1 package Text::Tradition::Collation;
3 use Encode qw( decode_utf8 );
7 use IPC::Run qw( run binary );
9 use Text::Tradition::Collation::Reading;
10 use Text::Tradition::Collation::RelationshipStore;
11 use Text::Tradition::Error;
13 use XML::LibXML::XPathContext;
19 default => sub { Graph->new() },
27 isa => 'Text::Tradition::Collation::RelationshipStore',
29 relationships => 'relationships',
30 related_readings => 'related_readings',
31 get_relationship => 'get_relationship',
32 del_relationship => 'del_relationship',
34 writer => '_set_relations',
39 isa => 'Text::Tradition',
44 isa => 'HashRef[Text::Tradition::Collation::Reading]',
48 _add_reading => 'set',
49 del_reading => 'delete',
50 has_reading => 'exists',
53 default => sub { {} },
56 has 'wit_list_separator' => (
65 default => 'base text',
82 isa => 'Text::Tradition::Collation::Reading',
83 writer => '_set_start',
89 isa => 'Text::Tradition::Collation::Reading',
97 predicate => 'has_cached_svg',
98 clearer => 'wipe_svg',
101 has 'cached_table' => (
104 predicate => 'has_cached_table',
105 clearer => 'wipe_table',
108 has '_graphcalc_done' => (
116 Text::Tradition::Collation - a software model for a text collation
121 my $t = Text::Tradition->new(
122 'name' => 'this is a text',
124 'file' => '/path/to/tei_parallel_seg_file.xml' );
126 my $c = $t->collation;
127 my @readings = $c->readings;
128 my @paths = $c->paths;
129 my @relationships = $c->relationships;
131 my $svg_variant_graph = $t->collation->as_svg();
135 Text::Tradition is a library for representation and analysis of collated
136 texts, particularly medieval ones. The Collation is the central feature of
137 a Tradition, where the text, its sequence of readings, and its relationships
138 between readings are actually kept.
144 The constructor. Takes a hash or hashref of the following arguments:
148 =item * tradition - The Text::Tradition object to which the collation
151 =item * linear - Whether the collation should be linear; that is, whether
152 transposed readings should be treated as two linked readings rather than one,
153 and therefore whether the collation graph is acyclic. Defaults to true.
155 =item * baselabel - The default label for the path taken by a base text
156 (if any). Defaults to 'base text'.
158 =item * wit_list_separator - The string to join a list of witnesses for
159 purposes of making labels in display graphs. Defaults to ', '.
161 =item * ac_label - The extra label to tack onto a witness sigil when
162 representing another layer of path for the given witness - that is, when
163 a text has more than one possible reading due to scribal corrections or
164 the like. Defaults to ' (a.c.)'.
174 =head2 wit_list_separator
180 Simple accessors for collation attributes.
184 The meta-reading at the start of every witness path.
188 The meta-reading at the end of every witness path.
192 Returns all Reading objects in the graph.
194 =head2 reading( $id )
196 Returns the Reading object corresponding to the given ID.
198 =head2 add_reading( $reading_args )
200 Adds a new reading object to the collation.
201 See L<Text::Tradition::Collation::Reading> for the available arguments.
203 =head2 del_reading( $object_or_id )
205 Removes the given reading from the collation, implicitly removing its
206 paths and relationships.
208 =head2 merge_readings( $main, $second )
210 Merges the $second reading into the $main one.
211 The arguments may be either readings or reading IDs.
213 =head2 has_reading( $id )
215 Predicate to see whether a given reading ID is in the graph.
217 =head2 reading_witnesses( $object_or_id )
219 Returns a list of sigils whose witnesses contain the reading.
223 Returns all reading paths within the document - that is, all edges in the
224 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
226 =head2 add_path( $source, $target, $sigil )
228 Links the given readings in the collation in sequence, under the given witness
229 sigil. The readings may be specified by object or ID.
231 =head2 del_path( $source, $target, $sigil )
233 Links the given readings in the collation in sequence, under the given witness
234 sigil. The readings may be specified by object or ID.
236 =head2 has_path( $source, $target );
238 Returns true if the two readings are linked in sequence in any witness.
239 The readings may be specified by object or ID.
243 Returns all Relationship objects in the collation.
245 =head2 add_relationship( $reading, $other_reading, $options )
247 Adds a new relationship of the type given in $options between the two readings,
248 which may be specified by object or ID. Returns a value of ( $status, @vectors)
249 where $status is true on success, and @vectors is a list of relationship edges
250 that were ultimately added.
