1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Data;
11 use Text::Tradition::Collation::Reading;
12 use Text::Tradition::Collation::RelationshipStore;
13 use Text::Tradition::Error;
14 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
16 use XML::LibXML::XPathContext;
20 isa => 'Text::Tradition::Collation::Data',
58 isa => 'Text::Tradition',
59 writer => '_set_tradition',
65 Text::Tradition::Collation - a software model for a text collation
70 my $t = Text::Tradition->new(
71 'name' => 'this is a text',
73 'file' => '/path/to/tei_parallel_seg_file.xml' );
75 my $c = $t->collation;
76 my @readings = $c->readings;
77 my @paths = $c->paths;
78 my @relationships = $c->relationships;
80 my $svg_variant_graph = $t->collation->as_svg();
84 Text::Tradition is a library for representation and analysis of collated
85 texts, particularly medieval ones. The Collation is the central feature of
86 a Tradition, where the text, its sequence of readings, and its relationships
87 between readings are actually kept.
93 The constructor. Takes a hash or hashref of the following arguments:
97 =item * tradition - The Text::Tradition object to which the collation
100 =item * linear - Whether the collation should be linear; that is, whether
101 transposed readings should be treated as two linked readings rather than one,
102 and therefore whether the collation graph is acyclic. Defaults to true.
104 =item * baselabel - The default label for the path taken by a base text
105 (if any). Defaults to 'base text'.
107 =item * wit_list_separator - The string to join a list of witnesses for
108 purposes of making labels in display graphs. Defaults to ', '.
110 =item * ac_label - The extra label to tack onto a witness sigil when
111 representing another layer of path for the given witness - that is, when
112 a text has more than one possible reading due to scribal corrections or
113 the like. Defaults to ' (a.c.)'.
115 =item * wordsep - The string used to separate words in the original text.
126 =head2 wit_list_separator
134 Simple accessors for collation attributes.
138 The meta-reading at the start of every witness path.
142 The meta-reading at the end of every witness path.
146 Returns all Reading objects in the graph.
148 =head2 reading( $id )
150 Returns the Reading object corresponding to the given ID.
152 =head2 add_reading( $reading_args )
154 Adds a new reading object to the collation.
155 See L<Text::Tradition::Collation::Reading> for the available arguments.
157 =head2 del_reading( $object_or_id )
159 Removes the given reading from the collation, implicitly removing its
160 paths and relationships.
162 =head2 merge_readings( $main, $second, $concatenate, $with_str )
164 Merges the $second reading into the $main one. If $concatenate is true, then
165 the merged node will carry the text of both readings, concatenated with either
166 $with_str (if specified) or a sensible default (the empty string if the
167 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
169 The first two arguments may be either readings or reading IDs.
171 =head2 has_reading( $id )
173 Predicate to see whether a given reading ID is in the graph.
175 =head2 reading_witnesses( $object_or_id )
177 Returns a list of sigils whose witnesses contain the reading.
181 Returns all reading paths within the document - that is, all edges in the
182 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
184 =head2 add_path( $source, $target, $sigil )
186 Links the given readings in the collation in sequence, under the given witness
187 sigil. The readings may be specified by object or ID.
189 =head2 del_path( $source, $target, $sigil )
191 Links the given readings in the collation in sequence, under the given witness
192 sigil. The readings may be specified by object or ID.
194 =head2 has_path( $source, $target );
196 Returns true if the two readings are linked in sequence in any witness.
197 The readings may be specified by object or ID.
201 Returns all Relationship objects in the collation.
203 =head2 add_relationship( $reading, $other_reading, $options )
205 Adds a new relationship of the type given in $options between the two readings,
206 which may be specified by object or ID. Returns a value of ( $status, @vectors)
207 where $status is true on success, and @vectors is a list of relationship edges
208 that were ultimately added.
209 See L<Text::Tradition::Collation::Relationship> for the available options.
214 my ( $class, @args ) = @_;
215 my %args = @args == 1 ? %{ $args[0] } : @args;
216 # TODO determine these from the Moose::Meta object
217 my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel
218 linear wordsep start end cached_table _graphcalc_done);
220 for my $attr (@delegate_attrs) {
221 $data_args{$attr} = delete $args{$attr} if exists $args{$attr};
223 $args{_data} = Text::Tradition::Collation::Data->new(%data_args);
229 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
230 $self->_set_start( $self->add_reading(
231 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
232 $self->_set_end( $self->add_reading(
233 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
236 sub register_relationship_type {
238 my %args = @_ == 1 ? %{$_[0]} : @_;
239 if( $self->relations->has_type( $args{name} ) ) {
240 throw( 'Relationship type ' . $args{name} . ' already registered' );
242 $self->relations->add_type( %args );
245 ### Reading construct/destruct functions
248 my( $self, $reading ) = @_;
249 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
250 my %args = %$reading;
251 if( $args{'init'} ) {
252 # If we are initializing an empty collation, don't assume that we
253 # have set a tradition.
254 delete $args{'init'};
255 } elsif( $self->tradition->can('language') && $self->tradition->has_language
256 && !exists $args{'language'} ) {
257 $args{'language'} = $self->tradition->language;
259 $reading = Text::Tradition::Collation::Reading->new(
260 'collation' => $self,
263 # First check to see if a reading with this ID exists.
264 if( $self->reading( $reading->id ) ) {
265 throw( "Collation already has a reading with id " . $reading->id );
267 $self->_graphcalc_done(0);
268 $self->_add_reading( $reading->id => $reading );
269 # Once the reading has been added, put it in both graphs.
270 $self->sequence->add_vertex( $reading->id );
271 $self->relations->add_reading( $reading->id );
275 around del_reading => sub {
280 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
283 # Remove the reading from the graphs.
