1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
6 use IPC::Run qw( run binary );
8 use Text::Tradition::Collation::Reading;
9 use Text::Tradition::Collation::RelationshipStore;
16 default => sub { Graph->new() },
24 isa => 'Text::Tradition::Collation::RelationshipStore',
26 relationships => 'relationships',
27 related_readings => 'related_readings',
29 writer => '_set_relations',
34 isa => 'Text::Tradition',
39 isa => 'HashRef[Text::Tradition::Collation::Reading]',
43 _add_reading => 'set',
44 del_reading => 'delete',
45 has_reading => 'exists',
48 default => sub { {} },
51 has 'wit_list_separator' => (
60 default => 'base text',
77 isa => 'Text::Tradition::Collation::Reading',
78 writer => '_set_start',
84 isa => 'Text::Tradition::Collation::Reading',
89 # The collation can be created two ways:
90 # 1. Collate a set of witnesses (with CollateX I guess) and process
91 # the results as in 2.
92 # 2. Read a pre-prepared collation in one of a variety of formats,
93 # and make the graph from that.
95 # The graph itself will (for now) be immutable, and the positions
96 # within the graph will also be immutable. We need to calculate those
97 # positions upon graph construction. The equivalences between graph
98 # nodes will be mutable, entirely determined by the user (or possibly
99 # by some semantic pre-processing provided by the user.) So the
100 # constructor should just make an empty equivalences object. The
101 # constructor will also need to make the witness objects, if we didn't
102 # come through option 1.
106 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
107 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
108 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
111 ### Reading construct/destruct functions
114 my( $self, $reading ) = @_;
115 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
116 my %args = %$reading;
117 $reading = Text::Tradition::Collation::Reading->new(
118 'collation' => $self,
121 # First check to see if a reading with this ID exists.
122 if( $self->reading( $reading->id ) ) {
123 warn "Collation already has a reading with id " . $reading->id;
126 $self->_add_reading( $reading->id => $reading );
127 # Once the reading has been added, put it in both graphs.
128 $self->sequence->add_vertex( $reading->id );
129 $self->relations->add_reading( $reading->id );
133 around del_reading => sub {
138 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
141 # Remove the reading from the graphs.
142 $self->sequence->delete_vertex( $arg );
143 $self->relations->delete_reading( $arg );
146 $self->$orig( $arg );
149 # merge_readings( $main, $to_be_deleted );
154 # We only need the IDs for adding paths to the graph, not the reading
155 # objects themselves.
156 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
158 # The kept reading should inherit the paths and the relationships
159 # of the deleted reading.
160 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
161 my @vector = ( $kept );
162 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
163 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
164 next if $vector[0] eq $vector[1]; # Don't add a self loop
165 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
166 $self->sequence->add_edge( @vector );
167 my $fwits = $self->sequence->get_edge_attributes( @vector );
168 @wits{keys %$fwits} = values %$fwits;
169 $self->sequence->set_edge_attributes( @vector, \%wits );
171 $self->relations->merge_readings( $kept, $deleted, $combine_char );
173 # Do the deletion deed.
174 if( $combine_char ) {
175 my $kept_obj = $self->reading( $kept );
176 my $new_text = join( $combine_char, $kept_obj->text,
177 $self->reading( $deleted )->text );
178 $kept_obj->alter_text( $new_text );
180 $self->del_reading( $deleted );
184 # Helper function for manipulating the graph.
185 sub _stringify_args {
186 my( $self, $first, $second, $arg ) = @_;
188 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
189 $second = $second->id
190 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
191 return( $first, $second, $arg );
199 # We only need the IDs for adding paths to the graph, not the reading
200 # objects themselves.
201 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
203 # Connect the readings
204 $self->sequence->add_edge( $source, $target );
205 # Note the witness in question
206 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
212 if( ref( $_[0] ) eq 'ARRAY' ) {
219 # We only need the IDs for adding paths to the graph, not the reading
220 # objects themselves.
