposition)$/ ) {
284 $options->{'colocated'} = 1;
287 # Make sure there is not another relationship between these two
289 if( $self->relations->has_edge( $source, $target ) ) {
290 return ( undef, "Relationship already exists between these readings" );
292 if( !$self->relationship_valid( $source, $target, $options->{'type'} ) ) {
293 return ( undef, 'Relationship creates witness loop' );
296 my @vector = ( $source, $target );
297 $self->relations->add_edge( @vector );
298 $self->relations->set_edge_attributes( @vector, $options );
300 # TODO Handle global relationship setting
302 return( 1, @vector );
305 sub relationship_valid {
306 my( $self, $source, $target, $rel ) = @_;
307 if( $rel eq 'repetition' ) {
309 } elsif ( $rel eq 'transposition' ) {
310 # Check that the two readings do not appear in the same witness.
312 map { $seen_wits{$_} = 1 } $self->reading_witnesses( $source );
313 foreach my $w ( $self->reading_witnesses( $target ) ) {
314 return 0 if $seen_wits{$w};
318 # Check that linking the source and target in a relationship won't lead
319 # to a path loop for any witness. First make a lookup table of all the
320 # readings related to either the source or the target.
321 my @proposed_related = ( $source, $target );
322 push( @proposed_related, $self->related_readings( $source, 'colocated' ) );
323 push( @proposed_related, $self->related_readings( $target, 'colocated' ) );
325 map { $pr_ids{ $_ } = 1 } @proposed_related;
327 # None of these proposed related readings should have a neighbor that
328 # is also in proposed_related.
329 foreach my $pr ( keys %pr_ids ) {
330 foreach my $neighbor( $self->sequence->neighbors( $pr ) ) {
331 return 0 if exists $pr_ids{$neighbor};
338 # Return a list of the witnesses in which the reading appears.
339 sub reading_witnesses {
340 my( $self, $reading ) = @_;
341 # We need only check either the incoming or the outgoing edges; I have
342 # arbitrarily chosen "incoming".
344 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
345 my $wits = $self->sequence->get_edge_attributes( @$e );
346 @all_witnesses{ keys %$wits } = 1;
348 return keys %all_witnesses;
351 sub related_readings {
352 my( $self, $reading, $colocated ) = @_;
354 if( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
355 $reading = $reading->id;
357 # print STDERR "Returning related objects\n";
359 # print STDERR "Returning related object names\n";
361 my @related = $self->relations->all_reachable( $reading );
363 my @colo = grep { $self->relations->has_edge_attribute( $reading, $_, 'colocated' ) } @related;
366 return $return_object ? map { $self->reading( $_ ) } @related : @related;
369 =head2 Output method(s)
375 print $graph->as_svg( $recalculate );
377 Returns an SVG string that represents the graph, via as_dot and graphviz.
384 my @cmd = qw/dot -Tsvg/;
386 my $dotfile = File::Temp->new();
388 # $dotfile->unlink_on_destroy(0);
389 binmode $dotfile, ':utf8';
390 print $dotfile $self->as_dot();
391 push( @cmd, $dotfile->filename );
392 run( \@cmd, ">", binary(), \$svg );
393 $svg = decode_utf8( $svg );
399 print $graph->as_dot( $view, $recalculate );
401 Returns a string that is the collation graph expressed in dot
402 (i.e. GraphViz) format. The 'view' argument determines what kind of
404 * 'path': a graph of witness paths through the collation (DEFAULT)
405 * 'relationship': a graph of how collation readings relate to
411 my( $self, $view ) = @_;
412 $view = 'sequence' unless $view;
413 # TODO consider making some of these things configurable
414 my $graph_name = $self->tradition->name;
415 $graph_name =~ s/[^\w\s]//g;
416 $graph_name = join( '_', split( /\s+/, $graph_name ) );
417 my $dot = sprintf( "digraph %s {\n", $graph_name );
418 $dot .= "\tedge [ arrowhead=open ];\n";
419 $dot .= "\tgraph [ rankdir=LR ];\n";
420 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
421 11, "white", "filled", "ellipse" );
423 foreach my $reading ( $self->readings ) {
424 # Need not output nodes without separate labels
425 next if $reading->id eq $reading->text;
426 my $label = $reading->text;
427 $label =~ s/\"/\\\"/g;
428 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->id, $label );
431 # TODO do something sensible for relationships
433 my @edges = $self->paths;
434 foreach my $edge ( @edges ) {
435 my %variables = ( 'color' => '#000000',
436 'fontcolor' => '#000000',
437 'label' => join( ', ', $self->path_display_label( $edge ) ),
439 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
440 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
441 $edge->[0], $edge->[1], $varopts );
448 my( $self, @edge ) = @_;
449 # If edge is an arrayref, cope.
