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 return keys %all_witnesses;
461 =head1 OUTPUT METHODS
463 =head2 as_svg( \%options )
465 Returns an SVG string that represents the graph, via as_dot and graphviz.
466 See as_dot for a list of options.
471 my( $self, $opts ) = @_;
472 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
473 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
474 my @cmd = qw/dot -Tsvg/;
476 my $dotfile = File::Temp->new();
478 # $dotfile->unlink_on_destroy(0);
479 binmode $dotfile, ':utf8';
480 print $dotfile $self->as_dot( $opts );
481 push( @cmd, $dotfile->filename );
482 run( \@cmd, ">", binary(), \$svg );
483 $svg = decode_utf8( $svg );
484 $self->cached_svg( $svg ) unless $want_subgraph;
487 return $self->cached_svg;
492 =head2 as_dot( \%options )
494 Returns a string that is the collation graph expressed in dot
495 (i.e. GraphViz) format. Options include:
510 my( $self, $opts ) = @_;
511 my $startrank = $opts->{'from'} if $opts;
512 my $endrank = $opts->{'to'} if $opts;
513 my $color_common = $opts->{'color_common'} if $opts;
514 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
515 && $self->end->rank > 100;
517 # Check the arguments
519 return if $endrank && $startrank > $endrank;
520 return if $startrank > $self->end->rank;
522 if( defined $endrank ) {
523 return if $endrank < 0;
524 $endrank = undef if $endrank == $self->end->rank;
527 my $graph_name = $self->tradition->name;
528 $graph_name =~ s/[^\w\s]//g;
529 $graph_name = join( '_', split( /\s+/, $graph_name ) );
537 'fillcolor' => 'white',
542 'arrowhead' => 'open',
543 'color' => '#000000',
544 'fontcolor' => '#000000',
547 my $dot = sprintf( "digraph %s {\n", $graph_name );
548 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
549 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
551 # Output substitute start/end readings if necessary
553 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
556 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
558 if( $STRAIGHTENHACK ) {
560 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
561 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
563 my %used; # Keep track of the readings that actually appear in the graph
564 # Sort the readings by rank if we have ranks; this speeds layout.
565 my @all_readings = $self->end->has_rank
566 ? sort { $a->rank <=> $b->rank } $self->readings
568 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
569 foreach my $reading ( @all_readings ) {
570 # Only output readings within our rank range.
571 next if $startrank && $reading->rank < $startrank;
572 next if $endrank && $reading->rank > $endrank;
573 $used{$reading->id} = 1;
574 # Need not output nodes without separate labels
575 next if $reading->id eq $reading->text;
577 my $label = $reading->text;
578 $label =~ s/\"/\\\"/g;
579 $rattrs->{'label'} = $label;
580 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
581 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
584 # Add the real edges. Need to weight one edge per rank jump, in a
586 # my $weighted = $self->_add_edge_weights;
587 my @edges = $self->paths;
588 my( %substart, %subend );
589 foreach my $edge ( @edges ) {
590 # Do we need to output this edge?
591 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
592 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
593 my $variables = { %edge_attrs, 'label' => $label };
595 # Account for the rank gap if necessary
596 my $rank0 = $self->reading( $edge->[0] )->rank
597 if $self->reading( $edge->[0] )->has_rank;
598 my $rank1 = $self->reading( $edge->[1] )->rank
599 if $self->reading( $edge->[1] )->has_rank;
600 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
601 $variables->{'minlen'} = $rank1 - $rank0;
604 # Add the calculated edge weights
605 # if( exists $weighted->{$edge->[0]}
606 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
607 # # $variables->{'color'} = 'red';
608 # $variables->{'weight'} = 3.0;
611 # EXPERIMENTAL: make edge width reflect no. of witnesses
612 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
613 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
615 my $varopts = _dot_attr_string( $variables );
616 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
617 $edge->[0], $edge->[1], $varopts );
618 } elsif( $used{$edge->[0]} ) {
619 $subend{$edge->[0]} = 1;
620 } elsif( $used{$edge->[1]} ) {
621 $substart{$edge->[1]} = 1;
624 # Add substitute start and end edges if necessary
625 foreach my $node ( keys %substart ) {
626 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
627 my $variables = { %edge_attrs, 'label' => $witstr };
628 my $varopts = _dot_attr_string( $variables );
629 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
631 foreach my $node ( keys %subend ) {
632 my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
633 my $variables = { %edge_attrs, 'label' => $witstr };
634 my $varopts = _dot_attr_string( $variables );
635 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
638 if( $STRAIGHTENHACK ) {
639 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
646 sub _dot_attr_string {
649 foreach my $k ( sort keys %$hash ) {
651 push( @attrs, $k.'="'.$v.'"' );
653 return( '[ ' . join( ', ', @attrs ) . ' ]' );
656 sub _add_edge_weights {
658 # Walk the graph from START to END, choosing the successor node with
659 # the largest number of witness paths each time.
