1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Reading;
11 use Text::Tradition::Collation::RelationshipStore;
12 use Text::Tradition::Error;
13 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
15 use XML::LibXML::XPathContext;
21 default => sub { Graph->new() },
29 isa => 'Text::Tradition::Collation::RelationshipStore',
31 relationships => 'relationships',
32 related_readings => 'related_readings',
33 get_relationship => 'get_relationship',
34 del_relationship => 'del_relationship',
35 equivalence => 'equivalence',
36 equivalence_graph => 'equivalence_graph',
38 writer => '_set_relations',
43 isa => 'Text::Tradition',
44 writer => '_set_tradition',
49 isa => 'HashRef[Text::Tradition::Collation::Reading]',
53 _add_reading => 'set',
54 del_reading => 'delete',
55 has_reading => 'exists',
58 default => sub { {} },
61 has 'wit_list_separator' => (
70 default => 'base text',
93 isa => 'Text::Tradition::Collation::Reading',
94 writer => '_set_start',
100 isa => 'Text::Tradition::Collation::Reading',
101 writer => '_set_end',
105 has 'cached_svg' => (
108 predicate => 'has_cached_svg',
109 clearer => 'wipe_svg',
112 has 'cached_table' => (
115 predicate => 'has_cached_table',
116 clearer => 'wipe_table',
119 has '_graphcalc_done' => (
127 Text::Tradition::Collation - a software model for a text collation
132 my $t = Text::Tradition->new(
133 'name' => 'this is a text',
135 'file' => '/path/to/tei_parallel_seg_file.xml' );
137 my $c = $t->collation;
138 my @readings = $c->readings;
139 my @paths = $c->paths;
140 my @relationships = $c->relationships;
142 my $svg_variant_graph = $t->collation->as_svg();
146 Text::Tradition is a library for representation and analysis of collated
147 texts, particularly medieval ones. The Collation is the central feature of
148 a Tradition, where the text, its sequence of readings, and its relationships
149 between readings are actually kept.
155 The constructor. Takes a hash or hashref of the following arguments:
159 =item * tradition - The Text::Tradition object to which the collation
162 =item * linear - Whether the collation should be linear; that is, whether
163 transposed readings should be treated as two linked readings rather than one,
164 and therefore whether the collation graph is acyclic. Defaults to true.
166 =item * baselabel - The default label for the path taken by a base text
167 (if any). Defaults to 'base text'.
169 =item * wit_list_separator - The string to join a list of witnesses for
170 purposes of making labels in display graphs. Defaults to ', '.
172 =item * ac_label - The extra label to tack onto a witness sigil when
173 representing another layer of path for the given witness - that is, when
174 a text has more than one possible reading due to scribal corrections or
175 the like. Defaults to ' (a.c.)'.
177 =item * wordsep - The string used to separate words in the original text.
188 =head2 wit_list_separator
196 Simple accessors for collation attributes.
200 The meta-reading at the start of every witness path.
204 The meta-reading at the end of every witness path.
208 Returns all Reading objects in the graph.
210 =head2 reading( $id )
212 Returns the Reading object corresponding to the given ID.
214 =head2 add_reading( $reading_args )
216 Adds a new reading object to the collation.
217 See L<Text::Tradition::Collation::Reading> for the available arguments.
219 =head2 del_reading( $object_or_id )
221 Removes the given reading from the collation, implicitly removing its
222 paths and relationships.
224 =head2 merge_readings( $main, $second, $concatenate, $with_str )
226 Merges the $second reading into the $main one. If $concatenate is true, then
227 the merged node will carry the text of both readings, concatenated with either
228 $with_str (if specified) or a sensible default (the empty string if the
229 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
231 The first two arguments may be either readings or reading IDs.
233 =head2 has_reading( $id )
235 Predicate to see whether a given reading ID is in the graph.
237 =head2 reading_witnesses( $object_or_id )
239 Returns a list of sigils whose witnesses contain the reading.
243 Returns all reading paths within the document - that is, all edges in the
244 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
246 =head2 add_path( $source, $target, $sigil )
248 Links the given readings in the collation in sequence, under the given witness
249 sigil. The readings may be specified by object or ID.
251 =head2 del_path( $source, $target, $sigil )
253 Links the given readings in the collation in sequence, under the given witness
254 sigil. The readings may be specified by object or ID.
256 =head2 has_path( $source, $target );
258 Returns true if the two readings are linked in sequence in any witness.
259 The readings may be specified by object or ID.
263 Returns all Relationship objects in the collation.
265 =head2 add_relationship( $reading, $other_reading, $options )
267 Adds a new relationship of the type given in $options between the two readings,
268 which may be specified by object or ID. Returns a value of ( $status, @vectors)
269 where $status is true on success, and @vectors is a list of relationship edges
270 that were ultimately added.
271 See L<Text::Tradition::Collation::Relationship> for the available options.
277 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
278 $self->_set_start( $self->add_reading(
279 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
280 $self->_set_end( $self->add_reading(
281 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
284 ### Reading construct/destruct functions
287 my( $self, $reading ) = @_;
288 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
289 my %args = %$reading;
290 if( $args{'init'} ) {
291 # If we are initializing an empty collation, don't assume that we
292 # have set a tradition.
293 delete $args{'init'};
294 } elsif( $self->tradition->has_language && !exists $args{'language'} ) {
295 $args{'language'} = $self->tradition->language;
297 $reading = Text::Tradition::Collation::Reading->new(
298 'collation' => $self,
301 # First check to see if a reading with this ID exists.
302 if( $self->reading( $reading->id ) ) {
303 throw( "Collation already has a reading with id " . $reading->id );
305 $self->_graphcalc_done(0);
306 $self->_add_reading( $reading->id => $reading );
307 # Once the reading has been added, put it in both graphs.