284 $self->_graphcalc_done(0);
285 $self->_clear_cache; # Explicitly clear caches to GC the reading
286 $self->sequence->delete_vertex( $arg );
287 $self->relations->delete_reading( $arg );
290 $self->$orig( $arg );
297 my $cxfile = 't/data/Collatex-16.xml';
298 my $t = Text::Tradition->new(
300 'input' => 'CollateX',
303 my $c = $t->collation;
305 my $rno = scalar $c->readings;
306 # Split n21 for testing purposes
307 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
308 my $old_r = $c->reading( 'n21' );
309 $old_r->alter_text( 'to' );
310 $c->del_path( 'n20', 'n21', 'A' );
311 $c->add_path( 'n20', 'n21p0', 'A' );
312 $c->add_path( 'n21p0', 'n21', 'A' );
314 ok( $c->reading( 'n21p0' ), "New reading exists" );
315 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
317 # Combine n3 and n4 ( with his )
318 $c->merge_readings( 'n3', 'n4', 1 );
319 ok( !$c->reading('n4'), "Reading n4 is gone" );
320 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
322 # Collapse n9 and n10 ( rood / root )
323 $c->merge_readings( 'n9', 'n10' );
324 ok( !$c->reading('n10'), "Reading n10 is gone" );
325 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
327 # Combine n21 and n21p0
328 my $remaining = $c->reading('n21');
329 $remaining ||= $c->reading('n22'); # one of these should still exist
330 $c->merge_readings( 'n21p0', $remaining, 1 );
331 ok( !$c->reading('n21'), "Reading $remaining is gone" );
332 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
342 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
343 my $mergemeta = $kept_obj->is_meta;
344 throw( "Cannot merge meta and non-meta reading" )
345 unless ( $mergemeta && $del_obj->is_meta )
346 || ( !$mergemeta && !$del_obj->is_meta );
348 throw( "Cannot merge with start or end node" )
349 if( $kept_obj eq $self->start || $kept_obj eq $self->end
350 || $del_obj eq $self->start || $del_obj eq $self->end );
351 throw( "Cannot combine text of meta readings" ) if $combine;
353 # We only need the IDs for adding paths to the graph, not the reading
354 # objects themselves.
355 my $kept = $kept_obj->id;
356 my $deleted = $del_obj->id;
357 $self->_graphcalc_done(0);
359 # The kept reading should inherit the paths and the relationships
360 # of the deleted reading.
361 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
362 my @vector = ( $kept );
363 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
364 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
365 next if $vector[0] eq $vector[1]; # Don't add a self loop
366 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
367 $self->sequence->add_edge( @vector );
368 my $fwits = $self->sequence->get_edge_attributes( @vector );
369 @wits{keys %$fwits} = values %$fwits;
370 $self->sequence->set_edge_attributes( @vector, \%wits );
372 $self->relations->merge_readings( $kept, $deleted, $combine );
374 # Do the deletion deed.
376 # Combine the text of the readings
377 my $joinstr = $combine_char;
378 unless( defined $joinstr ) {
379 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
380 $joinstr = $self->wordsep unless defined $joinstr;
382 $kept_obj->_combine( $del_obj, $joinstr );
384 $self->del_reading( $deleted );
387 =head2 compress_readings
389 Where possible in the graph, compresses plain sequences of readings into a
390 single reading. The sequences must consist of readings with no
391 relationships to other readings, with only a single witness path between
392 them and no other witness paths from either that would skip the other. The
393 readings must also not be marked as nonsense or bad grammar.
395 WARNING: This operation cannot be undone.
399 sub compress_readings {
401 # Anywhere in the graph that there is a reading that joins only to a single
402 # successor, and neither of these have any relationships, just join the two
404 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
405 # Now look for readings that can be joined to their successors.
406 next unless $rdg->is_combinable;
408 while( $self->sequence->successors( $rdg ) == 1 ) {
409 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
410 throw( "Infinite loop" ) if $seen{$next->id};
411 $seen{$next->id} = 1;
412 last if $self->sequence->predecessors( $next ) > 1;
413 last unless $next->is_combinable;
414 say "Joining readings $rdg and $next";
415 $self->merge_readings( $rdg, $next, 1 );
418 # Make sure we haven't screwed anything up
419 foreach my $wit ( $self->tradition->witnesses ) {
420 my $pathtext = $self->path_text( $wit->sigil );
421 my $origtext = join( ' ', @{$wit->text} );
422 throw( "Text differs for witness " . $wit->sigil )
423 unless $pathtext eq $origtext;
424 if( $wit->is_layered ) {
425 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
426 $origtext = join( ' ', @{$wit->layertext} );
427 throw( "Ante-corr text differs for witness " . $wit->sigil )
428 unless $pathtext eq $origtext;
432 $self->relations->rebuild_equivalence();
433 $self->calculate_ranks();
436 # Helper function for manipulating the graph.
437 sub _stringify_args {
438 my( $self, $first, $second, @args ) = @_;
440 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
441 $second = $second->id
442 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
443 return( $first, $second, @args );
446 # Helper function for manipulating the graph.
447 sub _objectify_args {
448 my( $self, $first, $second, $arg ) = @_;
449 $first = $self->reading( $first )
450 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
451 $second = $self->reading( $second )
452 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
453 return( $first, $second, $arg );
460 # We only need the IDs for adding paths to the graph, not the reading
461 # objects themselves.
462 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
464 $self->_graphcalc_done(0);
465 # Connect the readings
466 unless( $self->sequence->has_edge( $source, $target ) ) {
467 $self->sequence->add_edge( $source, $target );
468 $self->relations->add_equivalence_edge( $source, $target );
470 # Note the witness in question
471 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
477 if( ref( $_[0] ) eq 'ARRAY' ) {
484 # We only need the IDs for adding paths to the graph, not the reading
485 # objects themselves.