221 my( $source, $target, $wit ) = $self->_stringify_args( @args );
223 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
224 $self->sequence->delete_edge_attribute( $source, $target, $wit );
226 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
227 $self->sequence->delete_edge( $source, $target );
232 # Extra graph-alike utility
235 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
236 return undef unless $self->sequence->has_edge( $source, $target );
237 return $self->sequence->has_edge_attribute( $source, $target, $wit );
240 =head2 add_relationship( $reading1, $reading2, $definition )
242 Adds the specified relationship between the two readings. A relationship
243 is transitive (i.e. undirected); the options for its definition may be found
244 in Text::Tradition::Collation::Relationship.
248 # Wouldn't it be lovely if edges could be objects, and all this type checking
249 # and attribute management could be done via Moose?
251 sub add_relationship {
253 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
254 my( $ret, @vectors ) = $self->relations->add_relationship( $source,
255 $self->reading( $source ), $target, $self->reading( $target ), $opts );
256 # Force a full rank recalculation every time. Yuck.
257 $self->calculate_ranks() if $ret && $self->end->has_rank;
258 return( $ret, @vectors );
261 =head2 reading_witnesses( $reading )
263 Return a list of sigils corresponding to the witnesses in which the reading appears.
267 sub reading_witnesses {
268 my( $self, $reading ) = @_;
269 # We need only check either the incoming or the outgoing edges; I have
270 # arbitrarily chosen "incoming". Thus, special-case the start node.
271 if( $reading eq $self->start ) {
272 return map { $_->sigil } $self->tradition->witnesses;
275 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
276 my $wits = $self->sequence->get_edge_attributes( @$e );
277 @all_witnesses{ keys %$wits } = 1;
279 return keys %all_witnesses;
282 =head2 Output method(s)
288 print $collation->as_svg();
290 Returns an SVG string that represents the graph, via as_dot and graphviz.
297 my @cmd = qw/dot -Tsvg/;
299 my $dotfile = File::Temp->new();
301 # $dotfile->unlink_on_destroy(0);
302 binmode $dotfile, ':utf8';
303 print $dotfile $self->as_dot();
304 push( @cmd, $dotfile->filename );
305 run( \@cmd, ">", binary(), \$svg );
306 $svg = decode_utf8( $svg );
310 =item B<svg_subgraph>
312 print $collation->svg_subgraph( $from, $to )
314 Returns an SVG string that represents the portion of the graph given by the
315 specified range. The $from and $to variables refer to ranks within the graph.
320 my( $self, $from, $to ) = @_;
322 my $dot = $self->as_dot( $from, $to );
324 warn "Could not output a graph with range $from - $to";
328 my @cmd = qw/dot -Tsvg/;
330 my $dotfile = File::Temp->new();
332 # $dotfile->unlink_on_destroy(0);
333 binmode $dotfile, ':utf8';
335 push( @cmd, $dotfile->filename );
336 run( \@cmd, ">", binary(), \$svg );
337 $svg = decode_utf8( $svg );
344 print $collation->as_dot();
346 Returns a string that is the collation graph expressed in dot
347 (i.e. GraphViz) format. The 'view' argument determines what kind of
349 * 'path': a graph of witness paths through the collation (DEFAULT)
350 * 'relationship': a graph of how collation readings relate to
356 my( $self, $startrank, $endrank ) = @_;
358 # Check the arguments
360 return if $endrank && $startrank > $endrank;
361 return if $startrank > $self->end->rank;
363 if( defined $endrank ) {
364 return if $endrank < 0;
367 # TODO consider making some of these things configurable
368 my $graph_name = $self->tradition->name;
369 $graph_name =~ s/[^\w\s]//g;
370 $graph_name = join( '_', split( /\s+/, $graph_name ) );
371 my $dot = sprintf( "digraph %s {\n", $graph_name );
372 $dot .= "\tedge [ arrowhead=open ];\n";
373 $dot .= "\tgraph [ rankdir=LR,bgcolor=none ];\n";
374 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
375 11, "white", "filled", "ellipse" );
377 # Output substitute start/end readings if necessary
379 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
382 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
384 my %used; # Keep track of the readings that actually appear in the graph
387 foreach my $reading ( $self->readings ) {
388 # Only output readings within our rank range.
