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',
31 writer => '_set_relations',
36 isa => 'Text::Tradition',
41 isa => 'HashRef[Text::Tradition::Collation::Reading]',
45 _add_reading => 'set',
46 del_reading => 'delete',
47 has_reading => 'exists',
50 default => sub { {} },
53 has 'wit_list_separator' => (
62 default => 'base text',
79 isa => 'Text::Tradition::Collation::Reading',
80 writer => '_set_start',
86 isa => 'Text::Tradition::Collation::Reading',
94 predicate => 'has_cached_svg',
95 clearer => 'wipe_svg',
98 has 'cached_table' => (
101 predicate => 'has_cached_table',
102 clearer => 'wipe_table',
107 Text::Tradition::Collation - a software model for a text collation
112 my $t = Text::Tradition->new(
113 'name' => 'this is a text',
115 'file' => '/path/to/tei_parallel_seg_file.xml' );
117 my $c = $t->collation;
118 my @readings = $c->readings;
119 my @paths = $c->paths;
120 my @relationships = $c->relationships;
122 my $svg_variant_graph = $t->collation->as_svg();
126 Text::Tradition is a library for representation and analysis of collated
127 texts, particularly medieval ones. The Collation is the central feature of
128 a Tradition, where the text, its sequence of readings, and its relationships
129 between readings are actually kept.
135 The constructor. Takes a hash or hashref of the following arguments:
139 =item * tradition - The Text::Tradition object to which the collation
142 =item * linear - Whether the collation should be linear; that is, whether
143 transposed readings should be treated as two linked readings rather than one,
144 and therefore whether the collation graph is acyclic. Defaults to true.
146 =item * baselabel - The default label for the path taken by a base text
147 (if any). Defaults to 'base text'.
149 =item * wit_list_separator - The string to join a list of witnesses for
150 purposes of making labels in display graphs. Defaults to ', '.
152 =item * ac_label - The extra label to tack onto a witness sigil when
153 representing another layer of path for the given witness - that is, when
154 a text has more than one possible reading due to scribal corrections or
155 the like. Defaults to ' (a.c.)'.
165 =head2 wit_list_separator
171 Simple accessors for collation attributes.
175 The meta-reading at the start of every witness path.
179 The meta-reading at the end of every witness path.
183 Returns all Reading objects in the graph.
185 =head2 reading( $id )
187 Returns the Reading object corresponding to the given ID.
189 =head2 add_reading( $reading_args )
191 Adds a new reading object to the collation.
192 See L<Text::Tradition::Collation::Reading> for the available arguments.
194 =head2 del_reading( $object_or_id )
196 Removes the given reading from the collation, implicitly removing its
197 paths and relationships.
199 =head2 merge_readings( $main, $second )
201 Merges the $second reading into the $main one.
202 The arguments may be either readings or reading IDs.
204 =head2 has_reading( $id )
206 Predicate to see whether a given reading ID is in the graph.
208 =head2 reading_witnesses( $object_or_id )
210 Returns a list of sigils whose witnesses contain the reading.
214 Returns all reading paths within the document - that is, all edges in the
215 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
217 =head2 add_path( $source, $target, $sigil )
219 Links the given readings in the collation in sequence, under the given witness
220 sigil. The readings may be specified by object or ID.
222 =head2 del_path( $source, $target, $sigil )
224 Links the given readings in the collation in sequence, under the given witness
225 sigil. The readings may be specified by object or ID.
227 =head2 has_path( $source, $target );
229 Returns true if the two readings are linked in sequence in any witness.
230 The readings may be specified by object or ID.
234 Returns all Relationship objects in the collation.
236 =head2 add_relationship( $reading, $other_reading, $options )
238 Adds a new relationship of the type given in $options between the two readings,
239 which may be specified by object or ID. Returns a value of ( $status, @vectors)
240 where $status is true on success, and @vectors is a list of relationship edges
241 that were ultimately added.
242 See L<Text::Tradition::Collation::Relationship> for the available options.
248 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
249 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
250 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
253 ### Reading construct/destruct functions
257 $self->wipe_svg if $self->has_cached_svg;
258 $self->wipe_table if $self->has_cached_table;
262 my( $self, $reading ) = @_;
263 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
264 my %args = %$reading;
265 $reading = Text::Tradition::Collation::Reading->new(
266 'collation' => $self,
269 # First check to see if a reading with this ID exists.
270 if( $self->reading( $reading->id ) ) {
271 throw( "Collation already has a reading with id " . $reading->id );
274 $self->_add_reading( $reading->id => $reading );
275 # Once the reading has been added, put it in both graphs.
276 $self->sequence->add_vertex( $reading->id );
277 $self->relations->add_reading( $reading->id );
281 around del_reading => sub {
286 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
289 # Remove the reading from the graphs.
