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',
93 Text::Tradition::Collation - a software model for a text collation
98 my $t = Text::Tradition->new(
99 'name' => 'this is a text',
101 'file' => '/path/to/tei_parallel_seg_file.xml' );
103 my $c = $t->collation;
104 my @readings = $c->readings;
105 my @paths = $c->paths;
106 my @relationships = $c->relationships;
108 my $svg_variant_graph = $t->collation->as_svg();
112 Text::Tradition is a library for representation and analysis of collated
113 texts, particularly medieval ones. The Collation is the central feature of
114 a Tradition, where the text, its sequence of readings, and its relationships
115 between readings are actually kept.
121 The constructor. Takes a hash or hashref of the following arguments:
125 =item * tradition - The Text::Tradition object to which the collation
128 =item * linear - Whether the collation should be linear; that is, whether
129 transposed readings should be treated as two linked readings rather than one,
130 and therefore whether the collation graph is acyclic. Defaults to true.
132 =item * baselabel - The default label for the path taken by a base text
133 (if any). Defaults to 'base text'.
135 =item * wit_list_separator - The string to join a list of witnesses for
136 purposes of making labels in display graphs. Defaults to ', '.
138 =item * ac_label - The extra label to tack onto a witness sigil when
139 representing another layer of path for the given witness - that is, when
140 a text has more than one possible reading due to scribal corrections or
141 the like. Defaults to ' (a.c.)'.
151 =head2 wit_list_separator
157 Simple accessors for collation attributes.
161 The meta-reading at the start of every witness path.
165 The meta-reading at the end of every witness path.
169 Returns all Reading objects in the graph.
171 =head2 reading( $id )
173 Returns the Reading object corresponding to the given ID.
175 =head2 add_reading( $reading_args )
177 Adds a new reading object to the collation.
178 See L<Text::Tradition::Collation::Reading> for the available arguments.
180 =head2 del_reading( $object_or_id )
182 Removes the given reading from the collation, implicitly removing its
183 paths and relationships.
185 =head2 merge_readings( $main, $second )
187 Merges the $second reading into the $main one.
188 The arguments may be either readings or reading IDs.
190 =head2 has_reading( $id )
192 Predicate to see whether a given reading ID is in the graph.
194 =head2 reading_witnesses( $object_or_id )
196 Returns a list of sigils whose witnesses contain the reading.
200 Returns all reading paths within the document - that is, all edges in the
201 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
203 =head2 add_path( $source, $target, $sigil )
205 Links the given readings in the collation in sequence, under the given witness
206 sigil. The readings may be specified by object or ID.
208 =head2 del_path( $source, $target, $sigil )
210 Links the given readings in the collation in sequence, under the given witness
211 sigil. The readings may be specified by object or ID.
213 =head2 has_path( $source, $target );
215 Returns true if the two readings are linked in sequence in any witness.
216 The readings may be specified by object or ID.
220 Returns all Relationship objects in the collation.
222 =head2 add_relationship( $reading, $other_reading, $options )
224 Adds a new relationship of the type given in $options between the two readings,
225 which may be specified by object or ID. Returns a value of ( $status, @vectors)
226 where $status is true on success, and @vectors is a list of relationship edges
227 that were ultimately added.
228 See L<Text::Tradition::Collation::Relationship> for the available options.
234 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
235 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
236 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
239 ### Reading construct/destruct functions
242 my( $self, $reading ) = @_;
243 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
244 my %args = %$reading;
245 $reading = Text::Tradition::Collation::Reading->new(
246 'collation' => $self,
249 # First check to see if a reading with this ID exists.
250 if( $self->reading( $reading->id ) ) {
251 throw( "Collation already has a reading with id " . $reading->id );
253 $self->_add_reading( $reading->id => $reading );
254 # Once the reading has been added, put it in both graphs.
255 $self->sequence->add_vertex( $reading->id );
256 $self->relations->add_reading( $reading->id );
260 around del_reading => sub {
265 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
268 # Remove the reading from the graphs.
269 $self->sequence->delete_vertex( $arg );
270 $self->relations->delete_reading( $arg );
273 $self->$orig( $arg );
276 # merge_readings( $main, $to_be_deleted );
281 # We only need the IDs for adding paths to the graph, not the reading
282 # objects themselves.
