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;
10 use Text::Tradition::Error;
12 use XML::LibXML::XPathContext;
18 default => sub { Graph->new() },
26 isa => 'Text::Tradition::Collation::RelationshipStore',
28 relationships => 'relationships',
29 related_readings => 'related_readings',
30 del_relationship => 'del_relationship',
32 writer => '_set_relations',
37 isa => 'Text::Tradition',
42 isa => 'HashRef[Text::Tradition::Collation::Reading]',
46 _add_reading => 'set',
47 del_reading => 'delete',
48 has_reading => 'exists',
51 default => sub { {} },
54 has 'wit_list_separator' => (
63 default => 'base text',
80 isa => 'Text::Tradition::Collation::Reading',
81 writer => '_set_start',
87 isa => 'Text::Tradition::Collation::Reading',
95 predicate => 'has_cached_svg',
96 clearer => 'wipe_svg',
99 has 'cached_table' => (
102 predicate => 'has_cached_table',
103 clearer => 'wipe_table',
108 Text::Tradition::Collation - a software model for a text collation
113 my $t = Text::Tradition->new(
114 'name' => 'this is a text',
116 'file' => '/path/to/tei_parallel_seg_file.xml' );
118 my $c = $t->collation;
119 my @readings = $c->readings;
120 my @paths = $c->paths;
121 my @relationships = $c->relationships;
123 my $svg_variant_graph = $t->collation->as_svg();
127 Text::Tradition is a library for representation and analysis of collated
128 texts, particularly medieval ones. The Collation is the central feature of
129 a Tradition, where the text, its sequence of readings, and its relationships
130 between readings are actually kept.
136 The constructor. Takes a hash or hashref of the following arguments:
140 =item * tradition - The Text::Tradition object to which the collation
143 =item * linear - Whether the collation should be linear; that is, whether
144 transposed readings should be treated as two linked readings rather than one,
145 and therefore whether the collation graph is acyclic. Defaults to true.
147 =item * baselabel - The default label for the path taken by a base text
148 (if any). Defaults to 'base text'.
150 =item * wit_list_separator - The string to join a list of witnesses for
151 purposes of making labels in display graphs. Defaults to ', '.
153 =item * ac_label - The extra label to tack onto a witness sigil when
154 representing another layer of path for the given witness - that is, when
155 a text has more than one possible reading due to scribal corrections or
156 the like. Defaults to ' (a.c.)'.
166 =head2 wit_list_separator
172 Simple accessors for collation attributes.
176 The meta-reading at the start of every witness path.
180 The meta-reading at the end of every witness path.
184 Returns all Reading objects in the graph.
186 =head2 reading( $id )
188 Returns the Reading object corresponding to the given ID.
190 =head2 add_reading( $reading_args )
192 Adds a new reading object to the collation.
193 See L<Text::Tradition::Collation::Reading> for the available arguments.
195 =head2 del_reading( $object_or_id )
197 Removes the given reading from the collation, implicitly removing its
198 paths and relationships.
200 =head2 merge_readings( $main, $second )
202 Merges the $second reading into the $main one.
203 The arguments may be either readings or reading IDs.
205 =head2 has_reading( $id )
207 Predicate to see whether a given reading ID is in the graph.
209 =head2 reading_witnesses( $object_or_id )
211 Returns a list of sigils whose witnesses contain the reading.
215 Returns all reading paths within the document - that is, all edges in the
216 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
218 =head2 add_path( $source, $target, $sigil )
220 Links the given readings in the collation in sequence, under the given witness
221 sigil. The readings may be specified by object or ID.
223 =head2 del_path( $source, $target, $sigil )
225 Links the given readings in the collation in sequence, under the given witness
226 sigil. The readings may be specified by object or ID.
228 =head2 has_path( $source, $target );
230 Returns true if the two readings are linked in sequence in any witness.
231 The readings may be specified by object or ID.
235 Returns all Relationship objects in the collation.
237 =head2 add_relationship( $reading, $other_reading, $options )
239 Adds a new relationship of the type given in $options between the two readings,
240 which may be specified by object or ID. Returns a value of ( $status, @vectors)
241 where $status is true on success, and @vectors is a list of relationship edges
242 that were ultimately added.
243 See L<Text::Tradition::Collation::Relationship> for the available options.
249 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
250 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
251 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
254 ### Reading construct/destruct functions
258 $self->wipe_svg if $self->has_cached_svg;
259 $self->wipe_table if $self->has_cached_table;
263 my( $self, $reading ) = @_;
264 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
265 my %args = %$reading;
266 $reading = Text::Tradition::Collation::Reading->new(
267 'collation' => $self,
270 # First check to see if a reading with this ID exists.
