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',
69 has 'collapse_punctuation' => (
83 isa => 'Text::Tradition::Collation::Reading',
84 writer => '_set_start',
90 isa => 'Text::Tradition::Collation::Reading',
95 # The collation can be created two ways:
96 # 1. Collate a set of witnesses (with CollateX I guess) and process
97 # the results as in 2.
98 # 2. Read a pre-prepared collation in one of a variety of formats,
99 # and make the graph from that.
101 # The graph itself will (for now) be immutable, and the positions
102 # within the graph will also be immutable. We need to calculate those
103 # positions upon graph construction. The equivalences between graph
104 # nodes will be mutable, entirely determined by the user (or possibly
105 # by some semantic pre-processing provided by the user.) So the
106 # constructor should just make an empty equivalences object. The
107 # constructor will also need to make the witness objects, if we didn't
108 # come through option 1.
112 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
113 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
114 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
117 ### Reading construct/destruct functions
120 my( $self, $reading ) = @_;
121 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
122 my %args = %$reading;
123 $reading = Text::Tradition::Collation::Reading->new(
124 'collation' => $self,
127 # First check to see if a reading with this ID exists.
128 if( $self->reading( $reading->id ) ) {
129 warn "Collation already has a reading with id " . $reading->id;
132 $self->_add_reading( $reading->id => $reading );
133 # Once the reading has been added, put it in both graphs.
134 $self->sequence->add_vertex( $reading->id );
135 $self->relations->add_reading( $reading->id );
139 around del_reading => sub {
144 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
147 # Remove the reading from the graphs.
148 $self->sequence->delete_vertex( $arg );
149 $self->relations->delete_reading( $arg );
152 $self->$orig( $arg );
155 # merge_readings( $main, $to_be_deleted );
160 # We only need the IDs for adding paths to the graph, not the reading
161 # objects themselves.
162 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
164 # The kept reading should inherit the paths and the relationships
165 # of the deleted reading.
166 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
167 my @vector = ( $kept );
168 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
169 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
170 next if $vector[0] eq $vector[1]; # Don't add a self loop
171 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
172 $self->sequence->add_edge( @vector );
173 my $fwits = $self->sequence->get_edge_attributes( @vector );
174 @wits{keys %$fwits} = values %$fwits;
175 $self->sequence->set_edge_attributes( @vector, \%wits );
177 $self->relations->merge_readings( $kept, $deleted, $combine_char );
179 # Do the deletion deed.
180 if( $combine_char ) {
181 my $kept_obj = $self->reading( $kept );
182 my $new_text = join( $combine_char, $kept_obj->text,
183 $self->reading( $deleted )->text );
184 $kept_obj->alter_text( $new_text );
186 $self->del_reading( $deleted );
190 # Helper function for manipulating the graph.
191 sub _stringify_args {
192 my( $self, $first, $second, $arg ) = @_;
194 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
195 $second = $second->id
196 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
197 return( $first, $second, $arg );
205 # We only need the IDs for adding paths to the graph, not the reading
206 # objects themselves.
207 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
209 # Connect the readings
210 $self->sequence->add_edge( $source, $target );
211 # Note the witness in question
212 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
218 if( ref( $_[0] ) eq 'ARRAY' ) {
225 # We only need the IDs for adding paths to the graph, not the reading
226 # objects themselves.
227 my( $source, $target, $wit ) = $self->_stringify_args( @args );
229 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
230 $self->sequence->delete_edge_attribute( $source, $target, $wit );
232 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
233 $self->sequence->delete_edge( $source, $target );
238 # Extra graph-alike utility
241 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
242 return undef unless $self->sequence->has_edge( $source, $target );
243 return $self->sequence->has_edge_attribute( $source, $target, $wit );
246 =head2 add_relationship( $reading1, $reading2, $definition )
248 Adds the specified relationship between the two readings. A relationship
249 is transitive (i.e. undirected); the options for its definition may be found
250 in Text::Tradition::Collation::Relationship.
254 # Wouldn't it be lovely if edges could be objects, and all this type checking
255 # and attribute management could be done via Moose?
257 sub add_relationship {
259 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
260 my( $ret, @vectors ) = $self->relations->add_relationship( $source,
261 $self->reading( $source ), $target, $self->reading( $target ), $opts );
262 # Force a full rank recalculation every time. Yuck.