283 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
285 # The kept reading should inherit the paths and the relationships
286 # of the deleted reading.
287 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
288 my @vector = ( $kept );
289 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
290 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
291 next if $vector[0] eq $vector[1]; # Don't add a self loop
292 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
293 $self->sequence->add_edge( @vector );
294 my $fwits = $self->sequence->get_edge_attributes( @vector );
295 @wits{keys %$fwits} = values %$fwits;
296 $self->sequence->set_edge_attributes( @vector, \%wits );
298 $self->relations->merge_readings( $kept, $deleted, $combine_char );
300 # Do the deletion deed.
301 if( $combine_char ) {
302 my $kept_obj = $self->reading( $kept );
303 my $new_text = join( $combine_char, $kept_obj->text,
304 $self->reading( $deleted )->text );
305 $kept_obj->alter_text( $new_text );
307 $self->del_reading( $deleted );
311 # Helper function for manipulating the graph.
312 sub _stringify_args {
313 my( $self, $first, $second, $arg ) = @_;
315 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
316 $second = $second->id
317 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
318 return( $first, $second, $arg );
321 # Helper function for manipulating the graph.
322 sub _objectify_args {
323 my( $self, $first, $second, $arg ) = @_;
324 $first = $self->reading( $first )
325 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
326 $second = $self->reading( $second )
327 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
328 return( $first, $second, $arg );
335 # We only need the IDs for adding paths to the graph, not the reading
336 # objects themselves.
337 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
339 # Connect the readings
340 $self->sequence->add_edge( $source, $target );
341 # Note the witness in question
342 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
348 if( ref( $_[0] ) eq 'ARRAY' ) {
355 # We only need the IDs for adding paths to the graph, not the reading
356 # objects themselves.
357 my( $source, $target, $wit ) = $self->_stringify_args( @args );
359 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
360 $self->sequence->delete_edge_attribute( $source, $target, $wit );
362 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
363 $self->sequence->delete_edge( $source, $target );
368 # Extra graph-alike utility
371 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
372 return undef unless $self->sequence->has_edge( $source, $target );
373 return $self->sequence->has_edge_attribute( $source, $target, $wit );
376 =head2 clear_witness( @sigil_list )
378 Clear the given witnesses out of the collation entirely, removing references
379 to them in paths, and removing readings that belong only to them. Should only
380 be called via $tradition->del_witness.
385 my( $self, @sigils ) = @_;
387 # Clear the witness(es) out of the paths
388 foreach my $e ( $self->paths ) {
389 foreach my $sig ( @sigils ) {
390 $self->del_path( $e, $sig );
394 # Clear out the newly unused readings
395 foreach my $r ( $self->readings ) {
396 unless( $self->reading_witnesses( $r ) ) {
397 $self->del_reading( $r );
402 sub add_relationship {
404 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
405 my( @vectors ) = $self->relations->add_relationship( $source,
406 $self->reading( $source ), $target, $self->reading( $target ), $opts );
407 # Force a full rank recalculation every time. Yuck.
408 $self->calculate_ranks() if $self->end->has_rank;
412 =head2 reading_witnesses( $reading )
414 Return a list of sigils corresponding to the witnesses in which the reading appears.
418 sub reading_witnesses {
419 my( $self, $reading ) = @_;
420 # We need only check either the incoming or the outgoing edges; I have
421 # arbitrarily chosen "incoming". Thus, special-case the start node.
422 if( $reading eq $self->start ) {
423 return map { $_->sigil } $self->tradition->witnesses;
426 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
427 my $wits = $self->sequence->get_edge_attributes( @$e );
428 @all_witnesses{ keys %$wits } = 1;
430 return keys %all_witnesses;
433 =head1 OUTPUT METHODS
435 =head2 as_svg( \%options )