308 $self->sequence->add_vertex( $reading->id );
309 $self->relations->add_reading( $reading->id );
313 around del_reading => sub {
318 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
321 # Remove the reading from the graphs.
322 $self->_graphcalc_done(0);
323 $self->_clear_cache; # Explicitly clear caches to GC the reading
324 $self->sequence->delete_vertex( $arg );
325 $self->relations->delete_reading( $arg );
328 $self->$orig( $arg );
335 my $cxfile = 't/data/Collatex-16.xml';
336 my $t = Text::Tradition->new(
338 'input' => 'CollateX',
341 my $c = $t->collation;
343 my $rno = scalar $c->readings;
344 # Split n21 for testing purposes
345 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
346 my $old_r = $c->reading( 'n21' );
347 $old_r->alter_text( 'to' );
348 $c->del_path( 'n20', 'n21', 'A' );
349 $c->add_path( 'n20', 'n21p0', 'A' );
350 $c->add_path( 'n21p0', 'n21', 'A' );
352 ok( $c->reading( 'n21p0' ), "New reading exists" );
353 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
355 # Combine n3 and n4 ( with his )
356 $c->merge_readings( 'n3', 'n4', 1 );
357 ok( !$c->reading('n4'), "Reading n4 is gone" );
358 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
360 # Collapse n9 and n10 ( rood / root )
361 $c->merge_readings( 'n9', 'n10' );
362 ok( !$c->reading('n10'), "Reading n10 is gone" );
363 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
365 # Combine n21 and n21p0
366 my $remaining = $c->reading('n21');
367 $remaining ||= $c->reading('n22'); # one of these should still exist
368 $c->merge_readings( 'n21p0', $remaining, 1 );
369 ok( !$c->reading('n21'), "Reading $remaining is gone" );
370 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
380 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
381 my $mergemeta = $kept_obj->is_meta;
382 throw( "Cannot merge meta and non-meta reading" )
383 unless ( $mergemeta && $del_obj->is_meta )
384 || ( !$mergemeta && !$del_obj->is_meta );
386 throw( "Cannot merge with start or end node" )
387 if( $kept_obj eq $self->start || $kept_obj eq $self->end
388 || $del_obj eq $self->start || $del_obj eq $self->end );
390 # We only need the IDs for adding paths to the graph, not the reading
391 # objects themselves.
392 my $kept = $kept_obj->id;
393 my $deleted = $del_obj->id;
394 $self->_graphcalc_done(0);
396 # The kept reading should inherit the paths and the relationships
397 # of the deleted reading.
398 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
399 my @vector = ( $kept );
400 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
401 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
402 next if $vector[0] eq $vector[1]; # Don't add a self loop
403 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
404 $self->sequence->add_edge( @vector );
405 my $fwits = $self->sequence->get_edge_attributes( @vector );
406 @wits{keys %$fwits} = values %$fwits;
407 $self->sequence->set_edge_attributes( @vector, \%wits );
409 $self->relations->merge_readings( $kept, $deleted, $combine );
411 # Do the deletion deed.
413 # Combine the text of the readings
414 my $joinstr = $combine_char;
415 unless( defined $joinstr ) {
416 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
417 $joinstr = $self->wordsep unless defined $joinstr;
419 $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) );
420 $kept_obj->normal_form(
421 join( $joinstr, $kept_obj->normal_form, $del_obj->normal_form ) );
422 # Combine the lexemes present in the readings
423 if( $kept_obj->has_lexemes && $del_obj->has_lexemes ) {
424 $kept_obj->add_lexeme( $del_obj->lexemes );
427 $self->del_reading( $deleted );
430 =head2 compress_readings
432 Where possible in the graph, compresses plain sequences of readings into a
433 single reading. The sequences must consist of readings with no
434 relationships to other readings, with only a single witness path between
435 them and no other witness paths from either that would skip the other. The
436 readings must also not be marked as nonsense or bad grammar.
438 WARNING: This operation cannot be undone.
442 sub compress_readings {
444 # Anywhere in the graph that there is a reading that joins only to a single
445 # successor, and neither of these have any relationships, just join the two
448 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
449 next if $rdg->is_meta;
450 next if $gobbled{$rdg->id};
451 next if $rdg->grammar_invalid || $rdg->is_nonsense;
452 next if $rdg->related_readings();
454 while( $self->sequence->successors( $rdg ) == 1 ) {
455 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
456 throw( "Infinite loop" ) if $seen{$next->id};
457 $seen{$next->id} = 1;
458 last if $self->sequence->predecessors( $next ) > 1;
459 last if $next->is_meta;
460 last if $next->grammar_invalid || $next->is_nonsense;
461 last if $next->related_readings();
462 say "Joining readings $rdg and $next";
463 $self->merge_readings( $rdg, $next, 1 );
466 # Make sure we haven't screwed anything up
467 foreach my $wit ( $self->tradition->witnesses ) {
468 my $pathtext = $self->path_text( $wit->sigil );
469 my $origtext = join( ' ', @{$wit->text} );
470 throw( "Text differs for witness " . $wit->sigil )
471 unless $pathtext eq $origtext;
472 if( $wit->is_layered ) {
473 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
474 $origtext = join( ' ', @{$wit->layertext} );
475 throw( "Ante-corr text differs for witness " . $wit->sigil )
476 unless $pathtext eq $origtext;
480 $self->relations->rebuild_equivalence();
481 $self->calculate_ranks();
484 # Helper function for manipulating the graph.
485 sub _stringify_args {
486 my( $self, $first, $second, @args ) = @_;
488 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
489 $second = $second->id
490 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
491 return( $first, $second, @args );
494 # Helper function for manipulating the graph.
495 sub _objectify_args {
496 my( $self, $first, $second, $arg ) = @_;
497 $first = $self->reading( $first )
498 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
499 $second = $self->reading( $second )
500 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
501 return( $first, $second, $arg );
508 # We only need the IDs for adding paths to the graph, not the reading
509 # objects themselves.