450 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
454 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
458 sub path_display_label {
459 my( $self, $edge ) = @_;
460 my @wits = $self->path_witnesses( $edge );
461 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
462 if( scalar @wits > $maj ) {
465 return join( ', ', @wits );
472 print $graph->as_graphml( $recalculate )
474 Returns a GraphML representation of the collation graph, with
475 transposition information and position information. Unless
476 $recalculate is passed (and is a true value), the method will return a
477 cached copy of the SVG after the first call to the method.
485 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
486 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
487 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
488 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
490 # Create the document and root node
491 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
492 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
493 $graphml->setDocumentElement( $root );
494 $root->setNamespace( $xsi_ns, 'xsi', 0 );
495 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
497 # Add the data keys for the graph
500 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
501 foreach my $datum ( @graph_attributes ) {
502 $graph_data_keys{$datum} = 'dg'.$gdi++;
503 my $key = $root->addNewChild( $graphml_ns, 'key' );
504 $key->setAttribute( 'attr.name', $datum );
505 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
506 $key->setAttribute( 'for', 'graph' );
507 $key->setAttribute( 'id', $graph_data_keys{$datum} );
510 # Add the data keys for nodes
517 is_start => 'boolean',
519 is_lacuna => 'boolean',
521 foreach my $datum ( keys %node_data ) {
522 $node_data_keys{$datum} = 'dn'.$ndi++;
523 my $key = $root->addNewChild( $graphml_ns, 'key' );
524 $key->setAttribute( 'attr.name', $datum );
525 $key->setAttribute( 'attr.type', $node_data{$datum} );
526 $key->setAttribute( 'for', 'node' );
527 $key->setAttribute( 'id', $node_data_keys{$datum} );
530 # Add the data keys for edges, i.e. witnesses
534 class => 'string', # Path or relationship?
535 witness => 'string', # ID/label for a path
536 relationship => 'string', # ID/label for a relationship
537 extra => 'boolean', # Path key
538 colocated => 'boolean', # Relationship key
539 non_correctable => 'boolean', # Relationship key
540 non_independent => 'boolean', # Relationship key
542 foreach my $datum ( keys %edge_data ) {
543 $edge_data_keys{$datum} = 'de'.$edi++;
544 my $key = $root->addNewChild( $graphml_ns, 'key' );
545 $key->setAttribute( 'attr.name', $datum );
546 $key->setAttribute( 'attr.type', $edge_data{$datum} );
547 $key->setAttribute( 'for', 'edge' );
548 $key->setAttribute( 'id', $edge_data_keys{$datum} );
551 # Add the collation graph itself
552 my $graph = $root->addNewChild( $graphml_ns, 'graph' );
553 $graph->setAttribute( 'edgedefault', 'directed' );
554 $graph->setAttribute( 'id', $self->tradition->name );
555 $graph->setAttribute( 'parse.edgeids', 'canonical' );
556 $graph->setAttribute( 'parse.edges', scalar($self->paths) );
557 $graph->setAttribute( 'parse.nodeids', 'canonical' );
558 $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
559 $graph->setAttribute( 'parse.order', 'nodesfirst' );
561 # Collation attribute data
562 foreach my $datum ( @graph_attributes ) {
563 my $value = $datum eq 'version' ? '2.0' : $self->$datum;
564 _add_graphml_data( $graph, $graph_data_keys{$datum}, $value );
569 # Add our readings to the graph
570 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
571 my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
572 my $node_xmlid = 'n' . $node_ctr++;
573 $node_hash{ $n->id } = $node_xmlid;
574 $node_el->setAttribute( 'id', $node_xmlid );
575 foreach my $d ( keys %node_data ) {
577 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
584 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
585 # We add an edge in the graphml for every witness in $e.