437 Returns an SVG string that represents the graph, via as_dot and graphviz.
438 See as_dot for a list of options.
443 my( $self, $opts ) = @_;
445 my @cmd = qw/dot -Tsvg/;
447 my $dotfile = File::Temp->new();
449 # $dotfile->unlink_on_destroy(0);
450 binmode $dotfile, ':utf8';
451 print $dotfile $self->as_dot( $opts );
452 push( @cmd, $dotfile->filename );
453 run( \@cmd, ">", binary(), \$svg );
454 # HACK part 3 - remove silent node+edge
455 my $parser = XML::LibXML->new();
456 my $svgdom = $parser->parse_string( $svg );
457 my $xpc = XML::LibXML::XPathContext->new( $svgdom->documentElement );
458 $xpc->registerNs( 'svg', 'http://www.w3.org/2000/svg' );
459 my @hacknodes = $xpc->findnodes( '//svg:g[contains(child::svg:title, "#SILENT#")]' );
460 foreach my $h ( @hacknodes ) {
461 $h->parentNode->removeChild( $h );
463 return decode_utf8( $svgdom->toString() );
467 =head2 as_dot( \%options )
469 Returns a string that is the collation graph expressed in dot
470 (i.e. GraphViz) format. Options include:
485 my( $self, $opts ) = @_;
486 my $startrank = $opts->{'from'} if $opts;
487 my $endrank = $opts->{'to'} if $opts;
488 my $color_common = $opts->{'color_common'} if $opts;
489 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
490 && $self->end->rank > 100;
492 # Check the arguments
494 return if $endrank && $startrank > $endrank;
495 return if $startrank > $self->end->rank;
497 if( defined $endrank ) {
498 return if $endrank < 0;
499 $endrank = undef if $endrank == $self->end->rank;
502 # TODO consider making some of these things configurable
503 my $graph_name = $self->tradition->name;
504 $graph_name =~ s/[^\w\s]//g;
505 $graph_name = join( '_', split( /\s+/, $graph_name ) );
513 'fillcolor' => 'white',
518 'arrowhead' => 'open',
519 'color' => '#000000',
520 'fontcolor' => '#000000',
523 my $dot = sprintf( "digraph %s {\n", $graph_name );
524 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
525 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
527 # Output substitute start/end readings if necessary
529 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
532 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
534 if( $STRAIGHTENHACK ) {
536 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
537 $dot .= "\t\"#SILENT#\" [ color=white,penwidth=0,label=\"\" ];"
539 my %used; # Keep track of the readings that actually appear in the graph
540 foreach my $reading ( $self->readings ) {
541 # Only output readings within our rank range.
542 next if $startrank && $reading->rank < $startrank;
543 next if $endrank && $reading->rank > $endrank;
544 $used{$reading->id} = 1;
545 # Need not output nodes without separate labels
546 next if $reading->id eq $reading->text;
548 my $label = $reading->text;
549 $label =~ s/\"/\\\"/g;
550 $rattrs->{'label'} = $label;
551 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
552 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
556 my @edges = $self->paths;
557 my( %substart, %subend );
558 foreach my $edge ( @edges ) {
559 # Do we need to output this edge?
560 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
561 my $label = $self->path_display_label( $self->path_witnesses( $edge ) );
562 my $variables = { %edge_attrs, 'label' => $label };
563 # Account for the rank gap if necessary
564 if( $self->reading( $edge->[1] )->has_rank
565 && $self->reading( $edge->[0] )->has_rank
566 && $self->reading( $edge->[1] )->rank
567 - $self->reading( $edge->[0] )->rank > 1 ) {
568 $variables->{'minlen'} = $self->reading( $edge->[1] )->rank
569 - $self->reading( $edge->[0] )->rank;
571 # EXPERIMENTAL: make edge width reflect no. of witnesses
572 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
573 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
575 my $varopts = _dot_attr_string( $variables );
576 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
577 $edge->[0], $edge->[1], $varopts );
578 } elsif( $used{$edge->[0]} ) {
579 $subend{$edge->[0]} = 1;
580 } elsif( $used{$edge->[1]} ) {
581 $substart{$edge->[1]} = 1;
584 # Add substitute start and end edges if necessary
585 foreach my $node ( keys %substart ) {
586 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
587 my $variables = { %edge_attrs, 'label' => $witstr };
588 my $varopts = _dot_attr_string( $variables );
589 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
591 foreach my $node ( keys %subend ) {
592 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
593 my $variables = { %edge_attrs, 'label' => $witstr };
594 my $varopts = _dot_attr_string( $variables );
595 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
598 if( $STRAIGHTENHACK ) {
599 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
606 sub _dot_attr_string {
609 foreach my $k ( sort keys %$hash ) {
611 push( @attrs, $k.'="'.$v.'"' );
613 return( '[ ' . join( ', ', @attrs ) . ' ]' );
617 my( $self, @edge ) = @_;
618 # If edge is an arrayref, cope.
