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 ### Reading construct/destruct functions
239 my( $self, $reading ) = @_;
240 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
241 my %args = %$reading;
242 if( $args{'init'} ) {
243 # If we are initializing an empty collation, don't assume that we
244 # have set a tradition.
245 delete $args{'init'};
246 } elsif( $self->tradition->has_language && !exists $args{'language'} ) {
247 $args{'language'} = $self->tradition->language;
249 $reading = Text::Tradition::Collation::Reading->new(
250 'collation' => $self,
253 # First check to see if a reading with this ID exists.
254 if( $self->reading( $reading->id ) ) {
255 throw( "Collation already has a reading with id " . $reading->id );
257 $self->_graphcalc_done(0);
258 $self->_add_reading( $reading->id => $reading );
259 # Once the reading has been added, put it in both graphs.
260 $self->sequence->add_vertex( $reading->id );
261 $self->relations->add_reading( $reading->id );
265 around del_reading => sub {
270 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
273 # Remove the reading from the graphs.
274 $self->_graphcalc_done(0);
275 $self->_clear_cache; # Explicitly clear caches to GC the reading
276 $self->sequence->delete_vertex( $arg );
277 $self->relations->delete_reading( $arg );
280 $self->$orig( $arg );
287 my $cxfile = 't/data/Collatex-16.xml';
288 my $t = Text::Tradition->new(
290 'input' => 'CollateX',
293 my $c = $t->collation;
295 my $rno = scalar $c->readings;
296 # Split n21 for testing purposes
297 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
298 my $old_r = $c->reading( 'n21' );
299 $old_r->alter_text( 'to' );
300 $c->del_path( 'n20', 'n21', 'A' );
301 $c->add_path( 'n20', 'n21p0', 'A' );
302 $c->add_path( 'n21p0', 'n21', 'A' );
304 ok( $c->reading( 'n21p0' ), "New reading exists" );
305 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
307 # Combine n3 and n4 ( with his )
308 $c->merge_readings( 'n3', 'n4', 1 );
309 ok( !$c->reading('n4'), "Reading n4 is gone" );
310 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
312 # Collapse n9 and n10 ( rood / root )
313 $c->merge_readings( 'n9', 'n10' );
314 ok( !$c->reading('n10'), "Reading n10 is gone" );
315 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
317 # Combine n21 and n21p0
318 my $remaining = $c->reading('n21');
319 $remaining ||= $c->reading('n22'); # one of these should still exist
320 $c->merge_readings( 'n21p0', $remaining, 1 );
321 ok( !$c->reading('n21'), "Reading $remaining is gone" );
322 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
332 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
333 my $mergemeta = $kept_obj->is_meta;
334 throw( "Cannot merge meta and non-meta reading" )
335 unless ( $mergemeta && $del_obj->is_meta )
336 || ( !$mergemeta && !$del_obj->is_meta );
338 throw( "Cannot merge with start or end node" )
339 if( $kept_obj eq $self->start || $kept_obj eq $self->end
340 || $del_obj eq $self->start || $del_obj eq $self->end );
341 throw( "Cannot combine text of meta readings" ) if $combine;
343 # We only need the IDs for adding paths to the graph, not the reading
344 # objects themselves.
345 my $kept = $kept_obj->id;
346 my $deleted = $del_obj->id;
347 $self->_graphcalc_done(0);
349 # The kept reading should inherit the paths and the relationships
350 # of the deleted reading.
351 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
352 my @vector = ( $kept );
353 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
354 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
355 next if $vector[0] eq $vector[1]; # Don't add a self loop
356 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
357 $self->sequence->add_edge( @vector );
358 my $fwits = $self->sequence->get_edge_attributes( @vector );
359 @wits{keys %$fwits} = values %$fwits;
360 $self->sequence->set_edge_attributes( @vector, \%wits );
362 $self->relations->merge_readings( $kept, $deleted, $combine );
364 # Do the deletion deed.
366 # Combine the text of the readings
367 my $joinstr = $combine_char;
368 unless( defined $joinstr ) {
369 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
370 $joinstr = $self->wordsep unless defined $joinstr;
372 $kept_obj->_combine( $del_obj, $joinstr );
374 $self->del_reading( $deleted );
377 =head2 compress_readings
379 Where possible in the graph, compresses plain sequences of readings into a
380 single reading. The sequences must consist of readings with no
381 relationships to other readings, with only a single witness path between
382 them and no other witness paths from either that would skip the other. The
383 readings must also not be marked as nonsense or bad grammar.
385 WARNING: This operation cannot be undone.
389 sub compress_readings {
391 # Anywhere in the graph that there is a reading that joins only to a single
392 # successor, and neither of these have any relationships, just join the two
394 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
395 # Now look for readings that can be joined to their successors.