291 $self->sequence->delete_vertex( $arg );
292 $self->relations->delete_reading( $arg );
295 $self->$orig( $arg );
298 # merge_readings( $main, $to_be_deleted );
303 # We only need the IDs for adding paths to the graph, not the reading
304 # objects themselves.
305 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
308 # The kept reading should inherit the paths and the relationships
309 # of the deleted reading.
310 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
311 my @vector = ( $kept );
312 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
313 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
314 next if $vector[0] eq $vector[1]; # Don't add a self loop
315 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
316 $self->sequence->add_edge( @vector );
317 my $fwits = $self->sequence->get_edge_attributes( @vector );
318 @wits{keys %$fwits} = values %$fwits;
319 $self->sequence->set_edge_attributes( @vector, \%wits );
321 $self->relations->merge_readings( $kept, $deleted, $combine_char );
323 # Do the deletion deed.
324 if( $combine_char ) {
325 my $kept_obj = $self->reading( $kept );
326 my $new_text = join( $combine_char, $kept_obj->text,
327 $self->reading( $deleted )->text );
328 $kept_obj->alter_text( $new_text );
330 $self->del_reading( $deleted );
334 # Helper function for manipulating the graph.
335 sub _stringify_args {
336 my( $self, $first, $second, $arg ) = @_;
338 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
339 $second = $second->id
340 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
341 return( $first, $second, $arg );
344 # Helper function for manipulating the graph.
345 sub _objectify_args {
346 my( $self, $first, $second, $arg ) = @_;
347 $first = $self->reading( $first )
348 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
349 $second = $self->reading( $second )
350 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
351 return( $first, $second, $arg );
358 # We only need the IDs for adding paths to the graph, not the reading
359 # objects themselves.
360 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
363 # Connect the readings
364 $self->sequence->add_edge( $source, $target );
365 # Note the witness in question
366 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
372 if( ref( $_[0] ) eq 'ARRAY' ) {
379 # We only need the IDs for adding paths to the graph, not the reading
380 # objects themselves.
381 my( $source, $target, $wit ) = $self->_stringify_args( @args );
384 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
385 $self->sequence->delete_edge_attribute( $source, $target, $wit );
387 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
388 $self->sequence->delete_edge( $source, $target );
393 # Extra graph-alike utility
396 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
397 return undef unless $self->sequence->has_edge( $source, $target );
398 return $self->sequence->has_edge_attribute( $source, $target, $wit );
401 =head2 clear_witness( @sigil_list )
403 Clear the given witnesses out of the collation entirely, removing references
404 to them in paths, and removing readings that belong only to them. Should only
405 be called via $tradition->del_witness.
410 my( $self, @sigils ) = @_;
413 # Clear the witness(es) out of the paths
414 foreach my $e ( $self->paths ) {
415 foreach my $sig ( @sigils ) {
416 $self->del_path( $e, $sig );
420 # Clear out the newly unused readings
421 foreach my $r ( $self->readings ) {
422 unless( $self->reading_witnesses( $r ) ) {
423 $self->del_reading( $r );
428 sub add_relationship {
430 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
431 my( @vectors ) = $self->relations->add_relationship( $source,
432 $self->reading( $source ), $target, $self->reading( $target ), $opts );
433 # Force a full rank recalculation every time. Yuck.
434 $self->calculate_ranks() if $self->end->has_rank;
439 =head2 reading_witnesses( $reading )
441 Return a list of sigils corresponding to the witnesses in which the reading appears.
445 sub reading_witnesses {
446 my( $self, $reading ) = @_;
447 # We need only check either the incoming or the outgoing edges; I have
448 # arbitrarily chosen "incoming". Thus, special-case the start node.
449 if( $reading eq $self->start ) {
450 return map { $_->sigil } $self->tradition->witnesses;
453 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
454 my $wits = $self->sequence->get_edge_attributes( @$e );
455 @all_witnesses{ keys %$wits } = 1;
457 return keys %all_witnesses;
460 =head1 OUTPUT METHODS
462 =head2 as_svg( \%options )