263 $self->calculate_ranks() if $ret && $self->end->has_rank;
264 return( $ret, @vectors );
267 =head2 reading_witnesses( $reading )
269 Return a list of sigils corresponding to the witnesses in which the reading appears.
273 sub reading_witnesses {
274 my( $self, $reading ) = @_;
275 # We need only check either the incoming or the outgoing edges; I have
276 # arbitrarily chosen "incoming". Thus, special-case the start node.
277 if( $reading eq $self->start ) {
278 return map { $_->sigil } $self->tradition->witnesses;
281 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
282 my $wits = $self->sequence->get_edge_attributes( @$e );
283 @all_witnesses{ keys %$wits } = 1;
285 return keys %all_witnesses;
288 =head2 Output method(s)
294 print $collation->as_svg();
296 Returns an SVG string that represents the graph, via as_dot and graphviz.
303 my @cmd = qw/dot -Tsvg/;
305 my $dotfile = File::Temp->new();
307 # $dotfile->unlink_on_destroy(0);
308 binmode $dotfile, ':utf8';
309 print $dotfile $self->as_dot();
310 push( @cmd, $dotfile->filename );
311 run( \@cmd, ">", binary(), \$svg );
312 $svg = decode_utf8( $svg );
316 =item B<svg_subgraph>
318 print $collation->svg_subgraph( $from, $to )
320 Returns an SVG string that represents the portion of the graph given by the
321 specified range. The $from and $to variables refer to ranks within the graph.
326 my( $self, $from, $to ) = @_;
328 my $dot = $self->as_dot( $from, $to );
330 warn "Could not output a graph with range $from - $to";
334 my @cmd = qw/dot -Tsvg/;
336 my $dotfile = File::Temp->new();
338 # $dotfile->unlink_on_destroy(0);
339 binmode $dotfile, ':utf8';
341 push( @cmd, $dotfile->filename );
342 run( \@cmd, ">", binary(), \$svg );
343 $svg = decode_utf8( $svg );
350 print $collation->as_dot();
352 Returns a string that is the collation graph expressed in dot
353 (i.e. GraphViz) format. The 'view' argument determines what kind of
355 * 'path': a graph of witness paths through the collation (DEFAULT)
356 * 'relationship': a graph of how collation readings relate to
362 my( $self, $startrank, $endrank ) = @_;
364 # Check the arguments
366 return if $endrank && $startrank > $endrank;
367 return if $startrank > $self->end->rank;
369 if( defined $endrank ) {
370 return if $endrank < 0;
371 $endrank = undef if $endrank == $self->end->rank;
374 # TODO consider making some of these things configurable
375 my $graph_name = $self->tradition->name;
376 $graph_name =~ s/[^\w\s]//g;
377 $graph_name = join( '_', split( /\s+/, $graph_name ) );
378 my $dot = sprintf( "digraph %s {\n", $graph_name );
379 $dot .= "\tedge [ arrowhead=open ];\n";
380 $dot .= "\tgraph [ rankdir=LR,bgcolor=none ];\n";
381 $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
382 11, "white", "filled", "ellipse" );
384 # Output substitute start/end readings if necessary
386 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
389 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
391 my %used; # Keep track of the readings that actually appear in the graph
394 foreach my $reading ( $self->readings ) {
395 # Only output readings within our rank range.
396 next if $startrank && $reading->rank < $startrank;
397 next if $endrank && $reading->rank > $endrank;
398 $used{$reading->id} = 1;
399 $subedges{$reading->id} = '#SUBSTART#'
400 if $startrank && $startrank == $reading->rank;
401 $subedges{$reading->id} = '#SUBEND#'
402 if $endrank && $endrank == $reading->rank;
403 # Need not output nodes without separate labels
404 next if $reading->id eq $reading->text;
405 my $label = $reading->punctuated_form;
406 $label =~ s/\"/\\\"/g;
407 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->id, $label );
410 # Add substitute start and end edges if necessary
411 foreach my $node ( keys %subedges ) {
412 my @vector = ( $subedges{$node}, $node );
413 @vector = reverse( @vector ) if $vector[0] =~ /END/;
414 my $witstr = join( ', ', sort $self->reading_witnesses( $self->reading( $node ) ) );
415 my %variables = ( 'color' => '#000000',
416 'fontcolor' => '#000000',
419 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
420 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n", @vector, $varopts );
424 my @edges = $self->paths;
425 foreach my $edge ( @edges ) {
426 # Do we need to output this edge?