271 if( $self->reading( $reading->id ) ) {
272 throw( "Collation already has a reading with id " . $reading->id );
275 $self->_add_reading( $reading->id => $reading );
276 # Once the reading has been added, put it in both graphs.
277 $self->sequence->add_vertex( $reading->id );
278 $self->relations->add_reading( $reading->id );
282 around del_reading => sub {
287 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
290 # Remove the reading from the graphs.
292 $self->sequence->delete_vertex( $arg );
293 $self->relations->delete_reading( $arg );
296 $self->$orig( $arg );
299 # merge_readings( $main, $to_be_deleted );
304 # We only need the IDs for adding paths to the graph, not the reading
305 # objects themselves.
306 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
309 # The kept reading should inherit the paths and the relationships
310 # of the deleted reading.
311 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
312 my @vector = ( $kept );
313 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
314 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
315 next if $vector[0] eq $vector[1]; # Don't add a self loop
316 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
317 $self->sequence->add_edge( @vector );
318 my $fwits = $self->sequence->get_edge_attributes( @vector );
319 @wits{keys %$fwits} = values %$fwits;
320 $self->sequence->set_edge_attributes( @vector, \%wits );
322 $self->relations->merge_readings( $kept, $deleted, $combine_char );
324 # Do the deletion deed.
325 if( $combine_char ) {
326 my $kept_obj = $self->reading( $kept );
327 my $new_text = join( $combine_char, $kept_obj->text,
328 $self->reading( $deleted )->text );
329 $kept_obj->alter_text( $new_text );
331 $self->del_reading( $deleted );
335 # Helper function for manipulating the graph.
336 sub _stringify_args {
337 my( $self, $first, $second, $arg ) = @_;
339 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
340 $second = $second->id
341 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
342 return( $first, $second, $arg );
345 # Helper function for manipulating the graph.
346 sub _objectify_args {
347 my( $self, $first, $second, $arg ) = @_;
348 $first = $self->reading( $first )
349 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
350 $second = $self->reading( $second )
351 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
352 return( $first, $second, $arg );
359 # We only need the IDs for adding paths to the graph, not the reading
360 # objects themselves.
361 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
364 # Connect the readings
365 $self->sequence->add_edge( $source, $target );
366 # Note the witness in question
367 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
373 if( ref( $_[0] ) eq 'ARRAY' ) {
380 # We only need the IDs for adding paths to the graph, not the reading
381 # objects themselves.
382 my( $source, $target, $wit ) = $self->_stringify_args( @args );
385 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
386 $self->sequence->delete_edge_attribute( $source, $target, $wit );
388 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
389 $self->sequence->delete_edge( $source, $target );
394 # Extra graph-alike utility
397 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
398 return undef unless $self->sequence->has_edge( $source, $target );
399 return $self->sequence->has_edge_attribute( $source, $target, $wit );
402 =head2 clear_witness( @sigil_list )
404 Clear the given witnesses out of the collation entirely, removing references
405 to them in paths, and removing readings that belong only to them. Should only
406 be called via $tradition->del_witness.
411 my( $self, @sigils ) = @_;
414 # Clear the witness(es) out of the paths
415 foreach my $e ( $self->paths ) {
416 foreach my $sig ( @sigils ) {
417 $self->del_path( $e, $sig );
421 # Clear out the newly unused readings
422 foreach my $r ( $self->readings ) {
423 unless( $self->reading_witnesses( $r ) ) {
424 $self->del_reading( $r );
429 sub add_relationship {
431 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
432 my( @vectors ) = $self->relations->add_relationship( $source,
433 $self->reading( $source ), $target, $self->reading( $target ), $opts );
434 # Force a full rank recalculation every time. Yuck.
435 $self->calculate_ranks() if $self->end->has_rank;
440 =head2 reading_witnesses( $reading )
442 Return a list of sigils corresponding to the witnesses in which the reading appears.
446 sub reading_witnesses {
447 my( $self, $reading ) = @_;
448 # We need only check either the incoming or the outgoing edges; I have
449 # arbitrarily chosen "incoming". Thus, special-case the start node.
450 if( $reading eq $self->start ) {
451 return map { $_->sigil } $self->tradition->witnesses;
454 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
455 my $wits = $self->sequence->get_edge_attributes( @$e );
456 @all_witnesses{ keys %$wits } = 1;
458 my $acstr = $self->ac_label;
459 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
460 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
462 return keys %all_witnesses;
465 =head1 OUTPUT METHODS
467 =head2 as_svg( \%options )