396 next unless $rdg->is_combinable;
398 while( $self->sequence->successors( $rdg ) == 1 ) {
399 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
400 throw( "Infinite loop" ) if $seen{$next->id};
401 $seen{$next->id} = 1;
402 last if $self->sequence->predecessors( $next ) > 1;
403 last unless $next->is_combinable;
404 say "Joining readings $rdg and $next";
405 $self->merge_readings( $rdg, $next, 1 );
408 # Make sure we haven't screwed anything up
409 foreach my $wit ( $self->tradition->witnesses ) {
410 my $pathtext = $self->path_text( $wit->sigil );
411 my $origtext = join( ' ', @{$wit->text} );
412 throw( "Text differs for witness " . $wit->sigil )
413 unless $pathtext eq $origtext;
414 if( $wit->is_layered ) {
415 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
416 $origtext = join( ' ', @{$wit->layertext} );
417 throw( "Ante-corr text differs for witness " . $wit->sigil )
418 unless $pathtext eq $origtext;
422 $self->relations->rebuild_equivalence();
423 $self->calculate_ranks();
426 # Helper function for manipulating the graph.
427 sub _stringify_args {
428 my( $self, $first, $second, @args ) = @_;
430 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
431 $second = $second->id
432 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
433 return( $first, $second, @args );
436 # Helper function for manipulating the graph.
437 sub _objectify_args {
438 my( $self, $first, $second, $arg ) = @_;
439 $first = $self->reading( $first )
440 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
441 $second = $self->reading( $second )
442 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
443 return( $first, $second, $arg );
450 # We only need the IDs for adding paths to the graph, not the reading
451 # objects themselves.
452 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
454 $self->_graphcalc_done(0);
455 # Connect the readings
456 unless( $self->sequence->has_edge( $source, $target ) ) {
457 $self->sequence->add_edge( $source, $target );
458 $self->relations->add_equivalence_edge( $source, $target );
460 # Note the witness in question
461 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
467 if( ref( $_[0] ) eq 'ARRAY' ) {
474 # We only need the IDs for adding paths to the graph, not the reading
475 # objects themselves.
476 my( $source, $target, $wit ) = $self->_stringify_args( @args );
478 $self->_graphcalc_done(0);
479 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
480 $self->sequence->delete_edge_attribute( $source, $target, $wit );
482 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
483 $self->sequence->delete_edge( $source, $target );
484 $self->relations->delete_equivalence_edge( $source, $target );
489 # Extra graph-alike utility
492 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
493 return undef unless $self->sequence->has_edge( $source, $target );
494 return $self->sequence->has_edge_attribute( $source, $target, $wit );
497 =head2 clear_witness( @sigil_list )
499 Clear the given witnesses out of the collation entirely, removing references
500 to them in paths, and removing readings that belong only to them. Should only
501 be called via $tradition->del_witness.
506 my( $self, @sigils ) = @_;
508 $self->_graphcalc_done(0);
509 # Clear the witness(es) out of the paths
510 foreach my $e ( $self->paths ) {
511 foreach my $sig ( @sigils ) {
512 $self->del_path( $e, $sig );
516 # Clear out the newly unused readings
517 foreach my $r ( $self->readings ) {
518 unless( $self->reading_witnesses( $r ) ) {
519 $self->del_reading( $r );
524 sub add_relationship {
526 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
527 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
528 $self->_graphcalc_done(0);
532 around qw/ get_relationship del_relationship / => sub {
536 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
539 my( $source, $target ) = $self->_stringify_args( @args );
540 $self->$orig( $source, $target );
543 =head2 reading_witnesses( $reading )
545 Return a list of sigils corresponding to the witnesses in which the reading appears.
549 sub reading_witnesses {
550 my( $self, $reading ) = @_;
551 # We need only check either the incoming or the outgoing edges; I have
552 # arbitrarily chosen "incoming". Thus, special-case the start node.
553 if( $reading eq $self->start ) {
554 return map { $_->sigil } $self->tradition->witnesses;
557 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
558 my $wits = $self->sequence->get_edge_attributes( @$e );
559 @all_witnesses{ keys %$wits } = 1;
561 my $acstr = $self->ac_label;
562 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
563 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
565 return keys %all_witnesses;
568 =head1 OUTPUT METHODS
570 =head2 as_svg( \%options )
572 Returns an SVG string that represents the graph, via as_dot and graphviz.
573 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
578 my( $self, $opts ) = @_;
579 throw( "Need GraphViz installed to output SVG" )
580 unless File::Which::which( 'dot' );
581 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
582 $self->calculate_ranks()
583 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
584 my @cmd = qw/dot -Tsvg/;
586 my $dotfile = File::Temp->new();
588 # $dotfile->unlink_on_destroy(0);
589 binmode $dotfile, ':utf8';
590 print $dotfile $self->as_dot( $opts );
591 push( @cmd, $dotfile->filename );
592 run( \@cmd, ">", binary(), \$svg );
593 $svg = decode_utf8( $svg );
598 =head2 as_dot( \%options )
600 Returns a string that is the collation graph expressed in dot
601 (i.e. GraphViz) format. Options include:
616 my( $self, $opts ) = @_;
617 my $startrank = $opts->{'from'} if $opts;
618 my $endrank = $opts->{'to'} if $opts;
619 my $color_common = $opts->{'color_common'} if $opts;
620 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
621 && $self->end->rank > 100;
622 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
624 # Check the arguments
626 return if $endrank && $startrank > $endrank;
627 return if $startrank > $self->end->rank;
629 if( defined $endrank ) {
630 return if $endrank < 0;
631 $endrank = undef if $endrank == $self->end->rank;
634 my $graph_name = $self->tradition->name;
635 $graph_name =~ s/[^\w\s]//g;
636 $graph_name = join( '_', split( /\s+/, $graph_name ) );
644 'fillcolor' => 'white',
649 'arrowhead' => 'open',
650 'color' => '#000000',
651 'fontcolor' => '#000000',
654 my $dot = sprintf( "digraph %s {\n", $graph_name );
655 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
656 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
658 # Output substitute start/end readings if necessary
660 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
663 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
665 if( $STRAIGHTENHACK ) {
667 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
668 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
669 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
671 my %used; # Keep track of the readings that actually appear in the graph
672 # Sort the readings by rank if we have ranks; this speeds layout.