427 if( $used{$edge->[0]} && $used{$edge->[1]} ) {;
428 my %variables = ( 'color' => '#000000',
429 'fontcolor' => '#000000',
430 'label' => join( ', ', $self->path_display_label( $edge ) ),
432 my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
433 # Account for the rank gap if necessary
434 my $rankgap = $self->reading( $edge->[1] )->rank
435 - $self->reading( $edge->[0] )->rank;
436 $varopts .= ", minlen=$rankgap" if $rankgap > 1;
437 $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
438 $edge->[0], $edge->[1], $varopts );
447 my( $self, @edge ) = @_;
448 # If edge is an arrayref, cope.
449 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
453 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
457 sub path_display_label {
458 my( $self, $edge ) = @_;
459 my @wits = $self->path_witnesses( $edge );
460 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
461 if( scalar @wits > $maj ) {
464 return join( ', ', @wits );
471 print $collation->as_graphml( $recalculate )
473 Returns a GraphML representation of the collation graph, with
474 transposition information and position information. Unless
475 $recalculate is passed (and is a true value), the method will return a
476 cached copy of the SVG after the first call to the method.
484 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
485 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
486 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
487 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
489 # Create the document and root node
490 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
491 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
492 $graphml->setDocumentElement( $root );
493 $root->setNamespace( $xsi_ns, 'xsi', 0 );
494 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
496 # Add the data keys for the graph
499 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
500 foreach my $datum ( @graph_attributes ) {
501 $graph_data_keys{$datum} = 'dg'.$gdi++;
502 my $key = $root->addNewChild( $graphml_ns, 'key' );
503 $key->setAttribute( 'attr.name', $datum );
504 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
505 $key->setAttribute( 'for', 'graph' );
506 $key->setAttribute( 'id', $graph_data_keys{$datum} );
509 # Add the data keys for nodes
516 is_start => 'boolean',
518 is_lacuna => 'boolean',
520 foreach my $datum ( keys %node_data ) {
521 $node_data_keys{$datum} = 'dn'.$ndi++;
522 my $key = $root->addNewChild( $graphml_ns, 'key' );
523 $key->setAttribute( 'attr.name', $datum );
524 $key->setAttribute( 'attr.type', $node_data{$datum} );
525 $key->setAttribute( 'for', 'node' );
526 $key->setAttribute( 'id', $node_data_keys{$datum} );
529 # Add the data keys for edges, i.e. witnesses
533 class => 'string', # Class, deprecated soon
534 witness => 'string', # ID/label for a path
535 relationship => 'string', # ID/label for a relationship
536 extra => 'boolean', # Path key
537 scope => 'string', # Relationship key
538 non_correctable => 'boolean', # Relationship key
539 non_independent => 'boolean', # Relationship key
541 foreach my $datum ( keys %edge_data ) {
542 $edge_data_keys{$datum} = 'de'.$edi++;
543 my $key = $root->addNewChild( $graphml_ns, 'key' );
544 $key->setAttribute( 'attr.name', $datum );
545 $key->setAttribute( 'attr.type', $edge_data{$datum} );
546 $key->setAttribute( 'for', 'edge' );
547 $key->setAttribute( 'id', $edge_data_keys{$datum} );
550 # Add the collation graph itself
551 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
552 $sgraph->setAttribute( 'edgedefault', 'directed' );
553 $sgraph->setAttribute( 'id', $self->tradition->name );
554 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
555 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
556 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
557 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
558 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
560 # Collation attribute data
561 foreach my $datum ( @graph_attributes ) {
562 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
563 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
568 # Add our readings to the graph
569 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
570 # Add to the main graph
571 my $node_el = $sgraph->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 $nval = $n->punctuated_form if $d eq 'text';
578 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
583 # Add the path edges to the sequence graph
585 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
586 # We add an edge in the graphml for every witness in $e.
587 foreach my $wit ( $self->path_witnesses( $e ) ) {
588 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
589 $node_hash{ $e->[0] },
590 $node_hash{ $e->[1] } );
591 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
592 $edge_el->setAttribute( 'source', $from );
593 $edge_el->setAttribute( 'target', $to );
594 $edge_el->setAttribute( 'id', $id );
596 # It's a witness path, so add the witness
598 my $key = $edge_data_keys{'witness'};
599 # Is this an ante-corr witness?