469 Returns an SVG string that represents the graph, via as_dot and graphviz.
470 See as_dot for a list of options.
475 my( $self, $opts ) = @_;
476 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
477 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
478 my @cmd = qw/dot -Tsvg/;
480 my $dotfile = File::Temp->new();
482 # $dotfile->unlink_on_destroy(0);
483 binmode $dotfile, ':utf8';
484 print $dotfile $self->as_dot( $opts );
485 push( @cmd, $dotfile->filename );
486 run( \@cmd, ">", binary(), \$svg );
487 $svg = decode_utf8( $svg );
488 $self->cached_svg( $svg ) unless $want_subgraph;
491 return $self->cached_svg;
496 =head2 as_dot( \%options )
498 Returns a string that is the collation graph expressed in dot
499 (i.e. GraphViz) format. Options include:
514 my( $self, $opts ) = @_;
515 my $startrank = $opts->{'from'} if $opts;
516 my $endrank = $opts->{'to'} if $opts;
517 my $color_common = $opts->{'color_common'} if $opts;
518 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
519 && $self->end->rank > 100;
521 # Check the arguments
523 return if $endrank && $startrank > $endrank;
524 return if $startrank > $self->end->rank;
526 if( defined $endrank ) {
527 return if $endrank < 0;
528 $endrank = undef if $endrank == $self->end->rank;
531 my $graph_name = $self->tradition->name;
532 $graph_name =~ s/[^\w\s]//g;
533 $graph_name = join( '_', split( /\s+/, $graph_name ) );
541 'fillcolor' => 'white',
546 'arrowhead' => 'open',
547 'color' => '#000000',
548 'fontcolor' => '#000000',
551 my $dot = sprintf( "digraph %s {\n", $graph_name );
552 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
553 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
555 # Output substitute start/end readings if necessary
557 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
560 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
562 if( $STRAIGHTENHACK ) {
564 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
565 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
567 my %used; # Keep track of the readings that actually appear in the graph
568 # Sort the readings by rank if we have ranks; this speeds layout.
569 my @all_readings = $self->end->has_rank
570 ? sort { $a->rank <=> $b->rank } $self->readings
572 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
573 foreach my $reading ( @all_readings ) {
574 # Only output readings within our rank range.
575 next if $startrank && $reading->rank < $startrank;
576 next if $endrank && $reading->rank > $endrank;
577 $used{$reading->id} = 1;
578 # Need not output nodes without separate labels
579 next if $reading->id eq $reading->text;
581 my $label = $reading->text;
582 $label =~ s/\"/\\\"/g;
583 $rattrs->{'label'} = $label;
584 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
585 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
588 # Add the real edges. Need to weight one edge per rank jump, in a
590 # my $weighted = $self->_add_edge_weights;
591 my @edges = $self->paths;
592 my( %substart, %subend );
593 foreach my $edge ( @edges ) {
594 # Do we need to output this edge?
595 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
596 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
597 my $variables = { %edge_attrs, 'label' => $label };
599 # Account for the rank gap if necessary
600 my $rank0 = $self->reading( $edge->[0] )->rank
601 if $self->reading( $edge->[0] )->has_rank;
602 my $rank1 = $self->reading( $edge->[1] )->rank
603 if $self->reading( $edge->[1] )->has_rank;
604 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
605 $variables->{'minlen'} = $rank1 - $rank0;
608 # Add the calculated edge weights
609 # if( exists $weighted->{$edge->[0]}
610 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
611 # # $variables->{'color'} = 'red';
612 # $variables->{'weight'} = 3.0;
615 # EXPERIMENTAL: make edge width reflect no. of witnesses
616 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
617 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
619 my $varopts = _dot_attr_string( $variables );
620 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
621 $edge->[0], $edge->[1], $varopts );
622 } elsif( $used{$edge->[0]} ) {
623 $subend{$edge->[0]} = 1;
624 } elsif( $used{$edge->[1]} ) {
625 $substart{$edge->[1]} = 1;
628 # Add substitute start and end edges if necessary
629 foreach my $node ( keys %substart ) {
630 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
631 my $variables = { %edge_attrs, 'label' => $witstr };
632 my $varopts = _dot_attr_string( $variables );
633 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
635 foreach my $node ( keys %subend ) {
636 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
637 my $variables = { %edge_attrs, 'label' => $witstr };
638 my $varopts = _dot_attr_string( $variables );
639 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
642 if( $STRAIGHTENHACK ) {
643 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
650 sub _dot_attr_string {
653 foreach my $k ( sort keys %$hash ) {
655 push( @attrs, $k.'="'.$v.'"' );
657 return( '[ ' . join( ', ', @attrs ) . ' ]' );
660 sub _add_edge_weights {
662 # Walk the graph from START to END, choosing the successor node with
663 # the largest number of witness paths each time.
665 my $curr = $self->start->id;
666 my $ranked = $self->end->has_rank;
667 while( $curr ne $self->end->id ) {
668 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
669 my @succ = sort { $self->path_witnesses( $curr, $a )
670 <=> $self->path_witnesses( $curr, $b ) }
671 $self->sequence->successors( $curr );
672 my $next = pop @succ;
673 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
674 # Try to avoid lacunae in the weighted path.