673 my @all_readings = $self->end->has_rank
674 ? sort { $a->rank <=> $b->rank } $self->readings
676 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
677 foreach my $reading ( @all_readings ) {
678 # Only output readings within our rank range.
679 next if $startrank && $reading->rank < $startrank;
680 next if $endrank && $reading->rank > $endrank;
681 $used{$reading->id} = 1;
682 # Need not output nodes without separate labels
683 next if $reading->id eq $reading->text;
685 my $label = $reading->text;
686 $label .= '-' if $reading->join_next;
687 $label = "-$label" if $reading->join_prior;
688 $label =~ s/\"/\\\"/g;
689 $rattrs->{'label'} = $label;
690 $rattrs->{'id'} = $reading->id;
691 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
692 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
695 # Add the real edges. Need to weight one edge per rank jump, in a
697 # my $weighted = $self->_add_edge_weights;
698 my @edges = $self->paths;
699 my( %substart, %subend );
700 foreach my $edge ( @edges ) {
701 # Do we need to output this edge?
702 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
703 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
704 my $variables = { %edge_attrs, 'label' => $label };
706 # Account for the rank gap if necessary
707 my $rank0 = $self->reading( $edge->[0] )->rank
708 if $self->reading( $edge->[0] )->has_rank;
709 my $rank1 = $self->reading( $edge->[1] )->rank
710 if $self->reading( $edge->[1] )->has_rank;
711 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
712 $variables->{'minlen'} = $rank1 - $rank0;
715 # Add the calculated edge weights
716 # if( exists $weighted->{$edge->[0]}
717 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
718 # # $variables->{'color'} = 'red';
719 # $variables->{'weight'} = 3.0;
722 # EXPERIMENTAL: make edge width reflect no. of witnesses
723 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
724 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
726 my $varopts = _dot_attr_string( $variables );
727 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
728 $edge->[0], $edge->[1], $varopts );
729 } elsif( $used{$edge->[0]} ) {
730 $subend{$edge->[0]} = $edge->[1];
731 } elsif( $used{$edge->[1]} ) {
732 $substart{$edge->[1]} = $edge->[0];
735 # Add substitute start and end edges if necessary
736 foreach my $node ( keys %substart ) {
737 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
738 my $variables = { %edge_attrs, 'label' => $witstr };
739 my $nrdg = $self->reading( $node );
740 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
741 # Substart is actually one lower than $startrank
742 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
744 my $varopts = _dot_attr_string( $variables );
745 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
747 foreach my $node ( keys %subend ) {
748 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
749 my $variables = { %edge_attrs, 'label' => $witstr };
750 my $varopts = _dot_attr_string( $variables );
751 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
754 if( $STRAIGHTENHACK ) {
755 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
756 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
763 sub _dot_attr_string {
766 foreach my $k ( sort keys %$hash ) {
768 push( @attrs, $k.'="'.$v.'"' );
770 return( '[ ' . join( ', ', @attrs ) . ' ]' );
773 sub _add_edge_weights {
775 # Walk the graph from START to END, choosing the successor node with
776 # the largest number of witness paths each time.
778 my $curr = $self->start->id;
779 my $ranked = $self->end->has_rank;
780 while( $curr ne $self->end->id ) {
781 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
782 my @succ = sort { $self->path_witnesses( $curr, $a )
783 <=> $self->path_witnesses( $curr, $b ) }
784 $self->sequence->successors( $curr );
785 my $next = pop @succ;
786 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
787 # Try to avoid lacunae in the weighted path.
789 ( $self->reading( $next )->is_lacuna ||
790 $nextrank - $rank > 1 ) ){
793 $weighted->{$curr} = $next;
799 =head2 path_witnesses( $edge )
801 Returns the list of sigils whose witnesses are associated with the given edge.
802 The edge can be passed as either an array or an arrayref of ( $source, $target ).
807 my( $self, @edge ) = @_;
808 # If edge is an arrayref, cope.
809 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
813 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
817 # Helper function. Make a display label for the given witnesses, showing a.c.
818 # witnesses only where the main witness is not also in the list.
819 sub _path_display_label {
822 map { $wits{$_} = 1 } @_;
824 # If an a.c. wit is listed, remove it if the main wit is also listed.
825 # Otherwise keep it for explicit listing.
826 my $aclabel = $self->ac_label;
828 foreach my $w ( sort keys %wits ) {
829 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
830 if( exists $wits{$1} ) {
833 push( @disp_ac, $w );
838 # See if we are in a majority situation.
