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_table' => (
108 predicate => 'has_cached_table',
109 clearer => 'wipe_table',
112 has '_graphcalc_done' => (
120 Text::Tradition::Collation - a software model for a text collation
125 my $t = Text::Tradition->new(
126 'name' => 'this is a text',
128 'file' => '/path/to/tei_parallel_seg_file.xml' );
130 my $c = $t->collation;
131 my @readings = $c->readings;
132 my @paths = $c->paths;
133 my @relationships = $c->relationships;
135 my $svg_variant_graph = $t->collation->as_svg();
139 Text::Tradition is a library for representation and analysis of collated
140 texts, particularly medieval ones. The Collation is the central feature of
141 a Tradition, where the text, its sequence of readings, and its relationships
142 between readings are actually kept.
148 The constructor. Takes a hash or hashref of the following arguments:
152 =item * tradition - The Text::Tradition object to which the collation
155 =item * linear - Whether the collation should be linear; that is, whether
156 transposed readings should be treated as two linked readings rather than one,
157 and therefore whether the collation graph is acyclic. Defaults to true.
159 =item * baselabel - The default label for the path taken by a base text
160 (if any). Defaults to 'base text'.
162 =item * wit_list_separator - The string to join a list of witnesses for
163 purposes of making labels in display graphs. Defaults to ', '.
165 =item * ac_label - The extra label to tack onto a witness sigil when
166 representing another layer of path for the given witness - that is, when
167 a text has more than one possible reading due to scribal corrections or
168 the like. Defaults to ' (a.c.)'.
170 =item * wordsep - The string used to separate words in the original text.
181 =head2 wit_list_separator
189 Simple accessors for collation attributes.
193 The meta-reading at the start of every witness path.
197 The meta-reading at the end of every witness path.
201 Returns all Reading objects in the graph.
203 =head2 reading( $id )
205 Returns the Reading object corresponding to the given ID.
207 =head2 add_reading( $reading_args )
209 Adds a new reading object to the collation.
210 See L<Text::Tradition::Collation::Reading> for the available arguments.
212 =head2 del_reading( $object_or_id )
214 Removes the given reading from the collation, implicitly removing its
215 paths and relationships.
217 =head2 merge_readings( $main, $second, $concatenate, $with_str )
219 Merges the $second reading into the $main one. If $concatenate is true, then
220 the merged node will carry the text of both readings, concatenated with either
221 $with_str (if specified) or a sensible default (the empty string if the
222 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
224 The first two arguments may be either readings or reading IDs.
226 =head2 has_reading( $id )
228 Predicate to see whether a given reading ID is in the graph.
230 =head2 reading_witnesses( $object_or_id )
232 Returns a list of sigils whose witnesses contain the reading.
236 Returns all reading paths within the document - that is, all edges in the
237 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
239 =head2 add_path( $source, $target, $sigil )
241 Links the given readings in the collation in sequence, under the given witness
242 sigil. The readings may be specified by object or ID.
244 =head2 del_path( $source, $target, $sigil )
246 Links the given readings in the collation in sequence, under the given witness
247 sigil. The readings may be specified by object or ID.
249 =head2 has_path( $source, $target );
251 Returns true if the two readings are linked in sequence in any witness.
252 The readings may be specified by object or ID.
256 Returns all Relationship objects in the collation.
258 =head2 add_relationship( $reading, $other_reading, $options )
260 Adds a new relationship of the type given in $options between the two readings,
261 which may be specified by object or ID. Returns a value of ( $status, @vectors)
262 where $status is true on success, and @vectors is a list of relationship edges
263 that were ultimately added.
264 See L<Text::Tradition::Collation::Relationship> for the available options.
270 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
271 $self->_set_start( $self->add_reading(
272 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
273 $self->_set_end( $self->add_reading(
274 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
277 ### Reading construct/destruct functions
280 my( $self, $reading ) = @_;
281 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
282 my %args = %$reading;
283 if( $args{'init'} ) {
284 # If we are initializing an empty collation, don't assume that we
285 # have set a tradition.
286 delete $args{'init'};
287 } elsif( $self->tradition->has_language && !exists $args{'language'} ) {
288 $args{'language'} = $self->tradition->language;
290 $reading = Text::Tradition::Collation::Reading->new(
291 'collation' => $self,
294 # First check to see if a reading with this ID exists.
295 if( $self->reading( $reading->id ) ) {
296 throw( "Collation already has a reading with id " . $reading->id );
298 $self->_graphcalc_done(0);
299 $self->_add_reading( $reading->id => $reading );
300 # Once the reading has been added, put it in both graphs.
