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 return decode_utf8( $svg );
458 =head2 as_dot( \%options )
460 Returns a string that is the collation graph expressed in dot
461 (i.e. GraphViz) format. Options include:
476 my( $self, $opts ) = @_;
477 my $startrank = $opts->{'from'} if $opts;
478 my $endrank = $opts->{'to'} if $opts;
479 my $color_common = $opts->{'color_common'} if $opts;
480 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
481 && $self->end->rank > 100;
483 # Check the arguments
485 return if $endrank && $startrank > $endrank;
486 return if $startrank > $self->end->rank;
488 if( defined $endrank ) {
489 return if $endrank < 0;
490 $endrank = undef if $endrank == $self->end->rank;
493 # TODO consider making some of these things configurable
494 my $graph_name = $self->tradition->name;
495 $graph_name =~ s/[^\w\s]//g;
496 $graph_name = join( '_', split( /\s+/, $graph_name ) );
504 'fillcolor' => 'white',
509 'arrowhead' => 'open',
510 'color' => '#000000',
511 'fontcolor' => '#000000',
514 my $dot = sprintf( "digraph %s {\n", $graph_name );
515 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
516 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
518 # Output substitute start/end readings if necessary
520 $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n";
523 $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
525 if( $STRAIGHTENHACK ) {
527 $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n";
528 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
530 my %used; # Keep track of the readings that actually appear in the graph
531 foreach my $reading ( $self->readings ) {
532 # Only output readings within our rank range.
533 next if $startrank && $reading->rank < $startrank;
534 next if $endrank && $reading->rank > $endrank;
535 $used{$reading->id} = 1;
536 # Need not output nodes without separate labels
537 next if $reading->id eq $reading->text;
539 my $label = $reading->text;
540 $label =~ s/\"/\\\"/g;
541 $rattrs->{'label'} = $label;
542 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
543 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
547 my @edges = $self->paths;
548 my( %substart, %subend );
549 foreach my $edge ( @edges ) {
550 # Do we need to output this edge?
551 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
552 my $label = $self->path_display_label( $self->path_witnesses( $edge ) );
553 my $variables = { %edge_attrs, 'label' => $label };
554 # Account for the rank gap if necessary
555 if( $self->reading( $edge->[1] )->has_rank
556 && $self->reading( $edge->[0] )->has_rank
557 && $self->reading( $edge->[1] )->rank
558 - $self->reading( $edge->[0] )->rank > 1 ) {
559 $variables->{'minlen'} = $self->reading( $edge->[1] )->rank
560 - $self->reading( $edge->[0] )->rank;
562 # EXPERIMENTAL: make edge width reflect no. of witnesses
563 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
564 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
566 my $varopts = _dot_attr_string( $variables );
567 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
568 $edge->[0], $edge->[1], $varopts );
569 } elsif( $used{$edge->[0]} ) {
570 $subend{$edge->[0]} = 1;
571 } elsif( $used{$edge->[1]} ) {
572 $substart{$edge->[1]} = 1;
575 # Add substitute start and end edges if necessary
576 foreach my $node ( keys %substart ) {
577 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
578 my $variables = { %edge_attrs, 'label' => $witstr };
579 my $varopts = _dot_attr_string( $variables );
580 $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
582 foreach my $node ( keys %subend ) {
583 my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
584 my $variables = { %edge_attrs, 'label' => $witstr };
585 my $varopts = _dot_attr_string( $variables );
586 $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
589 if( $STRAIGHTENHACK ) {
590 $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
597 sub _dot_attr_string {
600 foreach my $k ( sort keys %$hash ) {
602 push( @attrs, $k.'="'.$v.'"' );
604 return( '[ ' . join( ', ', @attrs ) . ' ]' );
608 my( $self, @edge ) = @_;
609 # If edge is an arrayref, cope.
610 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
614 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
618 sub path_display_label {
621 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
622 if( scalar @wits > $maj ) {
623 # TODO break out a.c. wits
626 return join( ', ', @wits );
633 Returns a GraphML representation of the collation. The GraphML will contain
634 two graphs. The first expresses the attributes of the readings and the witness
635 paths that link them; the second expresses the relationships that link the
636 readings. This is the native transfer format for a tradition.
