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;
16 default => sub { Graph->new() },
24 isa => 'Text::Tradition::Collation::RelationshipStore',
26 relationships => 'relationships',
27 related_readings => 'related_readings',
29 writer => '_set_relations',
34 isa => 'Text::Tradition',
39 isa => 'HashRef[Text::Tradition::Collation::Reading]',
43 _add_reading => 'set',
44 del_reading => 'delete',
45 has_reading => 'exists',
48 default => sub { {} },
51 has 'wit_list_separator' => (
60 default => 'base text',
69 has 'collapse_punctuation' => (
83 isa => 'Text::Tradition::Collation::Reading',
84 writer => '_set_start',
90 isa => 'Text::Tradition::Collation::Reading',
97 Text::Tradition::Collation - a software model for a text collation
102 my $t = Text::Tradition->new(
103 'name' => 'this is a text',
105 'file' => '/path/to/tei_parallel_seg_file.xml' );
107 my $c = $t->collation;
108 my @readings = $c->readings;
109 my @paths = $c->paths;
110 my @relationships = $c->relationships;
112 my $svg_variant_graph = $t->collation->as_svg();
116 Text::Tradition is a library for representation and analysis of collated
117 texts, particularly medieval ones. The Collation is the central feature of
118 a Tradition, where the text, its sequence of readings, and its relationships
119 between readings are actually kept.
125 The constructor. Takes a hash or hashref of the following arguments:
129 =item * tradition - The Text::Tradition object to which the collation
132 =item * linear - Whether the collation should be linear; that is, whether
133 transposed readings should be treated as two linked readings rather than one,
134 and therefore whether the collation graph is acyclic. Defaults to true.
136 =item * collapse_punctuation - TODO
138 =item * baselabel - The default label for the path taken by a base text
139 (if any). Defaults to 'base text'.
141 =item * wit_list_separator - The string to join a list of witnesses for
142 purposes of making labels in display graphs. Defaults to ', '.
144 =item * ac_label - The extra label to tack onto a witness sigil when
145 representing another layer of path for the given witness - that is, when
146 a text has more than one possible reading due to scribal corrections or
147 the like. Defaults to ' (a.c.)'.
157 =head2 collapse_punctuation
159 =head2 wit_list_separator
165 Simple accessors for collation attributes.
169 The meta-reading at the start of every witness path.
173 The meta-reading at the end of every witness path.
177 Returns all Reading objects in the graph.
179 =head2 reading( $id )
181 Returns the Reading object corresponding to the given ID.
183 =head2 add_reading( $reading_args )
185 Adds a new reading object to the collation.
186 See L<Text::Tradition::Collation::Reading> for the available arguments.
188 =head2 del_reading( $object_or_id )
190 Removes the given reading from the collation, implicitly removing its
191 paths and relationships.
193 =head2 merge_readings( $main, $second )
195 Merges the $second reading into the $main one.
196 The arguments may be either readings or reading IDs.
198 =head2 has_reading( $id )
200 Predicate to see whether a given reading ID is in the graph.
202 =head2 reading_witnesses( $object_or_id )
204 Returns a list of sigils whose witnesses contain the reading.
208 Returns all reading paths within the document - that is, all edges in the
209 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
211 =head2 add_path( $source, $target, $sigil )
213 Links the given readings in the collation in sequence, under the given witness
214 sigil. The readings may be specified by object or ID.
216 =head2 del_path( $source, $target, $sigil )
218 Links the given readings in the collation in sequence, under the given witness
219 sigil. The readings may be specified by object or ID.
221 =head2 has_path( $source, $target );
223 Returns true if the two readings are linked in sequence in any witness.
224 The readings may be specified by object or ID.
228 Returns all Relationship objects in the collation.
230 =head2 add_relationship( $reading, $other_reading, $options )
232 Adds a new relationship of the type given in $options between the two readings,
233 which may be specified by object or ID. Returns a value of ( $status, @vectors)
234 where $status is true on success, and @vectors is a list of relationship edges
235 that were ultimately added.