510 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
512 $self->_graphcalc_done(0);
513 # Connect the readings
514 unless( $self->sequence->has_edge( $source, $target ) ) {
515 $self->sequence->add_edge( $source, $target );
516 $self->relations->add_equivalence_edge( $source, $target );
518 # Note the witness in question
519 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
525 if( ref( $_[0] ) eq 'ARRAY' ) {
532 # We only need the IDs for adding paths to the graph, not the reading
533 # objects themselves.
534 my( $source, $target, $wit ) = $self->_stringify_args( @args );
536 $self->_graphcalc_done(0);
537 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
538 $self->sequence->delete_edge_attribute( $source, $target, $wit );
540 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
541 $self->sequence->delete_edge( $source, $target );
542 $self->relations->delete_equivalence_edge( $source, $target );
547 # Extra graph-alike utility
550 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
551 return undef unless $self->sequence->has_edge( $source, $target );
552 return $self->sequence->has_edge_attribute( $source, $target, $wit );
555 =head2 clear_witness( @sigil_list )
557 Clear the given witnesses out of the collation entirely, removing references
558 to them in paths, and removing readings that belong only to them. Should only
559 be called via $tradition->del_witness.
564 my( $self, @sigils ) = @_;
566 $self->_graphcalc_done(0);
567 # Clear the witness(es) out of the paths
568 foreach my $e ( $self->paths ) {
569 foreach my $sig ( @sigils ) {
570 $self->del_path( $e, $sig );
574 # Clear out the newly unused readings
575 foreach my $r ( $self->readings ) {
576 unless( $self->reading_witnesses( $r ) ) {
577 $self->del_reading( $r );
582 sub add_relationship {
584 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
585 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
586 $self->_graphcalc_done(0);
590 around qw/ get_relationship del_relationship / => sub {
594 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
597 my( $source, $target ) = $self->_stringify_args( @args );
598 $self->$orig( $source, $target );
601 =head2 reading_witnesses( $reading )
603 Return a list of sigils corresponding to the witnesses in which the reading appears.
607 sub reading_witnesses {
608 my( $self, $reading ) = @_;
609 # We need only check either the incoming or the outgoing edges; I have
610 # arbitrarily chosen "incoming". Thus, special-case the start node.
611 if( $reading eq $self->start ) {
612 return map { $_->sigil } $self->tradition->witnesses;
615 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
616 my $wits = $self->sequence->get_edge_attributes( @$e );
617 @all_witnesses{ keys %$wits } = 1;
619 my $acstr = $self->ac_label;
620 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
621 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
623 return keys %all_witnesses;
626 =head1 OUTPUT METHODS
628 =head2 as_svg( \%options )
630 Returns an SVG string that represents the graph, via as_dot and graphviz.
631 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
636 my( $self, $opts ) = @_;
637 throw( "Need GraphViz installed to output SVG" )
638 unless File::Which::which( 'dot' );
639 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
640 $self->calculate_ranks()
641 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
642 if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) {
643 my @cmd = qw/dot -Tsvg/;
645 my $dotfile = File::Temp->new();
647 # $dotfile->unlink_on_destroy(0);
648 binmode $dotfile, ':utf8';
649 print $dotfile $self->as_dot( $opts );
650 push( @cmd, $dotfile->filename );
651 run( \@cmd, ">", binary(), \$svg );
652 $svg = decode_utf8( $svg );
653 $self->cached_svg( $svg ) unless $want_subgraph;
656 return $self->cached_svg;
661 =head2 as_dot( \%options )
663 Returns a string that is the collation graph expressed in dot
664 (i.e. GraphViz) format. Options include:
679 my( $self, $opts ) = @_;
680 my $startrank = $opts->{'from'} if $opts;
681 my $endrank = $opts->{'to'} if $opts;
682 my $color_common = $opts->{'color_common'} if $opts;
683 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
684 && $self->end->rank > 100;
685 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
687 # Check the arguments
689 return if $endrank && $startrank > $endrank;
690 return if $startrank > $self->end->rank;
692 if( defined $endrank ) {
693 return if $endrank < 0;
694 $endrank = undef if $endrank == $self->end->rank;
697 my $graph_name = $self->tradition->name;
698 $graph_name =~ s/[^\w\s]//g;
699 $graph_name = join( '_', split( /\s+/, $graph_name ) );
707 'fillcolor' => 'white',
712 'arrowhead' => 'open',
713 'color' => '#000000',
714 'fontcolor' => '#000000',
717 my $dot = sprintf( "digraph %s {\n", $graph_name );
718 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
719 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
721 # Output substitute start/end readings if necessary
723 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
726 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
728 if( $STRAIGHTENHACK ) {
730 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
731 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
732 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
734 my %used; # Keep track of the readings that actually appear in the graph
735 # Sort the readings by rank if we have ranks; this speeds layout.
736 my @all_readings = $self->end->has_rank
737 ? sort { $a->rank <=> $b->rank } $self->readings
739 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
740 foreach my $reading ( @all_readings ) {
741 # Only output readings within our rank range.
742 next if $startrank && $reading->rank < $startrank;
743 next if $endrank && $reading->rank > $endrank;
744 $used{$reading->id} = 1;
745 # Need not output nodes without separate labels
746 next if $reading->id eq $reading->text;
748 my $label = $reading->text;
749 $label .= '-' if $reading->join_next;
750 $label = "-$label" if $reading->join_prior;
751 $label =~ s/\"/\\\"/g;
752 $rattrs->{'label'} = $label;
753 $rattrs->{'id'} = $reading->id;
754 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
755 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
758 # Add the real edges. Need to weight one edge per rank jump, in a
760 # my $weighted = $self->_add_edge_weights;
761 my @edges = $self->paths;
762 my( %substart, %subend );
763 foreach my $edge ( @edges ) {
764 # Do we need to output this edge?