251 See L<Text::Tradition::Collation::Relationship> for the available options.
257 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
258 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
259 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
262 ### Reading construct/destruct functions
265 my( $self, $reading ) = @_;
266 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
267 my %args = %$reading;
268 $reading = Text::Tradition::Collation::Reading->new(
269 'collation' => $self,
272 # First check to see if a reading with this ID exists.
273 if( $self->reading( $reading->id ) ) {
274 throw( "Collation already has a reading with id " . $reading->id );
276 $self->_graphcalc_done(0);
277 $self->_add_reading( $reading->id => $reading );
278 # Once the reading has been added, put it in both graphs.
279 $self->sequence->add_vertex( $reading->id );
280 $self->relations->add_reading( $reading->id );
284 around del_reading => sub {
289 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
292 # Remove the reading from the graphs.
293 $self->_graphcalc_done(0);
294 $self->sequence->delete_vertex( $arg );
295 $self->relations->delete_reading( $arg );
298 $self->$orig( $arg );
301 # merge_readings( $main, $to_be_deleted );
306 # We only need the IDs for adding paths to the graph, not the reading
307 # objects themselves.
308 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
309 $self->_graphcalc_done(0);
311 # The kept reading should inherit the paths and the relationships
312 # of the deleted reading.
313 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
314 my @vector = ( $kept );
315 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
316 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
317 next if $vector[0] eq $vector[1]; # Don't add a self loop
318 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
319 $self->sequence->add_edge( @vector );
320 my $fwits = $self->sequence->get_edge_attributes( @vector );
321 @wits{keys %$fwits} = values %$fwits;
322 $self->sequence->set_edge_attributes( @vector, \%wits );
324 $self->relations->merge_readings( $kept, $deleted, $combine_char );
326 # Do the deletion deed.
327 if( $combine_char ) {
328 my $kept_obj = $self->reading( $kept );
329 my $new_text = join( $combine_char, $kept_obj->text,
330 $self->reading( $deleted )->text );
331 $kept_obj->alter_text( $new_text );
333 $self->del_reading( $deleted );
337 # Helper function for manipulating the graph.
338 sub _stringify_args {
339 my( $self, $first, $second, $arg ) = @_;
341 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
342 $second = $second->id
343 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
344 return( $first, $second, $arg );
347 # Helper function for manipulating the graph.
348 sub _objectify_args {
349 my( $self, $first, $second, $arg ) = @_;
350 $first = $self->reading( $first )
351 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
352 $second = $self->reading( $second )
353 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
354 return( $first, $second, $arg );
361 # We only need the IDs for adding paths to the graph, not the reading
362 # objects themselves.
363 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
365 $self->_graphcalc_done(0);
366 # Connect the readings
367 $self->sequence->add_edge( $source, $target );
368 # Note the witness in question
369 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
375 if( ref( $_[0] ) eq 'ARRAY' ) {
382 # We only need the IDs for adding paths to the graph, not the reading
383 # objects themselves.
384 my( $source, $target, $wit ) = $self->_stringify_args( @args );
386 $self->_graphcalc_done(0);
387 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
388 $self->sequence->delete_edge_attribute( $source, $target, $wit );
390 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
391 $self->sequence->delete_edge( $source, $target );
396 # Extra graph-alike utility
399 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
400 return undef unless $self->sequence->has_edge( $source, $target );
401 return $self->sequence->has_edge_attribute( $source, $target, $wit );
404 =head2 clear_witness( @sigil_list )
406 Clear the given witnesses out of the collation entirely, removing references
407 to them in paths, and removing readings that belong only to them. Should only
408 be called via $tradition->del_witness.
413 my( $self, @sigils ) = @_;
415 $self->_graphcalc_done(0);
416 # Clear the witness(es) out of the paths
417 foreach my $e ( $self->paths ) {
418 foreach my $sig ( @sigils ) {
419 $self->del_path( $e, $sig );
423 # Clear out the newly unused readings
424 foreach my $r ( $self->readings ) {
425 unless( $self->reading_witnesses( $r ) ) {
426 $self->del_reading( $r );
431 sub add_relationship {
433 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
434 my( @vectors ) = $self->relations->add_relationship( $source,
435 $self->reading( $source ), $target, $self->reading( $target ), $opts );
436 $self->_graphcalc_done(0);
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. Must have GraphViz (dot) installed to run.