619 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
623 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
627 sub path_display_label {
630 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
631 if( scalar @wits > $maj ) {
632 # TODO break out a.c. wits
635 return join( ', ', @wits );
642 Returns a GraphML representation of the collation. The GraphML will contain
643 two graphs. The first expresses the attributes of the readings and the witness
644 paths that link them; the second expresses the relationships that link the
645 readings. This is the native transfer format for a tradition.
654 my $datafile = 't/data/florilegium_tei_ps.xml';
655 my $tradition = Text::Tradition->new( 'input' => 'TEI',
660 ok( $tradition, "Got a tradition object" );
661 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
662 ok( $tradition->collation, "Tradition has a collation" );
664 my $c = $tradition->collation;
665 is( scalar $c->readings, $READINGS, "Collation has all readings" );
666 is( scalar $c->paths, $PATHS, "Collation has all paths" );
667 is( scalar $c->relationships, 0, "Collation has all relationships" );
669 # Add a few relationships
670 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
671 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
672 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
674 # Now write it to GraphML and parse it again.
676 my $graphml = $c->as_graphml;
677 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
678 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
679 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
680 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
690 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
691 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
692 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
693 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
695 # Create the document and root node
696 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
697 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
698 $graphml->setDocumentElement( $root );
699 $root->setNamespace( $xsi_ns, 'xsi', 0 );
700 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
702 # Add the data keys for the graph
705 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
706 foreach my $datum ( @graph_attributes ) {
707 $graph_data_keys{$datum} = 'dg'.$gdi++;
708 my $key = $root->addNewChild( $graphml_ns, 'key' );
709 $key->setAttribute( 'attr.name', $datum );
710 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
711 $key->setAttribute( 'for', 'graph' );
712 $key->setAttribute( 'id', $graph_data_keys{$datum} );
715 # Add the data keys for nodes
722 is_start => 'boolean',
724 is_lacuna => 'boolean',
726 foreach my $datum ( keys %node_data ) {
727 $node_data_keys{$datum} = 'dn'.$ndi++;
728 my $key = $root->addNewChild( $graphml_ns, 'key' );
729 $key->setAttribute( 'attr.name', $datum );
730 $key->setAttribute( 'attr.type', $node_data{$datum} );
731 $key->setAttribute( 'for', 'node' );
732 $key->setAttribute( 'id', $node_data_keys{$datum} );
735 # Add the data keys for edges, i.e. witnesses
739 class => 'string', # Class, deprecated soon
740 witness => 'string', # ID/label for a path
741 relationship => 'string', # ID/label for a relationship
742 extra => 'boolean', # Path key
743 scope => 'string', # Relationship key
744 non_correctable => 'boolean', # Relationship key
745 non_independent => 'boolean', # Relationship key
747 foreach my $datum ( keys %edge_data ) {
748 $edge_data_keys{$datum} = 'de'.$edi++;
749 my $key = $root->addNewChild( $graphml_ns, 'key' );
750 $key->setAttribute( 'attr.name', $datum );
751 $key->setAttribute( 'attr.type', $edge_data{$datum} );
752 $key->setAttribute( 'for', 'edge' );
753 $key->setAttribute( 'id', $edge_data_keys{$datum} );
756 # Add the collation graph itself
757 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
758 $sgraph->setAttribute( 'edgedefault', 'directed' );
759 $sgraph->setAttribute( 'id', $self->tradition->name );
760 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
761 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
762 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
763 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
764 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
766 # Collation attribute data
767 foreach my $datum ( @graph_attributes ) {
768 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
769 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
774 # Add our readings to the graph
775 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
776 # Add to the main graph
777 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
778 my $node_xmlid = 'n' . $node_ctr++;
779 $node_hash{ $n->id } = $node_xmlid;
780 $node_el->setAttribute( 'id', $node_xmlid );
781 foreach my $d ( keys %node_data ) {
783 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
788 # Add the path edges to the sequence graph
790 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
791 # We add an edge in the graphml for every witness in $e.