676 ( $self->reading( $next )->is_lacuna ||
677 $nextrank - $rank > 1 ) ){
680 $weighted->{$curr} = $next;
686 =head2 path_witnesses( $edge )
688 Returns the list of sigils whose witnesses are associated with the given edge.
689 The edge can be passed as either an array or an arrayref of ( $source, $target ).
694 my( $self, @edge ) = @_;
695 # If edge is an arrayref, cope.
696 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
700 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
704 sub _path_display_label {
707 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
708 if( scalar @wits > $maj ) {
709 # TODO break out a.c. wits
712 return join( ', ', @wits );
716 =head2 witnesses_at_rank
718 Returns a list of witnesses that are not lacunose, for a given rank.
722 sub witnesses_at_rank {
723 my( $self, $rank ) = @_;
728 Returns a GraphML representation of the collation. The GraphML will contain
729 two graphs. The first expresses the attributes of the readings and the witness
730 paths that link them; the second expresses the relationships that link the
731 readings. This is the native transfer format for a tradition.
740 my $datafile = 't/data/florilegium_tei_ps.xml';
741 my $tradition = Text::Tradition->new( 'input' => 'TEI',
746 ok( $tradition, "Got a tradition object" );
747 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
748 ok( $tradition->collation, "Tradition has a collation" );
750 my $c = $tradition->collation;
751 is( scalar $c->readings, $READINGS, "Collation has all readings" );
752 is( scalar $c->paths, $PATHS, "Collation has all paths" );
753 is( scalar $c->relationships, 0, "Collation has all relationships" );
755 # Add a few relationships
756 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
757 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
758 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
760 # Now write it to GraphML and parse it again.
762 my $graphml = $c->as_graphml;
763 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
764 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
765 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
766 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
776 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
777 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
778 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
779 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
781 # Create the document and root node
782 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
783 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
784 $graphml->setDocumentElement( $root );
785 $root->setNamespace( $xsi_ns, 'xsi', 0 );
786 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
788 # Add the data keys for the graph
791 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
792 foreach my $datum ( @graph_attributes ) {
793 $graph_data_keys{$datum} = 'dg'.$gdi++;
794 my $key = $root->addNewChild( $graphml_ns, 'key' );
795 $key->setAttribute( 'attr.name', $datum );
796 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
797 $key->setAttribute( 'for', 'graph' );
798 $key->setAttribute( 'id', $graph_data_keys{$datum} );
801 # Add the data keys for nodes
808 is_start => 'boolean',
810 is_lacuna => 'boolean',
812 foreach my $datum ( keys %node_data ) {
813 $node_data_keys{$datum} = 'dn'.$ndi++;
814 my $key = $root->addNewChild( $graphml_ns, 'key' );
815 $key->setAttribute( 'attr.name', $datum );
816 $key->setAttribute( 'attr.type', $node_data{$datum} );
817 $key->setAttribute( 'for', 'node' );
818 $key->setAttribute( 'id', $node_data_keys{$datum} );
821 # Add the data keys for edges, i.e. witnesses
825 class => 'string', # Class, deprecated soon
826 witness => 'string', # ID/label for a path
827 relationship => 'string', # ID/label for a relationship
828 extra => 'boolean', # Path key
829 scope => 'string', # Relationship key
830 non_correctable => 'boolean', # Relationship key
831 non_independent => 'boolean', # Relationship key
833 foreach my $datum ( keys %edge_data ) {
834 $edge_data_keys{$datum} = 'de'.$edi++;
835 my $key = $root->addNewChild( $graphml_ns, 'key' );
836 $key->setAttribute( 'attr.name', $datum );
837 $key->setAttribute( 'attr.type', $edge_data{$datum} );
838 $key->setAttribute( 'for', 'edge' );
839 $key->setAttribute( 'id', $edge_data_keys{$datum} );
842 # Add the collation graph itself
843 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
844 $sgraph->setAttribute( 'edgedefault', 'directed' );
845 $sgraph->setAttribute( 'id', $self->tradition->name );
846 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
847 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
848 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
849 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
850 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
852 # Collation attribute data
853 foreach my $datum ( @graph_attributes ) {
854 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
855 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
860 # Add our readings to the graph
861 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
862 # Add to the main graph
863 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
864 my $node_xmlid = 'n' . $node_ctr++;
865 $node_hash{ $n->id } = $node_xmlid;
866 $node_el->setAttribute( 'id', $node_xmlid );
867 foreach my $d ( keys %node_data ) {
869 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
874 # Add the path edges to the sequence graph
876 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
877 # We add an edge in the graphml for every witness in $e.