765 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
766 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
767 my $variables = { %edge_attrs, 'label' => $label };
769 # Account for the rank gap if necessary
770 my $rank0 = $self->reading( $edge->[0] )->rank
771 if $self->reading( $edge->[0] )->has_rank;
772 my $rank1 = $self->reading( $edge->[1] )->rank
773 if $self->reading( $edge->[1] )->has_rank;
774 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
775 $variables->{'minlen'} = $rank1 - $rank0;
778 # Add the calculated edge weights
779 # if( exists $weighted->{$edge->[0]}
780 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
781 # # $variables->{'color'} = 'red';
782 # $variables->{'weight'} = 3.0;
785 # EXPERIMENTAL: make edge width reflect no. of witnesses
786 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
787 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
789 my $varopts = _dot_attr_string( $variables );
790 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
791 $edge->[0], $edge->[1], $varopts );
792 } elsif( $used{$edge->[0]} ) {
793 $subend{$edge->[0]} = $edge->[1];
794 } elsif( $used{$edge->[1]} ) {
795 $substart{$edge->[1]} = $edge->[0];
798 # Add substitute start and end edges if necessary
799 foreach my $node ( keys %substart ) {
800 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
801 my $variables = { %edge_attrs, 'label' => $witstr };
802 my $nrdg = $self->reading( $node );
803 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
804 # Substart is actually one lower than $startrank
805 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
807 my $varopts = _dot_attr_string( $variables );
808 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
810 foreach my $node ( keys %subend ) {
811 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
812 my $variables = { %edge_attrs, 'label' => $witstr };
813 my $varopts = _dot_attr_string( $variables );
814 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
817 if( $STRAIGHTENHACK ) {
818 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
819 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
826 sub _dot_attr_string {
829 foreach my $k ( sort keys %$hash ) {
831 push( @attrs, $k.'="'.$v.'"' );
833 return( '[ ' . join( ', ', @attrs ) . ' ]' );
836 sub _add_edge_weights {
838 # Walk the graph from START to END, choosing the successor node with
839 # the largest number of witness paths each time.
841 my $curr = $self->start->id;
842 my $ranked = $self->end->has_rank;
843 while( $curr ne $self->end->id ) {
844 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
845 my @succ = sort { $self->path_witnesses( $curr, $a )
846 <=> $self->path_witnesses( $curr, $b ) }
847 $self->sequence->successors( $curr );
848 my $next = pop @succ;
849 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
850 # Try to avoid lacunae in the weighted path.
852 ( $self->reading( $next )->is_lacuna ||
853 $nextrank - $rank > 1 ) ){
856 $weighted->{$curr} = $next;
862 =head2 path_witnesses( $edge )
864 Returns the list of sigils whose witnesses are associated with the given edge.
865 The edge can be passed as either an array or an arrayref of ( $source, $target ).
870 my( $self, @edge ) = @_;
871 # If edge is an arrayref, cope.
872 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
876 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
880 # Helper function. Make a display label for the given witnesses, showing a.c.
881 # witnesses only where the main witness is not also in the list.
882 sub _path_display_label {
885 map { $wits{$_} = 1 } @_;
887 # If an a.c. wit is listed, remove it if the main wit is also listed.
888 # Otherwise keep it for explicit listing.
889 my $aclabel = $self->ac_label;
891 foreach my $w ( sort keys %wits ) {
892 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
893 if( exists $wits{$1} ) {
896 push( @disp_ac, $w );
901 # See if we are in a majority situation.
902 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
903 $maj = $maj > 5 ? $maj : 5;
904 if( scalar keys %wits > $maj ) {
905 unshift( @disp_ac, 'majority' );
906 return join( ', ', @disp_ac );
908 return join( ', ', sort keys %wits );
912 =head2 readings_at_rank( $rank )
914 Returns a list of readings at a given rank, taken from the alignment table.
918 sub readings_at_rank {
919 my( $self, $rank ) = @_;
920 my $table = $self->alignment_table;
921 # Table rank is real rank - 1.
922 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
924 foreach my $e ( @elements ) {
925 next unless ref( $e ) eq 'HASH';
926 next unless exists $e->{'t'};
927 $readings{$e->{'t'}->id} = $e->{'t'};
929 return values %readings;
934 Returns a GraphML representation of the collation. The GraphML will contain
935 two graphs. The first expresses the attributes of the readings and the witness
936 paths that link them; the second expresses the relationships that link the
937 readings. This is the native transfer format for a tradition.
946 my $datafile = 't/data/florilegium_tei_ps.xml';
947 my $tradition = Text::Tradition->new( 'input' => 'TEI',
952 ok( $tradition, "Got a tradition object" );
953 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
954 ok( $tradition->collation, "Tradition has a collation" );
956 my $c = $tradition->collation;
957 is( scalar $c->readings, $READINGS, "Collation has all readings" );
958 is( scalar $c->paths, $PATHS, "Collation has all paths" );
959 is( scalar $c->relationships, 0, "Collation has all relationships" );
961 # Add a few relationships
962 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
963 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
964 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
966 # Now write it to GraphML and parse it again.
968 my $graphml = $c->as_graphml;
969 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
970 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
971 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
972 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
974 # Now add a stemma, write to GraphML, and parse again.
975 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
976 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
977 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
978 $graphml = $c->as_graphml;
979 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
986 my( $self, $options ) = @_;
987 $self->calculate_ranks unless $self->_graphcalc_done;
989 my $start = $options->{'from'}
990 ? $self->reading( $options->{'from'} ) : $self->start;
991 my $end = $options->{'to'}
992 ? $self->reading( $options->{'to'} ) : $self->end;
993 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
994 throw( 'Start node must be before end node' );
996 # The readings need to be ranked for this to work.