792 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
793 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
794 $node_hash{ $e->[0] },
795 $node_hash{ $e->[1] } );
796 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
797 $edge_el->setAttribute( 'source', $from );
798 $edge_el->setAttribute( 'target', $to );
799 $edge_el->setAttribute( 'id', $id );
801 # It's a witness path, so add the witness
803 my $key = $edge_data_keys{'witness'};
804 # Is this an ante-corr witness?
805 my $aclabel = $self->ac_label;
806 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
807 # Keep the base witness
809 # ...and record that this is an 'extra' reading path
810 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
812 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
813 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
817 # Add the relationship graph to the XML
818 $self->relations->as_graphml( $graphml_ns, $root, \%node_hash,
819 $node_data_keys{'id'}, \%edge_data_keys );
821 # Save and return the thing
822 my $result = decode_utf8( $graphml->toString(1) );
826 sub _add_graphml_data {
827 my( $el, $key, $value ) = @_;
828 return unless defined $value;
829 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
830 $data_el->setAttribute( 'key', $key );
831 $data_el->appendText( $value );
836 Returns a CSV alignment table representation of the collation graph, one
837 row per witness (or witness uncorrected.)
843 my $table = $self->make_alignment_table;
844 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
846 # Make the header row
847 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
848 push( @result, decode_utf8( $csv->string ) );
849 # Make the rest of the rows
850 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
851 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
852 my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs;
853 $csv->combine( @row );
854 push( @result, decode_utf8( $csv->string ) );
856 return join( "\n", @result );
859 =head2 make_alignment_table( $use_refs, $include_witnesses )
861 Return a reference to an alignment table, in a slightly enhanced CollateX
862 format which looks like this:
864 $table = { alignment => [ { witness => "SIGIL",
865 tokens => [ { t => "TEXT" }, ... ] },
867 tokens => [ { t => "TEXT" }, ... ] },
871 If $use_refs is set to 1, the reading object is returned in the table
872 instead of READINGTEXT; if not, the text of the reading is returned.
874 If $include_witnesses is set to a hashref, only the witnesses whose sigil
875 keys have a true hash value will be included.
879 sub make_alignment_table {
880 my( $self, $noderefs, $include ) = @_;
881 # Make sure we can do this
882 throw( "Need a linear graph in order to make an alignment table" )
883 unless $self->linear;
884 $self->calculate_ranks unless $self->end->has_rank;
886 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
887 my @all_pos = ( 1 .. $self->end->rank - 1 );
888 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
890 next unless $include->{$wit->sigil};
892 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
893 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
894 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
895 push( @{$table->{'alignment'}},
896 { 'witness' => $wit->sigil, 'tokens' => \@row } );
897 if( $wit->is_layered ) {
898 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
899 $wit->sigil.$self->ac_label );
900 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
901 push( @{$table->{'alignment'}},
902 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
908 sub _make_witness_row {
909 my( $path, $positions, $noderefs ) = @_;
911 map { $char_hash{$_} = undef } @$positions;
913 foreach my $rdg ( @$path ) {
914 my $rtext = $rdg->text;
915 $rtext = '#LACUNA#' if $rdg->is_lacuna;
916 print STDERR "rank " . $rdg->rank . "\n" if $debug;
917 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
918 $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg }
921 my @row = map { $char_hash{$_} } @$positions;
922 # Fill in lacuna markers for undef spots in the row
923 my $last_el = shift @row;
924 my @filled_row = ( $last_el );
925 foreach my $el ( @row ) {
926 # If we are using node reference, make the lacuna node appear many times
927 # in the table. If not, use the lacuna tag.
928 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
929 $el = $noderefs ? $last_el : { 't' => '#LACUNA#' };
931 push( @filled_row, $el );
937 # Tiny utility function to say if a table element is a lacuna
940 return 1 if $el->{'t'} eq '#LACUNA#';
941 return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading'
942 && $el->{'t'}->is_lacuna;
946 # Helper to turn the witnesses along columns rather than rows. Assumes
951 return $result unless scalar @$table;
952 my $nrows = scalar @{$table->[0]};
953 foreach my $idx ( 0 .. $nrows - 1 ) {
954 foreach my $wit ( 0 .. $#{$table} ) {
955 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
961 =head1 NAVIGATION METHODS
963 =head2 reading_sequence( $first, $last, $sigil, $backup )
965 Returns the ordered list of readings, starting with $first and ending
966 with $last, for the witness given in $sigil. If a $backup sigil is
967 specified (e.g. when walking a layered witness), it will be used wherever
968 no $sigil path exists. If there is a base text reading, that will be
969 used wherever no path exists for $sigil or $backup.