301 $self->sequence->add_vertex( $reading->id );
302 $self->relations->add_reading( $reading->id );
306 around del_reading => sub {
311 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
314 # Remove the reading from the graphs.
315 $self->_graphcalc_done(0);
316 $self->_clear_cache; # Explicitly clear caches to GC the reading
317 $self->sequence->delete_vertex( $arg );
318 $self->relations->delete_reading( $arg );
321 $self->$orig( $arg );
328 my $cxfile = 't/data/Collatex-16.xml';
329 my $t = Text::Tradition->new(
331 'input' => 'CollateX',
334 my $c = $t->collation;
336 my $rno = scalar $c->readings;
337 # Split n21 for testing purposes
338 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
339 my $old_r = $c->reading( 'n21' );
340 $old_r->alter_text( 'to' );
341 $c->del_path( 'n20', 'n21', 'A' );
342 $c->add_path( 'n20', 'n21p0', 'A' );
343 $c->add_path( 'n21p0', 'n21', 'A' );
345 ok( $c->reading( 'n21p0' ), "New reading exists" );
346 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
348 # Combine n3 and n4 ( with his )
349 $c->merge_readings( 'n3', 'n4', 1 );
350 ok( !$c->reading('n4'), "Reading n4 is gone" );
351 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
353 # Collapse n9 and n10 ( rood / root )
354 $c->merge_readings( 'n9', 'n10' );
355 ok( !$c->reading('n10'), "Reading n10 is gone" );
356 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
358 # Combine n21 and n21p0
359 my $remaining = $c->reading('n21');
360 $remaining ||= $c->reading('n22'); # one of these should still exist
361 $c->merge_readings( 'n21p0', $remaining, 1 );
362 ok( !$c->reading('n21'), "Reading $remaining is gone" );
363 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
373 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
374 my $mergemeta = $kept_obj->is_meta;
375 throw( "Cannot merge meta and non-meta reading" )
376 unless ( $mergemeta && $del_obj->is_meta )
377 || ( !$mergemeta && !$del_obj->is_meta );
379 throw( "Cannot merge with start or end node" )
380 if( $kept_obj eq $self->start || $kept_obj eq $self->end
381 || $del_obj eq $self->start || $del_obj eq $self->end );
383 # We only need the IDs for adding paths to the graph, not the reading
384 # objects themselves.
385 my $kept = $kept_obj->id;
386 my $deleted = $del_obj->id;
387 $self->_graphcalc_done(0);
389 # The kept reading should inherit the paths and the relationships
390 # of the deleted reading.
391 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
392 my @vector = ( $kept );
393 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
394 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
395 next if $vector[0] eq $vector[1]; # Don't add a self loop
396 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
397 $self->sequence->add_edge( @vector );
398 my $fwits = $self->sequence->get_edge_attributes( @vector );
399 @wits{keys %$fwits} = values %$fwits;
400 $self->sequence->set_edge_attributes( @vector, \%wits );
402 $self->relations->merge_readings( $kept, $deleted, $combine );
404 # Do the deletion deed.
406 # Combine the text of the readings
407 my $joinstr = $combine_char;
408 unless( defined $joinstr ) {
409 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
410 $joinstr = $self->wordsep unless defined $joinstr;
412 $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) );
413 # Change this reading to a joining one if necessary
414 $kept_obj->_set_join_next( $del_obj->join_next );
415 $kept_obj->normal_form(
416 join( $joinstr, $kept_obj->normal_form, $del_obj->normal_form ) );
417 # Combine the lexemes present in the readings
418 if( $kept_obj->has_lexemes && $del_obj->has_lexemes ) {
419 $kept_obj->add_lexeme( $del_obj->lexemes );
422 $self->del_reading( $deleted );
425 =head2 compress_readings
427 Where possible in the graph, compresses plain sequences of readings into a
428 single reading. The sequences must consist of readings with no
429 relationships to other readings, with only a single witness path between
430 them and no other witness paths from either that would skip the other. The
431 readings must also not be marked as nonsense or bad grammar.