486 my( $source, $target, $wit ) = $self->_stringify_args( @args );
488 $self->_graphcalc_done(0);
489 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
490 $self->sequence->delete_edge_attribute( $source, $target, $wit );
492 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
493 $self->sequence->delete_edge( $source, $target );
494 $self->relations->delete_equivalence_edge( $source, $target );
499 # Extra graph-alike utility
502 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
503 return undef unless $self->sequence->has_edge( $source, $target );
504 return $self->sequence->has_edge_attribute( $source, $target, $wit );
507 =head2 clear_witness( @sigil_list )
509 Clear the given witnesses out of the collation entirely, removing references
510 to them in paths, and removing readings that belong only to them. Should only
511 be called via $tradition->del_witness.
516 my( $self, @sigils ) = @_;
518 $self->_graphcalc_done(0);
519 # Clear the witness(es) out of the paths
520 foreach my $e ( $self->paths ) {
521 foreach my $sig ( @sigils ) {
522 $self->del_path( $e, $sig );
526 # Clear out the newly unused readings
527 foreach my $r ( $self->readings ) {
528 unless( $self->reading_witnesses( $r ) ) {
529 $self->del_reading( $r );
534 sub add_relationship {
536 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
537 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
538 $self->_graphcalc_done(0);
542 around qw/ get_relationship del_relationship / => sub {
546 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
549 my( $source, $target ) = $self->_stringify_args( @args );
550 $self->$orig( $source, $target );
553 =head2 reading_witnesses( $reading )
555 Return a list of sigils corresponding to the witnesses in which the reading appears.
559 sub reading_witnesses {
560 my( $self, $reading ) = @_;
561 # We need only check either the incoming or the outgoing edges; I have
562 # arbitrarily chosen "incoming". Thus, special-case the start node.
563 if( $reading eq $self->start ) {
564 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
567 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
568 my $wits = $self->sequence->get_edge_attributes( @$e );
569 @all_witnesses{ keys %$wits } = 1;
571 my $acstr = $self->ac_label;
572 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
573 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
575 return keys %all_witnesses;
578 =head1 OUTPUT METHODS
580 =head2 as_svg( \%options )
582 Returns an SVG string that represents the graph, via as_dot and graphviz.
583 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
588 my( $self, $opts ) = @_;
589 throw( "Need GraphViz installed to output SVG" )
590 unless File::Which::which( 'dot' );
591 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
592 $self->calculate_ranks()
593 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
594 my @cmd = qw/dot -Tsvg/;
596 my $dotfile = File::Temp->new();
598 # $dotfile->unlink_on_destroy(0);
599 binmode $dotfile, ':utf8';
600 print $dotfile $self->as_dot( $opts );
601 push( @cmd, $dotfile->filename );
602 run( \@cmd, ">", binary(), \$svg );
603 $svg = decode_utf8( $svg );
608 =head2 as_dot( \%options )
610 Returns a string that is the collation graph expressed in dot
611 (i.e. GraphViz) format. Options include:
626 my( $self, $opts ) = @_;
627 my $startrank = $opts->{'from'} if $opts;
628 my $endrank = $opts->{'to'} if $opts;
629 my $color_common = $opts->{'color_common'} if $opts;
630 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
631 && $self->end->rank > 100;
632 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
634 # Check the arguments
636 return if $endrank && $startrank > $endrank;
637 return if $startrank > $self->end->rank;
639 if( defined $endrank ) {
640 return if $endrank < 0;
641 $endrank = undef if $endrank == $self->end->rank;
644 my $graph_name = $self->tradition->name;
645 $graph_name =~ s/[^\w\s]//g;
646 $graph_name = join( '_', split( /\s+/, $graph_name ) );
654 'fillcolor' => 'white',
659 'arrowhead' => 'open',
660 'color' => '#000000',
661 'fontcolor' => '#000000',
664 my $dot = sprintf( "digraph %s {\n", $graph_name );
665 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
666 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
668 # Output substitute start/end readings if necessary
670 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
673 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
675 if( $STRAIGHTENHACK ) {
677 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
678 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
679 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
681 my %used; # Keep track of the readings that actually appear in the graph
682 # Sort the readings by rank if we have ranks; this speeds layout.
683 my @all_readings = $self->end->has_rank
684 ? sort { $a->rank <=> $b->rank } $self->readings
686 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
687 foreach my $reading ( @all_readings ) {
688 # Only output readings within our rank range.
689 next if $startrank && $reading->rank < $startrank;
690 next if $endrank && $reading->rank > $endrank;
691 $used{$reading->id} = 1;
692 # Need not output nodes without separate labels
693 next if $reading->id eq $reading->text;
695 my $label = $reading->text;
696 $label .= '-' if $reading->join_next;
697 $label = "-$label" if $reading->join_prior;
698 $label =~ s/\"/\\\"/g;
699 $rattrs->{'label'} = $label;
700 $rattrs->{'id'} = $reading->id;
701 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
702 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
705 # Add the real edges. Need to weight one edge per rank jump, in a
707 # my $weighted = $self->_add_edge_weights;
708 my @edges = $self->paths;
709 my( %substart, %subend );
710 foreach my $edge ( @edges ) {
711 # Do we need to output this edge?