586 foreach my $wit ( $self->path_witnesses( $e ) ) {
587 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
588 $node_hash{ $e->[0] },
589 $node_hash{ $e->[1] } );
590 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
591 $edge_el->setAttribute( 'source', $from );
592 $edge_el->setAttribute( 'target', $to );
593 $edge_el->setAttribute( 'id', $id );
595 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
597 # It's a witness path, so add the witness
599 my $key = $edge_data_keys{'witness'};
600 # Is this an ante-corr witness?
601 my $aclabel = $self->ac_label;
602 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
603 # Keep the base witness
605 # ...and record that this is an 'extra' reading path
606 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
608 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
612 # Add the relationship edges
613 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->relationships ) {
614 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
615 $node_hash{ $e->[0] },
616 $node_hash{ $e->[1] } );
617 my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
618 $edge_el->setAttribute( 'source', $from );
619 $edge_el->setAttribute( 'target', $to );
620 $edge_el->setAttribute( 'id', $id );
622 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'relationship' );
624 my $data = $self->relations->get_edge_attributes( @$e );
625 # It's a relationship, so save the relationship data
626 _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $data->{type} );
627 _add_graphml_data( $edge_el, $edge_data_keys{'colocated'}, $data->{colocated} );
628 if( exists $data->{non_correctable} ) {
629 _add_graphml_data( $edge_el, $edge_data_keys{'non_correctable'},
630 $data->{non_correctable} );
632 if( exists $data->{non_independent} ) {
633 _add_graphml_data( $edge_el, $edge_data_keys{'non_independent'},
634 $data->{non_independent} );
638 # Save and return the thing
639 my $result = decode_utf8( $graphml->toString(1) );
643 sub _add_graphml_data {
644 my( $el, $key, $value ) = @_;
645 return unless defined $value;
646 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
647 $data_el->setAttribute( 'key', $key );
648 $data_el->appendText( $value );
653 print $graph->as_csv( $recalculate )
655 Returns a CSV alignment table representation of the collation graph, one
656 row per witness (or witness uncorrected.) Unless $recalculate is passed
657 (and is a true value), the method will return a cached copy of the CSV
658 after the first call to the method.
664 my $table = $self->make_alignment_table;
665 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
667 foreach my $row ( @$table ) {
668 $csv->combine( @$row );
669 push( @result, decode_utf8( $csv->string ) );
671 return join( "\n", @result );
674 # Make an alignment table - $noderefs controls whether the objects
675 # in the table are the nodes or simply their readings.
677 sub make_alignment_table {
678 my( $self, $noderefs, $include ) = @_;
679 unless( $self->linear ) {
680 warn "Need a linear graph in order to make an alignment table";
685 my @all_pos = ( 1 .. $self->end->rank - 1 );
686 foreach my $wit ( $self->tradition->witnesses ) {
687 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
688 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
689 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
690 unshift( @row, $wit->sigil );
691 push( @$table, \@row );
692 if( $wit->is_layered ) {
693 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
694 $wit->sigil.$self->ac_label, $wit->sigil );
695 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
696 unshift( @ac_row, $wit->sigil . $self->ac_label );
697 push( @$table, \@ac_row );
703 # Winnow out the rows for any witness not included.
704 foreach my $row ( @$table ) {
705 next unless $include->{$row->[0]};
706 push( @$winnowed, $row );
711 # Return a table where the witnesses read in columns rather than rows.
712 my $turned = _turn_table( $table );
713 # TODO We should really go through and delete empty rows.
717 sub _make_witness_row {
718 my( $path, $positions, $noderefs ) = @_;
720 map { $char_hash{$_} = undef } @$positions;
721 foreach my $rdg ( @$path ) {
722 my $rtext = $rdg->text;
723 $rtext = '#LACUNA#' if $rdg->is_lacuna;
724 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
725 $char_hash{$rdg->rank} = $noderefs ? $rdg : $rtext;
727 my @row = map { $char_hash{$_} } @$positions;
728 # Fill in lacuna markers for undef spots in the row
729 my $last_el = shift @row;
730 my @filled_row = ( $last_el );
731 foreach my $el ( @row ) {
732 # If we are using node reference, make the lacuna node appear many times
733 # in the table. If not, use the lacuna tag.