839 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
840 $maj = $maj > 5 ? $maj : 5;
841 if( scalar keys %wits > $maj ) {
842 unshift( @disp_ac, 'majority' );
843 return join( ', ', @disp_ac );
845 return join( ', ', sort keys %wits );
849 =head2 readings_at_rank( $rank )
851 Returns a list of readings at a given rank, taken from the alignment table.
855 sub readings_at_rank {
856 my( $self, $rank ) = @_;
857 my $table = $self->alignment_table;
858 # Table rank is real rank - 1.
859 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
861 foreach my $e ( @elements ) {
862 next unless ref( $e ) eq 'HASH';
863 next unless exists $e->{'t'};
864 $readings{$e->{'t'}->id} = $e->{'t'};
866 return values %readings;
871 Returns a GraphML representation of the collation. The GraphML will contain
872 two graphs. The first expresses the attributes of the readings and the witness
873 paths that link them; the second expresses the relationships that link the
874 readings. This is the native transfer format for a tradition.
884 my $datafile = 't/data/florilegium_tei_ps.xml';
885 my $tradition = Text::Tradition->new( 'input' => 'TEI',
890 ok( $tradition, "Got a tradition object" );
891 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
892 ok( $tradition->collation, "Tradition has a collation" );
894 my $c = $tradition->collation;
895 is( scalar $c->readings, $READINGS, "Collation has all readings" );
896 is( scalar $c->paths, $PATHS, "Collation has all paths" );
897 is( scalar $c->relationships, 0, "Collation has all relationships" );
899 # Add a few relationships
900 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
901 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
902 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
904 # Now write it to GraphML and parse it again.
906 my $graphml = $c->as_graphml;
907 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
908 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
909 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
910 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
912 # Now add a stemma, write to GraphML, and look at the output.
914 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
915 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
916 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
917 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
918 $graphml = $c->as_graphml;
919 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
922 # Now add a user, write to GraphML, and look at the output.
923 unlike( $graphml, qr/testuser/, "Test user name does not exist in GraphML yet" );
924 my $testuser = Text::Tradition::User->new(
925 id => 'testuser', password => 'testpass' );
926 is( ref( $testuser ), 'Text::Tradition::User', "Created test user object" );
927 $testuser->add_tradition( $tradition );
928 is( $tradition->user->id, $testuser->id, "Tradition assigned to test user" );
929 $graphml = $c->as_graphml;
930 like( $graphml, qr/testuser/, "Test user name now exists in GraphML" );
936 ## TODO MOVE this to Tradition.pm and modularize it better
938 my( $self, $options ) = @_;
939 $self->calculate_ranks unless $self->_graphcalc_done;
941 my $start = $options->{'from'}
942 ? $self->reading( $options->{'from'} ) : $self->start;
943 my $end = $options->{'to'}
944 ? $self->reading( $options->{'to'} ) : $self->end;
945 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
946 throw( 'Start node must be before end node' );
948 # The readings need to be ranked for this to work.
949 $start = $self->start unless $start->has_rank;
950 $end = $self->end unless $end->has_rank;
952 unless( $start eq $self->start ) {
953 $rankoffset = $start->rank - 1;
958 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
959 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
960 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
961 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
963 # Create the document and root node
965 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
966 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
967 $graphml->setDocumentElement( $root );
968 $root->setNamespace( $xsi_ns, 'xsi', 0 );
969 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
971 # List of attribute types to save on our objects and their corresponding
977 'ReadingID' => 'string',
978 'RelationshipType' => 'string',
979 'RelationshipScope' => 'string',
982 # Add the data keys for the graph. Include an extra key 'version' for the
983 # GraphML output version.
986 my %graph_attributes = ( 'version' => 'string' );
987 # Graph attributes include those of Tradition and those of Collation.
989 my $tmeta = $self->tradition->meta;
990 my $cmeta = $self->meta;
991 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
992 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
993 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
994 next if $attr->name =~ /^_/;
995 next unless $save_types{$attr->type_constraint->name};
996 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
998 # Extra custom keys for complex objects that should be saved in some form.
999 # The subroutine should return a string, or undef/empty.
1000 if( $tmeta->has_method('stemmata') ) {
1001 $graph_attributes{'stemmata'} = sub {
1003 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1004 $self->tradition->stemmata;
1005 join( "\n", @stemstrs );
1009 $graph_attributes{'user'} = sub {
1010 $self->tradition->user ? $self->tradition->user->id : undef
1013 foreach my $datum ( sort keys %graph_attributes ) {
1014 $graph_data_keys{$datum} = 'dg'.$gdi++;
1015 my $key = $root->addNewChild( $graphml_ns, 'key' );
1016 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1017 : $graph_attributes{$datum};
1018 $key->setAttribute( 'attr.name', $datum );
1019 $key->setAttribute( 'attr.type', $dtype );
1020 $key->setAttribute( 'for', 'graph' );
1021 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1024 # Add the data keys for reading nodes
1025 my %reading_attributes;
1026 my $rmeta = Text::Tradition::Collation::Reading->meta;
1027 foreach my $attr( $rmeta->get_all_attributes ) {
1028 next if $attr->name =~ /^_/;
1029 next unless $save_types{$attr->type_constraint->name};
1030 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1032 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1033 # Extra custom key for the reading morphology
1034 $reading_attributes{'lexemes'} = 'string';
1039 foreach my $datum ( sort keys %reading_attributes ) {
1040 $node_data_keys{$datum} = 'dn'.$ndi++;
1041 my $key = $root->addNewChild( $graphml_ns, 'key' );
1042 $key->setAttribute( 'attr.name', $datum );
1043 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1044 $key->setAttribute( 'for', 'node' );
1045 $key->setAttribute( 'id', $node_data_keys{$datum} );
1048 # Add the data keys for edges, that is, paths and relationships. Path
1049 # data does not come from a Moose class so is here manually.