973 # TODO Think about returning some lazy-eval iterator.
974 # TODO Get rid of backup; we should know from what witness is whether we need it.
976 sub reading_sequence {
977 my( $self, $start, $end, $witness ) = @_;
979 $witness = $self->baselabel unless $witness;
980 my @readings = ( $start );
983 while( $n && $n->id ne $end->id ) {
984 if( exists( $seen{$n->id} ) ) {
985 throw( "Detected loop for $witness at " . $n->id );
989 my $next = $self->next_reading( $n, $witness );
991 throw( "Did not find any path for $witness from reading " . $n->id );
993 push( @readings, $next );
996 # Check that the last reading is our end reading.
997 my $last = $readings[$#readings];
998 throw( "Last reading found from " . $start->text .
999 " for witness $witness is not the end!" ) # TODO do we get this far?
1000 unless $last->id eq $end->id;
1005 =head2 next_reading( $reading, $sigil );
1007 Returns the reading that follows the given reading along the given witness
1013 # Return the successor via the corresponding path.
1015 my $answer = $self->_find_linked_reading( 'next', @_ );
1016 return undef unless $answer;
1017 return $self->reading( $answer );
1020 =head2 prior_reading( $reading, $sigil )
1022 Returns the reading that precedes the given reading along the given witness
1028 # Return the predecessor via the corresponding path.
1030 my $answer = $self->_find_linked_reading( 'prior', @_ );
1031 return $self->reading( $answer );
1034 sub _find_linked_reading {
1035 my( $self, $direction, $node, $path ) = @_;
1037 # Get a backup if we are dealing with a layered witness
1039 my $aclabel = $self->ac_label;
1040 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1044 my @linked_paths = $direction eq 'next'
1045 ? $self->sequence->edges_from( $node )
1046 : $self->sequence->edges_to( $node );
1047 return undef unless scalar( @linked_paths );
1049 # We have to find the linked path that contains all of the
1050 # witnesses supplied in $path.
1051 my( @path_wits, @alt_path_wits );
1052 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1053 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1056 foreach my $le ( @linked_paths ) {
1057 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1060 my @le_wits = sort $self->path_witnesses( $le );
1061 if( _is_within( \@path_wits, \@le_wits ) ) {
1062 # This is the right path.
1063 return $direction eq 'next' ? $le->[1] : $le->[0];
1064 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1068 # Got this far? Return the alternate path if it exists.
1069 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1072 # Got this far? Return the base path if it exists.
1073 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1076 # Got this far? We have no appropriate path.
1077 warn "Could not find $direction node from " . $node->id
1078 . " along path $path";
1084 my( $set1, $set2 ) = @_;
1085 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1086 foreach my $el ( @$set1 ) {
1087 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1092 # Return the string that joins together a list of witnesses for
1093 # display on a single path.
1094 sub _witnesses_of_label {
1095 my( $self, $label ) = @_;
1096 my $regex = $self->wit_list_separator;
1097 my @answer = split( /\Q$regex\E/, $label );
1101 =head2 common_readings
1103 Returns the list of common readings in the graph (i.e. those readings that are
1104 shared by all non-lacunose witnesses.)
1108 sub common_readings {
1110 my @common = grep { $_->is_common } $self->readings;
1114 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1116 Returns the text of a witness (plus its backup, if we are using a layer)
1117 as stored in the collation. The text is returned as a string, where the
1118 individual readings are joined with spaces and the meta-readings (e.g.
1119 lacunae) are omitted. Optional specification of $start and $end allows
1120 the generation of a subset of the witness text.
1125 my( $self, $wit, $start, $end ) = @_;
1126 $start = $self->start unless $start;
1127 $end = $self->end unless $end;
1128 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1129 return join( ' ', map { $_->text } @path );
1132 =head1 INITIALIZATION METHODS
1134 These are mostly for use by parsers.
1136 =head2 make_witness_path( $witness )
1138 Link the array of readings contained in $witness->path (and in
1139 $witness->uncorrected_path if it exists) into collation paths.