464 Returns an SVG string that represents the graph, via as_dot and graphviz.
465 See as_dot for a list of options.
470 my( $self, $opts ) = @_;
471 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
472 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
473 my @cmd = qw/dot -Tsvg/;
475 my $dotfile = File::Temp->new();
477 # $dotfile->unlink_on_destroy(0);
478 binmode $dotfile, ':utf8';
479 print $dotfile $self->as_dot( $opts );
480 push( @cmd, $dotfile->filename );
481 run( \@cmd, ">", binary(), \$svg );
482 $svg = decode_utf8( $svg );
483 $self->cached_svg( $svg ) unless $want_subgraph;
486 return $self->cached_svg;
491 =head2 as_dot( \%options )
493 Returns a string that is the collation graph expressed in dot
494 (i.e. GraphViz) format. Options include:
509 my( $self, $opts ) = @_;
510 my $startrank = $opts->{'from'} if $opts;
511 my $endrank = $opts->{'to'} if $opts;
512 my $color_common = $opts->{'color_common'} if $opts;
513 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
514 && $self->end->rank > 100;
516 # Check the arguments
518 return if $endrank && $startrank > $endrank;
519 return if $startrank > $self->end->rank;
521 if( defined $endrank ) {
522 return if $endrank < 0;
523 $endrank = undef if $endrank == $self->end->rank;
526 my $graph_name = $self->tradition->name;
527 $graph_name =~ s/[^\w\s]//g;
528 $graph_name = join( '_', split( /\s+/, $graph_name ) );
536 'fillcolor' => 'white',
541 'arrowhead' => 'open',
542 'color' => '#000000',
543 'fontcolor' => '#000000',
546 my $dot = sprintf( "digraph %s {\n", $graph_name );
547 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
548 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
550 # Output substitute start/end readings if necessary
552 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
555 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
557 if( $STRAIGHTENHACK ) {
559 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
560 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
562 my %used; # Keep track of the readings that actually appear in the graph
563 # Sort the readings by rank if we have ranks; this speeds layout.
564 my @all_readings = $self->end->has_rank
565 ? sort { $a->rank <=> $b->rank } $self->readings
567 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
568 foreach my $reading ( @all_readings ) {
569 # Only output readings within our rank range.
570 next if $startrank && $reading->rank < $startrank;
571 next if $endrank && $reading->rank > $endrank;
572 $used{$reading->id} = 1;
573 # Need not output nodes without separate labels
574 next if $reading->id eq $reading->text;
576 my $label = $reading->text;
577 $label =~ s/\"/\\\"/g;
578 $rattrs->{'label'} = $label;
579 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
580 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
583 # Add the real edges. Need to weight one edge per rank jump, in a
585 # my $weighted = $self->_add_edge_weights;
586 my @edges = $self->paths;
587 my( %substart, %subend );
588 foreach my $edge ( @edges ) {
589 # Do we need to output this edge?
590 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
591 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
592 my $variables = { %edge_attrs, 'label' => $label };
594 # Account for the rank gap if necessary
595 my $rank0 = $self->reading( $edge->[0] )->rank
596 if $self->reading( $edge->[0] )->has_rank;
597 my $rank1 = $self->reading( $edge->[1] )->rank
598 if $self->reading( $edge->[1] )->has_rank;
599 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
600 $variables->{'minlen'} = $rank1 - $rank0;
603 # Add the calculated edge weights
604 # if( exists $weighted->{$edge->[0]}
605 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
606 # # $variables->{'color'} = 'red';
607 # $variables->{'weight'} = 3.0;
610 # EXPERIMENTAL: make edge width reflect no. of witnesses
611 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
612 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
614 my $varopts = _dot_attr_string( $variables );
615 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
616 $edge->[0], $edge->[1], $varopts );
617 } elsif( $used{$edge->[0]} ) {
618 $subend{$edge->[0]} = 1;
619 } elsif( $used{$edge->[1]} ) {
620 $substart{$edge->[1]} = 1;
623 # Add substitute start and end edges if necessary
624 foreach my $node ( keys %substart ) {
625 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
626 my $variables = { %edge_attrs, 'label' => $witstr };
627 my $varopts = _dot_attr_string( $variables );
628 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
630 foreach my $node ( keys %subend ) {
631 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
632 my $variables = { %edge_attrs, 'label' => $witstr };
633 my $varopts = _dot_attr_string( $variables );
634 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
637 if( $STRAIGHTENHACK ) {
638 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
645 sub _dot_attr_string {
648 foreach my $k ( sort keys %$hash ) {
650 push( @attrs, $k.'="'.$v.'"' );
652 return( '[ ' . join( ', ', @attrs ) . ' ]' );
655 sub _add_edge_weights {
657 # Walk the graph from START to END, choosing the successor node with
658 # the largest number of witness paths each time.
660 my $curr = $self->start->id;
661 my $ranked = $self->end->has_rank;
662 while( $curr ne $self->end->id ) {
663 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
664 my @succ = sort { $self->path_witnesses( $curr, $a )
665 <=> $self->path_witnesses( $curr, $b ) }
666 $self->sequence->successors( $curr );
667 my $next = pop @succ;
668 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
669 # Try to avoid lacunae in the weighted path.