1052 my %edge_attributes = (
1053 witness => 'string', # ID/label for a path
1054 extra => 'boolean', # Path key
1056 my @path_attributes = keys %edge_attributes; # track our manual additions
1057 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1058 foreach my $attr( $pmeta->get_all_attributes ) {
1059 next if $attr->name =~ /^_/;
1060 next unless $save_types{$attr->type_constraint->name};
1061 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1063 foreach my $datum ( sort keys %edge_attributes ) {
1064 $edge_data_keys{$datum} = 'de'.$edi++;
1065 my $key = $root->addNewChild( $graphml_ns, 'key' );
1066 $key->setAttribute( 'attr.name', $datum );
1067 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1068 $key->setAttribute( 'for', 'edge' );
1069 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1072 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1073 my $xmlidname = $self->tradition->name;
1074 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1075 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1076 $xmlidname = '_'.$xmlidname;
1078 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1079 $sgraph->setAttribute( 'edgedefault', 'directed' );
1080 $sgraph->setAttribute( 'id', $xmlidname );
1081 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1082 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1083 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1084 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1085 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1087 # Tradition/collation attribute data
1088 foreach my $datum ( keys %graph_attributes ) {
1090 if( $datum eq 'version' ) {
1092 } elsif( ref( $graph_attributes{$datum} ) ) {
1093 my $sub = $graph_attributes{$datum};
1095 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1096 $value = $self->tradition->$datum;
1098 $value = $self->$datum;
1100 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1105 # Add our readings to the graph
1106 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1107 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1108 ( $n->rank < $start->rank || $n->rank > $end->rank );
1109 $use_readings{$n->id} = 1;
1110 # Add to the main graph
1111 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1112 my $node_xmlid = 'n' . $node_ctr++;
1113 $node_hash{ $n->id } = $node_xmlid;
1114 $node_el->setAttribute( 'id', $node_xmlid );
1115 foreach my $d ( keys %reading_attributes ) {
1117 # Custom serialization
1118 if( $d eq 'lexemes' ) {
1119 # If nval is a true value, we have lexemes so we need to
1120 # serialize them. Otherwise set nval to undef so that the
1121 # key is excluded from this reading.
1122 $nval = $nval ? $n->_serialize_lexemes : undef;
1123 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1126 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1127 # Adjust the ranks within the subgraph.
1128 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1129 : $nval - $rankoffset;
1131 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1136 # Add the path edges to the sequence graph
1138 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1139 # We add an edge in the graphml for every witness in $e.
1140 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1141 my @edge_wits = sort $self->path_witnesses( $e );
1142 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1143 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1144 # Skip any path from start to end; that witness is not in the subgraph.
1145 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1146 foreach my $wit ( @edge_wits ) {
1147 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1148 $node_hash{ $e->[0] },
1149 $node_hash{ $e->[1] } );
1150 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1151 $edge_el->setAttribute( 'source', $from );
1152 $edge_el->setAttribute( 'target', $to );
1153 $edge_el->setAttribute( 'id', $id );
1155 # It's a witness path, so add the witness
1157 my $key = $edge_data_keys{'witness'};
1158 # Is this an ante-corr witness?
1159 my $aclabel = $self->ac_label;
1160 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1161 # Keep the base witness
1163 # ...and record that this is an 'extra' reading path
1164 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1166 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1170 # Report the actual number of nodes and edges that went in
1171 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1172 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1174 # Add the relationship graph to the XML
1175 map { delete $edge_data_keys{$_} } @path_attributes;
1176 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1177 $node_data_keys{'id'}, \%edge_data_keys );
1179 # Save and return the thing
1180 my $result = decode_utf8( $graphml->toString(1) );
1184 sub _add_graphml_data {
1185 my( $el, $key, $value ) = @_;
1186 return unless defined $value;
1187 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1188 $data_el->setAttribute( 'key', $key );
1189 $data_el->appendText( $value );
1194 Returns a CSV alignment table representation of the collation graph, one
1195 row per witness (or witness uncorrected.)
1201 my $table = $self->alignment_table;
1202 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1204 # Make the header row
1205 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1206 push( @result, decode_utf8( $csv->string ) );
1207 # Make the rest of the rows
1208 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1209 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1210 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1211 $csv->combine( @row );
1212 push( @result, decode_utf8( $csv->string ) );
1214 return join( "\n", @result );
1217 =head2 alignment_table( $use_refs, $include_witnesses )
1219 Return a reference to an alignment table, in a slightly enhanced CollateX
1220 format which looks like this:
1222 $table = { alignment => [ { witness => "SIGIL",
1223 tokens => [ { t => "TEXT" }, ... ] },
1224 { witness => "SIG2",
1225 tokens => [ { t => "TEXT" }, ... ] },
1227 length => TEXTLEN };
1229 If $use_refs is set to 1, the reading object is returned in the table
1230 instead of READINGTEXT; if not, the text of the reading is returned.