712 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
713 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
714 my $variables = { %edge_attrs, 'label' => $label };
716 # Account for the rank gap if necessary
717 my $rank0 = $self->reading( $edge->[0] )->rank
718 if $self->reading( $edge->[0] )->has_rank;
719 my $rank1 = $self->reading( $edge->[1] )->rank
720 if $self->reading( $edge->[1] )->has_rank;
721 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
722 $variables->{'minlen'} = $rank1 - $rank0;
725 # Add the calculated edge weights
726 # if( exists $weighted->{$edge->[0]}
727 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
728 # # $variables->{'color'} = 'red';
729 # $variables->{'weight'} = 3.0;
732 # EXPERIMENTAL: make edge width reflect no. of witnesses
733 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
734 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
736 my $varopts = _dot_attr_string( $variables );
737 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
738 $edge->[0], $edge->[1], $varopts );
739 } elsif( $used{$edge->[0]} ) {
740 $subend{$edge->[0]} = $edge->[1];
741 } elsif( $used{$edge->[1]} ) {
742 $substart{$edge->[1]} = $edge->[0];
746 # If we are asked to, add relationship links
747 if( exists $opts->{show_relations} ) {
748 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
749 if( $filter eq 'transposition' ) {
750 $filter =~ qr/^transposition$/;
752 foreach my $redge ( $self->relationships ) {
753 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
754 if( $filter ne 'all' ) {
755 my $rel = $self->get_relationship( $redge );
756 next unless $rel->type =~ /$filter/;
760 constraint => 'false',
761 label => uc( substr( $rel->type, 0, 4 ) ),
764 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
765 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
771 # Add substitute start and end edges if necessary
772 foreach my $node ( keys %substart ) {
773 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
774 my $variables = { %edge_attrs, 'label' => $witstr };
775 my $nrdg = $self->reading( $node );
776 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
777 # Substart is actually one lower than $startrank
778 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
780 my $varopts = _dot_attr_string( $variables );
781 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
783 foreach my $node ( keys %subend ) {
784 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
785 my $variables = { %edge_attrs, 'label' => $witstr };
786 my $varopts = _dot_attr_string( $variables );
787 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
790 if( $STRAIGHTENHACK ) {
791 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
792 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
799 sub _dot_attr_string {
802 foreach my $k ( sort keys %$hash ) {
804 push( @attrs, $k.'="'.$v.'"' );
806 return( '[ ' . join( ', ', @attrs ) . ' ]' );
809 sub _add_edge_weights {
811 # Walk the graph from START to END, choosing the successor node with
812 # the largest number of witness paths each time.
814 my $curr = $self->start->id;
815 my $ranked = $self->end->has_rank;
816 while( $curr ne $self->end->id ) {
817 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
818 my @succ = sort { $self->path_witnesses( $curr, $a )
819 <=> $self->path_witnesses( $curr, $b ) }
820 $self->sequence->successors( $curr );
821 my $next = pop @succ;
822 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
823 # Try to avoid lacunae in the weighted path.
825 ( $self->reading( $next )->is_lacuna ||
826 $nextrank - $rank > 1 ) ){
829 $weighted->{$curr} = $next;
835 =head2 path_witnesses( $edge )
837 Returns the list of sigils whose witnesses are associated with the given edge.
838 The edge can be passed as either an array or an arrayref of ( $source, $target ).
843 my( $self, @edge ) = @_;
844 # If edge is an arrayref, cope.
845 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
849 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
853 # Helper function. Make a display label for the given witnesses, showing a.c.
854 # witnesses only where the main witness is not also in the list.
855 sub _path_display_label {
858 map { $wits{$_} = 1 } @_;
860 # If an a.c. wit is listed, remove it if the main wit is also listed.
861 # Otherwise keep it for explicit listing.
862 my $aclabel = $self->ac_label;
864 foreach my $w ( sort keys %wits ) {
865 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
866 if( exists $wits{$1} ) {
869 push( @disp_ac, $w );
874 # See if we are in a majority situation.
875 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
876 $maj = $maj > 5 ? $maj : 5;
877 if( scalar keys %wits > $maj ) {
878 unshift( @disp_ac, 'majority' );
879 return join( ', ', @disp_ac );
881 return join( ', ', sort keys %wits );
885 =head2 readings_at_rank( $rank )
887 Returns a list of readings at a given rank, taken from the alignment table.
891 sub readings_at_rank {
892 my( $self, $rank ) = @_;
893 my $table = $self->alignment_table;
894 # Table rank is real rank - 1.
895 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
897 foreach my $e ( @elements ) {
898 next unless ref( $e ) eq 'HASH';
899 next unless exists $e->{'t'};
900 $readings{$e->{'t'}->id} = $e->{'t'};
902 return values %readings;
907 Returns a GraphML representation of the collation. The GraphML will contain
908 two graphs. The first expresses the attributes of the readings and the witness
909 paths that link them; the second expresses the relationships that link the
910 readings. This is the native transfer format for a tradition.
920 my $datafile = 't/data/florilegium_tei_ps.xml';
921 my $tradition = Text::Tradition->new( 'input' => 'TEI',
926 ok( $tradition, "Got a tradition object" );
927 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
928 ok( $tradition->collation, "Tradition has a collation" );
930 my $c = $tradition->collation;
931 is( scalar $c->readings, $READINGS, "Collation has all readings" );
932 is( scalar $c->paths, $PATHS, "Collation has all paths" );
933 is( scalar $c->relationships, 0, "Collation has all relationships" );
935 # Add a few relationships
936 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
937 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
938 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
940 # Now write it to GraphML and parse it again.
942 my $graphml = $c->as_graphml;
943 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
944 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
945 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
946 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
948 # Now add a stemma, write to GraphML, and look at the output.
950 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
951 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
952 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
953 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
954 $graphml = $c->as_graphml;
955 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
962 ## TODO MOVE this to Tradition.pm and modularize it better
964 my( $self, $options ) = @_;
965 $self->calculate_ranks unless $self->_graphcalc_done;
967 my $start = $options->{'from'}
968 ? $self->reading( $options->{'from'} ) : $self->start;
969 my $end = $options->{'to'}
970 ? $self->reading( $options->{'to'} ) : $self->end;
971 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
972 throw( 'Start node must be before end node' );
974 # The readings need to be ranked for this to work.