671 ( $self->reading( $next )->is_lacuna ||
672 $nextrank - $rank > 1 ) ){
675 $weighted->{$curr} = $next;
681 =head2 path_witnesses( $edge )
683 Returns the list of sigils whose witnesses are associated with the given edge.
684 The edge can be passed as either an array or an arrayref of ( $source, $target ).
689 my( $self, @edge ) = @_;
690 # If edge is an arrayref, cope.
691 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
695 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
699 sub _path_display_label {
702 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
703 if( scalar @wits > $maj ) {
704 # TODO break out a.c. wits
707 return join( ', ', @wits );
711 =head2 witnesses_at_rank
713 Returns a list of witnesses that are not lacunose, for a given rank.
717 sub witnesses_at_rank {
718 my( $self, $rank ) = @_;
723 Returns a GraphML representation of the collation. The GraphML will contain
724 two graphs. The first expresses the attributes of the readings and the witness
725 paths that link them; the second expresses the relationships that link the
726 readings. This is the native transfer format for a tradition.
735 my $datafile = 't/data/florilegium_tei_ps.xml';
736 my $tradition = Text::Tradition->new( 'input' => 'TEI',
741 ok( $tradition, "Got a tradition object" );
742 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
743 ok( $tradition->collation, "Tradition has a collation" );
745 my $c = $tradition->collation;
746 is( scalar $c->readings, $READINGS, "Collation has all readings" );
747 is( scalar $c->paths, $PATHS, "Collation has all paths" );
748 is( scalar $c->relationships, 0, "Collation has all relationships" );
750 # Add a few relationships
751 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
752 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
753 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
755 # Now write it to GraphML and parse it again.
757 my $graphml = $c->as_graphml;
758 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
759 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
760 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
761 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
771 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
772 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
773 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
774 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
776 # Create the document and root node
777 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
778 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
779 $graphml->setDocumentElement( $root );
780 $root->setNamespace( $xsi_ns, 'xsi', 0 );
781 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
783 # Add the data keys for the graph
786 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
787 foreach my $datum ( @graph_attributes ) {
788 $graph_data_keys{$datum} = 'dg'.$gdi++;
789 my $key = $root->addNewChild( $graphml_ns, 'key' );
790 $key->setAttribute( 'attr.name', $datum );
791 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
792 $key->setAttribute( 'for', 'graph' );
793 $key->setAttribute( 'id', $graph_data_keys{$datum} );
796 # Add the data keys for nodes
803 is_start => 'boolean',
805 is_lacuna => 'boolean',
807 foreach my $datum ( keys %node_data ) {
808 $node_data_keys{$datum} = 'dn'.$ndi++;
809 my $key = $root->addNewChild( $graphml_ns, 'key' );
810 $key->setAttribute( 'attr.name', $datum );
811 $key->setAttribute( 'attr.type', $node_data{$datum} );
812 $key->setAttribute( 'for', 'node' );
813 $key->setAttribute( 'id', $node_data_keys{$datum} );
816 # Add the data keys for edges, i.e. witnesses
820 class => 'string', # Class, deprecated soon
821 witness => 'string', # ID/label for a path
822 relationship => 'string', # ID/label for a relationship
823 extra => 'boolean', # Path key
824 scope => 'string', # Relationship key
825 non_correctable => 'boolean', # Relationship key
826 non_independent => 'boolean', # Relationship key
828 foreach my $datum ( keys %edge_data ) {
829 $edge_data_keys{$datum} = 'de'.$edi++;
830 my $key = $root->addNewChild( $graphml_ns, 'key' );
831 $key->setAttribute( 'attr.name', $datum );
832 $key->setAttribute( 'attr.type', $edge_data{$datum} );
833 $key->setAttribute( 'for', 'edge' );
834 $key->setAttribute( 'id', $edge_data_keys{$datum} );
837 # Add the collation graph itself
838 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
839 $sgraph->setAttribute( 'edgedefault', 'directed' );
840 $sgraph->setAttribute( 'id', $self->tradition->name );
841 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
842 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
843 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
844 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
845 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
847 # Collation attribute data
848 foreach my $datum ( @graph_attributes ) {
849 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
850 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
855 # Add our readings to the graph
856 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
857 # Add to the main graph
858 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
859 my $node_xmlid = 'n' . $node_ctr++;
860 $node_hash{ $n->id } = $node_xmlid;
861 $node_el->setAttribute( 'id', $node_xmlid );
862 foreach my $d ( keys %node_data ) {
864 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
869 # Add the path edges to the sequence graph
871 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
872 # We add an edge in the graphml for every witness in $e.
873 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
874 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
875 $node_hash{ $e->[0] },
876 $node_hash{ $e->[1] } );
877 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
878 $edge_el->setAttribute( 'source', $from );
879 $edge_el->setAttribute( 'target', $to );
880 $edge_el->setAttribute( 'id', $id );
882 # It's a witness path, so add the witness
884 my $key = $edge_data_keys{'witness'};
885 # Is this an ante-corr witness?