997 $start = $self->start unless $start->has_rank;
998 $end = $self->end unless $end->has_rank;
1000 unless( $start eq $self->start ) {
1001 $rankoffset = $start->rank - 1;
1006 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1007 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1008 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1009 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1011 # Create the document and root node
1012 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1013 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1014 $graphml->setDocumentElement( $root );
1015 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1016 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1018 # List of attribute types to save on our objects and their corresponding
1023 'Bool' => 'boolean',
1024 'ReadingID' => 'string',
1025 'RelationshipType' => 'string',
1026 'RelationshipScope' => 'string',
1029 # List of attribute names *not* to save on our objects.
1030 # We will also not save any attribute beginning with _.
1032 map { $skipsave{$_} = 1 } qw/ cached_svg /;
1034 # Add the data keys for the graph. Include an extra key 'version' for the
1035 # GraphML output version.
1036 my %graph_data_keys;
1038 my %graph_attributes = ( 'version' => 'string' );
1039 # Graph attributes include those of Tradition and those of Collation.
1041 my $tmeta = $self->tradition->meta;
1042 my $cmeta = $self->meta;
1043 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1044 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1045 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1046 next if $attr->name =~ /^_/;
1047 next if $skipsave{$attr->name};
1048 next unless $save_types{$attr->type_constraint->name};
1049 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1051 # Extra custom key for the tradition stemma(ta)
1052 $graph_attributes{'stemmata'} = 'string';
1054 foreach my $datum ( sort keys %graph_attributes ) {
1055 $graph_data_keys{$datum} = 'dg'.$gdi++;
1056 my $key = $root->addNewChild( $graphml_ns, 'key' );
1057 $key->setAttribute( 'attr.name', $datum );
1058 $key->setAttribute( 'attr.type', $graph_attributes{$datum} );
1059 $key->setAttribute( 'for', 'graph' );
1060 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1063 # Add the data keys for reading nodes
1064 my %reading_attributes;
1065 my $rmeta = Text::Tradition::Collation::Reading->meta;
1066 foreach my $attr( $rmeta->get_all_attributes ) {
1067 next if $attr->name =~ /^_/;
1068 next if $skipsave{$attr->name};
1069 next unless $save_types{$attr->type_constraint->name};
1070 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1072 # Extra custom key for the reading morphology
1073 $reading_attributes{'lexemes'} = 'string';
1077 foreach my $datum ( sort keys %reading_attributes ) {
1078 $node_data_keys{$datum} = 'dn'.$ndi++;
1079 my $key = $root->addNewChild( $graphml_ns, 'key' );
1080 $key->setAttribute( 'attr.name', $datum );
1081 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1082 $key->setAttribute( 'for', 'node' );
1083 $key->setAttribute( 'id', $node_data_keys{$datum} );
1086 # Add the data keys for edges, that is, paths and relationships. Path
1087 # data does not come from a Moose class so is here manually.
1090 my %edge_attributes = (
1091 witness => 'string', # ID/label for a path
1092 extra => 'boolean', # Path key
1094 my @path_attributes = keys %edge_attributes; # track our manual additions
1095 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1096 foreach my $attr( $pmeta->get_all_attributes ) {
1097 next if $attr->name =~ /^_/;
1098 next if $skipsave{$attr->name};
1099 next unless $save_types{$attr->type_constraint->name};
1100 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1102 foreach my $datum ( sort keys %edge_attributes ) {
1103 $edge_data_keys{$datum} = 'de'.$edi++;
1104 my $key = $root->addNewChild( $graphml_ns, 'key' );
1105 $key->setAttribute( 'attr.name', $datum );
1106 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1107 $key->setAttribute( 'for', 'edge' );
1108 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1111 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1112 my $xmlidname = $self->tradition->name;
1113 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1114 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1115 $xmlidname = '_'.$xmlidname;
1117 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1118 $sgraph->setAttribute( 'edgedefault', 'directed' );
1119 $sgraph->setAttribute( 'id', $xmlidname );
1120 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1121 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1122 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1123 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1124 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1126 # Tradition/collation attribute data
1127 foreach my $datum ( keys %graph_attributes ) {
1129 if( $datum eq 'version' ) {
1131 } elsif( $datum eq 'stemmata' ) {
1133 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1134 $self->tradition->stemmata;
1135 $value = join( "\n", @stemstrs );
1136 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1137 $value = $self->tradition->$datum;
1139 $value = $self->$datum;
1141 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1146 # Add our readings to the graph
1147 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1148 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1149 ( $n->rank < $start->rank || $n->rank > $end->rank );
1150 $use_readings{$n->id} = 1;
1151 # Add to the main graph
1152 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1153 my $node_xmlid = 'n' . $node_ctr++;
1154 $node_hash{ $n->id } = $node_xmlid;
1155 $node_el->setAttribute( 'id', $node_xmlid );
1156 foreach my $d ( keys %reading_attributes ) {
1158 # Custom serialization
1159 if( $d eq 'lexemes' ) {
1160 # If nval is a true value, we have lexemes so we need to
1161 # serialize them. Otherwise set nval to undef so that the
1162 # key is excluded from this reading.
1163 $nval = $nval ? $n->_serialize_lexemes : undef;
1164 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1167 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1168 # Adjust the ranks within the subgraph.
1169 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1170 : $nval - $rankoffset;
1172 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1177 # Add the path edges to the sequence graph
1179 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1180 # We add an edge in the graphml for every witness in $e.
1181 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1182 my @edge_wits = sort $self->path_witnesses( $e );
1183 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1184 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1185 # Skip any path from start to end; that witness is not in the subgraph.
1186 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1187 foreach my $wit ( @edge_wits ) {
1188 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1189 $node_hash{ $e->[0] },
1190 $node_hash{ $e->[1] } );
1191 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1192 $edge_el->setAttribute( 'source', $from );
1193 $edge_el->setAttribute( 'target', $to );
1194 $edge_el->setAttribute( 'id', $id );
1196 # It's a witness path, so add the witness
1198 my $key = $edge_data_keys{'witness'};
1199 # Is this an ante-corr witness?