975 $start = $self->start unless $start->has_rank;
976 $end = $self->end unless $end->has_rank;
978 unless( $start eq $self->start ) {
979 $rankoffset = $start->rank - 1;
984 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
985 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
986 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
987 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
989 # Create the document and root node
991 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
992 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
993 $graphml->setDocumentElement( $root );
994 $root->setNamespace( $xsi_ns, 'xsi', 0 );
995 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
997 # List of attribute types to save on our objects and their corresponding
1002 'Bool' => 'boolean',
1003 'ReadingID' => 'string',
1004 'RelationshipType' => 'string',
1005 'RelationshipScope' => 'string',
1008 # Add the data keys for the graph. Include an extra key 'version' for the
1009 # GraphML output version.
1010 my %graph_data_keys;
1012 my %graph_attributes = ( 'version' => 'string' );
1013 # Graph attributes include those of Tradition and those of Collation.
1015 my $tmeta = $self->tradition->meta;
1016 my $cmeta = $self->meta;
1017 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1018 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1019 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1020 next if $attr->name =~ /^_/;
1021 next unless $save_types{$attr->type_constraint->name};
1022 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1024 # Extra custom keys for complex objects that should be saved in some form.
1025 # The subroutine should return a string, or undef/empty.
1026 if( $tmeta->has_method('stemmata') ) {
1027 $graph_attributes{'stemmata'} = sub {
1029 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1030 $self->tradition->stemmata;
1031 join( "\n", @stemstrs );
1035 if( $tmeta->has_method('user') ) {
1036 $graph_attributes{'user'} = sub {
1037 $self->tradition->user ? $self->tradition->user->id : undef
1041 foreach my $datum ( sort keys %graph_attributes ) {
1042 $graph_data_keys{$datum} = 'dg'.$gdi++;
1043 my $key = $root->addNewChild( $graphml_ns, 'key' );
1044 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1045 : $graph_attributes{$datum};
1046 $key->setAttribute( 'attr.name', $datum );
1047 $key->setAttribute( 'attr.type', $dtype );
1048 $key->setAttribute( 'for', 'graph' );
1049 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1052 # Add the data keys for reading nodes
1053 my %reading_attributes;
1054 my $rmeta = Text::Tradition::Collation::Reading->meta;
1055 foreach my $attr( $rmeta->get_all_attributes ) {
1056 next if $attr->name =~ /^_/;
1057 next unless $save_types{$attr->type_constraint->name};
1058 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1060 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1061 # Extra custom key for the reading morphology
1062 $reading_attributes{'lexemes'} = 'string';
1067 foreach my $datum ( sort keys %reading_attributes ) {
1068 $node_data_keys{$datum} = 'dn'.$ndi++;
1069 my $key = $root->addNewChild( $graphml_ns, 'key' );
1070 $key->setAttribute( 'attr.name', $datum );
1071 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1072 $key->setAttribute( 'for', 'node' );
1073 $key->setAttribute( 'id', $node_data_keys{$datum} );
1076 # Add the data keys for edges, that is, paths and relationships. Path
1077 # data does not come from a Moose class so is here manually.
1080 my %edge_attributes = (
1081 witness => 'string', # ID/label for a path
1082 extra => 'boolean', # Path key
1084 my @path_attributes = keys %edge_attributes; # track our manual additions
1085 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1086 foreach my $attr( $pmeta->get_all_attributes ) {
1087 next if $attr->name =~ /^_/;
1088 next unless $save_types{$attr->type_constraint->name};
1089 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1091 foreach my $datum ( sort keys %edge_attributes ) {
1092 $edge_data_keys{$datum} = 'de'.$edi++;
1093 my $key = $root->addNewChild( $graphml_ns, 'key' );
1094 $key->setAttribute( 'attr.name', $datum );
1095 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1096 $key->setAttribute( 'for', 'edge' );
1097 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1100 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1101 my $xmlidname = $self->tradition->name;
1102 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1103 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1104 $xmlidname = '_'.$xmlidname;
1106 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1107 $sgraph->setAttribute( 'edgedefault', 'directed' );
1108 $sgraph->setAttribute( 'id', $xmlidname );
1109 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1110 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1111 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1112 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1113 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1115 # Tradition/collation attribute data
1116 foreach my $datum ( keys %graph_attributes ) {
1118 if( $datum eq 'version' ) {
1120 } elsif( ref( $graph_attributes{$datum} ) ) {
1121 my $sub = $graph_attributes{$datum};
1123 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1124 $value = $self->tradition->$datum;
1126 $value = $self->$datum;
1128 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1133 # Add our readings to the graph
1134 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1135 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1136 ( $n->rank < $start->rank || $n->rank > $end->rank );
1137 $use_readings{$n->id} = 1;
1138 # Add to the main graph
1139 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1140 my $node_xmlid = 'n' . $node_ctr++;
1141 $node_hash{ $n->id } = $node_xmlid;
1142 $node_el->setAttribute( 'id', $node_xmlid );
1143 foreach my $d ( keys %reading_attributes ) {
1145 # Custom serialization
1146 if( $d eq 'lexemes' ) {
1147 # If nval is a true value, we have lexemes so we need to
1148 # serialize them. Otherwise set nval to undef so that the
1149 # key is excluded from this reading.
1150 $nval = $nval ? $n->_serialize_lexemes : undef;
1151 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1154 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1155 # Adjust the ranks within the subgraph.
1156 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1157 : $nval - $rankoffset;
1159 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1164 # Add the path edges to the sequence graph
1166 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1167 # We add an edge in the graphml for every witness in $e.
1168 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1169 my @edge_wits = sort $self->path_witnesses( $e );
1170 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1171 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1172 # Skip any path from start to end; that witness is not in the subgraph.