878 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
879 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
880 $node_hash{ $e->[0] },
881 $node_hash{ $e->[1] } );
882 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
883 $edge_el->setAttribute( 'source', $from );
884 $edge_el->setAttribute( 'target', $to );
885 $edge_el->setAttribute( 'id', $id );
887 # It's a witness path, so add the witness
889 my $key = $edge_data_keys{'witness'};
890 # Is this an ante-corr witness?
891 my $aclabel = $self->ac_label;
892 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
893 # Keep the base witness
895 # ...and record that this is an 'extra' reading path
896 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
898 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
899 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
903 # Add the relationship graph to the XML
904 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
905 $node_data_keys{'id'}, \%edge_data_keys );
907 # Save and return the thing
908 my $result = decode_utf8( $graphml->toString(1) );
912 sub _add_graphml_data {
913 my( $el, $key, $value ) = @_;
914 return unless defined $value;
915 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
916 $data_el->setAttribute( 'key', $key );
917 $data_el->appendText( $value );
922 Returns a CSV alignment table representation of the collation graph, one
923 row per witness (or witness uncorrected.)
929 my $table = $self->alignment_table;
930 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
932 # Make the header row
933 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
934 push( @result, decode_utf8( $csv->string ) );
935 # Make the rest of the rows
936 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
937 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
938 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
939 $csv->combine( @row );
940 push( @result, decode_utf8( $csv->string ) );
942 return join( "\n", @result );
945 =head2 alignment_table( $use_refs, $include_witnesses )
947 Return a reference to an alignment table, in a slightly enhanced CollateX
948 format which looks like this:
950 $table = { alignment => [ { witness => "SIGIL",
951 tokens => [ { t => "TEXT" }, ... ] },
953 tokens => [ { t => "TEXT" }, ... ] },
957 If $use_refs is set to 1, the reading object is returned in the table
958 instead of READINGTEXT; if not, the text of the reading is returned.
960 If $include_witnesses is set to a hashref, only the witnesses whose sigil
961 keys have a true hash value will be included.
965 sub alignment_table {
967 my $include; # see if we can ditch this
968 return $self->cached_table if $self->has_cached_table;
970 # Make sure we can do this
971 throw( "Need a linear graph in order to make an alignment table" )
972 unless $self->linear;
973 $self->calculate_ranks unless $self->end->has_rank;
975 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
976 my @all_pos = ( 1 .. $self->end->rank - 1 );
977 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
979 next unless $include->{$wit->sigil};
981 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
982 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
983 my @row = _make_witness_row( \@wit_path, \@all_pos );
984 push( @{$table->{'alignment'}},
985 { 'witness' => $wit->sigil, 'tokens' => \@row } );
986 if( $wit->is_layered ) {
987 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
988 $wit->sigil.$self->ac_label );
989 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
990 push( @{$table->{'alignment'}},
991 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
994 $self->cached_table( $table );
998 sub _make_witness_row {
999 my( $path, $positions ) = @_;
1001 map { $char_hash{$_} = undef } @$positions;
1003 foreach my $rdg ( @$path ) {
1004 my $rtext = $rdg->text;
1005 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1006 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1007 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1008 $char_hash{$rdg->rank} = { 't' => $rdg };
1010 my @row = map { $char_hash{$_} } @$positions;
1011 # Fill in lacuna markers for undef spots in the row
1012 my $last_el = shift @row;
1013 my @filled_row = ( $last_el );
1014 foreach my $el ( @row ) {
1015 # If we are using node reference, make the lacuna node appear many times
1016 # in the table. If not, use the lacuna tag.
1017 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1020 push( @filled_row, $el );
1026 =head1 NAVIGATION METHODS
1028 =head2 reading_sequence( $first, $last, $sigil, $backup )
1030 Returns the ordered list of readings, starting with $first and ending
1031 with $last, for the witness given in $sigil. If a $backup sigil is
1032 specified (e.g. when walking a layered witness), it will be used wherever
1033 no $sigil path exists. If there is a base text reading, that will be
1034 used wherever no path exists for $sigil or $backup.
1038 # TODO Think about returning some lazy-eval iterator.
1039 # TODO Get rid of backup; we should know from what witness is whether we need it.
1041 sub reading_sequence {
1042 my( $self, $start, $end, $witness ) = @_;
1044 $witness = $self->baselabel unless $witness;
1045 my @readings = ( $start );
1048 while( $n && $n->id ne $end->id ) {
1049 if( exists( $seen{$n->id} ) ) {
1050 throw( "Detected loop for $witness at " . $n->id );
1054 my $next = $self->next_reading( $n, $witness );
1056 throw( "Did not find any path for $witness from reading " . $n->id );
1058 push( @readings, $next );
1061 # Check that the last reading is our end reading.
1062 my $last = $readings[$#readings];
1063 throw( "Last reading found from " . $start->text .