389 next if $startrank && $reading->rank < $startrank;
390 next if $endrank && $reading->rank > $endrank;
391 $used{$reading->id} = 1;
392 $subedges{$reading->id} = '#SUBSTART#'
393 if $startrank && $startrank == $reading->rank;
394 $subedges{$reading->id} = '#SUBEND#'
395 if $endrank && $endrank == $reading->rank;
396 # Need not output nodes without separate labels
397 next if $reading->id eq $reading->text;
398 my $label = $reading->text;
399 $label =~ s/\"/\\\"/g;
400 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->id, $label );
403 # Add substitute start and end edges if necessary
404 foreach my $node ( keys %subedges ) {
405 my @vector = ( $subedges{$node}, $node );
406 @vector = reverse( @vector ) if $vector[0] =~ /END/;
407 my $witstr = join( ', ', sort $self->reading_witnesses( $self->reading( $node ) ) );
408 my %variables = ( 'color' => '#000000',
409 'fontcolor' => '#000000',
412 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
413 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n", @vector, $varopts );
417 my @edges = $self->paths;
418 foreach my $edge ( @edges ) {
419 # Do we need to output this edge?
420 if( $used{$edge->[0]} && $used{$edge->[1]} ) {;
421 my %variables = ( 'color' => '#000000',
422 'fontcolor' => '#000000',
423 'label' => join( ', ', $self->path_display_label( $edge ) ),
425 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
426 # Account for the rank gap if necessary
427 my $rankgap = $self->reading( $edge->[1] )->rank
428 - $self->reading( $edge->[0] )->rank;
429 $varopts .= ", minlen=$rankgap" if $rankgap > 1;
430 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
431 $edge->[0], $edge->[1], $varopts );
440 my( $self, @edge ) = @_;
441 # If edge is an arrayref, cope.
442 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
446 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
450 sub path_display_label {
451 my( $self, $edge ) = @_;
452 my @wits = $self->path_witnesses( $edge );
453 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
454 if( scalar @wits > $maj ) {
457 return join( ', ', @wits );
464 print $collation->as_graphml( $recalculate )
466 Returns a GraphML representation of the collation graph, with
467 transposition information and position information. Unless
468 $recalculate is passed (and is a true value), the method will return a
469 cached copy of the SVG after the first call to the method.
477 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
478 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
479 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
480 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
482 # Create the document and root node
483 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
484 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
485 $graphml->setDocumentElement( $root );
486 $root->setNamespace( $xsi_ns, 'xsi', 0 );
487 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
489 # Add the data keys for the graph
492 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
493 foreach my $datum ( @graph_attributes ) {
494 $graph_data_keys{$datum} = 'dg'.$gdi++;
495 my $key = $root->addNewChild( $graphml_ns, 'key' );
496 $key->setAttribute( 'attr.name', $datum );
497 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
498 $key->setAttribute( 'for', 'graph' );
499 $key->setAttribute( 'id', $graph_data_keys{$datum} );
502 # Add the data keys for nodes
509 is_start => 'boolean',
511 is_lacuna => 'boolean',
513 foreach my $datum ( keys %node_data ) {
514 $node_data_keys{$datum} = 'dn'.$ndi++;
515 my $key = $root->addNewChild( $graphml_ns, 'key' );
516 $key->setAttribute( 'attr.name', $datum );
517 $key->setAttribute( 'attr.type', $node_data{$datum} );
518 $key->setAttribute( 'for', 'node' );
519 $key->setAttribute( 'id', $node_data_keys{$datum} );
522 # Add the data keys for edges, i.e. witnesses
526 class => 'string', # Class, deprecated soon
527 witness => 'string', # ID/label for a path
528 relationship => 'string', # ID/label for a relationship
529 extra => 'boolean', # Path key
530 colocated => 'boolean', # Relationship key
531 non_correctable => 'boolean', # Relationship key
532 non_independent => 'boolean', # Relationship key