600 my $aclabel = $self->ac_label;
601 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
602 # Keep the base witness
604 # ...and record that this is an 'extra' reading path
605 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
607 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
608 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
612 # Add the relationship graph to the XML
613 $self->relations->as_graphml( $graphml_ns, $root, \%node_hash,
614 $node_data_keys{'id'}, \%edge_data_keys );
616 # Save and return the thing
617 my $result = decode_utf8( $graphml->toString(1) );
621 sub _add_graphml_data {
622 my( $el, $key, $value ) = @_;
623 return unless defined $value;
624 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
625 $data_el->setAttribute( 'key', $key );
626 $data_el->appendText( $value );
631 print $collation->as_csv( $recalculate )
633 Returns a CSV alignment table representation of the collation graph, one
634 row per witness (or witness uncorrected.)
640 my $table = $self->make_alignment_table;
641 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
643 # Make the header row
644 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
645 push( @result, decode_utf8( $csv->string ) );
646 # Make the rest of the rows
647 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
648 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
649 my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs;
650 $csv->combine( @row );
651 push( @result, decode_utf8( $csv->string ) );
653 return join( "\n", @result );
656 =item B<make_alignment_table>
658 my $table = $collation->make_alignment_table( $use_refs, \@wits_to_include )
660 Return a reference to an alignment table, in a slightly enhanced CollateX
661 format which looks like this:
663 $table = { alignment => [ { witness => "SIGIL",
664 tokens => [ { t => "READINGTEXT" }, ... ] },
666 tokens => [ { t => "READINGTEXT" }, ... ] },
670 If $use_refs is set to 1, the reading object is returned in the table
671 instead of READINGTEXT; if not, the text of the reading is returned.
672 If $wits_to_include is set to a hashref, only the witnesses whose sigil
673 keys have a true hash value will be included.
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";
683 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
684 my @all_pos = ( 1 .. $self->end->rank - 1 );
685 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
687 next unless $include->{$wit->sigil};
689 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
690 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
691 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
692 push( @{$table->{'alignment'}},
693 { 'witness' => $wit->sigil, 'tokens' => \@row } );
694 if( $wit->is_layered ) {
695 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
696 $wit->sigil.$self->ac_label, $wit->sigil );
697 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
698 push( @{$table->{'alignment'}},
699 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
705 sub _make_witness_row {
706 my( $path, $positions, $noderefs ) = @_;
708 map { $char_hash{$_} = undef } @$positions;
710 foreach my $rdg ( @$path ) {
711 my $rtext = $rdg->text;
712 $rtext = '#LACUNA#' if $rdg->is_lacuna;
713 print STDERR "rank " . $rdg->rank . "\n" if $debug;
714 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
715 $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg }
718 my @row = map { $char_hash{$_} } @$positions;
719 # Fill in lacuna markers for undef spots in the row
720 my $last_el = shift @row;
721 my @filled_row = ( $last_el );
722 foreach my $el ( @row ) {
723 # If we are using node reference, make the lacuna node appear many times
724 # in the table. If not, use the lacuna tag.
725 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
726 $el = $noderefs ? $last_el : { 't' => '#LACUNA#' };
728 push( @filled_row, $el );
734 # Tiny utility function to say if a table element is a lacuna
737 return 1 if $el->{'t'} eq '#LACUNA#';
738 return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading'
739 && $el->{'t'}->is_lacuna;
743 # Helper to turn the witnesses along columns rather than rows. Assumes
748 return $result unless scalar @$table;
749 my $nrows = scalar @{$table->[0]};
750 foreach my $idx ( 0 .. $nrows - 1 ) {
751 foreach my $wit ( 0 .. $#{$table} ) {
752 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
760 =head2 Navigation methods
766 my $beginning = $collation->start();
768 Returns the beginning of the collation, a meta-reading with label '#START#'.
772 my $end = $collation->end();
774 Returns the end of the collation, a meta-reading with label '#END#'.
777 =item B<reading_sequence>
779 my @readings = $collation->reading_sequence( $first, $last, $path[, $alt_path] );
781 Returns the ordered list of readings, starting with $first and ending
782 with $last, along the given witness path. If no path is specified,
783 assume that the path is that of the base text (if any.)
787 # TODO Think about returning some lazy-eval iterator.