661 my $curr = $self->start->id;
662 my $ranked = $self->end->has_rank;
663 while( $curr ne $self->end->id ) {
664 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
665 my @succ = sort { $self->path_witnesses( $curr, $a )
666 <=> $self->path_witnesses( $curr, $b ) }
667 $self->sequence->successors( $curr );
668 my $next = pop @succ;
669 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
670 # Try to avoid lacunae in the weighted path.
672 ( $self->reading( $next )->is_lacuna ||
673 $nextrank - $rank > 1 ) ){
676 $weighted->{$curr} = $next;
682 =head2 path_witnesses( $edge )
684 Returns the list of sigils whose witnesses are associated with the given edge.
685 The edge can be passed as either an array or an arrayref of ( $source, $target ).
690 my( $self, @edge ) = @_;
691 # If edge is an arrayref, cope.
692 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
696 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
700 sub _path_display_label {
703 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
704 if( scalar @wits > $maj ) {
705 # TODO break out a.c. wits
708 return join( ', ', @wits );
712 =head2 witnesses_at_rank
714 Returns a list of witnesses that are not lacunose, for a given rank.
718 sub witnesses_at_rank {
719 my( $self, $rank ) = @_;
724 Returns a GraphML representation of the collation. The GraphML will contain
725 two graphs. The first expresses the attributes of the readings and the witness
726 paths that link them; the second expresses the relationships that link the
727 readings. This is the native transfer format for a tradition.
736 my $datafile = 't/data/florilegium_tei_ps.xml';
737 my $tradition = Text::Tradition->new( 'input' => 'TEI',
742 ok( $tradition, "Got a tradition object" );
743 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
744 ok( $tradition->collation, "Tradition has a collation" );
746 my $c = $tradition->collation;
747 is( scalar $c->readings, $READINGS, "Collation has all readings" );
748 is( scalar $c->paths, $PATHS, "Collation has all paths" );
749 is( scalar $c->relationships, 0, "Collation has all relationships" );
751 # Add a few relationships
752 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
753 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
754 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
756 # Now write it to GraphML and parse it again.
758 my $graphml = $c->as_graphml;
759 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
760 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
761 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
762 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
772 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
773 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
774 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
775 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
777 # Create the document and root node
778 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
779 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
780 $graphml->setDocumentElement( $root );
781 $root->setNamespace( $xsi_ns, 'xsi', 0 );
782 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
784 # Add the data keys for the graph
787 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
788 foreach my $datum ( @graph_attributes ) {
789 $graph_data_keys{$datum} = 'dg'.$gdi++;
790 my $key = $root->addNewChild( $graphml_ns, 'key' );
791 $key->setAttribute( 'attr.name', $datum );
792 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
793 $key->setAttribute( 'for', 'graph' );
794 $key->setAttribute( 'id', $graph_data_keys{$datum} );
797 # Add the data keys for nodes
804 is_start => 'boolean',
806 is_lacuna => 'boolean',
808 foreach my $datum ( keys %node_data ) {
809 $node_data_keys{$datum} = 'dn'.$ndi++;
810 my $key = $root->addNewChild( $graphml_ns, 'key' );
811 $key->setAttribute( 'attr.name', $datum );
812 $key->setAttribute( 'attr.type', $node_data{$datum} );
813 $key->setAttribute( 'for', 'node' );
814 $key->setAttribute( 'id', $node_data_keys{$datum} );
817 # Add the data keys for edges, i.e. witnesses
821 class => 'string', # Class, deprecated soon
822 witness => 'string', # ID/label for a path
823 relationship => 'string', # ID/label for a relationship
824 extra => 'boolean', # Path key