1200 my $aclabel = $self->ac_label;
1201 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1202 # Keep the base witness
1204 # ...and record that this is an 'extra' reading path
1205 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1207 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1211 # Report the actual number of nodes and edges that went in
1212 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1213 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1215 # Add the relationship graph to the XML
1216 map { delete $edge_data_keys{$_} } @path_attributes;
1217 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1218 $node_data_keys{'id'}, \%edge_data_keys );
1220 # Save and return the thing
1221 my $result = decode_utf8( $graphml->toString(1) );
1225 sub _add_graphml_data {
1226 my( $el, $key, $value ) = @_;
1227 return unless defined $value;
1228 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1229 $data_el->setAttribute( 'key', $key );
1230 $data_el->appendText( $value );
1235 Returns a CSV alignment table representation of the collation graph, one
1236 row per witness (or witness uncorrected.)
1242 my $table = $self->alignment_table;
1243 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1245 # Make the header row
1246 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1247 push( @result, decode_utf8( $csv->string ) );
1248 # Make the rest of the rows
1249 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1250 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1251 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1252 $csv->combine( @row );
1253 push( @result, decode_utf8( $csv->string ) );
1255 return join( "\n", @result );
1258 =head2 alignment_table( $use_refs, $include_witnesses )
1260 Return a reference to an alignment table, in a slightly enhanced CollateX
1261 format which looks like this:
1263 $table = { alignment => [ { witness => "SIGIL",
1264 tokens => [ { t => "TEXT" }, ... ] },
1265 { witness => "SIG2",
1266 tokens => [ { t => "TEXT" }, ... ] },
1268 length => TEXTLEN };
1270 If $use_refs is set to 1, the reading object is returned in the table
1271 instead of READINGTEXT; if not, the text of the reading is returned.
1273 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1274 keys have a true hash value will be included.
1278 sub alignment_table {
1280 $self->calculate_ranks() unless $self->_graphcalc_done;
1281 return $self->cached_table if $self->has_cached_table;
1283 # Make sure we can do this
1284 throw( "Need a linear graph in order to make an alignment table" )
1285 unless $self->linear;
1286 $self->calculate_ranks unless $self->end->has_rank;
1288 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1289 my @all_pos = ( 1 .. $self->end->rank - 1 );
1290 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1291 # say STDERR "Making witness row(s) for " . $wit->sigil;
1292 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1293 my @row = _make_witness_row( \@wit_path, \@all_pos );
1294 push( @{$table->{'alignment'}},
1295 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1296 if( $wit->is_layered ) {
1297 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1298 $wit->sigil.$self->ac_label );
1299 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1300 push( @{$table->{'alignment'}},
1301 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1304 $self->cached_table( $table );
1308 sub _make_witness_row {
1309 my( $path, $positions ) = @_;
1311 map { $char_hash{$_} = undef } @$positions;
1313 foreach my $rdg ( @$path ) {
1314 my $rtext = $rdg->text;
1315 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1316 say STDERR "rank " . $rdg->rank if $debug;
1317 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1318 $char_hash{$rdg->rank} = { 't' => $rdg };
1320 my @row = map { $char_hash{$_} } @$positions;
1321 # Fill in lacuna markers for undef spots in the row
1322 my $last_el = shift @row;
1323 my @filled_row = ( $last_el );
1324 foreach my $el ( @row ) {
1325 # If we are using node reference, make the lacuna node appear many times
1326 # in the table. If not, use the lacuna tag.
1327 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1330 push( @filled_row, $el );
1336 =head1 NAVIGATION METHODS
1338 =head2 reading_sequence( $first, $last, $sigil, $backup )
1340 Returns the ordered list of readings, starting with $first and ending
1341 with $last, for the witness given in $sigil. If a $backup sigil is
1342 specified (e.g. when walking a layered witness), it will be used wherever
1343 no $sigil path exists. If there is a base text reading, that will be
1344 used wherever no path exists for $sigil or $backup.
1348 # TODO Think about returning some lazy-eval iterator.
1349 # TODO Get rid of backup; we should know from what witness is whether we need it.
1351 sub reading_sequence {
1352 my( $self, $start, $end, $witness ) = @_;
1354 $witness = $self->baselabel unless $witness;
1355 my @readings = ( $start );
1358 while( $n && $n->id ne $end->id ) {
1359 if( exists( $seen{$n->id} ) ) {
1360 throw( "Detected loop for $witness at " . $n->id );
1364 my $next = $self->next_reading( $n, $witness );
1366 throw( "Did not find any path for $witness from reading " . $n->id );
1368 push( @readings, $next );
1371 # Check that the last reading is our end reading.
1372 my $last = $readings[$#readings];
1373 throw( "Last reading found from " . $start->text .
1374 " for witness $witness is not the end!" ) # TODO do we get this far?
1375 unless $last->id eq $end->id;
1380 =head2 next_reading( $reading, $sigil );
1382 Returns the reading that follows the given reading along the given witness
1388 # Return the successor via the corresponding path.
1390 my $answer = $self->_find_linked_reading( 'next', @_ );
1391 return undef unless $answer;
1392 return $self->reading( $answer );
1395 =head2 prior_reading( $reading, $sigil )
1397 Returns the reading that precedes the given reading along the given witness
1403 # Return the predecessor via the corresponding path.
1405 my $answer = $self->_find_linked_reading( 'prior', @_ );
1406 return $self->reading( $answer );
1409 sub _find_linked_reading {
1410 my( $self, $direction, $node, $path ) = @_;
1412 # Get a backup if we are dealing with a layered witness
1414 my $aclabel = $self->ac_label;
1415 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1419 my @linked_paths = $direction eq 'next'
1420 ? $self->sequence->edges_from( $node )
1421 : $self->sequence->edges_to( $node );
1422 return undef unless scalar( @linked_paths );
1424 # We have to find the linked path that contains all of the
1425 # witnesses supplied in $path.