433 WARNING: This operation cannot be undone.
437 sub compress_readings {
439 # Anywhere in the graph that there is a reading that joins only to a single
440 # successor, and neither of these have any relationships, just join the two
443 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
444 next if $rdg->is_meta;
445 next if $gobbled{$rdg->id};
446 next if $rdg->grammar_invalid || $rdg->is_nonsense;
447 next if $rdg->related_readings();
449 while( $self->sequence->successors( $rdg ) == 1 ) {
450 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
451 throw( "Infinite loop" ) if $seen{$next->id};
452 $seen{$next->id} = 1;
453 last if $self->sequence->predecessors( $next ) > 1;
454 last if $next->is_meta;
455 last if $next->grammar_invalid || $next->is_nonsense;
456 last if $next->related_readings();
457 say "Joining readings $rdg and $next";
458 $self->merge_readings( $rdg, $next, 1 );
461 # Make sure we haven't screwed anything up
462 foreach my $wit ( $self->tradition->witnesses ) {
463 my $pathtext = $self->path_text( $wit->sigil );
464 my $origtext = join( ' ', @{$wit->text} );
465 throw( "Text differs for witness " . $wit->sigil )
466 unless $pathtext eq $origtext;
467 if( $wit->is_layered ) {
468 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
469 $origtext = join( ' ', @{$wit->layertext} );
470 throw( "Ante-corr text differs for witness " . $wit->sigil )
471 unless $pathtext eq $origtext;
475 $self->relations->rebuild_equivalence();
476 $self->calculate_ranks();
479 # Helper function for manipulating the graph.
480 sub _stringify_args {
481 my( $self, $first, $second, @args ) = @_;
483 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
484 $second = $second->id
485 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
486 return( $first, $second, @args );
489 # Helper function for manipulating the graph.
490 sub _objectify_args {
491 my( $self, $first, $second, $arg ) = @_;
492 $first = $self->reading( $first )
493 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
494 $second = $self->reading( $second )
495 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
496 return( $first, $second, $arg );
503 # We only need the IDs for adding paths to the graph, not the reading
504 # objects themselves.
505 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
507 $self->_graphcalc_done(0);
508 # Connect the readings
509 unless( $self->sequence->has_edge( $source, $target ) ) {
510 $self->sequence->add_edge( $source, $target );
511 $self->relations->add_equivalence_edge( $source, $target );
513 # Note the witness in question
514 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
520 if( ref( $_[0] ) eq 'ARRAY' ) {
527 # We only need the IDs for adding paths to the graph, not the reading
528 # objects themselves.
529 my( $source, $target, $wit ) = $self->_stringify_args( @args );
531 $self->_graphcalc_done(0);
532 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
533 $self->sequence->delete_edge_attribute( $source, $target, $wit );
535 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
536 $self->sequence->delete_edge( $source, $target );
537 $self->relations->delete_equivalence_edge( $source, $target );
542 # Extra graph-alike utility
545 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
546 return undef unless $self->sequence->has_edge( $source, $target );
547 return $self->sequence->has_edge_attribute( $source, $target, $wit );
550 =head2 clear_witness( @sigil_list )
552 Clear the given witnesses out of the collation entirely, removing references
553 to them in paths, and removing readings that belong only to them. Should only
554 be called via $tradition->del_witness.
559 my( $self, @sigils ) = @_;
561 $self->_graphcalc_done(0);
562 # Clear the witness(es) out of the paths
563 foreach my $e ( $self->paths ) {
564 foreach my $sig ( @sigils ) {
565 $self->del_path( $e, $sig );
569 # Clear out the newly unused readings
570 foreach my $r ( $self->readings ) {
571 unless( $self->reading_witnesses( $r ) ) {
572 $self->del_reading( $r );
577 sub add_relationship {
579 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
580 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
581 $self->_graphcalc_done(0);
585 around qw/ get_relationship del_relationship / => sub {
589 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
592 my( $source, $target ) = $self->_stringify_args( @args );
593 $self->$orig( $source, $target );
596 =head2 reading_witnesses( $reading )
598 Return a list of sigils corresponding to the witnesses in which the reading appears.
602 sub reading_witnesses {
603 my( $self, $reading ) = @_;
604 # We need only check either the incoming or the outgoing edges; I have
605 # arbitrarily chosen "incoming". Thus, special-case the start node.