886 my $aclabel = $self->ac_label;
887 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
888 # Keep the base witness
890 # ...and record that this is an 'extra' reading path
891 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
893 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
894 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
898 # Add the relationship graph to the XML
899 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
900 $node_data_keys{'id'}, \%edge_data_keys );
902 # Save and return the thing
903 my $result = decode_utf8( $graphml->toString(1) );
907 sub _add_graphml_data {
908 my( $el, $key, $value ) = @_;
909 return unless defined $value;
910 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
911 $data_el->setAttribute( 'key', $key );
912 $data_el->appendText( $value );
917 Returns a CSV alignment table representation of the collation graph, one
918 row per witness (or witness uncorrected.)
924 my $table = $self->alignment_table;
925 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
927 # Make the header row
928 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
929 push( @result, decode_utf8( $csv->string ) );
930 # Make the rest of the rows
931 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
932 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
933 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
934 $csv->combine( @row );
935 push( @result, decode_utf8( $csv->string ) );
937 return join( "\n", @result );
940 =head2 alignment_table( $use_refs, $include_witnesses )
942 Return a reference to an alignment table, in a slightly enhanced CollateX
943 format which looks like this:
945 $table = { alignment => [ { witness => "SIGIL",
946 tokens => [ { t => "TEXT" }, ... ] },
948 tokens => [ { t => "TEXT" }, ... ] },
952 If $use_refs is set to 1, the reading object is returned in the table
953 instead of READINGTEXT; if not, the text of the reading is returned.
955 If $include_witnesses is set to a hashref, only the witnesses whose sigil
956 keys have a true hash value will be included.
960 sub alignment_table {
962 my $include; # see if we can ditch this
963 return $self->cached_table if $self->has_cached_table;
965 # Make sure we can do this
966 throw( "Need a linear graph in order to make an alignment table" )
967 unless $self->linear;
968 $self->calculate_ranks unless $self->end->has_rank;
970 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
971 my @all_pos = ( 1 .. $self->end->rank - 1 );
972 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
974 next unless $include->{$wit->sigil};
976 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
977 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
978 my @row = _make_witness_row( \@wit_path, \@all_pos );
979 push( @{$table->{'alignment'}},
980 { 'witness' => $wit->sigil, 'tokens' => \@row } );
981 if( $wit->is_layered ) {
982 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
983 $wit->sigil.$self->ac_label );
984 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
985 push( @{$table->{'alignment'}},
986 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
989 $self->cached_table( $table );
993 sub _make_witness_row {
994 my( $path, $positions ) = @_;
996 map { $char_hash{$_} = undef } @$positions;
998 foreach my $rdg ( @$path ) {
999 my $rtext = $rdg->text;
1000 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1001 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1002 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1003 $char_hash{$rdg->rank} = { 't' => $rdg };
1005 my @row = map { $char_hash{$_} } @$positions;
1006 # Fill in lacuna markers for undef spots in the row
1007 my $last_el = shift @row;
1008 my @filled_row = ( $last_el );
1009 foreach my $el ( @row ) {
1010 # If we are using node reference, make the lacuna node appear many times
1011 # in the table. If not, use the lacuna tag.
1012 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1015 push( @filled_row, $el );
1021 =head1 NAVIGATION METHODS
1023 =head2 reading_sequence( $first, $last, $sigil, $backup )
1025 Returns the ordered list of readings, starting with $first and ending
1026 with $last, for the witness given in $sigil. If a $backup sigil is
1027 specified (e.g. when walking a layered witness), it will be used wherever
1028 no $sigil path exists. If there is a base text reading, that will be
1029 used wherever no path exists for $sigil or $backup.
1033 # TODO Think about returning some lazy-eval iterator.
1034 # TODO Get rid of backup; we should know from what witness is whether we need it.
1036 sub reading_sequence {
1037 my( $self, $start, $end, $witness ) = @_;
1039 $witness = $self->baselabel unless $witness;
1040 my @readings = ( $start );
1043 while( $n && $n->id ne $end->id ) {
1044 if( exists( $seen{$n->id} ) ) {
1045 throw( "Detected loop for $witness at " . $n->id );
1049 my $next = $self->next_reading( $n, $witness );
1051 throw( "Did not find any path for $witness from reading " . $n->id );
1053 push( @readings, $next );
1056 # Check that the last reading is our end reading.
1057 my $last = $readings[$#readings];
1058 throw( "Last reading found from " . $start->text .
1059 " for witness $witness is not the end!" ) # TODO do we get this far?