1140 Clear out the arrays when finished.
1142 =head2 make_witness_paths
1144 Call make_witness_path for all witnesses in the tradition.
1148 # For use when a collation is constructed from a base text and an apparatus.
1149 # We have the sequences of readings and just need to add path edges.
1150 # When we are done, clear out the witness path attributes, as they are no
1152 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1154 sub make_witness_paths {
1156 foreach my $wit ( $self->tradition->witnesses ) {
1157 # print STDERR "Making path for " . $wit->sigil . "\n";
1158 $self->make_witness_path( $wit );
1162 sub make_witness_path {
1163 my( $self, $wit ) = @_;
1164 my @chain = @{$wit->path};
1165 my $sig = $wit->sigil;
1166 foreach my $idx ( 0 .. $#chain-1 ) {
1167 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1169 if( $wit->is_layered ) {
1170 @chain = @{$wit->uncorrected_path};
1171 foreach my $idx( 0 .. $#chain-1 ) {
1172 my $source = $chain[$idx];
1173 my $target = $chain[$idx+1];
1174 $self->add_path( $source, $target, $sig.$self->ac_label )
1175 unless $self->has_path( $source, $target, $sig );
1179 $wit->clear_uncorrected_path;
1182 =head2 calculate_ranks
1184 Calculate the reading ranks (that is, their aligned positions relative
1185 to each other) for the graph. This can only be called on linear collations.
1189 sub calculate_ranks {
1191 # Walk a version of the graph where every node linked by a relationship
1192 # edge is fundamentally the same node, and do a topological ranking on
1193 # the nodes in this graph.
1194 my $topo_graph = Graph->new();
1198 foreach my $r ( $self->readings ) {
1199 next if exists $rel_containers{$r->id};
1200 my @rels = $r->related_readings( 'colocated' );
1202 # Make a relationship container.
1204 my $rn = 'rel_container_' . $rel_ctr++;
1205 $topo_graph->add_vertex( $rn );
1207 $rel_containers{$_->id} = $rn;
1210 # Add a new node to mirror the old node.
1211 $rel_containers{$r->id} = $r->id;
1212 $topo_graph->add_vertex( $r->id );
1217 foreach my $r ( $self->readings ) {
1218 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1219 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1220 $rel_containers{$n} );
1221 # $DB::single = 1 unless $tfrom && $tto;
1222 $topo_graph->add_edge( $tfrom, $tto );
1226 # Now do the rankings, starting with the start node.
1227 my $topo_start = $rel_containers{$self->start->id};
1228 my $node_ranks = { $topo_start => 0 };
1229 my @curr_origin = ( $topo_start );
1230 # A little iterative function.
1231 while( @curr_origin ) {
1232 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1234 # Transfer our rankings from the topological graph to the real one.
1235 foreach my $r ( $self->readings ) {
1236 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1237 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1239 # Die. Find the last rank we calculated.
1240 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1241 <=> $node_ranks->{$rel_containers{$b->id}} }
1243 my $last = pop @all_defined;
1244 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1250 my( $graph, $node_ranks, @current_nodes ) = @_;
1251 # Look at each of the children of @current_nodes. If all the child's
1252 # parents have a rank, assign it the highest rank + 1 and add it to
1253 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1254 # parent gets a rank.
1256 foreach my $c ( @current_nodes ) {
1257 warn "Current reading $c has no rank!"
1258 unless exists $node_ranks->{$c};
1259 # print STDERR "Looking at child of node $c, rank "
1260 # . $node_ranks->{$c} . "\n";
1261 foreach my $child ( $graph->successors( $c ) ) {
1262 next if exists $node_ranks->{$child};
1263 my $highest_rank = -1;
1265 foreach my $parent ( $graph->predecessors( $child ) ) {
1266 if( exists $node_ranks->{$parent} ) {
1267 $highest_rank = $node_ranks->{$parent}
1268 if $highest_rank <= $node_ranks->{$parent};
1275 my $c_rank = $highest_rank + 1;
1276 # print STDERR "Assigning rank $c_rank to node $child \n";
1277 $node_ranks->{$child} = $c_rank;
1278 push( @next_nodes, $child );
1284 =head2 flatten_ranks
1286 A convenience method for parsing collation data. Searches the graph for readings
1287 with the same text at the same rank, and merges any that are found.