789 sub reading_sequence {
790 my( $self, $start, $end, $witness, $backup ) = @_;
792 $witness = $self->baselabel unless $witness;
793 my @readings = ( $start );
796 while( $n && $n->id ne $end->id ) {
797 if( exists( $seen{$n->id} ) ) {
798 warn "Detected loop at " . $n->id;
803 my $next = $self->next_reading( $n, $witness, $backup );
805 warn "Did not find any path for $witness from reading " . $n->id;
808 push( @readings, $next );
811 # Check that the last reading is our end reading.
812 my $last = $readings[$#readings];
813 warn "Last reading found from " . $start->text .
814 " for witness $witness is not the end!"
815 unless $last->id eq $end->id;
820 =item B<next_reading>
822 my $next_reading = $collation->next_reading( $reading, $witpath );
824 Returns the reading that follows the given reading along the given witness
830 # Return the successor via the corresponding path.
832 my $answer = $self->_find_linked_reading( 'next', @_ );
833 return undef unless $answer;
834 return $self->reading( $answer );
837 =item B<prior_reading>
839 my $prior_reading = $collation->prior_reading( $reading, $witpath );
841 Returns the reading that precedes the given reading along the given witness
847 # Return the predecessor via the corresponding path.
849 my $answer = $self->_find_linked_reading( 'prior', @_ );
850 return $self->reading( $answer );
853 sub _find_linked_reading {
854 my( $self, $direction, $node, $path, $alt_path ) = @_;
855 my @linked_paths = $direction eq 'next'
856 ? $self->sequence->edges_from( $node )
857 : $self->sequence->edges_to( $node );
858 return undef unless scalar( @linked_paths );
860 # We have to find the linked path that contains all of the
861 # witnesses supplied in $path.
862 my( @path_wits, @alt_path_wits );
863 @path_wits = sort( $self->witnesses_of_label( $path ) ) if $path;
864 @alt_path_wits = sort( $self->witnesses_of_label( $alt_path ) ) if $alt_path;
867 foreach my $le ( @linked_paths ) {
868 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
871 my @le_wits = $self->path_witnesses( $le );
872 if( _is_within( \@path_wits, \@le_wits ) ) {
873 # This is the right path.
874 return $direction eq 'next' ? $le->[1] : $le->[0];
875 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
879 # Got this far? Return the alternate path if it exists.
880 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
883 # Got this far? Return the base path if it exists.
884 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
887 # Got this far? We have no appropriate path.
888 warn "Could not find $direction node from " . $node->id
889 . " along path $path";
895 my( $set1, $set2 ) = @_;
896 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
897 foreach my $el ( @$set1 ) {
898 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
904 ## INITIALIZATION METHODS - for use by parsers
906 # For use when a collation is constructed from a base text and an apparatus.
907 # We have the sequences of readings and just need to add path edges.
908 # When we are done, clear out the witness path attributes, as they are no
910 # TODO Find a way to replace the witness path attributes with encapsulated functions?
912 sub make_witness_paths {
914 foreach my $wit ( $self->tradition->witnesses ) {
915 # print STDERR "Making path for " . $wit->sigil . "\n";
916 $self->make_witness_path( $wit );
920 sub make_witness_path {
921 my( $self, $wit ) = @_;
922 my @chain = @{$wit->path};
923 my $sig = $wit->sigil;
924 foreach my $idx ( 0 .. $#chain-1 ) {
925 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
927 if( $wit->is_layered ) {
928 @chain = @{$wit->uncorrected_path};
929 foreach my $idx( 0 .. $#chain-1 ) {
930 my $source = $chain[$idx];
931 my $target = $chain[$idx+1];
932 $self->add_path( $source, $target, $sig.$self->ac_label )
933 unless $self->has_path( $source, $target, $sig );
937 $wit->clear_uncorrected_path;
940 sub calculate_ranks {
942 # Walk a version of the graph where every node linked by a relationship
943 # edge is fundamentally the same node, and do a topological ranking on
944 # the nodes in this graph.
945 my $topo_graph = Graph->new();
949 foreach my $r ( $self->readings ) {
950 next if exists $rel_containers{$r->id};
951 my @rels = $r->related_readings( 'colocated' );
953 # Make a relationship container.
955 my $rn = 'rel_container_' . $rel_ctr++;
956 $topo_graph->add_vertex( $rn );
958 $rel_containers{$_->id} = $rn;
961 # Add a new node to mirror the old node.
962 $rel_containers{$r->id} = $r->id;
963 $topo_graph->add_vertex( $r->id );
968 foreach my $r ( $self->readings ) {
969 foreach my $n ( $self->sequence->successors( $r->id ) ) {
970 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
971 $rel_containers{$n} );
972 $DB::single = 1 unless $tfrom && $tto;
973 $topo_graph->add_edge( $tfrom, $tto );
977 # Now do the rankings, starting with the start node.