534 foreach my $datum ( keys %edge_data ) {
535 $edge_data_keys{$datum} = 'de'.$edi++;
536 my $key = $root->addNewChild( $graphml_ns, 'key' );
537 $key->setAttribute( 'attr.name', $datum );
538 $key->setAttribute( 'attr.type', $edge_data{$datum} );
539 $key->setAttribute( 'for', 'edge' );
540 $key->setAttribute( 'id', $edge_data_keys{$datum} );
543 # Add the collation graph itself
544 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
545 $sgraph->setAttribute( 'edgedefault', 'directed' );
546 $sgraph->setAttribute( 'id', $self->tradition->name );
547 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
548 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
549 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
550 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
551 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
553 # Collation attribute data
554 foreach my $datum ( @graph_attributes ) {
555 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
556 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
561 # Add our readings to the graph
562 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
563 # Add to the main graph
564 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
565 my $node_xmlid = 'n' . $node_ctr++;
566 $node_hash{ $n->id } = $node_xmlid;
567 $node_el->setAttribute( 'id', $node_xmlid );
568 foreach my $d ( keys %node_data ) {
570 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
575 # Add the path edges to the sequence graph
577 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
578 # We add an edge in the graphml for every witness in $e.
579 foreach my $wit ( $self->path_witnesses( $e ) ) {
580 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
581 $node_hash{ $e->[0] },
582 $node_hash{ $e->[1] } );
583 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
584 $edge_el->setAttribute( 'source', $from );
585 $edge_el->setAttribute( 'target', $to );
586 $edge_el->setAttribute( 'id', $id );
588 # It's a witness path, so add the witness
590 my $key = $edge_data_keys{'witness'};
591 # Is this an ante-corr witness?
592 my $aclabel = $self->ac_label;
593 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
594 # Keep the base witness
596 # ...and record that this is an 'extra' reading path
597 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
599 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
600 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
604 # Add the relationship graph to the XML
605 $self->relations->as_graphml( $root );
607 # Save and return the thing
608 my $result = decode_utf8( $graphml->toString(1) );
612 sub _add_graphml_data {
613 my( $el, $key, $value ) = @_;
614 return unless defined $value;
615 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
616 $data_el->setAttribute( 'key', $key );
617 $data_el->appendText( $value );
622 print $collation->as_csv( $recalculate )
624 Returns a CSV alignment table representation of the collation graph, one
625 row per witness (or witness uncorrected.)
631 my $table = $self->make_alignment_table;
632 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
634 # Make the header row
635 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
636 push( @result, decode_utf8( $csv->string ) );
637 # Make the rest of the rows
638 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
639 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
640 my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs;
641 $csv->combine( @row );
642 push( @result, decode_utf8( $csv->string ) );
644 return join( "\n", @result );
647 =item B<make_alignment_table>
649 my $table = $collation->make_alignment_table( $use_refs, \@wits_to_include )
651 Return a reference to an alignment table, in a slightly enhanced CollateX
652 format which looks like this:
654 $table = { alignment => [ { witness => "SIGIL",
655 tokens => [ { t => "READINGTEXT" }, ... ] },
657 tokens => [ { t => "READINGTEXT" }, ... ] },
661 If $use_refs is set to 1, the reading object is returned in the table
662 instead of READINGTEXT; if not, the text of the reading is returned.
663 If $wits_to_include is set to a hashref, only the witnesses whose sigil
664 keys have a true hash value will be included.