645 my $datafile = 't/data/florilegium_tei_ps.xml';
646 my $tradition = Text::Tradition->new( 'input' => 'TEI',
651 ok( $tradition, "Got a tradition object" );
652 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
653 ok( $tradition->collation, "Tradition has a collation" );
655 my $c = $tradition->collation;
656 is( scalar $c->readings, $READINGS, "Collation has all readings" );
657 is( scalar $c->paths, $PATHS, "Collation has all paths" );
658 is( scalar $c->relationships, 0, "Collation has all relationships" );
660 # Add a few relationships
661 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
662 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
663 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
665 # Now write it to GraphML and parse it again.
667 my $graphml = $c->as_graphml;
668 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
669 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
670 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
671 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
681 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
682 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
683 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
684 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
686 # Create the document and root node
687 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
688 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
689 $graphml->setDocumentElement( $root );
690 $root->setNamespace( $xsi_ns, 'xsi', 0 );
691 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
693 # Add the data keys for the graph
696 my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /;
697 foreach my $datum ( @graph_attributes ) {
698 $graph_data_keys{$datum} = 'dg'.$gdi++;
699 my $key = $root->addNewChild( $graphml_ns, 'key' );
700 $key->setAttribute( 'attr.name', $datum );
701 $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' );
702 $key->setAttribute( 'for', 'graph' );
703 $key->setAttribute( 'id', $graph_data_keys{$datum} );
706 # Add the data keys for nodes
713 is_start => 'boolean',
715 is_lacuna => 'boolean',
717 foreach my $datum ( keys %node_data ) {
718 $node_data_keys{$datum} = 'dn'.$ndi++;
719 my $key = $root->addNewChild( $graphml_ns, 'key' );
720 $key->setAttribute( 'attr.name', $datum );
721 $key->setAttribute( 'attr.type', $node_data{$datum} );
722 $key->setAttribute( 'for', 'node' );
723 $key->setAttribute( 'id', $node_data_keys{$datum} );
726 # Add the data keys for edges, i.e. witnesses
730 class => 'string', # Class, deprecated soon
731 witness => 'string', # ID/label for a path
732 relationship => 'string', # ID/label for a relationship
733 extra => 'boolean', # Path key
734 scope => 'string', # Relationship key
735 non_correctable => 'boolean', # Relationship key
736 non_independent => 'boolean', # Relationship key
738 foreach my $datum ( keys %edge_data ) {
739 $edge_data_keys{$datum} = 'de'.$edi++;
740 my $key = $root->addNewChild( $graphml_ns, 'key' );
741 $key->setAttribute( 'attr.name', $datum );
742 $key->setAttribute( 'attr.type', $edge_data{$datum} );
743 $key->setAttribute( 'for', 'edge' );
744 $key->setAttribute( 'id', $edge_data_keys{$datum} );
747 # Add the collation graph itself
748 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
749 $sgraph->setAttribute( 'edgedefault', 'directed' );
750 $sgraph->setAttribute( 'id', $self->tradition->name );
751 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
752 $sgraph->setAttribute( 'parse.edges', scalar($self->paths) );
753 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
754 $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) );
755 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
757 # Collation attribute data
758 foreach my $datum ( @graph_attributes ) {
759 my $value = $datum eq 'version' ? '3.0' : $self->$datum;
760 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
765 # Add our readings to the graph
766 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
767 # Add to the main graph
768 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
769 my $node_xmlid = 'n' . $node_ctr++;
770 $node_hash{ $n->id } = $node_xmlid;
771 $node_el->setAttribute( 'id', $node_xmlid );
772 foreach my $d ( keys %node_data ) {
774 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
779 # Add the path edges to the sequence graph
781 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
782 # We add an edge in the graphml for every witness in $e.
783 foreach my $wit ( sort $self->path_witnesses( $e ) ) {
784 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
785 $node_hash{ $e->[0] },
786 $node_hash{ $e->[1] } );
787 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
788 $edge_el->setAttribute( 'source', $from );
789 $edge_el->setAttribute( 'target', $to );
790 $edge_el->setAttribute( 'id', $id );
792 # It's a witness path, so add the witness
794 my $key = $edge_data_keys{'witness'};
795 # Is this an ante-corr witness?