475 my( $self, $opts ) = @_;
476 throw( "Need GraphViz installed to output SVG" )
477 unless File::Which::which( 'dot' );
478 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
479 $self->calculate_ranks() unless $self->_graphcalc_done;
480 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
481 my @cmd = qw/dot -Tsvg/;
483 my $dotfile = File::Temp->new();
485 # $dotfile->unlink_on_destroy(0);
486 binmode $dotfile, ':utf8';
487 print $dotfile $self->as_dot( $opts );
488 push( @cmd, $dotfile->filename );
489 run( \@cmd, ">", binary(), \$svg );
490 $svg = decode_utf8( $svg );
491 $self->cached_svg( $svg ) unless $want_subgraph;
494 return $self->cached_svg;
499 =head2 as_dot( \%options )
501 Returns a string that is the collation graph expressed in dot
502 (i.e. GraphViz) format. Options include:
517 my( $self, $opts ) = @_;
518 my $startrank = $opts->{'from'} if $opts;
519 my $endrank = $opts->{'to'} if $opts;
520 my $color_common = $opts->{'color_common'} if $opts;
521 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
522 && $self->end->rank > 100;
524 # Check the arguments
526 return if $endrank && $startrank > $endrank;
527 return if $startrank > $self->end->rank;
529 if( defined $endrank ) {
530 return if $endrank < 0;
531 $endrank = undef if $endrank == $self->end->rank;
534 my $graph_name = $self->tradition->name;
535 $graph_name =~ s/[^\w\s]//g;
536 $graph_name = join( '_', split( /\s+/, $graph_name ) );
544 'fillcolor' => 'white',
549 'arrowhead' => 'open',
550 'color' => '#000000',
551 'fontcolor' => '#000000',
554 my $dot = sprintf( "digraph %s {\n", $graph_name );
555 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
556 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
558 # Output substitute start/end readings if necessary
560 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
563 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
565 if( $STRAIGHTENHACK ) {
567 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
568 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
570 my %used; # Keep track of the readings that actually appear in the graph
571 # Sort the readings by rank if we have ranks; this speeds layout.
572 my @all_readings = $self->end->has_rank
573 ? sort { $a->rank <=> $b->rank } $self->readings
575 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
576 foreach my $reading ( @all_readings ) {
577 # Only output readings within our rank range.
578 next if $startrank && $reading->rank < $startrank;
579 next if $endrank && $reading->rank > $endrank;
580 $used{$reading->id} = 1;
581 # Need not output nodes without separate labels
582 next if $reading->id eq $reading->text;
584 my $label = $reading->text;
585 $label .= '-' if $reading->join_next;
586 $label = "-$label" if $reading->join_prior;
587 $label =~ s/\"/\\\"/g;
588 $rattrs->{'label'} = $label;
589 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
590 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
593 # Add the real edges. Need to weight one edge per rank jump, in a
595 # my $weighted = $self->_add_edge_weights;
596 my @edges = $self->paths;
597 my( %substart, %subend );
598 foreach my $edge ( @edges ) {
599 # Do we need to output this edge?
600 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
601 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
602 my $variables = { %edge_attrs, 'label' => $label };
604 # Account for the rank gap if necessary
605 my $rank0 = $self->reading( $edge->[0] )->rank
606 if $self->reading( $edge->[0] )->has_rank;
607 my $rank1 = $self->reading( $edge->[1] )->rank
608 if $self->reading( $edge->[1] )->has_rank;
609 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
610 $variables->{'minlen'} = $rank1 - $rank0;
613 # Add the calculated edge weights
614 # if( exists $weighted->{$edge->[0]}
615 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
616 # # $variables->{'color'} = 'red';
617 # $variables->{'weight'} = 3.0;
620 # EXPERIMENTAL: make edge width reflect no. of witnesses
621 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
622 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
624 my $varopts = _dot_attr_string( $variables );
625 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
626 $edge->[0], $edge->[1], $varopts );
627 } elsif( $used{$edge->[0]} ) {
628 $subend{$edge->[0]} = 1;
629 } elsif( $used{$edge->[1]} ) {
630 $substart{$edge->[1]} = 1;
633 # Add substitute start and end edges if necessary
634 foreach my $node ( keys %substart ) {
635 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
636 my $variables = { %edge_attrs, 'label' => $witstr };
637 my $varopts = _dot_attr_string( $variables );
638 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
640 foreach my $node ( keys %subend ) {
641 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
642 my $variables = { %edge_attrs, 'label' => $witstr };
643 my $varopts = _dot_attr_string( $variables );
644 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
647 if( $STRAIGHTENHACK ) {
648 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
655 sub _dot_attr_string {
658 foreach my $k ( sort keys %$hash ) {
660 push( @attrs, $k.'="'.$v.'"' );
662 return( '[ ' . join( ', ', @attrs ) . ' ]' );
665 sub _add_edge_weights {
667 # Walk the graph from START to END, choosing the successor node with
668 # the largest number of witness paths each time.