1173 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1174 foreach my $wit ( @edge_wits ) {
1175 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1176 $node_hash{ $e->[0] },
1177 $node_hash{ $e->[1] } );
1178 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1179 $edge_el->setAttribute( 'source', $from );
1180 $edge_el->setAttribute( 'target', $to );
1181 $edge_el->setAttribute( 'id', $id );
1183 # It's a witness path, so add the witness
1185 my $key = $edge_data_keys{'witness'};
1186 # Is this an ante-corr witness?
1187 my $aclabel = $self->ac_label;
1188 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1189 # Keep the base witness
1191 # ...and record that this is an 'extra' reading path
1192 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1194 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1198 # Report the actual number of nodes and edges that went in
1199 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1200 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1202 # Add the relationship graph to the XML
1203 map { delete $edge_data_keys{$_} } @path_attributes;
1204 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1205 $node_data_keys{'id'}, \%edge_data_keys );
1207 # Save and return the thing
1208 my $result = decode_utf8( $graphml->toString(1) );
1212 sub _add_graphml_data {
1213 my( $el, $key, $value ) = @_;
1214 return unless defined $value;
1215 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1216 $data_el->setAttribute( 'key', $key );
1217 $data_el->appendText( $value );
1222 Returns a CSV alignment table representation of the collation graph, one
1223 row per witness (or witness uncorrected.)
1229 my $table = $self->alignment_table;
1230 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1232 # Make the header row
1233 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1234 push( @result, decode_utf8( $csv->string ) );
1235 # Make the rest of the rows
1236 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1237 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1238 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1239 $csv->combine( @row );
1240 push( @result, decode_utf8( $csv->string ) );
1242 return join( "\n", @result );
1245 =head2 alignment_table( $use_refs, $include_witnesses )
1247 Return a reference to an alignment table, in a slightly enhanced CollateX
1248 format which looks like this:
1250 $table = { alignment => [ { witness => "SIGIL",
1251 tokens => [ { t => "TEXT" }, ... ] },
1252 { witness => "SIG2",
1253 tokens => [ { t => "TEXT" }, ... ] },
1255 length => TEXTLEN };
1257 If $use_refs is set to 1, the reading object is returned in the table
1258 instead of READINGTEXT; if not, the text of the reading is returned.
1260 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1261 keys have a true hash value will be included.
1265 sub alignment_table {
1267 $self->calculate_ranks() unless $self->_graphcalc_done;
1268 return $self->cached_table if $self->has_cached_table;
1270 # Make sure we can do this
1271 throw( "Need a linear graph in order to make an alignment table" )
1272 unless $self->linear;
1273 $self->calculate_ranks unless $self->end->has_rank;
1275 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1276 my @all_pos = ( 1 .. $self->end->rank - 1 );
1277 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1278 # say STDERR "Making witness row(s) for " . $wit->sigil;
1279 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1280 my @row = _make_witness_row( \@wit_path, \@all_pos );
1281 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1282 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1283 push( @{$table->{'alignment'}}, $witobj );
1284 if( $wit->is_layered ) {
1285 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1286 $wit->sigil.$self->ac_label );
1287 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1288 my $witacobj = { 'witness' => $wit->sigil.$self->ac_label,
1289 'tokens' => \@ac_row };
1290 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1291 push( @{$table->{'alignment'}}, $witacobj );
1294 $self->cached_table( $table );
1298 sub _make_witness_row {
1299 my( $path, $positions ) = @_;
1301 map { $char_hash{$_} = undef } @$positions;
1303 foreach my $rdg ( @$path ) {
1304 say STDERR "rank " . $rdg->rank if $debug;
1305 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1306 $char_hash{$rdg->rank} = { 't' => $rdg };
1308 my @row = map { $char_hash{$_} } @$positions;
1309 # Fill in lacuna markers for undef spots in the row
1310 my $last_el = shift @row;
1311 my @filled_row = ( $last_el );
1312 foreach my $el ( @row ) {
1313 # If we are using node reference, make the lacuna node appear many times
1314 # in the table. If not, use the lacuna tag.
1315 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1318 push( @filled_row, $el );
1324 =head1 NAVIGATION METHODS
1326 =head2 reading_sequence( $first, $last, $sigil, $backup )
1328 Returns the ordered list of readings, starting with $first and ending
1329 with $last, for the witness given in $sigil. If a $backup sigil is
1330 specified (e.g. when walking a layered witness), it will be used wherever
1331 no $sigil path exists. If there is a base text reading, that will be
1332 used wherever no path exists for $sigil or $backup.
1336 # TODO Think about returning some lazy-eval iterator.
1337 # TODO Get rid of backup; we should know from what witness is whether we need it.
1339 sub reading_sequence {
1340 my( $self, $start, $end, $witness ) = @_;
1342 $witness = $self->baselabel unless $witness;
1343 my @readings = ( $start );
1346 while( $n && $n->id ne $end->id ) {
1347 if( exists( $seen{$n->id} ) ) {
1348 throw( "Detected loop for $witness at " . $n->id );
1352 my $next = $self->next_reading( $n, $witness );
1354 throw( "Did not find any path for $witness from reading " . $n->id );
1356 push( @readings, $next );
1359 # Check that the last reading is our end reading.
1360 my $last = $readings[$#readings];
1361 throw( "Last reading found from " . $start->text .
1362 " for witness $witness is not the end!" ) # TODO do we get this far?
1363 unless $last->id eq $end->id;
1368 =head2 next_reading( $reading, $sigil );
1370 Returns the reading that follows the given reading along the given witness
1376 # Return the successor via the corresponding path.
1378 my $answer = $self->_find_linked_reading( 'next', @_ );
1379 return undef unless $answer;
1380 return $self->reading( $answer );
1383 =head2 prior_reading( $reading, $sigil )
1385 Returns the reading that precedes the given reading along the given witness
1391 # Return the predecessor via the corresponding path.