796 my $aclabel = $self->ac_label;
797 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
798 # Keep the base witness
800 # ...and record that this is an 'extra' reading path
801 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
803 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
804 _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' );
808 # Add the relationship graph to the XML
809 $self->relations->as_graphml( $graphml_ns, $root, \%node_hash,
810 $node_data_keys{'id'}, \%edge_data_keys );
812 # Save and return the thing
813 my $result = decode_utf8( $graphml->toString(1) );
817 sub _add_graphml_data {
818 my( $el, $key, $value ) = @_;
819 return unless defined $value;
820 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
821 $data_el->setAttribute( 'key', $key );
822 $data_el->appendText( $value );
827 Returns a CSV alignment table representation of the collation graph, one
828 row per witness (or witness uncorrected.)
834 my $table = $self->make_alignment_table;
835 my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );
837 # Make the header row
838 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
839 push( @result, decode_utf8( $csv->string ) );
840 # Make the rest of the rows
841 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
842 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
843 my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs;
844 $csv->combine( @row );
845 push( @result, decode_utf8( $csv->string ) );
847 return join( "\n", @result );
850 =head2 make_alignment_table( $use_refs, $include_witnesses )
852 Return a reference to an alignment table, in a slightly enhanced CollateX
853 format which looks like this:
855 $table = { alignment => [ { witness => "SIGIL",
856 tokens => [ { t => "TEXT" }, ... ] },
858 tokens => [ { t => "TEXT" }, ... ] },
862 If $use_refs is set to 1, the reading object is returned in the table
863 instead of READINGTEXT; if not, the text of the reading is returned.
865 If $include_witnesses is set to a hashref, only the witnesses whose sigil
866 keys have a true hash value will be included.
870 sub make_alignment_table {
871 my( $self, $noderefs, $include ) = @_;
872 # Make sure we can do this
873 throw( "Need a linear graph in order to make an alignment table" )
874 unless $self->linear;
875 $self->calculate_ranks unless $self->end->has_rank;
877 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
878 my @all_pos = ( 1 .. $self->end->rank - 1 );
879 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
881 next unless $include->{$wit->sigil};
883 # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
884 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
885 my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs );
886 push( @{$table->{'alignment'}},
887 { 'witness' => $wit->sigil, 'tokens' => \@row } );
888 if( $wit->is_layered ) {
889 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
890 $wit->sigil.$self->ac_label );
891 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
892 push( @{$table->{'alignment'}},
893 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
899 sub _make_witness_row {
900 my( $path, $positions, $noderefs ) = @_;
902 map { $char_hash{$_} = undef } @$positions;
904 foreach my $rdg ( @$path ) {
905 my $rtext = $rdg->text;
906 $rtext = '#LACUNA#' if $rdg->is_lacuna;
907 print STDERR "rank " . $rdg->rank . "\n" if $debug;
908 # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank;
909 $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg }
912 my @row = map { $char_hash{$_} } @$positions;
913 # Fill in lacuna markers for undef spots in the row
914 my $last_el = shift @row;
915 my @filled_row = ( $last_el );
916 foreach my $el ( @row ) {
917 # If we are using node reference, make the lacuna node appear many times
918 # in the table. If not, use the lacuna tag.
919 if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) {
920 $el = $noderefs ? $last_el : { 't' => '#LACUNA#' };
922 push( @filled_row, $el );
928 # Tiny utility function to say if a table element is a lacuna
931 return 1 if $el->{'t'} eq '#LACUNA#';
932 return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading'
933 && $el->{'t'}->is_lacuna;
937 # Helper to turn the witnesses along columns rather than rows. Assumes
942 return $result unless scalar @$table;
943 my $nrows = scalar @{$table->[0]};
944 foreach my $idx ( 0 .. $nrows - 1 ) {
945 foreach my $wit ( 0 .. $#{$table} ) {
946 $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
952 =head1 NAVIGATION METHODS
954 =head2 reading_sequence( $first, $last, $sigil, $backup )
956 Returns the ordered list of readings, starting with $first and ending
957 with $last, for the witness given in $sigil. If a $backup sigil is
958 specified (e.g. when walking a layered witness), it will be used wherever
959 no $sigil path exists. If there is a base text reading, that will be
960 used wherever no path exists for $sigil or $backup.
964 # TODO Think about returning some lazy-eval iterator.
965 # TODO Get rid of backup; we should know from what witness is whether we need it.