670 my $curr = $self->start->id;
671 my $ranked = $self->end->has_rank;
672 while( $curr ne $self->end->id ) {
673 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
674 my @succ = sort { $self->path_witnesses( $curr, $a )
675 <=> $self->path_witnesses( $curr, $b ) }
676 $self->sequence->successors( $curr );
677 my $next = pop @succ;
678 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
679 # Try to avoid lacunae in the weighted path.
681 ( $self->reading( $next )->is_lacuna ||
682 $nextrank - $rank > 1 ) ){
685 $weighted->{$curr} = $next;
691 =head2 path_witnesses( $edge )
693 Returns the list of sigils whose witnesses are associated with the given edge.
694 The edge can be passed as either an array or an arrayref of ( $source, $target ).
699 my( $self, @edge ) = @_;
700 # If edge is an arrayref, cope.
701 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
705 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
709 sub _path_display_label {
712 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
713 if( scalar @wits > $maj ) {
714 # TODO break out a.c. wits
717 return join( ', ', @wits );
721 =head2 witnesses_at_rank
723 Returns a list of witnesses that are not lacunose, for a given rank.
727 sub witnesses_at_rank {
728 my( $self, $rank ) = @_;
733 Returns a GraphML representation of the collation. The GraphML will contain
734 two graphs. The first expresses the attributes of the readings and the witness
735 paths that link them; the second expresses the relationships that link the
736 readings. This is the native transfer format for a tradition.
745 my $datafile = 't/data/florilegium_tei_ps.xml';
746 my $tradition = Text::Tradition->new( 'input' => 'TEI',
751 ok( $tradition, "Got a tradition object" );
752 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
753 ok( $tradition->collation, "Tradition has a collation" );
755 my $c = $tradition->collation;
756 is( scalar $c->readings, $READINGS, "Collation has all readings" );
757 is( scalar $c->paths, $PATHS, "Collation has all paths" );
758 is( scalar $c->relationships, 0, "Collation has all relationships" );
760 # Add a few relationships
761 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
762 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
763 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
765 # Now write it to GraphML and parse it again.
767 my $graphml = $c->as_graphml;
768 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
769 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
770 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
771 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
779 $self->calculate_ranks unless $self->_graphcalc_done;
782 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
783 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
784 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
785 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
787 # Create the document and root node
788 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
789 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
790 $graphml->setDocumentElement( $root );
791 $root->setNamespace( $xsi_ns, 'xsi', 0 );
792 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
794 # Add the data keys for the graph
797 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
798 foreach my $datum ( @graph_attributes ) {
799 $graph_data_keys{$datum} = 'dg'.$gdi++;
800 my $key = $root->addNewChild( $graphml_ns, 'key' );
801 $key->setAttribute( 'attr.name', $datum );
802 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
803 $key->setAttribute( 'for', 'graph' );
804 $key->setAttribute( 'id', $graph_data_keys{$datum} );
807 # Add the data keys for nodes
814 is_start => 'boolean',
816 is_lacuna => 'boolean',
817 is_common => 'boolean',
818 join_prior => 'boolean',
819 join_next => 'boolean',
821 foreach my $datum ( keys %node_data ) {
822 $node_data_keys{$datum} = 'dn'.$ndi++;
823 my $key = $root->addNewChild( $graphml_ns, 'key' );
824 $key->setAttribute( 'attr.name', $datum );
825 $key->setAttribute( 'attr.type', $node_data{$datum} );
826 $key->setAttribute( 'for', 'node' );
827 $key->setAttribute( 'id', $node_data_keys{$datum} );
830 # Add the data keys for edges, i.e. witnesses
834 class => 'string', # Class, deprecated soon
835 witness => 'string', # ID/label for a path
836 relationship => 'string', # ID/label for a relationship
837 extra => 'boolean', # Path key
838 scope => 'string', # Relationship key
839 annotation => 'string', # Relationship key
840 non_correctable => 'boolean', # Relationship key
841 non_independent => 'boolean', # Relationship key
843 foreach my $datum ( keys %edge_data ) {
844 $edge_data_keys{$datum} = 'de'.$edi++;
845 my $key = $root->addNewChild( $graphml_ns, 'key' );
846 $key->setAttribute( 'attr.name', $datum );
847 $key->setAttribute( 'attr.type', $edge_data{$datum} );
848 $key->setAttribute( 'for', 'edge' );
849 $key->setAttribute( 'id', $edge_data_keys{$datum} );
852 # Add the collation graph itself
853 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
854 $sgraph->setAttribute( 'edgedefault', 'directed' );
855 $sgraph->setAttribute( 'id', $self->tradition->name );
856 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
857 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
858 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
859 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
860 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
862 # Collation attribute data
863 foreach my $datum ( @graph_attributes ) {
864 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
865 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
870 # Add our readings to the graph
871 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
872 # Add to the main graph
873 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
874 my $node_xmlid = 'n' . $node_ctr++;
875 $node_hash{ $n->id } = $node_xmlid;
876 $node_el->setAttribute( 'id', $node_xmlid );
877 foreach my $d ( keys %node_data ) {
879 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
884 # Add the path edges to the sequence graph
886 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
887 # We add an edge in the graphml for every witness in $e.