236 See L<Text::Tradition::Collation::Relationship> for the available options.
242 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
243 $self->_set_start( $self->add_reading( { 'collation' => $self, 'is_start' => 1 } ) );
244 $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) );
247 ### Reading construct/destruct functions
250 my( $self, $reading ) = @_;
251 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
252 my %args = %$reading;
253 $reading = Text::Tradition::Collation::Reading->new(
254 'collation' => $self,
257 # First check to see if a reading with this ID exists.
258 if( $self->reading( $reading->id ) ) {
259 warn "Collation already has a reading with id " . $reading->id;
262 $self->_add_reading( $reading->id => $reading );
263 # Once the reading has been added, put it in both graphs.
264 $self->sequence->add_vertex( $reading->id );
265 $self->relations->add_reading( $reading->id );
269 around del_reading => sub {
274 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
277 # Remove the reading from the graphs.
278 $self->sequence->delete_vertex( $arg );
279 $self->relations->delete_reading( $arg );
282 $self->$orig( $arg );
285 # merge_readings( $main, $to_be_deleted );
290 # We only need the IDs for adding paths to the graph, not the reading
291 # objects themselves.
292 my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ );
294 # The kept reading should inherit the paths and the relationships
295 # of the deleted reading.
296 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
297 my @vector = ( $kept );
298 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
299 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
300 next if $vector[0] eq $vector[1]; # Don't add a self loop
301 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
302 $self->sequence->add_edge( @vector );
303 my $fwits = $self->sequence->get_edge_attributes( @vector );
304 @wits{keys %$fwits} = values %$fwits;
305 $self->sequence->set_edge_attributes( @vector, \%wits );
307 $self->relations->merge_readings( $kept, $deleted, $combine_char );
309 # Do the deletion deed.
310 if( $combine_char ) {
311 my $kept_obj = $self->reading( $kept );
312 my $new_text = join( $combine_char, $kept_obj->text,
313 $self->reading( $deleted )->text );
314 $kept_obj->alter_text( $new_text );
316 $self->del_reading( $deleted );
320 # Helper function for manipulating the graph.
321 sub _stringify_args {
322 my( $self, $first, $second, $arg ) = @_;
324 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
325 $second = $second->id
326 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
327 return( $first, $second, $arg );
330 # Helper function for manipulating the graph.
331 sub _objectify_args {
332 my( $self, $first, $second, $arg ) = @_;
333 $first = $self->reading( $first )
334 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
335 $second = $self->reading( $second )
336 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
337 return( $first, $second, $arg );
344 # We only need the IDs for adding paths to the graph, not the reading
345 # objects themselves.
346 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
348 # Connect the readings
349 $self->sequence->add_edge( $source, $target );
350 # Note the witness in question
351 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
357 if( ref( $_[0] ) eq 'ARRAY' ) {
364 # We only need the IDs for adding paths to the graph, not the reading
365 # objects themselves.
366 my( $source, $target, $wit ) = $self->_stringify_args( @args );
368 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
369 $self->sequence->delete_edge_attribute( $source, $target, $wit );
371 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
372 $self->sequence->delete_edge( $source, $target );
377 # Extra graph-alike utility
380 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
381 return undef unless $self->sequence->has_edge( $source, $target );
382 return $self->sequence->has_edge_attribute( $source, $target, $wit );
385 =head2 clear_witness( @sigil_list )
387 Clear the given witnesses out of the collation entirely, removing references
388 to them in paths, and removing readings that belong only to them. Should only
389 be called via $tradition->del_witness.
394 my( $self, @sigils ) = @_;
396 # Clear the witness(es) out of the paths
397 foreach my $e ( $self->paths ) {
398 foreach my $sig ( @sigils ) {
399 $self->del_path( $e, $sig );
403 # Clear out the newly unused readings
404 foreach my $r ( $self->readings ) {
405 unless( $self->reading_witnesses( $r ) ) {
406 $self->del_reading( $r );
411 sub add_relationship {
413 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
414 my( $ret, @vectors ) = $self->relations->add_relationship( $source,
415 $self->reading( $source ), $target, $self->reading( $target ), $opts );
416 # Force a full rank recalculation every time. Yuck.