1064 " for witness $witness is not the end!" ) # TODO do we get this far?
1065 unless $last->id eq $end->id;
1070 =head2 next_reading( $reading, $sigil );
1072 Returns the reading that follows the given reading along the given witness
1078 # Return the successor via the corresponding path.
1080 my $answer = $self->_find_linked_reading( 'next', @_ );
1081 return undef unless $answer;
1082 return $self->reading( $answer );
1085 =head2 prior_reading( $reading, $sigil )
1087 Returns the reading that precedes the given reading along the given witness
1093 # Return the predecessor via the corresponding path.
1095 my $answer = $self->_find_linked_reading( 'prior', @_ );
1096 return $self->reading( $answer );
1099 sub _find_linked_reading {
1100 my( $self, $direction, $node, $path ) = @_;
1102 # Get a backup if we are dealing with a layered witness
1104 my $aclabel = $self->ac_label;
1105 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1109 my @linked_paths = $direction eq 'next'
1110 ? $self->sequence->edges_from( $node )
1111 : $self->sequence->edges_to( $node );
1112 return undef unless scalar( @linked_paths );
1114 # We have to find the linked path that contains all of the
1115 # witnesses supplied in $path.
1116 my( @path_wits, @alt_path_wits );
1117 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1118 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1121 foreach my $le ( @linked_paths ) {
1122 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1125 my @le_wits = sort $self->path_witnesses( $le );
1126 if( _is_within( \@path_wits, \@le_wits ) ) {
1127 # This is the right path.
1128 return $direction eq 'next' ? $le->[1] : $le->[0];
1129 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1133 # Got this far? Return the alternate path if it exists.
1134 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1137 # Got this far? Return the base path if it exists.
1138 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1141 # Got this far? We have no appropriate path.
1142 warn "Could not find $direction node from " . $node->id
1143 . " along path $path";
1149 my( $set1, $set2 ) = @_;
1150 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1151 foreach my $el ( @$set1 ) {
1152 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1157 # Return the string that joins together a list of witnesses for
1158 # display on a single path.
1159 sub _witnesses_of_label {
1160 my( $self, $label ) = @_;
1161 my $regex = $self->wit_list_separator;
1162 my @answer = split( /\Q$regex\E/, $label );
1166 =head2 common_readings
1168 Returns the list of common readings in the graph (i.e. those readings that are
1169 shared by all non-lacunose witnesses.)
1173 sub common_readings {
1175 my @common = grep { $_->is_common } $self->readings;
1179 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1181 Returns the text of a witness (plus its backup, if we are using a layer)
1182 as stored in the collation. The text is returned as a string, where the
1183 individual readings are joined with spaces and the meta-readings (e.g.
1184 lacunae) are omitted. Optional specification of $start and $end allows
1185 the generation of a subset of the witness text.
1190 my( $self, $wit, $start, $end ) = @_;
1191 $start = $self->start unless $start;
1192 $end = $self->end unless $end;
1193 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1194 return join( ' ', map { $_->text } @path );
1197 =head1 INITIALIZATION METHODS
1199 These are mostly for use by parsers.
1201 =head2 make_witness_path( $witness )
1203 Link the array of readings contained in $witness->path (and in
1204 $witness->uncorrected_path if it exists) into collation paths.
1205 Clear out the arrays when finished.
1207 =head2 make_witness_paths
1209 Call make_witness_path for all witnesses in the tradition.
1213 # For use when a collation is constructed from a base text and an apparatus.
1214 # We have the sequences of readings and just need to add path edges.
1215 # When we are done, clear out the witness path attributes, as they are no
1217 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1219 sub make_witness_paths {
1221 foreach my $wit ( $self->tradition->witnesses ) {
1222 # print STDERR "Making path for " . $wit->sigil . "\n";
1223 $self->make_witness_path( $wit );
1227 sub make_witness_path {
1228 my( $self, $wit ) = @_;
1229 my @chain = @{$wit->path};
1230 my $sig = $wit->sigil;
1231 foreach my $idx ( 0 .. $#chain-1 ) {
1232 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1234 if( $wit->is_layered ) {
1235 @chain = @{$wit->uncorrected_path};
1236 foreach my $idx( 0 .. $#chain-1 ) {
1237 my $source = $chain[$idx];
1238 my $target = $chain[$idx+1];
1239 $self->add_path( $source, $target, $sig.$self->ac_label )
1240 unless $self->has_path( $source, $target, $sig );
1244 $wit->clear_uncorrected_path;
1247 =head2 calculate_ranks
1249 Calculate the reading ranks (that is, their aligned positions relative
1250 to each other) for the graph. This can only be called on linear collations.