888 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
889 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
890 $node_hash{ $e->[0] },
891 $node_hash{ $e->[1] } );
892 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
893 $edge_el->setAttribute( 'source', $from );
894 $edge_el->setAttribute( 'target', $to );
895 $edge_el->setAttribute( 'id', $id );
897 # It's a witness path, so add the witness
899 my $key = $edge_data_keys{'witness'};
900 # Is this an ante-corr witness?
901 my $aclabel = $self->ac_label;
902 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
903 # Keep the base witness
905 # ...and record that this is an 'extra' reading path
906 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
908 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
909 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
913 # Add the relationship graph to the XML
914 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
915 $node_data_keys{'id'}, \%edge_data_keys );
917 # Save and return the thing
918 my $result = decode_utf8( $graphml->toString(1) );
922 sub _add_graphml_data {
923 my( $el, $key, $value ) = @_;
924 return unless defined $value;
925 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
926 $data_el->setAttribute( 'key', $key );
927 $data_el->appendText( $value );
932 Returns a CSV alignment table representation of the collation graph, one
933 row per witness (or witness uncorrected.)
939 my $table = $self->alignment_table;
940 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
942 # Make the header row
943 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
944 push( @result, decode_utf8( $csv->string ) );
945 # Make the rest of the rows
946 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
947 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
948 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
949 $csv->combine( @row );
950 push( @result, decode_utf8( $csv->string ) );
952 return join( "\n", @result );
955 =head2 alignment_table( $use_refs, $include_witnesses )
957 Return a reference to an alignment table, in a slightly enhanced CollateX
958 format which looks like this:
960 $table = { alignment => [ { witness => "SIGIL",
961 tokens => [ { t => "TEXT" }, ... ] },
963 tokens => [ { t => "TEXT" }, ... ] },
967 If $use_refs is set to 1, the reading object is returned in the table
968 instead of READINGTEXT; if not, the text of the reading is returned.
970 If $include_witnesses is set to a hashref, only the witnesses whose sigil
971 keys have a true hash value will be included.
975 sub alignment_table {
977 $self->calculate_ranks() unless $self->_graphcalc_done;
978 return $self->cached_table if $self->has_cached_table;
980 # Make sure we can do this
981 throw( "Need a linear graph in order to make an alignment table" )
982 unless $self->linear;
983 $self->calculate_ranks unless $self->end->has_rank;
985 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
986 my @all_pos = ( 1 .. $self->end->rank - 1 );
987 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
988 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
989 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
990 my @row = _make_witness_row( \@wit_path, \@all_pos );
991 push( @{$table->{'alignment'}},
992 { 'witness' => $wit->sigil, 'tokens' => \@row } );
993 if( $wit->is_layered ) {
994 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
995 $wit->sigil.$self->ac_label );
996 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
997 push( @{$table->{'alignment'}},
998 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1001 $self->cached_table( $table );
1005 sub _make_witness_row {
1006 my( $path, $positions ) = @_;
1008 map { $char_hash{$_} = undef } @$positions;
1010 foreach my $rdg ( @$path ) {
1011 my $rtext = $rdg->text;
1012 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1013 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1014 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1015 $char_hash{$rdg->rank} = { 't' => $rdg };
1017 my @row = map { $char_hash{$_} } @$positions;
1018 # Fill in lacuna markers for undef spots in the row
1019 my $last_el = shift @row;
1020 my @filled_row = ( $last_el );
1021 foreach my $el ( @row ) {
1022 # If we are using node reference, make the lacuna node appear many times
1023 # in the table. If not, use the lacuna tag.
1024 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1027 push( @filled_row, $el );
1033 =head1 NAVIGATION METHODS
1035 =head2 reading_sequence( $first, $last, $sigil, $backup )
1037 Returns the ordered list of readings, starting with $first and ending
1038 with $last, for the witness given in $sigil. If a $backup sigil is
1039 specified (e.g. when walking a layered witness), it will be used wherever
1040 no $sigil path exists. If there is a base text reading, that will be
1041 used wherever no path exists for $sigil or $backup.
1045 # TODO Think about returning some lazy-eval iterator.
1046 # TODO Get rid of backup; we should know from what witness is whether we need it.