417 $self->calculate_ranks() if $ret && $self->end->has_rank;
418 return( $ret, @vectors );
421 =head2 reading_witnesses( $reading )
423 Return a list of sigils corresponding to the witnesses in which the reading appears.
427 sub reading_witnesses {
428 my( $self, $reading ) = @_;
429 # We need only check either the incoming or the outgoing edges; I have
430 # arbitrarily chosen "incoming". Thus, special-case the start node.
431 if( $reading eq $self->start ) {
432 return map { $_->sigil } $self->tradition->witnesses;
435 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
436 my $wits = $self->sequence->get_edge_attributes( @$e );
437 @all_witnesses{ keys %$wits } = 1;
439 return keys %all_witnesses;
442 =head1 OUTPUT METHODS
446 Returns an SVG string that represents the graph, via as_dot and graphviz.
453 my @cmd = qw/dot -Tsvg/;
455 my $dotfile = File::Temp->new();
457 # $dotfile->unlink_on_destroy(0);
458 binmode $dotfile, ':utf8';
459 print $dotfile $self->as_dot();
460 push( @cmd, $dotfile->filename );
461 run( \@cmd, ">", binary(), \$svg );
462 $svg = decode_utf8( $svg );
466 =head2 svg_subgraph( $from, $to )
468 Returns an SVG string that represents the portion of the graph given by the
469 specified range. The $from and $to variables refer to ranks within the graph.
474 my( $self, $from, $to ) = @_;
476 my $dot = $self->as_dot( $from, $to );
478 warn "Could not output a graph with range $from - $to";
482 my @cmd = qw/dot -Tsvg/;
484 my $dotfile = File::Temp->new();
486 # $dotfile->unlink_on_destroy(0);
487 binmode $dotfile, ':utf8';
489 push( @cmd, $dotfile->filename );
490 run( \@cmd, ">", binary(), \$svg );
491 $svg = decode_utf8( $svg );
496 =head2 as_dot( $from, $to )
498 Returns a string that is the collation graph expressed in dot
499 (i.e. GraphViz) format. If $from or $to is passed, as_dot creates
500 a subgraph rather than the entire graph.
505 my( $self, $startrank, $endrank ) = @_;
507 # Check the arguments
509 return if $endrank && $startrank > $endrank;
510 return if $startrank > $self->end->rank;
512 if( defined $endrank ) {
513 return if $endrank < 0;
514 $endrank = undef if $endrank == $self->end->rank;
517 # TODO consider making some of these things configurable
518 my $graph_name = $self->tradition->name;
519 $graph_name =~ s/[^\w\s]//g;
520 $graph_name = join( '_', split( /\s+/, $graph_name ) );
528 'fillcolor' => 'white',
533 'arrowhead' => 'open',
534 'color' => '#000000',
535 'fontcolor' => '#000000',
538 my $dot = sprintf( "digraph %s {\n", $graph_name );
539 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
540 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
542 # Output substitute start/end readings if necessary
544 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
547 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
549 my %used; # Keep track of the readings that actually appear in the graph
550 foreach my $reading ( $self->readings ) {
551 # Only output readings within our rank range.
552 next if $startrank && $reading->rank < $startrank;
553 next if $endrank && $reading->rank > $endrank;
554 $used{$reading->id} = 1;
555 # Need not output nodes without separate labels
556 next if $reading->id eq $reading->text;
557 my $label = $reading->punctuated_form;
558 $label =~ s/\"/\\\"/g;
559 $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->id, $label );
563 my @edges = $self->paths;
564 my( %substart, %subend );
565 foreach my $edge ( @edges ) {
566 # Do we need to output this edge?