606 if( $reading eq $self->start ) {
607 return map { $_->sigil } $self->tradition->witnesses;
610 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
611 my $wits = $self->sequence->get_edge_attributes( @$e );
612 @all_witnesses{ keys %$wits } = 1;
614 my $acstr = $self->ac_label;
615 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
616 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
618 return keys %all_witnesses;
621 =head1 OUTPUT METHODS
623 =head2 as_svg( \%options )
625 Returns an SVG string that represents the graph, via as_dot and graphviz.
626 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
631 my( $self, $opts ) = @_;
632 throw( "Need GraphViz installed to output SVG" )
633 unless File::Which::which( 'dot' );
634 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
635 $self->calculate_ranks()
636 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
637 my @cmd = qw/dot -Tsvg/;
639 my $dotfile = File::Temp->new();
641 # $dotfile->unlink_on_destroy(0);
642 binmode $dotfile, ':utf8';
643 print $dotfile $self->as_dot( $opts );
644 push( @cmd, $dotfile->filename );
645 run( \@cmd, ">", binary(), \$svg );
646 $svg = decode_utf8( $svg );
651 =head2 as_dot( \%options )
653 Returns a string that is the collation graph expressed in dot
654 (i.e. GraphViz) format. Options include:
669 my( $self, $opts ) = @_;
670 my $startrank = $opts->{'from'} if $opts;
671 my $endrank = $opts->{'to'} if $opts;
672 my $color_common = $opts->{'color_common'} if $opts;
673 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
674 && $self->end->rank > 100;
675 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
677 # Check the arguments
679 return if $endrank && $startrank > $endrank;
680 return if $startrank > $self->end->rank;
682 if( defined $endrank ) {
683 return if $endrank < 0;
684 $endrank = undef if $endrank == $self->end->rank;
687 my $graph_name = $self->tradition->name;
688 $graph_name =~ s/[^\w\s]//g;
689 $graph_name = join( '_', split( /\s+/, $graph_name ) );
697 'fillcolor' => 'white',
702 'arrowhead' => 'open',
703 'color' => '#000000',
704 'fontcolor' => '#000000',
707 my $dot = sprintf( "digraph %s {\n", $graph_name );
708 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
709 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
711 # Output substitute start/end readings if necessary
713 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
716 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
718 if( $STRAIGHTENHACK ) {
720 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
721 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
722 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
724 my %used; # Keep track of the readings that actually appear in the graph
725 # Sort the readings by rank if we have ranks; this speeds layout.
726 my @all_readings = $self->end->has_rank
727 ? sort { $a->rank <=> $b->rank } $self->readings
729 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
730 foreach my $reading ( @all_readings ) {
731 # Only output readings within our rank range.
732 next if $startrank && $reading->rank < $startrank;
733 next if $endrank && $reading->rank > $endrank;
734 $used{$reading->id} = 1;
735 # Need not output nodes without separate labels
736 next if $reading->id eq $reading->text;
738 my $label = $reading->text;
739 $label .= '-' if $reading->join_next;
740 $label = "-$label" if $reading->join_prior;
741 $label =~ s/\"/\\\"/g;
742 $rattrs->{'label'} = $label;
743 $rattrs->{'id'} = $reading->id;
744 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
745 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
748 # Add the real edges. Need to weight one edge per rank jump, in a
750 # my $weighted = $self->_add_edge_weights;
751 my @edges = $self->paths;
752 my( %substart, %subend );
753 foreach my $edge ( @edges ) {
754 # Do we need to output this edge?
755 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
756 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
757 my $variables = { %edge_attrs, 'label' => $label };
759 # Account for the rank gap if necessary
760 my $rank0 = $self->reading( $edge->[0] )->rank
761 if $self->reading( $edge->[0] )->has_rank;
762 my $rank1 = $self->reading( $edge->[1] )->rank
763 if $self->reading( $edge->[1] )->has_rank;
764 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
765 $variables->{'minlen'} = $rank1 - $rank0;
768 # Add the calculated edge weights
769 # if( exists $weighted->{$edge->[0]}
770 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
771 # # $variables->{'color'} = 'red';
772 # $variables->{'weight'} = 3.0;
775 # EXPERIMENTAL: make edge width reflect no. of witnesses
776 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
777 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
779 my $varopts = _dot_attr_string( $variables );
780 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
781 $edge->[0], $edge->[1], $varopts );
782 } elsif( $used{$edge->[0]} ) {
783 $subend{$edge->[0]} = $edge->[1];
784 } elsif( $used{$edge->[1]} ) {
785 $substart{$edge->[1]} = $edge->[0];
788 # Add substitute start and end edges if necessary
789 foreach my $node ( keys %substart ) {
790 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
791 my $variables = { %edge_attrs, 'label' => $witstr };
792 my $nrdg = $self->reading( $node );
793 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
794 # Substart is actually one lower than $startrank
795 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
797 my $varopts = _dot_attr_string( $variables );
798 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
800 foreach my $node ( keys %subend ) {
801 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
802 my $variables = { %edge_attrs, 'label' => $witstr };
803 my $varopts = _dot_attr_string( $variables );
804 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
807 if( $STRAIGHTENHACK ) {
808 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
809 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
816 sub _dot_attr_string {
819 foreach my $k ( sort keys %$hash ) {
821 push( @attrs, $k.'="'.$v.'"' );
823 return( '[ ' . join( ', ', @attrs ) . ' ]' );
826 sub _add_edge_weights {
828 # Walk the graph from START to END, choosing the successor node with
829 # the largest number of witness paths each time.