1048 sub reading_sequence {
1049 my( $self, $start, $end, $witness ) = @_;
1051 $witness = $self->baselabel unless $witness;
1052 my @readings = ( $start );
1055 while( $n && $n->id ne $end->id ) {
1056 if( exists( $seen{$n->id} ) ) {
1057 throw( "Detected loop for $witness at " . $n->id );
1061 my $next = $self->next_reading( $n, $witness );
1063 throw( "Did not find any path for $witness from reading " . $n->id );
1065 push( @readings, $next );
1068 # Check that the last reading is our end reading.
1069 my $last = $readings[$#readings];
1070 throw( "Last reading found from " . $start->text .
1071 " for witness $witness is not the end!" ) # TODO do we get this far?
1072 unless $last->id eq $end->id;
1077 =head2 next_reading( $reading, $sigil );
1079 Returns the reading that follows the given reading along the given witness
1085 # Return the successor via the corresponding path.
1087 my $answer = $self->_find_linked_reading( 'next', @_ );
1088 return undef unless $answer;
1089 return $self->reading( $answer );
1092 =head2 prior_reading( $reading, $sigil )
1094 Returns the reading that precedes the given reading along the given witness
1100 # Return the predecessor via the corresponding path.
1102 my $answer = $self->_find_linked_reading( 'prior', @_ );
1103 return $self->reading( $answer );
1106 sub _find_linked_reading {
1107 my( $self, $direction, $node, $path ) = @_;
1109 # Get a backup if we are dealing with a layered witness
1111 my $aclabel = $self->ac_label;
1112 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1116 my @linked_paths = $direction eq 'next'
1117 ? $self->sequence->edges_from( $node )
1118 : $self->sequence->edges_to( $node );
1119 return undef unless scalar( @linked_paths );
1121 # We have to find the linked path that contains all of the
1122 # witnesses supplied in $path.
1123 my( @path_wits, @alt_path_wits );
1124 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1125 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1128 foreach my $le ( @linked_paths ) {
1129 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1132 my @le_wits = sort $self->path_witnesses( $le );
1133 if( _is_within( \@path_wits, \@le_wits ) ) {
1134 # This is the right path.
1135 return $direction eq 'next' ? $le->[1] : $le->[0];
1136 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1140 # Got this far? Return the alternate path if it exists.
1141 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1144 # Got this far? Return the base path if it exists.
1145 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1148 # Got this far? We have no appropriate path.
1149 warn "Could not find $direction node from " . $node->id
1150 . " along path $path";
1156 my( $set1, $set2 ) = @_;
1157 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1158 foreach my $el ( @$set1 ) {
1159 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1164 # Return the string that joins together a list of witnesses for
1165 # display on a single path.
1166 sub _witnesses_of_label {
1167 my( $self, $label ) = @_;
1168 my $regex = $self->wit_list_separator;
1169 my @answer = split( /\Q$regex\E/, $label );
1173 =head2 common_readings
1175 Returns the list of common readings in the graph (i.e. those readings that are
1176 shared by all non-lacunose witnesses.)
1180 sub common_readings {
1182 my @common = grep { $_->is_common } $self->readings;
1186 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1188 Returns the text of a witness (plus its backup, if we are using a layer)
1189 as stored in the collation. The text is returned as a string, where the
1190 individual readings are joined with spaces and the meta-readings (e.g.
1191 lacunae) are omitted. Optional specification of $start and $end allows
1192 the generation of a subset of the witness text.
1197 my( $self, $wit, $start, $end ) = @_;
1198 $start = $self->start unless $start;
1199 $end = $self->end unless $end;
1200 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1203 foreach my $r ( @path ) {
1204 if( $r->join_prior || !$last || $last->join_next ) {
1205 $pathtext .= $r->text;
1207 $pathtext .= ' ' . $r->text;
1214 =head1 INITIALIZATION METHODS
1216 These are mostly for use by parsers.
1218 =head2 make_witness_path( $witness )
1220 Link the array of readings contained in $witness->path (and in
1221 $witness->uncorrected_path if it exists) into collation paths.
1222 Clear out the arrays when finished.
1224 =head2 make_witness_paths
1226 Call make_witness_path for all witnesses in the tradition.
1230 # For use when a collation is constructed from a base text and an apparatus.
1231 # We have the sequences of readings and just need to add path edges.