1426 my( @path_wits, @alt_path_wits );
1427 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1428 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1431 foreach my $le ( @linked_paths ) {
1432 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1435 my @le_wits = sort $self->path_witnesses( $le );
1436 if( _is_within( \@path_wits, \@le_wits ) ) {
1437 # This is the right path.
1438 return $direction eq 'next' ? $le->[1] : $le->[0];
1439 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1443 # Got this far? Return the alternate path if it exists.
1444 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1447 # Got this far? Return the base path if it exists.
1448 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1451 # Got this far? We have no appropriate path.
1452 warn "Could not find $direction node from " . $node->id
1453 . " along path $path";
1459 my( $set1, $set2 ) = @_;
1460 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1461 foreach my $el ( @$set1 ) {
1462 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1467 # Return the string that joins together a list of witnesses for
1468 # display on a single path.
1469 sub _witnesses_of_label {
1470 my( $self, $label ) = @_;
1471 my $regex = $self->wit_list_separator;
1472 my @answer = split( /\Q$regex\E/, $label );
1476 =head2 common_readings
1478 Returns the list of common readings in the graph (i.e. those readings that are
1479 shared by all non-lacunose witnesses.)
1483 sub common_readings {
1485 my @common = grep { $_->is_common } $self->readings;
1489 =head2 path_text( $sigil, [, $start, $end ] )
1491 Returns the text of a witness (plus its backup, if we are using a layer)
1492 as stored in the collation. The text is returned as a string, where the
1493 individual readings are joined with spaces and the meta-readings (e.g.
1494 lacunae) are omitted. Optional specification of $start and $end allows
1495 the generation of a subset of the witness text.
1500 my( $self, $wit, $start, $end ) = @_;
1501 $start = $self->start unless $start;
1502 $end = $self->end unless $end;
1503 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1506 foreach my $r ( @path ) {
1507 unless ( $r->join_prior || !$last || $last->join_next ) {
1510 $pathtext .= $r->text;
1516 =head1 INITIALIZATION METHODS
1518 These are mostly for use by parsers.
1520 =head2 make_witness_path( $witness )
1522 Link the array of readings contained in $witness->path (and in
1523 $witness->uncorrected_path if it exists) into collation paths.
1524 Clear out the arrays when finished.
1526 =head2 make_witness_paths
1528 Call make_witness_path for all witnesses in the tradition.
1532 # For use when a collation is constructed from a base text and an apparatus.
1533 # We have the sequences of readings and just need to add path edges.
1534 # When we are done, clear out the witness path attributes, as they are no
1536 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1538 sub make_witness_paths {
1540 foreach my $wit ( $self->tradition->witnesses ) {
1541 # say STDERR "Making path for " . $wit->sigil;
1542 $self->make_witness_path( $wit );
1546 sub make_witness_path {
1547 my( $self, $wit ) = @_;
1548 my @chain = @{$wit->path};
1549 my $sig = $wit->sigil;
1550 # Add start and end if necessary
1551 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1552 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1553 foreach my $idx ( 0 .. $#chain-1 ) {
1554 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1556 if( $wit->is_layered ) {
1557 @chain = @{$wit->uncorrected_path};
1558 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1559 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1560 foreach my $idx( 0 .. $#chain-1 ) {
1561 my $source = $chain[$idx];
1562 my $target = $chain[$idx+1];
1563 $self->add_path( $source, $target, $sig.$self->ac_label )
1564 unless $self->has_path( $source, $target, $sig );
1568 $wit->clear_uncorrected_path;
1571 =head2 calculate_ranks
1573 Calculate the reading ranks (that is, their aligned positions relative
1574 to each other) for the graph. This can only be called on linear collations.
1578 use Text::Tradition;
1580 my $cxfile = 't/data/Collatex-16.xml';
1581 my $t = Text::Tradition->new(
1583 'input' => 'CollateX',
1586 my $c = $t->collation;
1589 my $table = $c->alignment_table;
1590 ok( $c->has_cached_table, "Alignment table was cached" );
1591 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1592 $c->calculate_ranks;
1593 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1594 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1595 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1601 sub calculate_ranks {
1603 # Save the existing ranks, in case we need to invalidate the cached SVG.
1605 map { $existing_ranks{$_} = $_->rank } $self->readings;
1607 # Do the rankings based on the relationship equivalence graph, starting
1608 # with the start node.
1609 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1611 # Transfer our rankings from the topological graph to the real one.
1612 foreach my $r ( $self->readings ) {
1613 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1614 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1616 # Die. Find the last rank we calculated.
1617 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1618 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1620 my $last = pop @all_defined;
1621 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1624 # Do we need to invalidate the cached data?
1625 if( $self->has_cached_svg || $self->has_cached_table ) {
1626 foreach my $r ( $self->readings ) {
1627 next if defined( $existing_ranks{$r} )
1628 && $existing_ranks{$r} == $r->rank;
1629 # Something has changed, so clear the cache
1630 $self->_clear_cache;
1631 # ...and recalculate the common readings.
1632 $self->calculate_common_readings();
1636 # The graph calculation information is now up to date.
1637 $self->_graphcalc_done(1);
1642 $self->wipe_svg if $self->has_cached_svg;
1643 $self->wipe_table if $self->has_cached_table;
1647 =head2 flatten_ranks
1649 A convenience method for parsing collation data. Searches the graph for readings
1650 with the same text at the same rank, and merges any that are found.