668 sub make_alignment_table {
669 my( $self, $noderefs, $include ) = @_;
670 unless( $self->linear ) {
671 warn "Need a linear graph in order to make an alignment table";
674 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
675 my @all_pos = ( 1 .. $self->end->rank - 1 );
676 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
678 next unless $include->{$wit->sigil};
680 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
681 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
682 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
683 push( @{$table->{'alignment'}},
684 { 'witness' => $wit->sigil, 'tokens' => \@row } );
685 if( $wit->is_layered ) {
686 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
687 $wit->sigil.$self->ac_label, $wit->sigil );
688 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
689 push( @{$table->{'alignment'}},
690 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
696 sub _make_witness_row {
697 my( $path, $positions, $noderefs ) = @_;
699 map { $char_hash{$_} = undef } @$positions;
701 foreach my $rdg ( @$path ) {
702 my $rtext = $rdg->text;
703 $rtext = '#LACUNA#' if $rdg->is_lacuna;
704 print STDERR "rank " . $rdg->rank . "\n" if $debug;
705 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
706 $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg }
709 my @row = map { $char_hash{$_} } @$positions;
710 # Fill in lacuna markers for undef spots in the row
711 my $last_el = shift @row;
712 my @filled_row = ( $last_el );
713 foreach my $el ( @row ) {
714 # If we are using node reference, make the lacuna node appear many times
715 # in the table. If not, use the lacuna tag.
716 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
717 $el = $noderefs ? $last_el : { 't' => '#LACUNA#' };
719 push( @filled_row, $el );
725 # Tiny utility function to say if a table element is a lacuna
728 return 1 if $el->{'t'} eq '#LACUNA#';
729 return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading'
730 && $el->{'t'}->is_lacuna;
734 # Helper to turn the witnesses along columns rather than rows. Assumes
739 return $result unless scalar @$table;
740 my $nrows = scalar @{$table->[0]};
741 foreach my $idx ( 0 .. $nrows - 1 ) {
742 foreach my $wit ( 0 .. $#{$table} ) {
743 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
751 =head2 Navigation methods
757 my $beginning = $collation->start();
759 Returns the beginning of the collation, a meta-reading with label '#START#'.
763 my $end = $collation->end();
765 Returns the end of the collation, a meta-reading with label '#END#'.
768 =item B<reading_sequence>
770 my @readings = $collation->reading_sequence( $first, $last, $path[, $alt_path] );
772 Returns the ordered list of readings, starting with $first and ending
773 with $last, along the given witness path. If no path is specified,
774 assume that the path is that of the base text (if any.)
778 # TODO Think about returning some lazy-eval iterator.
780 sub reading_sequence {
781 my( $self, $start, $end, $witness, $backup ) = @_;
783 $witness = $self->baselabel unless $witness;
784 my @readings = ( $start );
787 while( $n && $n->id ne $end->id ) {
788 if( exists( $seen{$n->id} ) ) {
789 warn "Detected loop at " . $n->id;
794 my $next = $self->next_reading( $n, $witness, $backup );
796 warn "Did not find any path for $witness from reading " . $n->id;
799 push( @readings, $next );
802 # Check that the last reading is our end reading.
803 my $last = $readings[$#readings];
804 warn "Last reading found from " . $start->text .
805 " for witness $witness is not the end!"
806 unless $last->id eq $end->id;
811 =item B<next_reading>
813 my $next_reading = $collation->next_reading( $reading, $witpath );
815 Returns the reading that follows the given reading along the given witness
821 # Return the successor via the corresponding path.
823 my $answer = $self->_find_linked_reading( 'next', @_ );
824 return undef unless $answer;
825 return $self->reading( $answer );
828 =item B<prior_reading>
830 my $prior_reading = $collation->prior_reading( $reading, $witpath );
832 Returns the reading that precedes the given reading along the given witness
838 # Return the predecessor via the corresponding path.