1232 # When we are done, clear out the witness path attributes, as they are no
1234 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1236 sub make_witness_paths {
1238 foreach my $wit ( $self->tradition->witnesses ) {
1239 # print STDERR "Making path for " . $wit->sigil . "\n";
1240 $self->make_witness_path( $wit );
1244 sub make_witness_path {
1245 my( $self, $wit ) = @_;
1246 my @chain = @{$wit->path};
1247 my $sig = $wit->sigil;
1248 foreach my $idx ( 0 .. $#chain-1 ) {
1249 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1251 if( $wit->is_layered ) {
1252 @chain = @{$wit->uncorrected_path};
1253 foreach my $idx( 0 .. $#chain-1 ) {
1254 my $source = $chain[$idx];
1255 my $target = $chain[$idx+1];
1256 $self->add_path( $source, $target, $sig.$self->ac_label )
1257 unless $self->has_path( $source, $target, $sig );
1261 $wit->clear_uncorrected_path;
1264 =head2 calculate_ranks
1266 Calculate the reading ranks (that is, their aligned positions relative
1267 to each other) for the graph. This can only be called on linear collations.
1271 use Text::Tradition;
1273 my $cxfile = 't/data/Collatex-16.xml';
1274 my $t = Text::Tradition->new(
1276 'input' => 'CollateX',
1279 my $c = $t->collation;
1282 my $table = $c->alignment_table;
1283 ok( $c->has_cached_table, "Alignment table was cached" );
1284 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1285 $c->calculate_ranks;
1286 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1287 $c->add_relationship( 'n9', 'n23', { 'type' => 'spelling' } );
1288 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1294 sub calculate_ranks {
1296 # Save the existing ranks, in case we need to invalidate the cached SVG.
1298 # Walk a version of the graph where every node linked by a relationship
1299 # edge is fundamentally the same node, and do a topological ranking on
1300 # the nodes in this graph.
1301 my $topo_graph = Graph->new();
1305 foreach my $r ( $self->readings ) {
1306 next if exists $rel_containers{$r->id};
1307 my @rels = $r->related_readings( 'colocated' );
1309 # Make a relationship container.
1311 my $rn = 'rel_container_' . $rel_ctr++;
1312 $topo_graph->add_vertex( $rn );
1314 $rel_containers{$_->id} = $rn;
1317 # Add a new node to mirror the old node.
1318 $rel_containers{$r->id} = $r->id;
1319 $topo_graph->add_vertex( $r->id );
1324 foreach my $r ( $self->readings ) {
1325 $existing_ranks{$r} = $r->rank;
1326 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1327 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1328 $rel_containers{$n} );
1329 # $DB::single = 1 unless $tfrom && $tto;
1330 $topo_graph->add_edge( $tfrom, $tto );
1334 # Now do the rankings, starting with the start node.
1335 my $topo_start = $rel_containers{$self->start->id};
1336 my $node_ranks = { $topo_start => 0 };
1337 my @curr_origin = ( $topo_start );
1338 # A little iterative function.
1339 while( @curr_origin ) {
1340 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1342 # Transfer our rankings from the topological graph to the real one.
1343 foreach my $r ( $self->readings ) {
1344 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1345 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1347 # Die. Find the last rank we calculated.
1348 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1349 <=> $node_ranks->{$rel_containers{$b->id}} }
1351 my $last = pop @all_defined;
1352 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1355 # Do we need to invalidate the cached data?
1356 if( $self->has_cached_svg || $self->has_cached_table ) {
1357 foreach my $r ( $self->readings ) {
1358 next if $existing_ranks{$r} == $r->rank;
1359 # Something has changed, so clear the cache
1360 $self->_clear_cache;
1361 # ...and recalculate the common readings.
1362 $self->calculate_common_readings();
1366 # The graph calculation information is now up to date.
1367 $self->_graphcalc_done(1);
1371 my( $graph, $node_ranks, @current_nodes ) = @_;
1372 # Look at each of the children of @current_nodes. If all the child's
1373 # parents have a rank, assign it the highest rank + 1 and add it to
1374 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1375 # parent gets a rank.
1377 foreach my $c ( @current_nodes ) {
1378 warn "Current reading $c has no rank!"
1379 unless exists $node_ranks->{$c};
1380 # print STDERR "Looking at child of node $c, rank "
1381 # . $node_ranks->{$c} . "\n";
1382 foreach my $child ( $graph->successors( $c ) ) {
1383 next if exists $node_ranks->{$child};
1384 my $highest_rank = -1;
1386 foreach my $parent ( $graph->predecessors( $child ) ) {
1387 if( exists $node_ranks->{$parent} ) {
1388 $highest_rank = $node_ranks->{$parent}
1389 if $highest_rank <= $node_ranks->{$parent};
1396 my $c_rank = $highest_rank + 1;
1397 # print STDERR "Assigning rank $c_rank to node $child \n";
1398 $node_ranks->{$child} = $c_rank;
1399 push( @next_nodes, $child );
1407 $self->wipe_svg if $self->has_cached_svg;
1408 $self->wipe_table if $self->has_cached_table;
1412 =head2 flatten_ranks
1414 A convenience method for parsing collation data. Searches the graph for readings
1415 with the same text at the same rank, and merges any that are found.