1293 my %unique_rank_rdg;
1294 foreach my $rdg ( $self->readings ) {
1295 next unless $rdg->has_rank;
1296 my $key = $rdg->rank . "||" . $rdg->text;
1297 if( exists $unique_rank_rdg{$key} ) {
1299 # print STDERR "Combining readings at same rank: $key\n";
1300 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1301 # TODO see if this now makes a common point.
1303 $unique_rank_rdg{$key} = $rdg;
1308 =head2 calculate_common_readings
1310 Goes through the graph identifying the readings that appear in every witness
1311 (apart from those with lacunae at that spot.) Marks them as common and returns
1316 use Text::Tradition;
1318 my $cxfile = 't/data/Collatex-16.xml';
1319 my $t = Text::Tradition->new(
1321 'input' => 'CollateX',
1324 my $c = $t->collation;
1326 my @common = $c->calculate_common_readings();
1327 is( scalar @common, 8, "Found correct number of common readings" );
1328 my @marked = sort $c->common_readings();
1329 is( scalar @common, 8, "All common readings got marked as such" );
1330 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1331 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1337 sub calculate_common_readings {
1340 my $table = $self->make_alignment_table( 1 );
1341 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1342 my @row = map { $_->{'tokens'}->[$idx]->{'t'} } @{$table->{'alignment'}};
1344 foreach my $r ( @row ) {
1346 $hash{$r->id} = $r unless $r->is_meta;
1348 $hash{'UNDEF'} = $r;
1351 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1352 my( $r ) = values %hash;
1354 push( @common, $r );
1360 =head2 text_from_paths
1362 Calculate the text array for all witnesses from the path, for later consistency
1363 checking. Only to be used if there is no non-graph-based way to know the
1368 sub text_from_paths {
1370 foreach my $wit ( $self->tradition->witnesses ) {
1371 my @text = split( /\s+/,
1372 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1373 $wit->text( \@text );
1374 if( $wit->is_layered ) {
1375 my @uctext = split( /\s+/,
1376 $self->reading_sequence( $self->start, $self->end,
1377 $wit->sigil.$self->ac_label ) );
1378 $wit->text( \@uctext );
1383 =head1 UTILITY FUNCTIONS
1385 =head2 common_predecessor( $reading_a, $reading_b )
1387 Find the last reading that occurs in sequence before both the given readings.
1389 =head2 common_successor( $reading_a, $reading_b )
1391 Find the first reading that occurs in sequence after both the given readings.
1395 use Text::Tradition;
1397 my $cxfile = 't/data/Collatex-16.xml';
1398 my $t = Text::Tradition->new(
1400 'input' => 'CollateX',
1403 my $c = $t->collation;
1405 is( $c->common_predecessor( 'n9', 'n23' )->id,
1406 'n20', "Found correct common predecessor" );
1407 is( $c->common_successor( 'n9', 'n23' )->id,
1408 '#END#', "Found correct common successor" );
1410 is( $c->common_predecessor( 'n19', 'n17' )->id,
1411 'n16', "Found correct common predecessor for readings on same path" );
1412 is( $c->common_successor( 'n21', 'n26' )->id,
1413 '#END#', "Found correct common successor for readings on same path" );
1419 ## Return the closest reading that is a predecessor of both the given readings.
1420 sub common_predecessor {
1422 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1423 return $self->common_in_path( $r1, $r2, 'predecessors' );
1426 sub common_successor {
1428 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1429 return $self->common_in_path( $r1, $r2, 'successors' );
1432 sub common_in_path {
1433 my( $self, $r1, $r2, $dir ) = @_;
1434 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1435 $iter = $self->end->rank - $iter if $dir eq 'successors';
1437 my @last_checked = ( $r1, $r2 );
1439 while( !@candidates ) {
1441 foreach my $lc ( @last_checked ) {
1442 foreach my $p ( $lc->$dir ) {
1443 if( $all_seen{$p->id} ) {
1444 push( @candidates, $p );
1446 $all_seen{$p->id} = 1;
1447 push( @new_lc, $p );
1451 @last_checked = @new_lc;
1453 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1454 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1458 Text::Tradition::Error->throw(
1459 'ident' => 'Collation error',
1465 __PACKAGE__->meta->make_immutable;
1471 =item * Get rid of $backup in reading_sequence