1254 use Text::Tradition;
1256 my $cxfile = 't/data/Collatex-16.xml';
1257 my $t = Text::Tradition->new(
1259 'input' => 'CollateX',
1262 my $c = $t->collation;
1265 my $svg = $c->as_svg;
1266 is( substr( $svg, 0, 5 ), '<?xml', "Got XML doc for svg" );
1267 ok( $c->has_cached_svg, "SVG was cached" );
1268 is( $c->as_svg, $svg, "Cached SVG returned upon second call" );
1269 $c->calculate_ranks;
1270 is( $c->as_svg, $svg, "Cached SVG retained with no rank change" );
1271 $c->add_relationship( 'n9', 'n23', { 'type' => 'spelling' } );
1272 isnt( $c->as_svg, $svg, "SVG changed after relationship add" );
1278 sub calculate_ranks {
1280 # Save the existing ranks, in case we need to invalidate the cached SVG.
1282 # Walk a version of the graph where every node linked by a relationship
1283 # edge is fundamentally the same node, and do a topological ranking on
1284 # the nodes in this graph.
1285 my $topo_graph = Graph->new();
1289 foreach my $r ( $self->readings ) {
1290 next if exists $rel_containers{$r->id};
1291 my @rels = $r->related_readings( 'colocated' );
1293 # Make a relationship container.
1295 my $rn = 'rel_container_' . $rel_ctr++;
1296 $topo_graph->add_vertex( $rn );
1298 $rel_containers{$_->id} = $rn;
1301 # Add a new node to mirror the old node.
1302 $rel_containers{$r->id} = $r->id;
1303 $topo_graph->add_vertex( $r->id );
1308 foreach my $r ( $self->readings ) {
1309 $existing_ranks{$r} = $r->rank;
1310 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1311 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1312 $rel_containers{$n} );
1313 # $DB::single = 1 unless $tfrom && $tto;
1314 $topo_graph->add_edge( $tfrom, $tto );
1318 # Now do the rankings, starting with the start node.
1319 my $topo_start = $rel_containers{$self->start->id};
1320 my $node_ranks = { $topo_start => 0 };
1321 my @curr_origin = ( $topo_start );
1322 # A little iterative function.
1323 while( @curr_origin ) {
1324 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1326 # Transfer our rankings from the topological graph to the real one.
1327 foreach my $r ( $self->readings ) {
1328 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1329 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1331 # Die. Find the last rank we calculated.
1332 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1333 <=> $node_ranks->{$rel_containers{$b->id}} }
1335 my $last = pop @all_defined;
1336 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1339 # Do we need to invalidate the cached SVG?
1340 if( $self->has_cached_svg ) {
1341 foreach my $r ( $self->readings ) {
1342 next if $existing_ranks{$r} == $r->rank;
1350 my( $graph, $node_ranks, @current_nodes ) = @_;
1351 # Look at each of the children of @current_nodes. If all the child's
1352 # parents have a rank, assign it the highest rank + 1 and add it to
1353 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1354 # parent gets a rank.
1356 foreach my $c ( @current_nodes ) {
1357 warn "Current reading $c has no rank!"
1358 unless exists $node_ranks->{$c};
1359 # print STDERR "Looking at child of node $c, rank "
1360 # . $node_ranks->{$c} . "\n";
1361 foreach my $child ( $graph->successors( $c ) ) {
1362 next if exists $node_ranks->{$child};
1363 my $highest_rank = -1;
1365 foreach my $parent ( $graph->predecessors( $child ) ) {
1366 if( exists $node_ranks->{$parent} ) {
1367 $highest_rank = $node_ranks->{$parent}
1368 if $highest_rank <= $node_ranks->{$parent};
1375 my $c_rank = $highest_rank + 1;
1376 # print STDERR "Assigning rank $c_rank to node $child \n";
1377 $node_ranks->{$child} = $c_rank;
1378 push( @next_nodes, $child );
1384 =head2 flatten_ranks
1386 A convenience method for parsing collation data. Searches the graph for readings
1387 with the same text at the same rank, and merges any that are found.
1393 my %unique_rank_rdg;
1394 foreach my $rdg ( $self->readings ) {
1395 next unless $rdg->has_rank;
1396 my $key = $rdg->rank . "||" . $rdg->text;
1397 if( exists $unique_rank_rdg{$key} ) {
1399 # print STDERR "Combining readings at same rank: $key\n";
1400 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1401 # TODO see if this now makes a common point.
1403 $unique_rank_rdg{$key} = $rdg;
1408 =head2 remove_collations
1410 Another convenience method for parsing. Removes all 'collation' relationships
1411 that were defined in order to get the reading ranks to be correct.