978 my $topo_start = $rel_containers{$self->start->id};
979 my $node_ranks = { $topo_start => 0 };
980 my @curr_origin = ( $topo_start );
981 # A little iterative function.
982 while( @curr_origin ) {
983 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
985 # Transfer our rankings from the topological graph to the real one.
986 foreach my $r ( $self->readings ) {
987 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
988 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
991 die "No rank calculated for node " . $r->id
992 . " - do you have a cycle in the graph?";
998 my( $graph, $node_ranks, @current_nodes ) = @_;
999 # Look at each of the children of @current_nodes. If all the child's
1000 # parents have a rank, assign it the highest rank + 1 and add it to
1001 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1002 # parent gets a rank.
1004 foreach my $c ( @current_nodes ) {
1005 warn "Current reading $c has no rank!"
1006 unless exists $node_ranks->{$c};
1007 # print STDERR "Looking at child of node $c, rank "
1008 # . $node_ranks->{$c} . "\n";
1009 foreach my $child ( $graph->successors( $c ) ) {
1010 next if exists $node_ranks->{$child};
1011 my $highest_rank = -1;
1013 foreach my $parent ( $graph->predecessors( $child ) ) {
1014 if( exists $node_ranks->{$parent} ) {
1015 $highest_rank = $node_ranks->{$parent}
1016 if $highest_rank <= $node_ranks->{$parent};
1023 my $c_rank = $highest_rank + 1;
1024 # print STDERR "Assigning rank $c_rank to node $child \n";
1025 $node_ranks->{$child} = $c_rank;
1026 push( @next_nodes, $child );
1032 # Another method to make up for rough collation methods. If the same reading
1033 # appears multiple times at the same rank, collapse the nodes.
1036 my %unique_rank_rdg;
1037 foreach my $rdg ( $self->readings ) {
1038 next unless $rdg->has_rank;
1039 my $key = $rdg->rank . "||" . $rdg->text;
1040 if( exists $unique_rank_rdg{$key} ) {
1042 # print STDERR "Combining readings at same rank: $key\n";
1043 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1045 $unique_rank_rdg{$key} = $rdg;
1051 ## Utility functions
1053 # Return the string that joins together a list of witnesses for
1054 # display on a single path.
1055 sub witnesses_of_label {
1056 my( $self, $label ) = @_;
1057 my $regex = $self->wit_list_separator;
1058 my @answer = split( /\Q$regex\E/, $label );
1064 use Text::Tradition;
1066 my $cxfile = 't/data/Collatex-16.xml';
1067 my $t = Text::Tradition->new(
1069 'input' => 'CollateX',
1072 my $c = $t->collation;
1074 is( $c->common_predecessor( $c->reading('n9'), $c->reading('n23') )->id,
1075 'n20', "Found correct common predecessor" );
1076 is( $c->common_successor( $c->reading('n9'), $c->reading('n23') )->id,
1077 '#END#', "Found correct common successor" );
1079 is( $c->common_predecessor( $c->reading('n19'), $c->reading('n17') )->id,
1080 'n16', "Found correct common predecessor for readings on same path" );
1081 is( $c->common_successor( $c->reading('n21'), $c->reading('n26') )->id,
1082 '#END#', "Found correct common successor for readings on same path" );
1088 ## Return the closest reading that is a predecessor of both the given readings.
1089 sub common_predecessor {
1091 return $self->common_in_path( @_, 'predecessors' );
1094 sub common_successor {
1096 return $self->common_in_path( @_, 'successors' );
1099 sub common_in_path {
1100 my( $self, $r1, $r2, $dir ) = @_;
1101 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1102 $iter = $self->end->rank - $iter if $dir eq 'successors';
1104 my @last_checked = ( $r1, $r2 );
1106 while( !@candidates ) {
1108 foreach my $lc ( @last_checked ) {
1109 foreach my $p ( $lc->$dir ) {
1110 if( $all_seen{$p->id} ) {
1111 push( @candidates, $p );
1113 $all_seen{$p->id} = 1;
1114 push( @new_lc, $p );
1118 @last_checked = @new_lc;
1120 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1121 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1125 __PACKAGE__->meta->make_immutable;
1131 =item * Think about making Relationship objects again