967 sub reading_sequence {
968 my( $self, $start, $end, $witness ) = @_;
970 $witness = $self->baselabel unless $witness;
971 my @readings = ( $start );
974 while( $n && $n->id ne $end->id ) {
975 if( exists( $seen{$n->id} ) ) {
976 throw( "Detected loop for $witness at " . $n->id );
980 my $next = $self->next_reading( $n, $witness );
982 throw( "Did not find any path for $witness from reading " . $n->id );
984 push( @readings, $next );
987 # Check that the last reading is our end reading.
988 my $last = $readings[$#readings];
989 throw( "Last reading found from " . $start->text .
990 " for witness $witness is not the end!" ) # TODO do we get this far?
991 unless $last->id eq $end->id;
996 =head2 next_reading( $reading, $sigil );
998 Returns the reading that follows the given reading along the given witness
1004 # Return the successor via the corresponding path.
1006 my $answer = $self->_find_linked_reading( 'next', @_ );
1007 return undef unless $answer;
1008 return $self->reading( $answer );
1011 =head2 prior_reading( $reading, $sigil )
1013 Returns the reading that precedes the given reading along the given witness
1019 # Return the predecessor via the corresponding path.
1021 my $answer = $self->_find_linked_reading( 'prior', @_ );
1022 return $self->reading( $answer );
1025 sub _find_linked_reading {
1026 my( $self, $direction, $node, $path ) = @_;
1028 # Get a backup if we are dealing with a layered witness
1030 my $aclabel = $self->ac_label;
1031 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1035 my @linked_paths = $direction eq 'next'
1036 ? $self->sequence->edges_from( $node )
1037 : $self->sequence->edges_to( $node );
1038 return undef unless scalar( @linked_paths );
1040 # We have to find the linked path that contains all of the
1041 # witnesses supplied in $path.
1042 my( @path_wits, @alt_path_wits );
1043 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1044 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1047 foreach my $le ( @linked_paths ) {
1048 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1051 my @le_wits = sort $self->path_witnesses( $le );
1052 if( _is_within( \@path_wits, \@le_wits ) ) {
1053 # This is the right path.
1054 return $direction eq 'next' ? $le->[1] : $le->[0];
1055 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1059 # Got this far? Return the alternate path if it exists.
1060 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1063 # Got this far? Return the base path if it exists.
1064 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1067 # Got this far? We have no appropriate path.
1068 warn "Could not find $direction node from " . $node->id
1069 . " along path $path";
1075 my( $set1, $set2 ) = @_;
1076 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1077 foreach my $el ( @$set1 ) {
1078 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1083 # Return the string that joins together a list of witnesses for
1084 # display on a single path.
1085 sub _witnesses_of_label {
1086 my( $self, $label ) = @_;
1087 my $regex = $self->wit_list_separator;
1088 my @answer = split( /\Q$regex\E/, $label );
1092 =head2 common_readings
1094 Returns the list of common readings in the graph (i.e. those readings that are
1095 shared by all non-lacunose witnesses.)
1099 sub common_readings {
1101 my @common = grep { $_->is_common } $self->readings;
1105 =head2 path_text( $sigil, $mainsigil [, $start, $end ] )
1107 Returns the text of a witness (plus its backup, if we are using a layer)
1108 as stored in the collation. The text is returned as a string, where the
1109 individual readings are joined with spaces and the meta-readings (e.g.
1110 lacunae) are omitted. Optional specification of $start and $end allows
1111 the generation of a subset of the witness text.
1116 my( $self, $wit, $start, $end ) = @_;
1117 $start = $self->start unless $start;
1118 $end = $self->end unless $end;
1119 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1120 return join( ' ', map { $_->text } @path );
1123 =head1 INITIALIZATION METHODS
1125 These are mostly for use by parsers.
1127 =head2 make_witness_path( $witness )
1129 Link the array of readings contained in $witness->path (and in
1130 $witness->uncorrected_path if it exists) into collation paths.
1131 Clear out the arrays when finished.
1133 =head2 make_witness_paths
1135 Call make_witness_path for all witnesses in the tradition.
1139 # For use when a collation is constructed from a base text and an apparatus.
1140 # We have the sequences of readings and just need to add path edges.