840 my $answer = $self->_find_linked_reading( 'prior', @_ );
841 return $self->reading( $answer );
844 sub _find_linked_reading {
845 my( $self, $direction, $node, $path, $alt_path ) = @_;
846 my @linked_paths = $direction eq 'next'
847 ? $self->sequence->edges_from( $node )
848 : $self->sequence->edges_to( $node );
849 return undef unless scalar( @linked_paths );
851 # We have to find the linked path that contains all of the
852 # witnesses supplied in $path.
853 my( @path_wits, @alt_path_wits );
854 @path_wits = sort( $self->witnesses_of_label( $path ) ) if $path;
855 @alt_path_wits = sort( $self->witnesses_of_label( $alt_path ) ) if $alt_path;
858 foreach my $le ( @linked_paths ) {
859 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
862 my @le_wits = $self->path_witnesses( $le );
863 if( _is_within( \@path_wits, \@le_wits ) ) {
864 # This is the right path.
865 return $direction eq 'next' ? $le->[1] : $le->[0];
866 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
870 # Got this far? Return the alternate path if it exists.
871 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
874 # Got this far? Return the base path if it exists.
875 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
878 # Got this far? We have no appropriate path.
879 warn "Could not find $direction node from " . $node->id
880 . " along path $path";
886 my( $set1, $set2 ) = @_;
887 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
888 foreach my $el ( @$set1 ) {
889 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
895 ## INITIALIZATION METHODS - for use by parsers
897 # For use when a collation is constructed from a base text and an apparatus.
898 # We have the sequences of readings and just need to add path edges.
899 # When we are done, clear out the witness path attributes, as they are no
901 # TODO Find a way to replace the witness path attributes with encapsulated functions?
903 sub make_witness_paths {
905 foreach my $wit ( $self->tradition->witnesses ) {
906 # print STDERR "Making path for " . $wit->sigil . "\n";
907 $self->make_witness_path( $wit );
911 sub make_witness_path {
912 my( $self, $wit ) = @_;
913 my @chain = @{$wit->path};
914 my $sig = $wit->sigil;
915 foreach my $idx ( 0 .. $#chain-1 ) {
916 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
918 if( $wit->is_layered ) {
919 @chain = @{$wit->uncorrected_path};
920 foreach my $idx( 0 .. $#chain-1 ) {
921 my $source = $chain[$idx];
922 my $target = $chain[$idx+1];
923 $self->add_path( $source, $target, $sig.$self->ac_label )
924 unless $self->has_path( $source, $target, $sig );
928 $wit->clear_uncorrected_path;
931 sub calculate_ranks {
933 # Walk a version of the graph where every node linked by a relationship
934 # edge is fundamentally the same node, and do a topological ranking on
935 # the nodes in this graph.
936 my $topo_graph = Graph->new();
940 foreach my $r ( $self->readings ) {
941 next if exists $rel_containers{$r->id};
942 my @rels = $r->related_readings( 'colocated' );
944 # Make a relationship container.
946 my $rn = 'rel_container_' . $rel_ctr++;
947 $topo_graph->add_vertex( $rn );
949 $rel_containers{$_->id} = $rn;
952 # Add a new node to mirror the old node.
953 $rel_containers{$r->id} = $r->id;
954 $topo_graph->add_vertex( $r->id );
959 foreach my $r ( $self->readings ) {
960 foreach my $n ( $self->sequence->successors( $r->id ) ) {
961 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
962 $rel_containers{$n} );
963 $DB::single = 1 unless $tfrom && $tto;
964 $topo_graph->add_edge( $tfrom, $tto );
968 # Now do the rankings, starting with the start node.
969 my $topo_start = $rel_containers{$self->start->id};
970 my $node_ranks = { $topo_start => 0 };
971 my @curr_origin = ( $topo_start );
972 # A little iterative function.
973 while( @curr_origin ) {
974 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
976 # Transfer our rankings from the topological graph to the real one.