1060 unless $last->id eq $end->id;
1065 =head2 next_reading( $reading, $sigil );
1067 Returns the reading that follows the given reading along the given witness
1073 # Return the successor via the corresponding path.
1075 my $answer = $self->_find_linked_reading( 'next', @_ );
1076 return undef unless $answer;
1077 return $self->reading( $answer );
1080 =head2 prior_reading( $reading, $sigil )
1082 Returns the reading that precedes the given reading along the given witness
1088 # Return the predecessor via the corresponding path.
1090 my $answer = $self->_find_linked_reading( 'prior', @_ );
1091 return $self->reading( $answer );
1094 sub _find_linked_reading {
1095 my( $self, $direction, $node, $path ) = @_;
1097 # Get a backup if we are dealing with a layered witness
1099 my $aclabel = $self->ac_label;
1100 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1104 my @linked_paths = $direction eq 'next'
1105 ? $self->sequence->edges_from( $node )
1106 : $self->sequence->edges_to( $node );
1107 return undef unless scalar( @linked_paths );
1109 # We have to find the linked path that contains all of the
1110 # witnesses supplied in $path.
1111 my( @path_wits, @alt_path_wits );
1112 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1113 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1116 foreach my $le ( @linked_paths ) {
1117 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1120 my @le_wits = sort $self->path_witnesses( $le );
1121 if( _is_within( \@path_wits, \@le_wits ) ) {
1122 # This is the right path.
1123 return $direction eq 'next' ? $le->[1] : $le->[0];
1124 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1128 # Got this far? Return the alternate path if it exists.
1129 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1132 # Got this far? Return the base path if it exists.
1133 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1136 # Got this far? We have no appropriate path.
1137 warn "Could not find $direction node from " . $node->id
1138 . " along path $path";
1144 my( $set1, $set2 ) = @_;
1145 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1146 foreach my $el ( @$set1 ) {
1147 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1152 # Return the string that joins together a list of witnesses for
1153 # display on a single path.
1154 sub _witnesses_of_label {
1155 my( $self, $label ) = @_;
1156 my $regex = $self->wit_list_separator;
1157 my @answer = split( /\Q$regex\E/, $label );
1161 =head2 common_readings
1163 Returns the list of common readings in the graph (i.e. those readings that are
1164 shared by all non-lacunose witnesses.)
1168 sub common_readings {
1170 my @common = grep { $_->is_common } $self->readings;
1174 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1176 Returns the text of a witness (plus its backup, if we are using a layer)
1177 as stored in the collation. The text is returned as a string, where the
1178 individual readings are joined with spaces and the meta-readings (e.g.
1179 lacunae) are omitted. Optional specification of $start and $end allows
1180 the generation of a subset of the witness text.
1185 my( $self, $wit, $start, $end ) = @_;
1186 $start = $self->start unless $start;
1187 $end = $self->end unless $end;
1188 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1189 return join( ' ', map { $_->text } @path );
1192 =head1 INITIALIZATION METHODS
1194 These are mostly for use by parsers.
1196 =head2 make_witness_path( $witness )
1198 Link the array of readings contained in $witness->path (and in
1199 $witness->uncorrected_path if it exists) into collation paths.
1200 Clear out the arrays when finished.
1202 =head2 make_witness_paths
1204 Call make_witness_path for all witnesses in the tradition.
1208 # For use when a collation is constructed from a base text and an apparatus.
1209 # We have the sequences of readings and just need to add path edges.
1210 # When we are done, clear out the witness path attributes, as they are no
1212 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1214 sub make_witness_paths {
1216 foreach my $wit ( $self->tradition->witnesses ) {
1217 # print STDERR "Making path for " . $wit->sigil . "\n";
1218 $self->make_witness_path( $wit );
1222 sub make_witness_path {
1223 my( $self, $wit ) = @_;
1224 my @chain = @{$wit->path};
1225 my $sig = $wit->sigil;
1226 foreach my $idx ( 0 .. $#chain-1 ) {
1227 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1229 if( $wit->is_layered ) {
1230 @chain = @{$wit->uncorrected_path};
1231 foreach my $idx( 0 .. $#chain-1 ) {
1232 my $source = $chain[$idx];
1233 my $target = $chain[$idx+1];
1234 $self->add_path( $source, $target, $sig.$self->ac_label )
1235 unless $self->has_path( $source, $target, $sig );
1239 $wit->clear_uncorrected_path;
1242 =head2 calculate_ranks
1244 Calculate the reading ranks (that is, their aligned positions relative
1245 to each other) for the graph. This can only be called on linear collations.