734 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
735 $el = $noderefs ? $last_el : '#LACUNA#';
737 push( @filled_row, $el );
743 # Tiny utility function to say if a table element is a lacuna
746 return 1 if $el eq '#LACUNA#';
747 return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading'
752 # Helper to turn the witnesses along columns rather than rows. Assumes
757 return $result unless scalar @$table;
758 my $nrows = scalar @{$table->[0]};
759 foreach my $idx ( 0 .. $nrows - 1 ) {
760 foreach my $wit ( 0 .. $#{$table} ) {
761 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
769 =head2 Navigation methods
775 my $beginning = $collation->start();
777 Returns the beginning of the collation, a meta-reading with label '#START#'.
781 my $end = $collation->end();
783 Returns the end of the collation, a meta-reading with label '#END#'.
786 =item B<reading_sequence>
788 my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );
790 Returns the ordered list of readings, starting with $first and ending
791 with $last, along the given witness path. If no path is specified,
792 assume that the path is that of the base text (if any.)
796 # TODO Think about returning some lazy-eval iterator.
798 sub reading_sequence {
799 my( $self, $start, $end, $witness, $backup ) = @_;
801 $witness = $self->baselabel unless $witness;
802 my @readings = ( $start );
805 while( $n && $n->id ne $end->id ) {
806 if( exists( $seen{$n->id} ) ) {
807 warn "Detected loop at " . $n->id;
812 my $next = $self->next_reading( $n, $witness, $backup );
814 warn "Did not find any path for $witness from reading " . $n->id;
817 push( @readings, $next );
820 # Check that the last reading is our end reading.
821 my $last = $readings[$#readings];
822 warn "Last reading found from " . $start->text .
823 " for witness $witness is not the end!"
824 unless $last->id eq $end->id;
829 =item B<next_reading>
831 my $next_reading = $graph->next_reading( $reading, $witpath );
833 Returns the reading that follows the given reading along the given witness
839 # Return the successor via the corresponding path.
841 my $answer = $self->_find_linked_reading( 'next', @_ );
842 return $self->reading( $answer );
845 =item B<prior_reading>
847 my $prior_reading = $graph->prior_reading( $reading, $witpath );
849 Returns the reading that precedes the given reading along the given witness
855 # Return the predecessor via the corresponding path.
857 my $answer = $self->_find_linked_reading( 'prior', @_ );
858 return $self->reading( $answer );
861 sub _find_linked_reading {
862 my( $self, $direction, $node, $path, $alt_path ) = @_;
863 my @linked_paths = $direction eq 'next'
864 ? $self->sequence->edges_from( $node )
865 : $self->sequence->edges_to( $node );
866 return undef unless scalar( @linked_paths );
868 # We have to find the linked path that contains all of the
869 # witnesses supplied in $path.
870 my( @path_wits, @alt_path_wits );
871 @path_wits = sort( $self->witnesses_of_label( $path ) ) if $path;
872 @alt_path_wits = sort( $self->witnesses_of_label( $alt_path ) ) if $alt_path;
875 foreach my $le ( @linked_paths ) {
876 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
879 my @le_wits = $self->path_witnesses( $le );
880 if( _is_within( \@path_wits, \@le_wits ) ) {
881 # This is the right path.
882 return $direction eq 'next' ? $le->[1] : $le->[0];
883 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
887 # Got this far? Return the alternate path if it exists.
888 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
891 # Got this far? Return the base path if it exists.
892 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
895 # Got this far? We have no appropriate path.
896 warn "Could not find $direction node from " . $node->label
897 . " along path $path";
903 my( $set1, $set2 ) = @_;
904 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
905 foreach my $el ( @$set1 ) {
906 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
912 ## INITIALIZATION METHODS - for use by parsers
914 # For use when a collation is constructed from a base text and an apparatus.
915 # We have the sequences of readings and just need to add path edges.
916 # When we are done, clear out the witness path attributes, as they are no
918 # TODO Find a way to replace the witness path attributes with encapsulated functions?