1141 # When we are done, clear out the witness path attributes, as they are no
1143 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1145 sub make_witness_paths {
1147 foreach my $wit ( $self->tradition->witnesses ) {
1148 # print STDERR "Making path for " . $wit->sigil . "\n";
1149 $self->make_witness_path( $wit );
1153 sub make_witness_path {
1154 my( $self, $wit ) = @_;
1155 my @chain = @{$wit->path};
1156 my $sig = $wit->sigil;
1157 foreach my $idx ( 0 .. $#chain-1 ) {
1158 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1160 if( $wit->is_layered ) {
1161 @chain = @{$wit->uncorrected_path};
1162 foreach my $idx( 0 .. $#chain-1 ) {
1163 my $source = $chain[$idx];
1164 my $target = $chain[$idx+1];
1165 $self->add_path( $source, $target, $sig.$self->ac_label )
1166 unless $self->has_path( $source, $target, $sig );
1170 $wit->clear_uncorrected_path;
1173 =head2 calculate_ranks
1175 Calculate the reading ranks (that is, their aligned positions relative
1176 to each other) for the graph. This can only be called on linear collations.
1180 sub calculate_ranks {
1182 # Walk a version of the graph where every node linked by a relationship
1183 # edge is fundamentally the same node, and do a topological ranking on
1184 # the nodes in this graph.
1185 my $topo_graph = Graph->new();
1189 foreach my $r ( $self->readings ) {
1190 next if exists $rel_containers{$r->id};
1191 my @rels = $r->related_readings( 'colocated' );
1193 # Make a relationship container.
1195 my $rn = 'rel_container_' . $rel_ctr++;
1196 $topo_graph->add_vertex( $rn );
1198 $rel_containers{$_->id} = $rn;
1201 # Add a new node to mirror the old node.
1202 $rel_containers{$r->id} = $r->id;
1203 $topo_graph->add_vertex( $r->id );
1208 foreach my $r ( $self->readings ) {
1209 foreach my $n ( $self->sequence->successors( $r->id ) ) {
1210 my( $tfrom, $tto ) = ( $rel_containers{$r->id},
1211 $rel_containers{$n} );
1212 # $DB::single = 1 unless $tfrom && $tto;
1213 $topo_graph->add_edge( $tfrom, $tto );
1217 # Now do the rankings, starting with the start node.
1218 my $topo_start = $rel_containers{$self->start->id};
1219 my $node_ranks = { $topo_start => 0 };
1220 my @curr_origin = ( $topo_start );
1221 # A little iterative function.
1222 while( @curr_origin ) {
1223 @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin );
1225 # Transfer our rankings from the topological graph to the real one.
1226 foreach my $r ( $self->readings ) {
1227 if( defined $node_ranks->{$rel_containers{$r->id}} ) {
1228 $r->rank( $node_ranks->{$rel_containers{$r->id}} );
1230 # Die. Find the last rank we calculated.
1231 my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
1232 <=> $node_ranks->{$rel_containers{$b->id}} }
1234 my $last = pop @all_defined;
1235 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1241 my( $graph, $node_ranks, @current_nodes ) = @_;
1242 # Look at each of the children of @current_nodes. If all the child's
1243 # parents have a rank, assign it the highest rank + 1 and add it to
1244 # @next_nodes. Otherwise skip it; we will return when the highest-ranked
1245 # parent gets a rank.
1247 foreach my $c ( @current_nodes ) {
1248 warn "Current reading $c has no rank!"
1249 unless exists $node_ranks->{$c};
1250 # print STDERR "Looking at child of node $c, rank "
1251 # . $node_ranks->{$c} . "\n";
1252 foreach my $child ( $graph->successors( $c ) ) {
1253 next if exists $node_ranks->{$child};
1254 my $highest_rank = -1;
1256 foreach my $parent ( $graph->predecessors( $child ) ) {
1257 if( exists $node_ranks->{$parent} ) {
1258 $highest_rank = $node_ranks->{$parent}
1259 if $highest_rank <= $node_ranks->{$parent};
1266 my $c_rank = $highest_rank + 1;
1267 # print STDERR "Assigning rank $c_rank to node $child \n";
1268 $node_ranks->{$child} = $c_rank;
1269 push( @next_nodes, $child );
1275 =head2 flatten_ranks
1277 A convenience method for parsing collation data. Searches the graph for readings
1278 with the same text at the same rank, and merges any that are found.