567 if( $used{$edge->[0]} && $used{$edge->[1]} ) {;
568 my $label = $self->path_display_label( $self->path_witnesses( $edge ) );
569 my $variables = { %edge_attrs, 'label' => $label };
570 # Account for the rank gap if necessary
571 my $rankgap = $self->reading( $edge->[1] )->rank
572 - $self->reading( $edge->[0] )->rank;
573 $variables->{'minlen'} = $rankgap if $rankgap > 1;
574 my $varopts = _dot_attr_string( $variables );
575 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
576 $edge->[0], $edge->[1], $varopts );
577 } elsif( $used{$edge->[0]} ) {
578 $subend{$edge->[0]} = 1;
579 } elsif( $used{$edge->[1]} ) {
580 $substart{$edge->[1]} = 1;
583 # Add substitute start and end edges if necessary
584 foreach my $node ( keys %substart ) {
585 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
586 my $variables = { %edge_attrs, 'label' => $witstr };
587 my $varopts = _dot_attr_string( $variables );
588 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
590 foreach my $node ( keys %subend ) {
591 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
592 my $variables = { %edge_attrs, 'label' => $witstr };
593 my $varopts = _dot_attr_string( $variables );
594 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
601 sub _dot_attr_string {
604 foreach my $k ( sort keys %$hash ) {
606 push( @attrs, $k.'="'.$v.'"' );
608 return( '[ ' . join( ', ', @attrs ) . ' ]' );
612 my( $self, @edge ) = @_;
613 # If edge is an arrayref, cope.
614 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
618 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
622 sub path_display_label {
623 my( $self, @wits ) = @_;
624 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
625 if( scalar @wits > $maj ) {
626 # TODO break out a.c. wits
629 return join( ', ', @wits );
636 Returns a GraphML representation of the collation. The GraphML will contain
637 two graphs. The first expresses the attributes of the readings and the witness
638 paths that link them; the second expresses the relationships that link the
639 readings. This is the native transfer format for a tradition.
648 my $datafile = 't/data/florilegium_tei_ps.xml';
649 my $tradition = Text::Tradition->new( 'input' => 'TEI',
654 ok( $tradition, "Got a tradition object" );
655 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
656 ok( $tradition->collation, "Tradition has a collation" );
658 my $c = $tradition->collation;
659 is( scalar $c->readings, $READINGS, "Collation has all readings" );
660 is( scalar $c->paths, $PATHS, "Collation has all paths" );
661 is( scalar $c->relationships, 0, "Collation has all relationships" );
663 # Add a few relationships
664 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
665 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
666 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
668 # Now write it to GraphML and parse it again.
670 my $graphml = $c->as_graphml;
671 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
672 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
673 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
674 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
684 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
685 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
686 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
687 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
689 # Create the document and root node
690 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
691 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
692 $graphml->setDocumentElement( $root );
693 $root->setNamespace( $xsi_ns, 'xsi', 0 );
694 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
696 # Add the data keys for the graph
699 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
700 foreach my $datum ( @graph_attributes ) {
701 $graph_data_keys{$datum} = 'dg'.$gdi++;
702 my $key = $root->addNewChild( $graphml_ns, 'key' );
703 $key->setAttribute( 'attr.name', $datum );
704 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
705 $key->setAttribute( 'for', 'graph' );
706 $key->setAttribute( 'id', $graph_data_keys{$datum} );
709 # Add the data keys for nodes
716 is_start => 'boolean',
718 is_lacuna => 'boolean',
720 foreach my $datum ( keys %node_data ) {
721 $node_data_keys{$datum} = 'dn'.$ndi++;
722 my $key = $root->addNewChild( $graphml_ns, 'key' );
723 $key->setAttribute( 'attr.name', $datum );
724 $key->setAttribute( 'attr.type', $node_data{$datum} );
725 $key->setAttribute( 'for', 'node' );
726 $key->setAttribute( 'id', $node_data_keys{$datum} );
729 # Add the data keys for edges, i.e. witnesses