920 sub make_witness_paths {
922 foreach my $wit ( $self->tradition->witnesses ) {
923 # print STDERR "Making path for " . $wit->sigil . "\n";
924 $self->make_witness_path( $wit );
928 sub make_witness_path {
929 my( $self, $wit ) = @_;
930 my @chain = @{$wit->path};
931 my $sig = $wit->sigil;
932 foreach my $idx ( 0 .. $#chain-1 ) {
933 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
935 if( $wit->is_layered ) {
936 @chain = @{$wit->uncorrected_path};
937 foreach my $idx( 0 .. $#chain-1 ) {
938 my $source = $chain[$idx];
939 my $target = $chain[$idx+1];
940 $self->add_path( $source, $target, $sig.$self->ac_label )
941 unless $self->has_path( $source, $target, $sig );
945 $wit->clear_uncorrected_path;
948 sub calculate_ranks {
950 # Walk a version of the graph where every node linked by a relationship
951 # edge is fundamentally the same node, and do a topological ranking on
952 # the nodes in this graph.
953 my $topo_graph = Graph->new();
957 foreach my $r ( $self->readings ) {
958 next if exists $rel_containers{$r->id};
959 my @rels = $r->related_readings( 'colocated' );
961 # Make a relationship container.
963 my $rn = 'rel_container_' . $rel_ctr++;
964 $topo_graph->add_vertex( $rn );
966 $rel_containers{$_->id} = $rn;
969 # Add a new node to mirror the old node.
970 $rel_containers{$r->id} = $r->id;
971 $topo_graph->add_vertex( $r->id );
976 foreach my $r ( $self->readings ) {
977 foreach my $n ( $self->sequence->successors( $r->id ) ) {
978 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
979 $rel_containers{$n} );
980 $DB::single = 1 unless $tfrom && $tto;
981 $topo_graph->add_edge( $tfrom, $tto );
985 # Now do the rankings, starting with the start node.
986 my $topo_start = $rel_containers{$self->start->id};
987 my $node_ranks = { $topo_start => 0 };
988 my @curr_origin = ( $topo_start );
989 # A little iterative function.
990 while( @curr_origin ) {
991 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
993 # Transfer our rankings from the topological graph to the real one.
994 foreach my $r ( $self->readings ) {
995 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
996 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
999 die "No rank calculated for node " . $r->id
1000 . " - do you have a cycle in the graph?";
1006 my( $graph, $node_ranks, @current_nodes ) = @_;
1007 # Look at each of the children of @current_nodes. If all the child's
1008 # parents have a rank, assign it the highest rank + 1 and add it to
1009 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1010 # parent gets a rank.
1012 foreach my $c ( @current_nodes ) {
1013 warn "Current reading $c has no rank!"
1014 unless exists $node_ranks->{$c};
1015 # print STDERR "Looking at child of node $c, rank "
1016 # . $node_ranks->{$c} . "\n";
1017 foreach my $child ( $graph->successors( $c ) ) {
1018 next if exists $node_ranks->{$child};
1019 my $highest_rank = -1;
1021 foreach my $parent ( $graph->predecessors( $child ) ) {
1022 if( exists $node_ranks->{$parent} ) {
1023 $highest_rank = $node_ranks->{$parent}
1024 if $highest_rank <= $node_ranks->{$parent};
1031 my $c_rank = $highest_rank + 1;
1032 # print STDERR "Assigning rank $c_rank to node $child \n";
1033 $node_ranks->{$child} = $c_rank;
1034 push( @next_nodes, $child );
1040 # Another method to make up for rough collation methods. If the same reading
1041 # appears multiple times at the same rank, collapse the nodes.
1044 my %unique_rank_rdg;
1045 foreach my $rdg ( $self->readings ) {
1046 next unless $rdg->has_rank;
1047 my $key = $rdg->rank . "||" . $rdg->text;
1048 if( exists $unique_rank_rdg{$key} ) {
1050 # print STDERR "Combining readings at same rank: $key\n";
1051 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1053 $unique_rank_rdg{$key} = $rdg;
1059 ## Utility functions
1061 # Return the string that joins together a list of witnesses for
1062 # display on a single path.
1063 sub witnesses_of_label {
1064 my( $self, $label ) = @_;
1065 my $regex = $self->wit_list_separator;
1066 my @answer = split( /\Q$regex\E/, $label );
1071 __PACKAGE__->meta->make_immutable;
1077 =item * Think about making Relationship objects again