1421 my %unique_rank_rdg;
1422 foreach my $rdg ( $self->readings ) {
1423 next unless $rdg->has_rank;
1424 my $key = $rdg->rank . "||" . $rdg->text;
1425 if( exists $unique_rank_rdg{$key} ) {
1427 # print STDERR "Combining readings at same rank: $key\n";
1428 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1429 # TODO see if this now makes a common point.
1431 $unique_rank_rdg{$key} = $rdg;
1437 =head2 calculate_common_readings
1439 Goes through the graph identifying the readings that appear in every witness
1440 (apart from those with lacunae at that spot.) Marks them as common and returns
1445 use Text::Tradition;
1447 my $cxfile = 't/data/Collatex-16.xml';
1448 my $t = Text::Tradition->new(
1450 'input' => 'CollateX',
1453 my $c = $t->collation;
1455 my @common = $c->calculate_common_readings();
1456 is( scalar @common, 8, "Found correct number of common readings" );
1457 my @marked = sort $c->common_readings();
1458 is( scalar @common, 8, "All common readings got marked as such" );
1459 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1460 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1466 sub calculate_common_readings {
1469 map { $_->is_common( 0 ) } $self->readings;
1470 # Implicitly calls calculate_ranks
1471 my $table = $self->alignment_table;
1472 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1473 my @row = map { $_->{'tokens'}->[$idx]
1474 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1475 @{$table->{'alignment'}};
1477 foreach my $r ( @row ) {
1479 $hash{$r->id} = $r unless $r->is_meta;
1481 $hash{'UNDEF'} = $r;
1484 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1485 my( $r ) = values %hash;
1487 push( @common, $r );
1493 =head2 text_from_paths
1495 Calculate the text array for all witnesses from the path, for later consistency
1496 checking. Only to be used if there is no non-graph-based way to know the
1501 sub text_from_paths {
1503 foreach my $wit ( $self->tradition->witnesses ) {
1504 my @text = split( /\s+/,
1505 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1506 $wit->text( \@text );
1507 if( $wit->is_layered ) {
1508 my @uctext = split( /\s+/,
1509 $self->reading_sequence( $self->start, $self->end,
1510 $wit->sigil.$self->ac_label ) );
1511 $wit->text( \@uctext );
1516 =head1 UTILITY FUNCTIONS
1518 =head2 common_predecessor( $reading_a, $reading_b )
1520 Find the last reading that occurs in sequence before both the given readings.
1522 =head2 common_successor( $reading_a, $reading_b )
1524 Find the first reading that occurs in sequence after both the given readings.
1528 use Text::Tradition;
1530 my $cxfile = 't/data/Collatex-16.xml';
1531 my $t = Text::Tradition->new(
1533 'input' => 'CollateX',
1536 my $c = $t->collation;
1538 is( $c->common_predecessor( 'n9', 'n23' )->id,
1539 'n20', "Found correct common predecessor" );
1540 is( $c->common_successor( 'n9', 'n23' )->id,
1541 '#END#', "Found correct common successor" );
1543 is( $c->common_predecessor( 'n19', 'n17' )->id,
1544 'n16', "Found correct common predecessor for readings on same path" );
1545 is( $c->common_successor( 'n21', 'n26' )->id,
1546 '#END#', "Found correct common successor for readings on same path" );
1552 ## Return the closest reading that is a predecessor of both the given readings.
1553 sub common_predecessor {
1555 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1556 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1559 sub common_successor {
1561 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1562 return $self->_common_in_path( $r1, $r2, 'successors' );
1565 sub _common_in_path {
1566 my( $self, $r1, $r2, $dir ) = @_;
1567 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1568 $iter = $self->end->rank - $iter if $dir eq 'successors';
1570 my @last_checked = ( $r1, $r2 );
1572 while( !@candidates ) {
1574 foreach my $lc ( @last_checked ) {
1575 foreach my $p ( $lc->$dir ) {
1576 if( $all_seen{$p->id} ) {
1577 push( @candidates, $p );
1579 $all_seen{$p->id} = 1;
1580 push( @new_lc, $p );
1584 @last_checked = @new_lc;
1586 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1587 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1591 Text::Tradition::Error->throw(
1592 'ident' => 'Collation error',
1598 __PACKAGE__->meta->make_immutable;
1602 This package is free software and is provided "as is" without express
1603 or implied warranty. You can redistribute it and/or modify it under
1604 the same terms as Perl itself.
1608 Tara L Andrews E<lt>aurum@cpan.orgE<gt>