1415 use Text::Tradition;
1417 my $cxfile = 't/data/Collatex-16.xml';
1418 my $t = Text::Tradition->new(
1420 'input' => 'CollateX',
1423 my $c = $t->collation;
1425 isnt( $c->reading('n23')->rank, $c->reading('n9')->rank, "Rank skew exists" );
1426 $c->add_relationship( 'n23', 'n9', { 'type' => 'collated', 'scope' => 'local' } );
1427 is( scalar $c->relationships, 4, "Found all expected relationships" );
1428 $c->remove_collations;
1429 is( scalar $c->relationships, 3, "Collated relationships now gone" );
1430 is( $c->reading('n23')->rank, $c->reading('n9')->rank, "Aligned ranks were preserved" );
1436 sub remove_collations {
1438 foreach my $reledge ( $self->relationships ) {
1439 my $relobj = $self->relations->get_relationship( $reledge );
1440 if( $relobj && $relobj->type eq 'collated' ) {
1441 $self->relations->delete_relationship( $reledge );
1447 =head2 calculate_common_readings
1449 Goes through the graph identifying the readings that appear in every witness
1450 (apart from those with lacunae at that spot.) Marks them as common and returns
1455 use Text::Tradition;
1457 my $cxfile = 't/data/Collatex-16.xml';
1458 my $t = Text::Tradition->new(
1460 'input' => 'CollateX',
1463 my $c = $t->collation;
1465 my @common = $c->calculate_common_readings();
1466 is( scalar @common, 8, "Found correct number of common readings" );
1467 my @marked = sort $c->common_readings();
1468 is( scalar @common, 8, "All common readings got marked as such" );
1469 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1470 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1476 sub calculate_common_readings {
1479 my $table = $self->alignment_table;
1480 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1481 my @row = map { $_->{'tokens'}->[$idx]
1482 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1483 @{$table->{'alignment'}};
1485 foreach my $r ( @row ) {
1487 $hash{$r->id} = $r unless $r->is_meta;
1489 $hash{'UNDEF'} = $r;
1492 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1493 my( $r ) = values %hash;
1495 push( @common, $r );
1501 =head2 text_from_paths
1503 Calculate the text array for all witnesses from the path, for later consistency
1504 checking. Only to be used if there is no non-graph-based way to know the
1509 sub text_from_paths {
1511 foreach my $wit ( $self->tradition->witnesses ) {
1512 my @text = split( /\s+/,
1513 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1514 $wit->text( \@text );
1515 if( $wit->is_layered ) {
1516 my @uctext = split( /\s+/,
1517 $self->reading_sequence( $self->start, $self->end,
1518 $wit->sigil.$self->ac_label ) );
1519 $wit->text( \@uctext );
1524 =head1 UTILITY FUNCTIONS
1526 =head2 common_predecessor( $reading_a, $reading_b )
1528 Find the last reading that occurs in sequence before both the given readings.
1530 =head2 common_successor( $reading_a, $reading_b )
1532 Find the first reading that occurs in sequence after both the given readings.
1536 use Text::Tradition;
1538 my $cxfile = 't/data/Collatex-16.xml';
1539 my $t = Text::Tradition->new(
1541 'input' => 'CollateX',
1544 my $c = $t->collation;
1546 is( $c->common_predecessor( 'n9', 'n23' )->id,
1547 'n20', "Found correct common predecessor" );
1548 is( $c->common_successor( 'n9', 'n23' )->id,
1549 '#END#', "Found correct common successor" );
1551 is( $c->common_predecessor( 'n19', 'n17' )->id,
1552 'n16', "Found correct common predecessor for readings on same path" );
1553 is( $c->common_successor( 'n21', 'n26' )->id,
1554 '#END#', "Found correct common successor for readings on same path" );
1560 ## Return the closest reading that is a predecessor of both the given readings.
1561 sub common_predecessor {
1563 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1564 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1567 sub common_successor {
1569 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1570 return $self->_common_in_path( $r1, $r2, 'successors' );
1573 sub _common_in_path {
1574 my( $self, $r1, $r2, $dir ) = @_;
1575 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1576 $iter = $self->end->rank - $iter if $dir eq 'successors';
1578 my @last_checked = ( $r1, $r2 );
1580 while( !@candidates ) {
1582 foreach my $lc ( @last_checked ) {
1583 foreach my $p ( $lc->$dir ) {
1584 if( $all_seen{$p->id} ) {
1585 push( @candidates, $p );
1587 $all_seen{$p->id} = 1;
1588 push( @new_lc, $p );
1592 @last_checked = @new_lc;
1594 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1595 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1599 Text::Tradition::Error->throw(
1600 'ident' => 'Collation error',
1606 __PACKAGE__->meta->make_immutable;
1610 This package is free software and is provided "as is" without express
1611 or implied warranty. You can redistribute it and/or modify it under
1612 the same terms as Perl itself.
1616 Tara L Andrews E<lt>aurum@cpan.orgE<gt>