825 scope => 'string', # Relationship key
826 non_correctable => 'boolean', # Relationship key
827 non_independent => 'boolean', # Relationship key
829 foreach my $datum ( keys %edge_data ) {
830 $edge_data_keys{$datum} = 'de'.$edi++;
831 my $key = $root->addNewChild( $graphml_ns, 'key' );
832 $key->setAttribute( 'attr.name', $datum );
833 $key->setAttribute( 'attr.type', $edge_data{$datum} );
834 $key->setAttribute( 'for', 'edge' );
835 $key->setAttribute( 'id', $edge_data_keys{$datum} );
838 # Add the collation graph itself
839 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
840 $sgraph->setAttribute( 'edgedefault', 'directed' );
841 $sgraph->setAttribute( 'id', $self->tradition->name );
842 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
843 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
844 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
845 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
846 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
848 # Collation attribute data
849 foreach my $datum ( @graph_attributes ) {
850 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
851 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
856 # Add our readings to the graph
857 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
858 # Add to the main graph
859 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
860 my $node_xmlid = 'n' . $node_ctr++;
861 $node_hash{ $n->id } = $node_xmlid;
862 $node_el->setAttribute( 'id', $node_xmlid );
863 foreach my $d ( keys %node_data ) {
865 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
870 # Add the path edges to the sequence graph
872 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
873 # We add an edge in the graphml for every witness in $e.
874 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
875 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
876 $node_hash{ $e->[0] },
877 $node_hash{ $e->[1] } );
878 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
879 $edge_el->setAttribute( 'source', $from );
880 $edge_el->setAttribute( 'target', $to );
881 $edge_el->setAttribute( 'id', $id );
883 # It's a witness path, so add the witness
885 my $key = $edge_data_keys{'witness'};
886 # Is this an ante-corr witness?
887 my $aclabel = $self->ac_label;
888 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
889 # Keep the base witness
891 # ...and record that this is an 'extra' reading path
892 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
894 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
895 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
899 # Add the relationship graph to the XML
900 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
901 $node_data_keys{'id'}, \%edge_data_keys );
903 # Save and return the thing
904 my $result = decode_utf8( $graphml->toString(1) );
908 sub _add_graphml_data {
909 my( $el, $key, $value ) = @_;
910 return unless defined $value;
911 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
912 $data_el->setAttribute( 'key', $key );
913 $data_el->appendText( $value );
918 Returns a CSV alignment table representation of the collation graph, one
919 row per witness (or witness uncorrected.)
925 my $table = $self->alignment_table;
926 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
928 # Make the header row
929 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
930 push( @result, decode_utf8( $csv->string ) );
931 # Make the rest of the rows
932 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
933 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
934 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
935 $csv->combine( @row );
936 push( @result, decode_utf8( $csv->string ) );
938 return join( "\n", @result );
941 =head2 alignment_table( $use_refs, $include_witnesses )
943 Return a reference to an alignment table, in a slightly enhanced CollateX
944 format which looks like this:
946 $table = { alignment => [ { witness => "SIGIL",
947 tokens => [ { t => "TEXT" }, ... ] },
949 tokens => [ { t => "TEXT" }, ... ] },
953 If $use_refs is set to 1, the reading object is returned in the table
954 instead of READINGTEXT; if not, the text of the reading is returned.
956 If $include_witnesses is set to a hashref, only the witnesses whose sigil
957 keys have a true hash value will be included.