977 foreach my $r ( $self->readings ) {
978 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
979 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
982 die "No rank calculated for node " . $r->id
983 . " - do you have a cycle in the graph?";
989 my( $graph, $node_ranks, @current_nodes ) = @_;
990 # Look at each of the children of @current_nodes. If all the child's
991 # parents have a rank, assign it the highest rank + 1 and add it to
992 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
993 # parent gets a rank.
995 foreach my $c ( @current_nodes ) {
996 warn "Current reading $c has no rank!"
997 unless exists $node_ranks->{$c};
998 # print STDERR "Looking at child of node $c, rank "
999 # . $node_ranks->{$c} . "\n";
1000 foreach my $child ( $graph->successors( $c ) ) {
1001 next if exists $node_ranks->{$child};
1002 my $highest_rank = -1;
1004 foreach my $parent ( $graph->predecessors( $child ) ) {
1005 if( exists $node_ranks->{$parent} ) {
1006 $highest_rank = $node_ranks->{$parent}
1007 if $highest_rank <= $node_ranks->{$parent};
1014 my $c_rank = $highest_rank + 1;
1015 # print STDERR "Assigning rank $c_rank to node $child \n";
1016 $node_ranks->{$child} = $c_rank;
1017 push( @next_nodes, $child );
1023 # Another method to make up for rough collation methods. If the same reading
1024 # appears multiple times at the same rank, collapse the nodes.
1027 my %unique_rank_rdg;
1028 foreach my $rdg ( $self->readings ) {
1029 next unless $rdg->has_rank;
1030 my $key = $rdg->rank . "||" . $rdg->text;
1031 if( exists $unique_rank_rdg{$key} ) {
1033 # print STDERR "Combining readings at same rank: $key\n";
1034 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1036 $unique_rank_rdg{$key} = $rdg;
1042 ## Utility functions
1044 # Return the string that joins together a list of witnesses for
1045 # display on a single path.
1046 sub witnesses_of_label {
1047 my( $self, $label ) = @_;
1048 my $regex = $self->wit_list_separator;
1049 my @answer = split( /\Q$regex\E/, $label );
1055 use Text::Tradition;
1057 my $cxfile = 't/data/Collatex-16.xml';
1058 my $t = Text::Tradition->new(
1060 'input' => 'CollateX',
1063 my $c = $t->collation;
1065 is( $c->common_predecessor( $c->reading('n9'), $c->reading('n23') )->id,
1066 'n20', "Found correct common predecessor" );
1067 is( $c->common_successor( $c->reading('n9'), $c->reading('n23') )->id,
1068 '#END#', "Found correct common successor" );
1070 is( $c->common_predecessor( $c->reading('n19'), $c->reading('n17') )->id,
1071 'n16', "Found correct common predecessor for readings on same path" );
1072 is( $c->common_successor( $c->reading('n21'), $c->reading('n26') )->id,
1073 '#END#', "Found correct common successor for readings on same path" );
1079 ## Return the closest reading that is a predecessor of both the given readings.
1080 sub common_predecessor {
1082 return $self->common_in_path( @_, 'predecessors' );
1085 sub common_successor {
1087 return $self->common_in_path( @_, 'successors' );
1090 sub common_in_path {
1091 my( $self, $r1, $r2, $dir ) = @_;
1092 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1093 $iter = $self->end->rank - $iter if $dir eq 'successors';
1095 my @last_checked = ( $r1, $r2 );
1097 while( !@candidates ) {
1099 foreach my $lc ( @last_checked ) {
1100 foreach my $p ( $lc->$dir ) {
1101 if( $all_seen{$p->id} ) {
1102 push( @candidates, $p );
1104 $all_seen{$p->id} = 1;
1105 push( @new_lc, $p );
1109 @last_checked = @new_lc;
1111 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1112 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1116 __PACKAGE__->meta->make_immutable;
1122 =item * Think about making Relationship objects again