1249 use Text::Tradition;
1251 my $cxfile = 't/data/Collatex-16.xml';
1252 my $t = Text::Tradition->new(
1254 'input' => 'CollateX',
1257 my $c = $t->collation;
1260 my $svg = $c->as_svg;
1261 is( substr( $svg, 0, 5 ), '<?xml', "Got XML doc for svg" );
1262 ok( $c->has_cached_svg, "SVG was cached" );
1263 is( $c->as_svg, $svg, "Cached SVG returned upon second call" );
1264 $c->calculate_ranks;
1265 is( $c->as_svg, $svg, "Cached SVG retained with no rank change" );
1266 $c->add_relationship( 'n9', 'n23', { 'type' => 'spelling' } );
1267 isnt( $c->as_svg, $svg, "SVG changed after relationship add" );
1273 sub calculate_ranks {
1275 # Save the existing ranks, in case we need to invalidate the cached SVG.
1277 # Walk a version of the graph where every node linked by a relationship
1278 # edge is fundamentally the same node, and do a topological ranking on
1279 # the nodes in this graph.
1280 my $topo_graph = Graph->new();
1284 foreach my $r ( $self->readings ) {
1285 next if exists $rel_containers{$r->id};
1286 my @rels = $r->related_readings( 'colocated' );
1288 # Make a relationship container.
1290 my $rn = 'rel_container_' . $rel_ctr++;
1291 $topo_graph->add_vertex( $rn );
1293 $rel_containers{$_->id} = $rn;
1296 # Add a new node to mirror the old node.
1297 $rel_containers{$r->id} = $r->id;
1298 $topo_graph->add_vertex( $r->id );
1303 foreach my $r ( $self->readings ) {
1304 $existing_ranks{$r} = $r->rank;
1305 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1306 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1307 $rel_containers{$n} );
1308 # $DB::single = 1 unless $tfrom && $tto;
1309 $topo_graph->add_edge( $tfrom, $tto );
1313 # Now do the rankings, starting with the start node.
1314 my $topo_start = $rel_containers{$self->start->id};
1315 my $node_ranks = { $topo_start => 0 };
1316 my @curr_origin = ( $topo_start );
1317 # A little iterative function.
1318 while( @curr_origin ) {
1319 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1321 # Transfer our rankings from the topological graph to the real one.
1322 foreach my $r ( $self->readings ) {
1323 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1324 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1326 # Die. Find the last rank we calculated.
1327 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1328 <=> $node_ranks->{$rel_containers{$b->id}} }
1330 my $last = pop @all_defined;
1331 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1334 # Do we need to invalidate the cached SVG?
1335 if( $self->has_cached_svg ) {
1336 foreach my $r ( $self->readings ) {
1337 next if $existing_ranks{$r} == $r->rank;
1345 my( $graph, $node_ranks, @current_nodes ) = @_;
1346 # Look at each of the children of @current_nodes. If all the child's
1347 # parents have a rank, assign it the highest rank + 1 and add it to
1348 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1349 # parent gets a rank.
1351 foreach my $c ( @current_nodes ) {
1352 warn "Current reading $c has no rank!"
1353 unless exists $node_ranks->{$c};
1354 # print STDERR "Looking at child of node $c, rank "
1355 # . $node_ranks->{$c} . "\n";
1356 foreach my $child ( $graph->successors( $c ) ) {
1357 next if exists $node_ranks->{$child};
1358 my $highest_rank = -1;
1360 foreach my $parent ( $graph->predecessors( $child ) ) {
1361 if( exists $node_ranks->{$parent} ) {
1362 $highest_rank = $node_ranks->{$parent}
1363 if $highest_rank <= $node_ranks->{$parent};
1370 my $c_rank = $highest_rank + 1;
1371 # print STDERR "Assigning rank $c_rank to node $child \n";
1372 $node_ranks->{$child} = $c_rank;
1373 push( @next_nodes, $child );
1379 =head2 flatten_ranks
1381 A convenience method for parsing collation data. Searches the graph for readings
1382 with the same text at the same rank, and merges any that are found.
1388 my %unique_rank_rdg;
1389 foreach my $rdg ( $self->readings ) {
1390 next unless $rdg->has_rank;
1391 my $key = $rdg->rank . "||" . $rdg->text;
1392 if( exists $unique_rank_rdg{$key} ) {
1394 # print STDERR "Combining readings at same rank: $key\n";
1395 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1396 # TODO see if this now makes a common point.
1398 $unique_rank_rdg{$key} = $rdg;
1403 =head2 remove_collations
1405 Another convenience method for parsing. Removes all 'collation' relationships
1406 that were defined in order to get the reading ranks to be correct.