733 class => 'string', # Class, deprecated soon
734 witness => 'string', # ID/label for a path
735 relationship => 'string', # ID/label for a relationship
736 extra => 'boolean', # Path key
737 scope => 'string', # Relationship key
738 non_correctable => 'boolean', # Relationship key
739 non_independent => 'boolean', # Relationship key
741 foreach my $datum ( keys %edge_data ) {
742 $edge_data_keys{$datum} = 'de'.$edi++;
743 my $key = $root->addNewChild( $graphml_ns, 'key' );
744 $key->setAttribute( 'attr.name', $datum );
745 $key->setAttribute( 'attr.type', $edge_data{$datum} );
746 $key->setAttribute( 'for', 'edge' );
747 $key->setAttribute( 'id', $edge_data_keys{$datum} );
750 # Add the collation graph itself
751 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
752 $sgraph->setAttribute( 'edgedefault', 'directed' );
753 $sgraph->setAttribute( 'id', $self->tradition->name );
754 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
755 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
756 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
757 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
758 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
760 # Collation attribute data
761 foreach my $datum ( @graph_attributes ) {
762 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
763 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
768 # Add our readings to the graph
769 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
770 # Add to the main graph
771 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
772 my $node_xmlid = 'n' . $node_ctr++;
773 $node_hash{ $n->id } = $node_xmlid;
774 $node_el->setAttribute( 'id', $node_xmlid );
775 foreach my $d ( keys %node_data ) {
777 $nval = $n->punctuated_form if $d eq 'text';
778 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
783 # Add the path edges to the sequence graph
785 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
786 # We add an edge in the graphml for every witness in $e.
787 foreach my $wit ( $self->path_witnesses( $e ) ) {
788 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
789 $node_hash{ $e->[0] },
790 $node_hash{ $e->[1] } );
791 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
792 $edge_el->setAttribute( 'source', $from );
793 $edge_el->setAttribute( 'target', $to );
794 $edge_el->setAttribute( 'id', $id );
796 # It's a witness path, so add the witness
798 my $key = $edge_data_keys{'witness'};
799 # Is this an ante-corr witness?
800 my $aclabel = $self->ac_label;
801 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
802 # Keep the base witness
804 # ...and record that this is an 'extra' reading path
805 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
807 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
808 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
812 # Add the relationship graph to the XML
813 $self->relations->as_graphml( $graphml_ns, $root, \%node_hash,
814 $node_data_keys{'id'}, \%edge_data_keys );
816 # Save and return the thing
817 my $result = decode_utf8( $graphml->toString(1) );
821 sub _add_graphml_data {
822 my( $el, $key, $value ) = @_;
823 return unless defined $value;
824 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
825 $data_el->setAttribute( 'key', $key );
826 $data_el->appendText( $value );
831 Returns a CSV alignment table representation of the collation graph, one
832 row per witness (or witness uncorrected.)
838 my $table = $self->make_alignment_table;
839 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
841 # Make the header row
842 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
843 push( @result, decode_utf8( $csv->string ) );
844 # Make the rest of the rows
845 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
846 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
847 my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs;
848 $csv->combine( @row );
849 push( @result, decode_utf8( $csv->string ) );
851 return join( "\n", @result );
854 =head2 make_alignment_table( $use_refs, $include_witnesses )
856 Return a reference to an alignment table, in a slightly enhanced CollateX
857 format which looks like this:
859 $table = { alignment => [ { witness => "SIGIL",
860 tokens => [ { t => "TEXT" }, ... ] },
862 tokens => [ { t => "TEXT" }, ... ] },
866 If $use_refs is set to 1, the reading object is returned in the table
867 instead of READINGTEXT; if not, the text of the reading is returned.
869 If $include_witnesses is set to a hashref, only the witnesses whose sigil
870 keys have a true hash value will be included.