1284 my %unique_rank_rdg;
1285 foreach my $rdg ( $self->readings ) {
1286 next unless $rdg->has_rank;
1287 my $key = $rdg->rank . "||" . $rdg->text;
1288 if( exists $unique_rank_rdg{$key} ) {
1290 # print STDERR "Combining readings at same rank: $key\n";
1291 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1292 # TODO see if this now makes a common point.
1294 $unique_rank_rdg{$key} = $rdg;
1299 =head2 calculate_common_readings
1301 Goes through the graph identifying the readings that appear in every witness
1302 (apart from those with lacunae at that spot.) Marks them as common and returns
1307 use Text::Tradition;
1309 my $cxfile = 't/data/Collatex-16.xml';
1310 my $t = Text::Tradition->new(
1312 'input' => 'CollateX',
1315 my $c = $t->collation;
1317 my @common = $c->calculate_common_readings();
1318 is( scalar @common, 8, "Found correct number of common readings" );
1319 my @marked = sort $c->common_readings();
1320 is( scalar @common, 8, "All common readings got marked as such" );
1321 my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
1322 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1328 sub calculate_common_readings {
1331 my $table = $self->make_alignment_table( 1 );
1332 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1333 my @row = map { $_->{'tokens'}->[$idx]->{'t'} } @{$table->{'alignment'}};
1335 foreach my $r ( @row ) {
1337 $hash{$r->id} = $r unless $r->is_meta;
1339 $hash{'UNDEF'} = $r;
1342 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1343 my( $r ) = values %hash;
1345 push( @common, $r );
1351 =head2 text_from_paths
1353 Calculate the text array for all witnesses from the path, for later consistency
1354 checking. Only to be used if there is no non-graph-based way to know the
1359 sub text_from_paths {
1361 foreach my $wit ( $self->tradition->witnesses ) {
1362 my @text = split( /\s+/,
1363 $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
1364 $wit->text( \@text );
1365 if( $wit->is_layered ) {
1366 my @uctext = split( /\s+/,
1367 $self->reading_sequence( $self->start, $self->end,
1368 $wit->sigil.$self->ac_label ) );
1369 $wit->text( \@uctext );
1374 =head1 UTILITY FUNCTIONS
1376 =head2 common_predecessor( $reading_a, $reading_b )
1378 Find the last reading that occurs in sequence before both the given readings.
1380 =head2 common_successor( $reading_a, $reading_b )
1382 Find the first reading that occurs in sequence after both the given readings.
1386 use Text::Tradition;
1388 my $cxfile = 't/data/Collatex-16.xml';
1389 my $t = Text::Tradition->new(
1391 'input' => 'CollateX',
1394 my $c = $t->collation;
1396 is( $c->common_predecessor( 'n9', 'n23' )->id,
1397 'n20', "Found correct common predecessor" );
1398 is( $c->common_successor( 'n9', 'n23' )->id,
1399 '#END#', "Found correct common successor" );
1401 is( $c->common_predecessor( 'n19', 'n17' )->id,
1402 'n16', "Found correct common predecessor for readings on same path" );
1403 is( $c->common_successor( 'n21', 'n26' )->id,
1404 '#END#', "Found correct common successor for readings on same path" );
1410 ## Return the closest reading that is a predecessor of both the given readings.
1411 sub common_predecessor {
1413 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1414 return $self->common_in_path( $r1, $r2, 'predecessors' );
1417 sub common_successor {
1419 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1420 return $self->common_in_path( $r1, $r2, 'successors' );
1423 sub common_in_path {
1424 my( $self, $r1, $r2, $dir ) = @_;
1425 my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank;
1426 $iter = $self->end->rank - $iter if $dir eq 'successors';
1428 my @last_checked = ( $r1, $r2 );
1430 while( !@candidates ) {
1432 foreach my $lc ( @last_checked ) {
1433 foreach my $p ( $lc->$dir ) {
1434 if( $all_seen{$p->id} ) {
1435 push( @candidates, $p );
1437 $all_seen{$p->id} = 1;
1438 push( @new_lc, $p );
1442 @last_checked = @new_lc;
1444 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1445 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1449 Text::Tradition::Error->throw(
1450 'ident' => 'Collation error',
1456 __PACKAGE__->meta->make_immutable;
1462 =item * Get rid of $backup in reading_sequence