1656 my %unique_rank_rdg;
1658 foreach my $rdg ( $self->readings ) {
1659 next unless $rdg->has_rank;
1660 my $key = $rdg->rank . "||" . $rdg->text;
1661 if( exists $unique_rank_rdg{$key} ) {
1662 # Make sure they don't have different grammatical forms
1663 my $ur = $unique_rank_rdg{$key};
1664 if( $rdg->disambiguated && $ur->disambiguated ) {
1665 my $rform = join( '//', map { $_->form->to_string } $rdg->lexemes );
1666 my $uform = join( '//', map { $_->form->to_string } $ur->lexemes );
1667 next unless $rform eq $uform;
1668 } elsif( $rdg->disambiguated xor $ur->disambiguated ) {
1672 #say STDERR "Combining readings at same rank: $key";
1674 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1675 # TODO see if this now makes a common point.
1677 $unique_rank_rdg{$key} = $rdg;
1680 # If we merged readings, the ranks are still fine but the alignment
1681 # table is wrong. Wipe it.
1682 $self->wipe_table() if $changed;
1686 =head2 calculate_common_readings
1688 Goes through the graph identifying the readings that appear in every witness
1689 (apart from those with lacunae at that spot.) Marks them as common and returns
1694 use Text::Tradition;
1696 my $cxfile = 't/data/Collatex-16.xml';
1697 my $t = Text::Tradition->new(
1699 'input' => 'CollateX',
1702 my $c = $t->collation;
1704 my @common = $c->calculate_common_readings();
1705 is( scalar @common, 8, "Found correct number of common readings" );
1706 my @marked = sort $c->common_readings();
1707 is( scalar @common, 8, "All common readings got marked as such" );
1708 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1709 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1715 sub calculate_common_readings {
1718 map { $_->is_common( 0 ) } $self->readings;
1719 # Implicitly calls calculate_ranks
1720 my $table = $self->alignment_table;
1721 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1722 my @row = map { $_->{'tokens'}->[$idx]
1723 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1724 @{$table->{'alignment'}};
1726 foreach my $r ( @row ) {
1728 $hash{$r->id} = $r unless $r->is_meta;
1730 $hash{'UNDEF'} = $r;
1733 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1734 my( $r ) = values %hash;
1736 push( @common, $r );
1742 =head2 text_from_paths
1744 Calculate the text array for all witnesses from the path, for later consistency
1745 checking. Only to be used if there is no non-graph-based way to know the
1750 sub text_from_paths {
1752 foreach my $wit ( $self->tradition->witnesses ) {
1753 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1755 foreach my $r ( @readings ) {
1756 next if $r->is_meta;
1757 push( @text, $r->text );
1759 $wit->text( \@text );
1760 if( $wit->is_layered ) {
1761 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1762 $wit->sigil.$self->ac_label );
1764 foreach my $r ( @ucrdgs ) {
1765 next if $r->is_meta;
1766 push( @uctext, $r->text );
1768 $wit->layertext( \@uctext );
1773 =head1 UTILITY FUNCTIONS
1775 =head2 common_predecessor( $reading_a, $reading_b )
1777 Find the last reading that occurs in sequence before both the given readings.
1778 At the very least this should be $self->start.
1780 =head2 common_successor( $reading_a, $reading_b )
1782 Find the first reading that occurs in sequence after both the given readings.
1783 At the very least this should be $self->end.
1787 use Text::Tradition;
1789 my $cxfile = 't/data/Collatex-16.xml';
1790 my $t = Text::Tradition->new(
1792 'input' => 'CollateX',
1795 my $c = $t->collation;
1797 is( $c->common_predecessor( 'n24', 'n23' )->id,
1798 'n20', "Found correct common predecessor" );
1799 is( $c->common_successor( 'n24', 'n23' )->id,
1800 '__END__', "Found correct common successor" );
1802 is( $c->common_predecessor( 'n19', 'n17' )->id,
1803 'n16', "Found correct common predecessor for readings on same path" );
1804 is( $c->common_successor( 'n21', 'n10' )->id,
1805 '__END__', "Found correct common successor for readings on same path" );
1811 ## Return the closest reading that is a predecessor of both the given readings.
1812 sub common_predecessor {
1814 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1815 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1818 sub common_successor {
1820 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1821 return $self->_common_in_path( $r1, $r2, 'successors' );
1825 # TODO think about how to do this without ranks...
1826 sub _common_in_path {
1827 my( $self, $r1, $r2, $dir ) = @_;
1828 my $iter = $self->end->rank;
1830 my @last_r1 = ( $r1 );
1831 my @last_r2 = ( $r2 );
1832 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1834 # say STDERR "Finding common $dir for $r1, $r2";
1835 while( !@candidates ) {
1836 last unless $iter--; # Avoid looping infinitely
1837 # Iterate separately down the graph from r1 and r2
1838 my( @new_lc1, @new_lc2 );
1839 foreach my $lc ( @last_r1 ) {
1840 foreach my $p ( $lc->$dir ) {
1841 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1842 # say STDERR "Path candidate $p from $lc";
1843 push( @candidates, $p );
1844 } elsif( !$all_seen{$p->id} ) {
1845 $all_seen{$p->id} = 'r1';
1846 push( @new_lc1, $p );
1850 foreach my $lc ( @last_r2 ) {
1851 foreach my $p ( $lc->$dir ) {
1852 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1853 # say STDERR "Path candidate $p from $lc";
1854 push( @candidates, $p );
1855 } elsif( !$all_seen{$p->id} ) {
1856 $all_seen{$p->id} = 'r2';
1857 push( @new_lc2, $p );
1861 @last_r1 = @new_lc1;
1862 @last_r2 = @new_lc2;
1864 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1865 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1869 Text::Tradition::Error->throw(
1870 'ident' => 'Collation error',
1876 __PACKAGE__->meta->make_immutable;
1880 This package is free software and is provided "as is" without express
1881 or implied warranty. You can redistribute it and/or modify it under
1882 the same terms as Perl itself.
1886 Tara L Andrews E<lt>aurum@cpan.orgE<gt>