874 sub make_alignment_table {
875 my( $self, $noderefs, $include ) = @_;
876 unless( $self->linear ) {
877 warn "Need a linear graph in order to make an alignment table";
880 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
881 my @all_pos = ( 1 .. $self->end->rank - 1 );
882 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
884 next unless $include->{$wit->sigil};
886 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
887 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
888 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
889 push( @{$table->{'alignment'}},
890 { 'witness' => $wit->sigil, 'tokens' => \@row } );
891 if( $wit->is_layered ) {
892 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
893 $wit->sigil.$self->ac_label, $wit->sigil );
894 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
895 push( @{$table->{'alignment'}},
896 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
902 sub _make_witness_row {
903 my( $path, $positions, $noderefs ) = @_;
905 map { $char_hash{$_} = undef } @$positions;
907 foreach my $rdg ( @$path ) {
908 my $rtext = $rdg->text;
909 $rtext = '#LACUNA#' if $rdg->is_lacuna;
910 print STDERR "rank " . $rdg->rank . "\n" if $debug;
911 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
912 $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg }
915 my @row = map { $char_hash{$_} } @$positions;
916 # Fill in lacuna markers for undef spots in the row
917 my $last_el = shift @row;
918 my @filled_row = ( $last_el );
919 foreach my $el ( @row ) {
920 # If we are using node reference, make the lacuna node appear many times
921 # in the table. If not, use the lacuna tag.
922 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
923 $el = $noderefs ? $last_el : { 't' => '#LACUNA#' };
925 push( @filled_row, $el );
931 # Tiny utility function to say if a table element is a lacuna
934 return 1 if $el->{'t'} eq '#LACUNA#';
935 return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading'
936 && $el->{'t'}->is_lacuna;
940 # Helper to turn the witnesses along columns rather than rows. Assumes
945 return $result unless scalar @$table;
946 my $nrows = scalar @{$table->[0]};
947 foreach my $idx ( 0 .. $nrows - 1 ) {
948 foreach my $wit ( 0 .. $#{$table} ) {
949 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
955 =head1 NAVIGATION METHODS
957 =head2 reading_sequence( $first, $last, $sigil, $backup )
959 Returns the ordered list of readings, starting with $first and ending
960 with $last, for the witness given in $sigil. If a $backup sigil is
961 specified (e.g. when walking a layered witness), it will be used wherever
962 no $sigil path exists. If there is a base text reading, that will be
963 used wherever no path exists for $sigil or $backup.
967 # TODO Think about returning some lazy-eval iterator.
969 sub reading_sequence {
970 my( $self, $start, $end, $witness, $backup ) = @_;
972 $witness = $self->baselabel unless $witness;
973 my @readings = ( $start );
976 while( $n && $n->id ne $end->id ) {
977 if( exists( $seen{$n->id} ) ) {
978 warn "Detected loop at " . $n->id;
983 my $next = $self->next_reading( $n, $witness, $backup );
985 warn "Did not find any path for $witness from reading " . $n->id;
988 push( @readings, $next );
991 # Check that the last reading is our end reading.
992 my $last = $readings[$#readings];
993 warn "Last reading found from " . $start->text .
994 " for witness $witness is not the end!"
995 unless $last->id eq $end->id;
1000 =head2 next_reading( $reading, $sigil );
1002 Returns the reading that follows the given reading along the given witness
1008 # Return the successor via the corresponding path.
1010 my $answer = $self->_find_linked_reading( 'next', @_ );
1011 return undef unless $answer;
1012 return $self->reading( $answer );
1015 =head2 prior_reading( $reading, $sigil )
1017 Returns the reading that precedes the given reading along the given witness
1023 # Return the predecessor via the corresponding path.
1025 my $answer = $self->_find_linked_reading( 'prior', @_ );
1026 return $self->reading( $answer );
1029 sub _find_linked_reading {
1030 my( $self, $direction, $node, $path, $alt_path ) = @_;
1031 my @linked_paths = $direction eq 'next'
1032 ? $self->sequence->edges_from( $node )
1033 : $self->sequence->edges_to( $node );
1034 return undef unless scalar( @linked_paths );
1036 # We have to find the linked path that contains all of the
1037 # witnesses supplied in $path.