1232 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1233 keys have a true hash value will be included.
1237 sub alignment_table {
1239 $self->calculate_ranks() unless $self->_graphcalc_done;
1240 return $self->cached_table if $self->has_cached_table;
1242 # Make sure we can do this
1243 throw( "Need a linear graph in order to make an alignment table" )
1244 unless $self->linear;
1245 $self->calculate_ranks unless $self->end->has_rank;
1247 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1248 my @all_pos = ( 1 .. $self->end->rank - 1 );
1249 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1250 # say STDERR "Making witness row(s) for " . $wit->sigil;
1251 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1252 my @row = _make_witness_row( \@wit_path, \@all_pos );
1253 push( @{$table->{'alignment'}},
1254 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1255 if( $wit->is_layered ) {
1256 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1257 $wit->sigil.$self->ac_label );
1258 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1259 push( @{$table->{'alignment'}},
1260 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1263 $self->cached_table( $table );
1267 sub _make_witness_row {
1268 my( $path, $positions ) = @_;
1270 map { $char_hash{$_} = undef } @$positions;
1272 foreach my $rdg ( @$path ) {
1273 say STDERR "rank " . $rdg->rank if $debug;
1274 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1275 $char_hash{$rdg->rank} = { 't' => $rdg };
1277 my @row = map { $char_hash{$_} } @$positions;
1278 # Fill in lacuna markers for undef spots in the row
1279 my $last_el = shift @row;
1280 my @filled_row = ( $last_el );
1281 foreach my $el ( @row ) {
1282 # If we are using node reference, make the lacuna node appear many times
1283 # in the table. If not, use the lacuna tag.
1284 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1287 push( @filled_row, $el );
1293 =head1 NAVIGATION METHODS
1295 =head2 reading_sequence( $first, $last, $sigil, $backup )
1297 Returns the ordered list of readings, starting with $first and ending
1298 with $last, for the witness given in $sigil. If a $backup sigil is
1299 specified (e.g. when walking a layered witness), it will be used wherever
1300 no $sigil path exists. If there is a base text reading, that will be
1301 used wherever no path exists for $sigil or $backup.
1305 # TODO Think about returning some lazy-eval iterator.
1306 # TODO Get rid of backup; we should know from what witness is whether we need it.
1308 sub reading_sequence {
1309 my( $self, $start, $end, $witness ) = @_;
1311 $witness = $self->baselabel unless $witness;
1312 my @readings = ( $start );
1315 while( $n && $n->id ne $end->id ) {
1316 if( exists( $seen{$n->id} ) ) {
1317 throw( "Detected loop for $witness at " . $n->id );
1321 my $next = $self->next_reading( $n, $witness );
1323 throw( "Did not find any path for $witness from reading " . $n->id );
1325 push( @readings, $next );
1328 # Check that the last reading is our end reading.
1329 my $last = $readings[$#readings];
1330 throw( "Last reading found from " . $start->text .
1331 " for witness $witness is not the end!" ) # TODO do we get this far?
1332 unless $last->id eq $end->id;
1337 =head2 next_reading( $reading, $sigil );
1339 Returns the reading that follows the given reading along the given witness
1345 # Return the successor via the corresponding path.
1347 my $answer = $self->_find_linked_reading( 'next', @_ );
1348 return undef unless $answer;
1349 return $self->reading( $answer );
1352 =head2 prior_reading( $reading, $sigil )
1354 Returns the reading that precedes the given reading along the given witness
1360 # Return the predecessor via the corresponding path.
1362 my $answer = $self->_find_linked_reading( 'prior', @_ );
1363 return $self->reading( $answer );
1366 sub _find_linked_reading {
1367 my( $self, $direction, $node, $path ) = @_;
1369 # Get a backup if we are dealing with a layered witness
1371 my $aclabel = $self->ac_label;
1372 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1376 my @linked_paths = $direction eq 'next'
1377 ? $self->sequence->edges_from( $node )
1378 : $self->sequence->edges_to( $node );
1379 return undef unless scalar( @linked_paths );
1381 # We have to find the linked path that contains all of the
1382 # witnesses supplied in $path.
1383 my( @path_wits, @alt_path_wits );
1384 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1385 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1388 foreach my $le ( @linked_paths ) {
1389 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1392 my @le_wits = sort $self->path_witnesses( $le );
1393 if( _is_within( \@path_wits, \@le_wits ) ) {
1394 # This is the right path.
1395 return $direction eq 'next' ? $le->[1] : $le->[0];
1396 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1400 # Got this far? Return the alternate path if it exists.
1401 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1404 # Got this far? Return the base path if it exists.
1405 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1408 # Got this far? We have no appropriate path.
1409 warn "Could not find $direction node from " . $node->id
1410 . " along path $path";
1416 my( $set1, $set2 ) = @_;
1417 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1418 foreach my $el ( @$set1 ) {
1419 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1424 # Return the string that joins together a list of witnesses for
1425 # display on a single path.