961 sub alignment_table {
963 my $include; # see if we can ditch this
965 return $self->cached_table if $self->has_cached_table;
967 # Make sure we can do this
968 throw( "Need a linear graph in order to make an alignment table" )
969 unless $self->linear;
970 $self->calculate_ranks unless $self->end->has_rank;
972 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
973 my @all_pos = ( 1 .. $self->end->rank - 1 );
974 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
976 next unless $include->{$wit->sigil};
978 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
979 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
980 my @row = _make_witness_row( \@wit_path, \@all_pos );
981 push( @{$table->{'alignment'}},
982 { 'witness' => $wit->sigil, 'tokens' => \@row } );
983 if( $wit->is_layered ) {
984 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
985 $wit->sigil.$self->ac_label );
986 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
987 push( @{$table->{'alignment'}},
988 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
991 $self->cached_table( $table );
995 sub _make_witness_row {
996 my( $path, $positions ) = @_;
998 map { $char_hash{$_} = undef } @$positions;
1000 foreach my $rdg ( @$path ) {
1001 my $rtext = $rdg->text;
1002 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1003 print STDERR "rank " . $rdg->rank . "\n" if $debug;
1004 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
1005 $char_hash{$rdg->rank} = { 't' => $rdg };
1007 my @row = map { $char_hash{$_} } @$positions;
1008 # Fill in lacuna markers for undef spots in the row
1009 my $last_el = shift @row;
1010 my @filled_row = ( $last_el );
1011 foreach my $el ( @row ) {
1012 # If we are using node reference, make the lacuna node appear many times
1013 # in the table. If not, use the lacuna tag.
1014 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1017 push( @filled_row, $el );
1023 =head1 NAVIGATION METHODS
1025 =head2 reading_sequence( $first, $last, $sigil, $backup )
1027 Returns the ordered list of readings, starting with $first and ending
1028 with $last, for the witness given in $sigil. If a $backup sigil is
1029 specified (e.g. when walking a layered witness), it will be used wherever
1030 no $sigil path exists. If there is a base text reading, that will be
1031 used wherever no path exists for $sigil or $backup.
1035 # TODO Think about returning some lazy-eval iterator.
1036 # TODO Get rid of backup; we should know from what witness is whether we need it.
1038 sub reading_sequence {
1039 my( $self, $start, $end, $witness ) = @_;
1041 $witness = $self->baselabel unless $witness;
1042 my @readings = ( $start );
1045 while( $n && $n->id ne $end->id ) {
1046 if( exists( $seen{$n->id} ) ) {
1047 throw( "Detected loop for $witness at " . $n->id );
1051 my $next = $self->next_reading( $n, $witness );
1053 throw( "Did not find any path for $witness from reading " . $n->id );
1055 push( @readings, $next );
1058 # Check that the last reading is our end reading.
1059 my $last = $readings[$#readings];
1060 throw( "Last reading found from " . $start->text .
1061 " for witness $witness is not the end!" ) # TODO do we get this far?
1062 unless $last->id eq $end->id;
1067 =head2 next_reading( $reading, $sigil );
1069 Returns the reading that follows the given reading along the given witness
1075 # Return the successor via the corresponding path.
1077 my $answer = $self->_find_linked_reading( 'next', @_ );
1078 return undef unless $answer;
1079 return $self->reading( $answer );
1082 =head2 prior_reading( $reading, $sigil )
1084 Returns the reading that precedes the given reading along the given witness
1090 # Return the predecessor via the corresponding path.
1092 my $answer = $self->_find_linked_reading( 'prior', @_ );
1093 return $self->reading( $answer );
1096 sub _find_linked_reading {
1097 my( $self, $direction, $node, $path ) = @_;
1099 # Get a backup if we are dealing with a layered witness
1101 my $aclabel = $self->ac_label;
1102 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1106 my @linked_paths = $direction eq 'next'
1107 ? $self->sequence->edges_from( $node )
1108 : $self->sequence->edges_to( $node );
1109 return undef unless scalar( @linked_paths );
1111 # We have to find the linked path that contains all of the
1112 # witnesses supplied in $path.
1113 my( @path_wits, @alt_path_wits );
1114 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1115 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1118 foreach my $le ( @linked_paths ) {
1119 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1122 my @le_wits = sort $self->path_witnesses( $le );
1123 if( _is_within( \@path_wits, \@le_wits ) ) {
1124 # This is the right path.
1125 return $direction eq 'next' ? $le->[1] : $le->[0];
1126 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1130 # Got this far? Return the alternate path if it exists.
1131 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1134 # Got this far? Return the base path if it exists.
1135 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1138 # Got this far? We have no appropriate path.