1038 my( @path_wits, @alt_path_wits );
1039 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1040 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1043 foreach my $le ( @linked_paths ) {
1044 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1047 my @le_wits = $self->path_witnesses( $le );
1048 if( _is_within( \@path_wits, \@le_wits ) ) {
1049 # This is the right path.
1050 return $direction eq 'next' ? $le->[1] : $le->[0];
1051 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1055 # Got this far? Return the alternate path if it exists.
1056 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1059 # Got this far? Return the base path if it exists.
1060 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1063 # Got this far? We have no appropriate path.
1064 warn "Could not find $direction node from " . $node->id
1065 . " along path $path";
1071 my( $set1, $set2 ) = @_;
1072 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1073 foreach my $el ( @$set1 ) {
1074 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1079 # Return the string that joins together a list of witnesses for
1080 # display on a single path.
1081 sub _witnesses_of_label {
1082 my( $self, $label ) = @_;
1083 my $regex = $self->wit_list_separator;
1084 my @answer = split( /\Q$regex\E/, $label );
1089 =head1 INITIALIZATION METHODS
1091 These are mostly for use by parsers.
1093 =head2 make_witness_path( $witness )
1095 Link the array of readings contained in $witness->path (and in
1096 $witness->uncorrected_path if it exists) into collation paths.
1097 Clear out the arrays when finished.
1099 =head2 make_witness_paths
1101 Call make_witness_path for all witnesses in the tradition.
1105 # For use when a collation is constructed from a base text and an apparatus.
1106 # We have the sequences of readings and just need to add path edges.
1107 # When we are done, clear out the witness path attributes, as they are no
1109 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1111 sub make_witness_paths {
1113 foreach my $wit ( $self->tradition->witnesses ) {
1114 # print STDERR "Making path for " . $wit->sigil . "\n";
1115 $self->make_witness_path( $wit );
1119 sub make_witness_path {
1120 my( $self, $wit ) = @_;
1121 my @chain = @{$wit->path};
1122 my $sig = $wit->sigil;
1123 foreach my $idx ( 0 .. $#chain-1 ) {
1124 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1126 if( $wit->is_layered ) {
1127 @chain = @{$wit->uncorrected_path};
1128 foreach my $idx( 0 .. $#chain-1 ) {
1129 my $source = $chain[$idx];
1130 my $target = $chain[$idx+1];
1131 $self->add_path( $source, $target, $sig.$self->ac_label )
1132 unless $self->has_path( $source, $target, $sig );
1136 $wit->clear_uncorrected_path;
1139 =head2 calculate_ranks
1141 Calculate the reading ranks (that is, their aligned positions relative
1142 to each other) for the graph. This can only be called on linear collations.
1146 sub calculate_ranks {
1148 # Walk a version of the graph where every node linked by a relationship
1149 # edge is fundamentally the same node, and do a topological ranking on
1150 # the nodes in this graph.
1151 my $topo_graph = Graph->new();
1155 foreach my $r ( $self->readings ) {
1156 next if exists $rel_containers{$r->id};
1157 my @rels = $r->related_readings( 'colocated' );
1159 # Make a relationship container.
1161 my $rn = 'rel_container_' . $rel_ctr++;
1162 $topo_graph->add_vertex( $rn );
1164 $rel_containers{$_->id} = $rn;
1167 # Add a new node to mirror the old node.
1168 $rel_containers{$r->id} = $r->id;
1169 $topo_graph->add_vertex( $r->id );
1174 foreach my $r ( $self->readings ) {
1175 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1176 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1177 $rel_containers{$n} );
1178 # $DB::single = 1 unless $tfrom && $tto;
1179 $topo_graph->add_edge( $tfrom, $tto );
1183 # Now do the rankings, starting with the start node.
1184 my $topo_start = $rel_containers{$self->start->id};
1185 my $node_ranks = { $topo_start => 0 };
1186 my @curr_origin = ( $topo_start );
1187 # A little iterative function.
1188 while( @curr_origin ) {
1189 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1191 # Transfer our rankings from the topological graph to the real one.