1410 use Text::Tradition;
1412 my $cxfile = 't/data/Collatex-16.xml';
1413 my $t = Text::Tradition->new(
1415 'input' => 'CollateX',
1418 my $c = $t->collation;
1420 isnt( $c->reading('n23')->rank, $c->reading('n9')->rank, "Rank skew exists" );
1421 $c->add_relationship( 'n23', 'n9', { 'type' => 'collated', 'scope' => 'local' } );
1422 is( scalar $c->relationships, 4, "Found all expected relationships" );
1423 $c->remove_collations;
1424 is( scalar $c->relationships, 3, "Collated relationships now gone" );
1425 is( $c->reading('n23')->rank, $c->reading('n9')->rank, "Aligned ranks were preserved" );
1431 sub remove_collations {
1433 foreach my $reledge ( $self->relationships ) {
1434 my $relobj = $self->relations->get_relationship( $reledge );
1435 if( $relobj && $relobj->type eq 'collated' ) {
1436 $self->relations->delete_relationship( $reledge );
1442 =head2 calculate_common_readings
1444 Goes through the graph identifying the readings that appear in every witness
1445 (apart from those with lacunae at that spot.) Marks them as common and returns
1450 use Text::Tradition;
1452 my $cxfile = 't/data/Collatex-16.xml';
1453 my $t = Text::Tradition->new(
1455 'input' => 'CollateX',
1458 my $c = $t->collation;
1460 my @common = $c->calculate_common_readings();
1461 is( scalar @common, 8, "Found correct number of common readings" );
1462 my @marked = sort $c->common_readings();
1463 is( scalar @common, 8, "All common readings got marked as such" );
1464 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1465 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1471 sub calculate_common_readings {
1474 my $table = $self->alignment_table;
1475 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1476 my @row = map { $_->{'tokens'}->[$idx]->{'t'} } @{$table->{'alignment'}};
1478 foreach my $r ( @row ) {
1480 $hash{$r->id} = $r unless $r->is_meta;
1482 $hash{'UNDEF'} = $r;
1485 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1486 my( $r ) = values %hash;
1488 push( @common, $r );
1494 =head2 text_from_paths
1496 Calculate the text array for all witnesses from the path, for later consistency
1497 checking. Only to be used if there is no non-graph-based way to know the
1502 sub text_from_paths {
1504 foreach my $wit ( $self->tradition->witnesses ) {
1505 my @text = split( /\s+/,
1506 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1507 $wit->text( \@text );
1508 if( $wit->is_layered ) {
1509 my @uctext = split( /\s+/,
1510 $self->reading_sequence( $self->start, $self->end,
1511 $wit->sigil.$self->ac_label ) );
1512 $wit->text( \@uctext );
1517 =head1 UTILITY FUNCTIONS
1519 =head2 common_predecessor( $reading_a, $reading_b )
1521 Find the last reading that occurs in sequence before both the given readings.
1523 =head2 common_successor( $reading_a, $reading_b )
1525 Find the first reading that occurs in sequence after both the given readings.
1529 use Text::Tradition;
1531 my $cxfile = 't/data/Collatex-16.xml';
1532 my $t = Text::Tradition->new(
1534 'input' => 'CollateX',
1537 my $c = $t->collation;
1539 is( $c->common_predecessor( 'n9', 'n23' )->id,
1540 'n20', "Found correct common predecessor" );
1541 is( $c->common_successor( 'n9', 'n23' )->id,
1542 '#END#', "Found correct common successor" );
1544 is( $c->common_predecessor( 'n19', 'n17' )->id,
1545 'n16', "Found correct common predecessor for readings on same path" );
1546 is( $c->common_successor( 'n21', 'n26' )->id,
1547 '#END#', "Found correct common successor for readings on same path" );
1553 ## Return the closest reading that is a predecessor of both the given readings.
1554 sub common_predecessor {
1556 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1557 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1560 sub common_successor {
1562 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1563 return $self->_common_in_path( $r1, $r2, 'successors' );
1566 sub _common_in_path {
1567 my( $self, $r1, $r2, $dir ) = @_;
1568 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1569 $iter = $self->end->rank - $iter if $dir eq 'successors';
1571 my @last_checked = ( $r1, $r2 );
1573 while( !@candidates ) {
1575 foreach my $lc ( @last_checked ) {
1576 foreach my $p ( $lc->$dir ) {
1577 if( $all_seen{$p->id} ) {
1578 push( @candidates, $p );
1580 $all_seen{$p->id} = 1;
1581 push( @new_lc, $p );
1585 @last_checked = @new_lc;
1587 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1588 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1592 Text::Tradition::Error->throw(
1593 'ident' => 'Collation error',
1599 __PACKAGE__->meta->make_immutable;
1603 This package is free software and is provided "as is" without express
1604 or implied warranty. You can redistribute it and/or modify it under
1605 the same terms as Perl itself.
1609 Tara L Andrews E<lt>aurum@cpan.orgE<gt>