1139 warn "Could not find $direction node from " . $node->id
1140 . " along path $path";
1146 my( $set1, $set2 ) = @_;
1147 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1148 foreach my $el ( @$set1 ) {
1149 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1154 # Return the string that joins together a list of witnesses for
1155 # display on a single path.
1156 sub _witnesses_of_label {
1157 my( $self, $label ) = @_;
1158 my $regex = $self->wit_list_separator;
1159 my @answer = split( /\Q$regex\E/, $label );
1163 =head2 common_readings
1165 Returns the list of common readings in the graph (i.e. those readings that are
1166 shared by all non-lacunose witnesses.)
1170 sub common_readings {
1172 my @common = grep { $_->is_common } $self->readings;
1176 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1178 Returns the text of a witness (plus its backup, if we are using a layer)
1179 as stored in the collation. The text is returned as a string, where the
1180 individual readings are joined with spaces and the meta-readings (e.g.
1181 lacunae) are omitted. Optional specification of $start and $end allows
1182 the generation of a subset of the witness text.
1187 my( $self, $wit, $start, $end ) = @_;
1188 $start = $self->start unless $start;
1189 $end = $self->end unless $end;
1190 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1191 return join( ' ', map { $_->text } @path );
1194 =head1 INITIALIZATION METHODS
1196 These are mostly for use by parsers.
1198 =head2 make_witness_path( $witness )
1200 Link the array of readings contained in $witness->path (and in
1201 $witness->uncorrected_path if it exists) into collation paths.
1202 Clear out the arrays when finished.
1204 =head2 make_witness_paths
1206 Call make_witness_path for all witnesses in the tradition.
1210 # For use when a collation is constructed from a base text and an apparatus.
1211 # We have the sequences of readings and just need to add path edges.
1212 # When we are done, clear out the witness path attributes, as they are no
1214 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1216 sub make_witness_paths {
1218 foreach my $wit ( $self->tradition->witnesses ) {
1219 # print STDERR "Making path for " . $wit->sigil . "\n";
1220 $self->make_witness_path( $wit );
1224 sub make_witness_path {
1225 my( $self, $wit ) = @_;
1226 my @chain = @{$wit->path};
1227 my $sig = $wit->sigil;
1228 foreach my $idx ( 0 .. $#chain-1 ) {
1229 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1231 if( $wit->is_layered ) {
1232 @chain = @{$wit->uncorrected_path};
1233 foreach my $idx( 0 .. $#chain-1 ) {
1234 my $source = $chain[$idx];
1235 my $target = $chain[$idx+1];
1236 $self->add_path( $source, $target, $sig.$self->ac_label )
1237 unless $self->has_path( $source, $target, $sig );
1241 $wit->clear_uncorrected_path;
1244 =head2 calculate_ranks
1246 Calculate the reading ranks (that is, their aligned positions relative
1247 to each other) for the graph. This can only be called on linear collations.
1251 use Text::Tradition;
1253 my $cxfile = 't/data/Collatex-16.xml';
1254 my $t = Text::Tradition->new(
1256 'input' => 'CollateX',
1259 my $c = $t->collation;
1262 my $svg = $c->as_svg;
1263 is( substr( $svg, 0, 5 ), '<?xml', "Got XML doc for svg" );
1264 ok( $c->has_cached_svg, "SVG was cached" );
1265 is( $c->as_svg, $svg, "Cached SVG returned upon second call" );
1266 $c->calculate_ranks;
1267 is( $c->as_svg, $svg, "Cached SVG retained with no rank change" );
1268 $c->add_relationship( 'n9', 'n23', { 'type' => 'spelling' } );
1269 isnt( $c->as_svg, $svg, "SVG changed after relationship add" );
1275 sub calculate_ranks {
1277 # Save the existing ranks, in case we need to invalidate the cached SVG.
1279 # Walk a version of the graph where every node linked by a relationship
1280 # edge is fundamentally the same node, and do a topological ranking on
1281 # the nodes in this graph.
1282 my $topo_graph = Graph->new();
1286 foreach my $r ( $self->readings ) {
1287 next if exists $rel_containers{$r->id};
1288 my @rels = $r->related_readings( 'colocated' );
1290 # Make a relationship container.