1426 sub _witnesses_of_label {
1427 my( $self, $label ) = @_;
1428 my $regex = $self->wit_list_separator;
1429 my @answer = split( /\Q$regex\E/, $label );
1433 =head2 common_readings
1435 Returns the list of common readings in the graph (i.e. those readings that are
1436 shared by all non-lacunose witnesses.)
1440 sub common_readings {
1442 my @common = grep { $_->is_common } $self->readings;
1446 =head2 path_text( $sigil, [, $start, $end ] )
1448 Returns the text of a witness (plus its backup, if we are using a layer)
1449 as stored in the collation. The text is returned as a string, where the
1450 individual readings are joined with spaces and the meta-readings (e.g.
1451 lacunae) are omitted. Optional specification of $start and $end allows
1452 the generation of a subset of the witness text.
1457 my( $self, $wit, $start, $end ) = @_;
1458 $start = $self->start unless $start;
1459 $end = $self->end unless $end;
1460 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1463 foreach my $r ( @path ) {
1464 unless ( $r->join_prior || !$last || $last->join_next ) {
1467 $pathtext .= $r->text;
1473 =head1 INITIALIZATION METHODS
1475 These are mostly for use by parsers.
1477 =head2 make_witness_path( $witness )
1479 Link the array of readings contained in $witness->path (and in
1480 $witness->uncorrected_path if it exists) into collation paths.
1481 Clear out the arrays when finished.
1483 =head2 make_witness_paths
1485 Call make_witness_path for all witnesses in the tradition.
1489 # For use when a collation is constructed from a base text and an apparatus.
1490 # We have the sequences of readings and just need to add path edges.
1491 # When we are done, clear out the witness path attributes, as they are no
1493 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1495 sub make_witness_paths {
1497 foreach my $wit ( $self->tradition->witnesses ) {
1498 # say STDERR "Making path for " . $wit->sigil;
1499 $self->make_witness_path( $wit );
1503 sub make_witness_path {
1504 my( $self, $wit ) = @_;
1505 my @chain = @{$wit->path};
1506 my $sig = $wit->sigil;
1507 # Add start and end if necessary
1508 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1509 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1510 foreach my $idx ( 0 .. $#chain-1 ) {
1511 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1513 if( $wit->is_layered ) {
1514 @chain = @{$wit->uncorrected_path};
1515 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1516 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1517 foreach my $idx( 0 .. $#chain-1 ) {
1518 my $source = $chain[$idx];
1519 my $target = $chain[$idx+1];
1520 $self->add_path( $source, $target, $sig.$self->ac_label )
1521 unless $self->has_path( $source, $target, $sig );
1525 $wit->clear_uncorrected_path;
1528 =head2 calculate_ranks
1530 Calculate the reading ranks (that is, their aligned positions relative
1531 to each other) for the graph. This can only be called on linear collations.
1535 use Text::Tradition;
1537 my $cxfile = 't/data/Collatex-16.xml';
1538 my $t = Text::Tradition->new(
1540 'input' => 'CollateX',
1543 my $c = $t->collation;
1546 my $table = $c->alignment_table;
1547 ok( $c->has_cached_table, "Alignment table was cached" );
1548 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1549 $c->calculate_ranks;
1550 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1551 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1552 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1558 sub calculate_ranks {
1560 # Save the existing ranks, in case we need to invalidate the cached SVG.
1562 map { $existing_ranks{$_} = $_->rank } $self->readings;
1564 # Do the rankings based on the relationship equivalence graph, starting
1565 # with the start node.
1566 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1568 # Transfer our rankings from the topological graph to the real one.
1569 foreach my $r ( $self->readings ) {
1570 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1571 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1573 # Die. Find the last rank we calculated.
1574 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1575 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1577 my $last = pop @all_defined;
1578 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1581 # Do we need to invalidate the cached data?
1582 if( $self->has_cached_table ) {
1583 foreach my $r ( $self->readings ) {
1584 next if defined( $existing_ranks{$r} )
1585 && $existing_ranks{$r} == $r->rank;
1586 # Something has changed, so clear the cache
1587 $self->_clear_cache;
1588 # ...and recalculate the common readings.
1589 $self->calculate_common_readings();
1593 # The graph calculation information is now up to date.
1594 $self->_graphcalc_done(1);
1599 $self->wipe_table if $self->has_cached_table;
1603 =head2 flatten_ranks
1605 A convenience method for parsing collation data. Searches the graph for readings
1606 with the same text at the same rank, and merges any that are found.
1612 my %unique_rank_rdg;
1614 foreach my $rdg ( $self->readings ) {
1615 next unless $rdg->has_rank;
1616 my $key = $rdg->rank . "||" . $rdg->text;
1617 if( exists $unique_rank_rdg{$key} ) {
1618 # Make sure they don't have different grammatical forms
1619 my $ur = $unique_rank_rdg{$key};
1620 if( $rdg->is_identical( $ur ) ) {
1622 #say STDERR "Combining readings at same rank: $key";
1624 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1625 # TODO see if this now makes a common point.
1628 $unique_rank_rdg{$key} = $rdg;
1631 # If we merged readings, the ranks are still fine but the alignment
1632 # table is wrong. Wipe it.