1192 foreach my $r ( $self->readings ) {
1193 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1194 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1196 die "No rank calculated for node " . $r->id
1197 . " - do you have a cycle in the graph?";
1203 my( $graph, $node_ranks, @current_nodes ) = @_;
1204 # Look at each of the children of @current_nodes. If all the child's
1205 # parents have a rank, assign it the highest rank + 1 and add it to
1206 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1207 # parent gets a rank.
1209 foreach my $c ( @current_nodes ) {
1210 warn "Current reading $c has no rank!"
1211 unless exists $node_ranks->{$c};
1212 # print STDERR "Looking at child of node $c, rank "
1213 # . $node_ranks->{$c} . "\n";
1214 foreach my $child ( $graph->successors( $c ) ) {
1215 next if exists $node_ranks->{$child};
1216 my $highest_rank = -1;
1218 foreach my $parent ( $graph->predecessors( $child ) ) {
1219 if( exists $node_ranks->{$parent} ) {
1220 $highest_rank = $node_ranks->{$parent}
1221 if $highest_rank <= $node_ranks->{$parent};
1228 my $c_rank = $highest_rank + 1;
1229 # print STDERR "Assigning rank $c_rank to node $child \n";
1230 $node_ranks->{$child} = $c_rank;
1231 push( @next_nodes, $child );
1237 =head2 flatten_ranks
1239 A convenience method for parsing collation data. Searches the graph for readings
1240 with the same text at the same rank, and merges any that are found.
1246 my %unique_rank_rdg;
1247 foreach my $rdg ( $self->readings ) {
1248 next unless $rdg->has_rank;
1249 my $key = $rdg->rank . "||" . $rdg->text;
1250 if( exists $unique_rank_rdg{$key} ) {
1252 # print STDERR "Combining readings at same rank: $key\n";
1253 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1255 $unique_rank_rdg{$key} = $rdg;
1261 =head1 UTILITY FUNCTIONS
1263 =head2 common_predecessor( $reading_a, $reading_b )
1265 Find the last reading that occurs in sequence before both the given readings.
1267 =head2 common_successor( $reading_a, $reading_b )
1269 Find the first reading that occurs in sequence after both the given readings.
1273 use Text::Tradition;
1275 my $cxfile = 't/data/Collatex-16.xml';
1276 my $t = Text::Tradition->new(
1278 'input' => 'CollateX',
1281 my $c = $t->collation;
1283 is( $c->common_predecessor( 'n9', 'n23' )->id,
1284 'n20', "Found correct common predecessor" );
1285 is( $c->common_successor( 'n9', 'n23' )->id,
1286 '#END#', "Found correct common successor" );
1288 is( $c->common_predecessor( 'n19', 'n17' )->id,
1289 'n16', "Found correct common predecessor for readings on same path" );
1290 is( $c->common_successor( 'n21', 'n26' )->id,
1291 '#END#', "Found correct common successor for readings on same path" );
1297 ## Return the closest reading that is a predecessor of both the given readings.
1298 sub common_predecessor {
1300 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1301 return $self->common_in_path( $r1, $r2, 'predecessors' );
1304 sub common_successor {
1306 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1307 return $self->common_in_path( $r1, $r2, 'successors' );
1310 sub common_in_path {
1311 my( $self, $r1, $r2, $dir ) = @_;
1312 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1313 $iter = $self->end->rank - $iter if $dir eq 'successors';
1315 my @last_checked = ( $r1, $r2 );
1317 while( !@candidates ) {
1319 foreach my $lc ( @last_checked ) {
1320 foreach my $p ( $lc->$dir ) {
1321 if( $all_seen{$p->id} ) {
1322 push( @candidates, $p );
1324 $all_seen{$p->id} = 1;
1325 push( @new_lc, $p );
1329 @last_checked = @new_lc;
1331 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1332 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1336 __PACKAGE__->meta->make_immutable;
1342 =item * Get rid of $backup in reading_sequence