1292 my $rn = 'rel_container_' . $rel_ctr++;
1293 $topo_graph->add_vertex( $rn );
1295 $rel_containers{$_->id} = $rn;
1298 # Add a new node to mirror the old node.
1299 $rel_containers{$r->id} = $r->id;
1300 $topo_graph->add_vertex( $r->id );
1305 foreach my $r ( $self->readings ) {
1306 $existing_ranks{$r} = $r->rank;
1307 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1308 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1309 $rel_containers{$n} );
1310 # $DB::single = 1 unless $tfrom && $tto;
1311 $topo_graph->add_edge( $tfrom, $tto );
1315 # Now do the rankings, starting with the start node.
1316 my $topo_start = $rel_containers{$self->start->id};
1317 my $node_ranks = { $topo_start => 0 };
1318 my @curr_origin = ( $topo_start );
1319 # A little iterative function.
1320 while( @curr_origin ) {
1321 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1323 # Transfer our rankings from the topological graph to the real one.
1324 foreach my $r ( $self->readings ) {
1325 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1326 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1328 # Die. Find the last rank we calculated.
1329 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1330 <=> $node_ranks->{$rel_containers{$b->id}} }
1332 my $last = pop @all_defined;
1333 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1336 # Do we need to invalidate the cached SVG?
1337 if( $self->has_cached_svg ) {
1338 foreach my $r ( $self->readings ) {
1339 next if $existing_ranks{$r} == $r->rank;
1347 my( $graph, $node_ranks, @current_nodes ) = @_;
1348 # Look at each of the children of @current_nodes. If all the child's
1349 # parents have a rank, assign it the highest rank + 1 and add it to
1350 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1351 # parent gets a rank.
1353 foreach my $c ( @current_nodes ) {
1354 warn "Current reading $c has no rank!"
1355 unless exists $node_ranks->{$c};
1356 # print STDERR "Looking at child of node $c, rank "
1357 # . $node_ranks->{$c} . "\n";
1358 foreach my $child ( $graph->successors( $c ) ) {
1359 next if exists $node_ranks->{$child};
1360 my $highest_rank = -1;
1362 foreach my $parent ( $graph->predecessors( $child ) ) {
1363 if( exists $node_ranks->{$parent} ) {
1364 $highest_rank = $node_ranks->{$parent}
1365 if $highest_rank <= $node_ranks->{$parent};
1372 my $c_rank = $highest_rank + 1;
1373 # print STDERR "Assigning rank $c_rank to node $child \n";
1374 $node_ranks->{$child} = $c_rank;
1375 push( @next_nodes, $child );
1381 =head2 flatten_ranks
1383 A convenience method for parsing collation data. Searches the graph for readings
1384 with the same text at the same rank, and merges any that are found.
1390 my %unique_rank_rdg;
1391 foreach my $rdg ( $self->readings ) {
1392 next unless $rdg->has_rank;
1393 my $key = $rdg->rank . "||" . $rdg->text;
1394 if( exists $unique_rank_rdg{$key} ) {
1396 # print STDERR "Combining readings at same rank: $key\n";
1397 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1398 # TODO see if this now makes a common point.
1400 $unique_rank_rdg{$key} = $rdg;
1405 =head2 remove_collations
1407 Another convenience method for parsing. Removes all 'collation' relationships
1408 that were defined in order to get the reading ranks to be correct.