1393 my $answer = $self->_find_linked_reading( 'prior', @_ );
1394 return $self->reading( $answer );
1397 sub _find_linked_reading {
1398 my( $self, $direction, $node, $path ) = @_;
1400 # Get a backup if we are dealing with a layered witness
1402 my $aclabel = $self->ac_label;
1403 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1407 my @linked_paths = $direction eq 'next'
1408 ? $self->sequence->edges_from( $node )
1409 : $self->sequence->edges_to( $node );
1410 return undef unless scalar( @linked_paths );
1412 # We have to find the linked path that contains all of the
1413 # witnesses supplied in $path.
1414 my( @path_wits, @alt_path_wits );
1415 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1416 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1419 foreach my $le ( @linked_paths ) {
1420 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1423 my @le_wits = sort $self->path_witnesses( $le );
1424 if( _is_within( \@path_wits, \@le_wits ) ) {
1425 # This is the right path.
1426 return $direction eq 'next' ? $le->[1] : $le->[0];
1427 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1431 # Got this far? Return the alternate path if it exists.
1432 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1435 # Got this far? Return the base path if it exists.
1436 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1439 # Got this far? We have no appropriate path.
1440 warn "Could not find $direction node from " . $node->id
1441 . " along path $path";
1447 my( $set1, $set2 ) = @_;
1448 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1449 foreach my $el ( @$set1 ) {
1450 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1455 # Return the string that joins together a list of witnesses for
1456 # display on a single path.
1457 sub _witnesses_of_label {
1458 my( $self, $label ) = @_;
1459 my $regex = $self->wit_list_separator;
1460 my @answer = split( /\Q$regex\E/, $label );
1464 =head2 common_readings
1466 Returns the list of common readings in the graph (i.e. those readings that are
1467 shared by all non-lacunose witnesses.)
1471 sub common_readings {
1473 my @common = grep { $_->is_common } $self->readings;
1477 =head2 path_text( $sigil, [, $start, $end ] )
1479 Returns the text of a witness (plus its backup, if we are using a layer)
1480 as stored in the collation. The text is returned as a string, where the
1481 individual readings are joined with spaces and the meta-readings (e.g.
1482 lacunae) are omitted. Optional specification of $start and $end allows
1483 the generation of a subset of the witness text.
1488 my( $self, $wit, $start, $end ) = @_;
1489 $start = $self->start unless $start;
1490 $end = $self->end unless $end;
1491 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1494 foreach my $r ( @path ) {
1495 unless ( $r->join_prior || !$last || $last->join_next ) {
1498 $pathtext .= $r->text;
1504 =head1 INITIALIZATION METHODS
1506 These are mostly for use by parsers.
1508 =head2 make_witness_path( $witness )
1510 Link the array of readings contained in $witness->path (and in
1511 $witness->uncorrected_path if it exists) into collation paths.
1512 Clear out the arrays when finished.
1514 =head2 make_witness_paths
1516 Call make_witness_path for all witnesses in the tradition.
1520 # For use when a collation is constructed from a base text and an apparatus.
1521 # We have the sequences of readings and just need to add path edges.
1522 # When we are done, clear out the witness path attributes, as they are no
1524 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1526 sub make_witness_paths {
1528 foreach my $wit ( $self->tradition->witnesses ) {
1529 # say STDERR "Making path for " . $wit->sigil;
1530 $self->make_witness_path( $wit );
1534 sub make_witness_path {
1535 my( $self, $wit ) = @_;
1536 my @chain = @{$wit->path};
1537 my $sig = $wit->sigil;
1538 # Add start and end if necessary
1539 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1540 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1541 foreach my $idx ( 0 .. $#chain-1 ) {
1542 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1544 if( $wit->is_layered ) {
1545 @chain = @{$wit->uncorrected_path};
1546 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1547 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1548 foreach my $idx( 0 .. $#chain-1 ) {
1549 my $source = $chain[$idx];
1550 my $target = $chain[$idx+1];
1551 $self->add_path( $source, $target, $sig.$self->ac_label )
1552 unless $self->has_path( $source, $target, $sig );
1556 $wit->clear_uncorrected_path;
1559 =head2 calculate_ranks
1561 Calculate the reading ranks (that is, their aligned positions relative
1562 to each other) for the graph. This can only be called on linear collations.
1566 use Text::Tradition;
1568 my $cxfile = 't/data/Collatex-16.xml';
1569 my $t = Text::Tradition->new(
1571 'input' => 'CollateX',
1574 my $c = $t->collation;
1577 my $table = $c->alignment_table;
1578 ok( $c->has_cached_table, "Alignment table was cached" );
1579 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1580 $c->calculate_ranks;
1581 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1582 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1583 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1589 sub calculate_ranks {
1591 # Save the existing ranks, in case we need to invalidate the cached SVG.
1593 map { $existing_ranks{$_} = $_->rank } $self->readings;
1595 # Do the rankings based on the relationship equivalence graph, starting
1596 # with the start node.
1597 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1599 # Transfer our rankings from the topological graph to the real one.
1600 foreach my $r ( $self->readings ) {
1601 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1602 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1604 # Die. Find the last rank we calculated.
1605 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1606 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1608 my $last = pop @all_defined;
1609 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1612 # Do we need to invalidate the cached data?
1613 if( $self->has_cached_table ) {
1614 foreach my $r ( $self->readings ) {
1615 next if defined( $existing_ranks{$r} )
1616 && $existing_ranks{$r} == $r->rank;
1617 # Something has changed, so clear the cache
1618 $self->_clear_cache;
1619 # ...and recalculate the common readings.
1620 $self->calculate_common_readings();
1624 # The graph calculation information is now up to date.
1625 $self->_graphcalc_done(1);
1630 $self->wipe_table if $self->has_cached_table;
1634 =head2 flatten_ranks
1636 A convenience method for parsing collation data. Searches the graph for readings
1637 with the same text at the same rank, and merges any that are found.