1633 $self->wipe_table() if $changed;
1637 =head2 calculate_common_readings
1639 Goes through the graph identifying the readings that appear in every witness
1640 (apart from those with lacunae at that spot.) Marks them as common and returns
1645 use Text::Tradition;
1647 my $cxfile = 't/data/Collatex-16.xml';
1648 my $t = Text::Tradition->new(
1650 'input' => 'CollateX',
1653 my $c = $t->collation;
1655 my @common = $c->calculate_common_readings();
1656 is( scalar @common, 8, "Found correct number of common readings" );
1657 my @marked = sort $c->common_readings();
1658 is( scalar @common, 8, "All common readings got marked as such" );
1659 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1660 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1666 sub calculate_common_readings {
1669 map { $_->is_common( 0 ) } $self->readings;
1670 # Implicitly calls calculate_ranks
1671 my $table = $self->alignment_table;
1672 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1673 my @row = map { $_->{'tokens'}->[$idx]
1674 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1675 @{$table->{'alignment'}};
1677 foreach my $r ( @row ) {
1679 $hash{$r->id} = $r unless $r->is_meta;
1681 $hash{'UNDEF'} = $r;
1684 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1685 my( $r ) = values %hash;
1687 push( @common, $r );
1693 =head2 text_from_paths
1695 Calculate the text array for all witnesses from the path, for later consistency
1696 checking. Only to be used if there is no non-graph-based way to know the
1701 sub text_from_paths {
1703 foreach my $wit ( $self->tradition->witnesses ) {
1704 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1706 foreach my $r ( @readings ) {
1707 next if $r->is_meta;
1708 push( @text, $r->text );
1710 $wit->text( \@text );
1711 if( $wit->is_layered ) {
1712 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1713 $wit->sigil.$self->ac_label );
1715 foreach my $r ( @ucrdgs ) {
1716 next if $r->is_meta;
1717 push( @uctext, $r->text );
1719 $wit->layertext( \@uctext );
1724 =head1 UTILITY FUNCTIONS
1726 =head2 common_predecessor( $reading_a, $reading_b )
1728 Find the last reading that occurs in sequence before both the given readings.
1729 At the very least this should be $self->start.
1731 =head2 common_successor( $reading_a, $reading_b )
1733 Find the first reading that occurs in sequence after both the given readings.
1734 At the very least this should be $self->end.
1738 use Text::Tradition;
1740 my $cxfile = 't/data/Collatex-16.xml';
1741 my $t = Text::Tradition->new(
1743 'input' => 'CollateX',
1746 my $c = $t->collation;
1748 is( $c->common_predecessor( 'n24', 'n23' )->id,
1749 'n20', "Found correct common predecessor" );
1750 is( $c->common_successor( 'n24', 'n23' )->id,
1751 '__END__', "Found correct common successor" );
1753 is( $c->common_predecessor( 'n19', 'n17' )->id,
1754 'n16', "Found correct common predecessor for readings on same path" );
1755 is( $c->common_successor( 'n21', 'n10' )->id,
1756 '__END__', "Found correct common successor for readings on same path" );
1762 ## Return the closest reading that is a predecessor of both the given readings.
1763 sub common_predecessor {
1765 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1766 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1769 sub common_successor {
1771 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1772 return $self->_common_in_path( $r1, $r2, 'successors' );
1776 # TODO think about how to do this without ranks...
1777 sub _common_in_path {
1778 my( $self, $r1, $r2, $dir ) = @_;
1779 my $iter = $self->end->rank;
1781 my @last_r1 = ( $r1 );
1782 my @last_r2 = ( $r2 );
1783 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1785 # say STDERR "Finding common $dir for $r1, $r2";
1786 while( !@candidates ) {
1787 last unless $iter--; # Avoid looping infinitely
1788 # Iterate separately down the graph from r1 and r2
1789 my( @new_lc1, @new_lc2 );
1790 foreach my $lc ( @last_r1 ) {
1791 foreach my $p ( $lc->$dir ) {
1792 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1793 # say STDERR "Path candidate $p from $lc";
1794 push( @candidates, $p );
1795 } elsif( !$all_seen{$p->id} ) {
1796 $all_seen{$p->id} = 'r1';
1797 push( @new_lc1, $p );
1801 foreach my $lc ( @last_r2 ) {
1802 foreach my $p ( $lc->$dir ) {
1803 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1804 # say STDERR "Path candidate $p from $lc";
1805 push( @candidates, $p );
1806 } elsif( !$all_seen{$p->id} ) {
1807 $all_seen{$p->id} = 'r2';
1808 push( @new_lc2, $p );
1812 @last_r1 = @new_lc1;
1813 @last_r2 = @new_lc2;
1815 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1816 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1820 Text::Tradition::Error->throw(
1821 'ident' => 'Collation error',
1827 __PACKAGE__->meta->make_immutable;
1833 =item * Rework XML serialization in a more modular way
1839 This package is free software and is provided "as is" without express
1840 or implied warranty. You can redistribute it and/or modify it under
1841 the same terms as Perl itself.
1845 Tara L Andrews E<lt>aurum@cpan.orgE<gt>