1412 use Text::Tradition;
1414 my $cxfile = 't/data/Collatex-16.xml';
1415 my $t = Text::Tradition->new(
1417 'input' => 'CollateX',
1420 my $c = $t->collation;
1422 isnt( $c->reading('n23')->rank, $c->reading('n9')->rank, "Rank skew exists" );
1423 $c->add_relationship( 'n23', 'n9', { 'type' => 'collated', 'scope' => 'local' } );
1424 is( scalar $c->relationships, 4, "Found all expected relationships" );
1425 $c->remove_collations;
1426 is( scalar $c->relationships, 3, "Collated relationships now gone" );
1427 is( $c->reading('n23')->rank, $c->reading('n9')->rank, "Aligned ranks were preserved" );
1433 sub remove_collations {
1435 foreach my $reledge ( $self->relationships ) {
1436 my $relobj = $self->relations->get_relationship( $reledge );
1437 if( $relobj && $relobj->type eq 'collated' ) {
1438 $self->relations->delete_relationship( $reledge );
1444 =head2 calculate_common_readings
1446 Goes through the graph identifying the readings that appear in every witness
1447 (apart from those with lacunae at that spot.) Marks them as common and returns
1452 use Text::Tradition;
1454 my $cxfile = 't/data/Collatex-16.xml';
1455 my $t = Text::Tradition->new(
1457 'input' => 'CollateX',
1460 my $c = $t->collation;
1462 my @common = $c->calculate_common_readings();
1463 is( scalar @common, 8, "Found correct number of common readings" );
1464 my @marked = sort $c->common_readings();
1465 is( scalar @common, 8, "All common readings got marked as such" );
1466 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1467 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1473 sub calculate_common_readings {
1476 my $table = $self->alignment_table;
1477 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1478 my @row = map { $_->{'tokens'}->[$idx]
1479 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1480 @{$table->{'alignment'}};
1482 foreach my $r ( @row ) {
1484 $hash{$r->id} = $r unless $r->is_meta;
1486 $hash{'UNDEF'} = $r;
1489 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1490 my( $r ) = values %hash;
1492 push( @common, $r );
1498 =head2 text_from_paths
1500 Calculate the text array for all witnesses from the path, for later consistency
1501 checking. Only to be used if there is no non-graph-based way to know the
1506 sub text_from_paths {
1508 foreach my $wit ( $self->tradition->witnesses ) {
1509 my @text = split( /\s+/,
1510 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1511 $wit->text( \@text );
1512 if( $wit->is_layered ) {
1513 my @uctext = split( /\s+/,
1514 $self->reading_sequence( $self->start, $self->end,
1515 $wit->sigil.$self->ac_label ) );
1516 $wit->text( \@uctext );
1521 =head1 UTILITY FUNCTIONS
1523 =head2 common_predecessor( $reading_a, $reading_b )
1525 Find the last reading that occurs in sequence before both the given readings.
1527 =head2 common_successor( $reading_a, $reading_b )
1529 Find the first reading that occurs in sequence after both the given readings.
1533 use Text::Tradition;
1535 my $cxfile = 't/data/Collatex-16.xml';
1536 my $t = Text::Tradition->new(
1538 'input' => 'CollateX',
1541 my $c = $t->collation;
1543 is( $c->common_predecessor( 'n9', 'n23' )->id,
1544 'n20', "Found correct common predecessor" );
1545 is( $c->common_successor( 'n9', 'n23' )->id,
1546 '#END#', "Found correct common successor" );
1548 is( $c->common_predecessor( 'n19', 'n17' )->id,
1549 'n16', "Found correct common predecessor for readings on same path" );
1550 is( $c->common_successor( 'n21', 'n26' )->id,
1551 '#END#', "Found correct common successor for readings on same path" );
1557 ## Return the closest reading that is a predecessor of both the given readings.
1558 sub common_predecessor {
1560 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1561 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1564 sub common_successor {
1566 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1567 return $self->_common_in_path( $r1, $r2, 'successors' );
1570 sub _common_in_path {
1571 my( $self, $r1, $r2, $dir ) = @_;
1572 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1573 $iter = $self->end->rank - $iter if $dir eq 'successors';
1575 my @last_checked = ( $r1, $r2 );
1577 while( !@candidates ) {
1579 foreach my $lc ( @last_checked ) {
1580 foreach my $p ( $lc->$dir ) {
1581 if( $all_seen{$p->id} ) {
1582 push( @candidates, $p );
1584 $all_seen{$p->id} = 1;
1585 push( @new_lc, $p );
1589 @last_checked = @new_lc;
1591 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1592 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1596 Text::Tradition::Error->throw(
1597 'ident' => 'Collation error',
1603 __PACKAGE__->meta->make_immutable;
1607 This package is free software and is provided "as is" without express
1608 or implied warranty. You can redistribute it and/or modify it under
1609 the same terms as Perl itself.
1613 Tara L Andrews E<lt>aurum@cpan.orgE<gt>