1643 my %unique_rank_rdg;
1645 foreach my $rdg ( $self->readings ) {
1646 next unless $rdg->has_rank;
1647 my $key = $rdg->rank . "||" . $rdg->text;
1648 if( exists $unique_rank_rdg{$key} ) {
1649 # Make sure they don't have different grammatical forms
1650 my $ur = $unique_rank_rdg{$key};
1651 if( $rdg->is_identical( $ur ) ) {
1653 #say STDERR "Combining readings at same rank: $key";
1655 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1656 # TODO see if this now makes a common point.
1659 $unique_rank_rdg{$key} = $rdg;
1662 # If we merged readings, the ranks are still fine but the alignment
1663 # table is wrong. Wipe it.
1664 $self->wipe_table() if $changed;
1668 =head2 calculate_common_readings
1670 Goes through the graph identifying the readings that appear in every witness
1671 (apart from those with lacunae at that spot.) Marks them as common and returns
1676 use Text::Tradition;
1678 my $cxfile = 't/data/Collatex-16.xml';
1679 my $t = Text::Tradition->new(
1681 'input' => 'CollateX',
1684 my $c = $t->collation;
1686 my @common = $c->calculate_common_readings();
1687 is( scalar @common, 8, "Found correct number of common readings" );
1688 my @marked = sort $c->common_readings();
1689 is( scalar @common, 8, "All common readings got marked as such" );
1690 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1691 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1697 sub calculate_common_readings {
1700 map { $_->is_common( 0 ) } $self->readings;
1701 # Implicitly calls calculate_ranks
1702 my $table = $self->alignment_table;
1703 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1704 my @row = map { $_->{'tokens'}->[$idx]
1705 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1706 @{$table->{'alignment'}};
1708 foreach my $r ( @row ) {
1710 $hash{$r->id} = $r unless $r->is_meta;
1712 $hash{'UNDEF'} = $r;
1715 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1716 my( $r ) = values %hash;
1718 push( @common, $r );
1724 =head2 text_from_paths
1726 Calculate the text array for all witnesses from the path, for later consistency
1727 checking. Only to be used if there is no non-graph-based way to know the
1732 sub text_from_paths {
1734 foreach my $wit ( $self->tradition->witnesses ) {
1735 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1737 foreach my $r ( @readings ) {
1738 next if $r->is_meta;
1739 push( @text, $r->text );
1741 $wit->text( \@text );
1742 if( $wit->is_layered ) {
1743 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1744 $wit->sigil.$self->ac_label );
1746 foreach my $r ( @ucrdgs ) {
1747 next if $r->is_meta;
1748 push( @uctext, $r->text );
1750 $wit->layertext( \@uctext );
1755 =head1 UTILITY FUNCTIONS
1757 =head2 common_predecessor( $reading_a, $reading_b )
1759 Find the last reading that occurs in sequence before both the given readings.
1760 At the very least this should be $self->start.
1762 =head2 common_successor( $reading_a, $reading_b )
1764 Find the first reading that occurs in sequence after both the given readings.
1765 At the very least this should be $self->end.
1769 use Text::Tradition;
1771 my $cxfile = 't/data/Collatex-16.xml';
1772 my $t = Text::Tradition->new(
1774 'input' => 'CollateX',
1777 my $c = $t->collation;
1779 is( $c->common_predecessor( 'n24', 'n23' )->id,
1780 'n20', "Found correct common predecessor" );
1781 is( $c->common_successor( 'n24', 'n23' )->id,
1782 '__END__', "Found correct common successor" );
1784 is( $c->common_predecessor( 'n19', 'n17' )->id,
1785 'n16', "Found correct common predecessor for readings on same path" );
1786 is( $c->common_successor( 'n21', 'n10' )->id,
1787 '__END__', "Found correct common successor for readings on same path" );
1793 ## Return the closest reading that is a predecessor of both the given readings.
1794 sub common_predecessor {
1796 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1797 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1800 sub common_successor {
1802 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1803 return $self->_common_in_path( $r1, $r2, 'successors' );
1807 # TODO think about how to do this without ranks...
1808 sub _common_in_path {
1809 my( $self, $r1, $r2, $dir ) = @_;
1810 my $iter = $self->end->rank;
1812 my @last_r1 = ( $r1 );
1813 my @last_r2 = ( $r2 );
1814 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1816 # say STDERR "Finding common $dir for $r1, $r2";
1817 while( !@candidates ) {
1818 last unless $iter--; # Avoid looping infinitely
1819 # Iterate separately down the graph from r1 and r2
1820 my( @new_lc1, @new_lc2 );
1821 foreach my $lc ( @last_r1 ) {
1822 foreach my $p ( $lc->$dir ) {
1823 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1824 # say STDERR "Path candidate $p from $lc";
1825 push( @candidates, $p );
1826 } elsif( !$all_seen{$p->id} ) {
1827 $all_seen{$p->id} = 'r1';
1828 push( @new_lc1, $p );
1832 foreach my $lc ( @last_r2 ) {
1833 foreach my $p ( $lc->$dir ) {
1834 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1835 # say STDERR "Path candidate $p from $lc";
1836 push( @candidates, $p );
1837 } elsif( !$all_seen{$p->id} ) {
1838 $all_seen{$p->id} = 'r2';
1839 push( @new_lc2, $p );
1843 @last_r1 = @new_lc1;
1844 @last_r2 = @new_lc2;
1846 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1847 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1851 Text::Tradition::Error->throw(
1852 'ident' => 'Collation error',
1858 __PACKAGE__->meta->make_immutable;
1864 =item * Rework XML serialization in a more modular way
1870 This package is free software and is provided "as is" without express
1871 or implied warranty. You can redistribute it and/or modify it under
1872 the same terms as Perl itself.
1876 Tara L Andrews E<lt>aurum@cpan.orgE<gt>