831 my $curr = $self->start->id;
832 my $ranked = $self->end->has_rank;
833 while( $curr ne $self->end->id ) {
834 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
835 my @succ = sort { $self->path_witnesses( $curr, $a )
836 <=> $self->path_witnesses( $curr, $b ) }
837 $self->sequence->successors( $curr );
838 my $next = pop @succ;
839 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
840 # Try to avoid lacunae in the weighted path.
842 ( $self->reading( $next )->is_lacuna ||
843 $nextrank - $rank > 1 ) ){
846 $weighted->{$curr} = $next;
852 =head2 path_witnesses( $edge )
854 Returns the list of sigils whose witnesses are associated with the given edge.
855 The edge can be passed as either an array or an arrayref of ( $source, $target ).
860 my( $self, @edge ) = @_;
861 # If edge is an arrayref, cope.
862 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
866 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
870 # Helper function. Make a display label for the given witnesses, showing a.c.
871 # witnesses only where the main witness is not also in the list.
872 sub _path_display_label {
875 map { $wits{$_} = 1 } @_;
877 # If an a.c. wit is listed, remove it if the main wit is also listed.
878 # Otherwise keep it for explicit listing.
879 my $aclabel = $self->ac_label;
881 foreach my $w ( sort keys %wits ) {
882 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
883 if( exists $wits{$1} ) {
886 push( @disp_ac, $w );
891 # See if we are in a majority situation.
892 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
893 $maj = $maj > 5 ? $maj : 5;
894 if( scalar keys %wits > $maj ) {
895 unshift( @disp_ac, 'majority' );
896 return join( ', ', @disp_ac );
898 return join( ', ', sort keys %wits );
902 =head2 readings_at_rank( $rank )
904 Returns a list of readings at a given rank, taken from the alignment table.
908 sub readings_at_rank {
909 my( $self, $rank ) = @_;
910 my $table = $self->alignment_table;
911 # Table rank is real rank - 1.
912 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
914 foreach my $e ( @elements ) {
915 next unless ref( $e ) eq 'HASH';
916 next unless exists $e->{'t'};
917 $readings{$e->{'t'}->id} = $e->{'t'};
919 return values %readings;
924 Returns a GraphML representation of the collation. The GraphML will contain
925 two graphs. The first expresses the attributes of the readings and the witness
926 paths that link them; the second expresses the relationships that link the
927 readings. This is the native transfer format for a tradition.
936 my $datafile = 't/data/florilegium_tei_ps.xml';
937 my $tradition = Text::Tradition->new( 'input' => 'TEI',
942 ok( $tradition, "Got a tradition object" );
943 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
944 ok( $tradition->collation, "Tradition has a collation" );
946 my $c = $tradition->collation;
947 is( scalar $c->readings, $READINGS, "Collation has all readings" );
948 is( scalar $c->paths, $PATHS, "Collation has all paths" );
949 is( scalar $c->relationships, 0, "Collation has all relationships" );
951 # Add a few relationships
952 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
953 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
954 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
956 # Now write it to GraphML and parse it again.
958 my $graphml = $c->as_graphml;
959 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
960 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
961 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
962 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
964 # Now add a stemma, write to GraphML, and look at the output.
965 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
966 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
967 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
968 $graphml = $c->as_graphml;
969 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
971 # Now add a user, write to GraphML, and look at the output.
972 unlike( $graphml, qr/testuser/, "Test user name does not exist in GraphML yet" );
973 my $testuser = Text::Tradition::User->new(
974 id => 'testuser', password => 'testpass' );
975 is( ref( $testuser ), 'Text::Tradition::User', "Created test user object" );
976 $testuser->add_tradition( $tradition );
977 is( $tradition->user->id, $testuser->id, "Tradition assigned to test user" );
978 $graphml = $c->as_graphml;
979 like( $graphml, qr/testuser/, "Test user name now 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 # Add the data keys for the graph. Include an extra key 'version' for the
1030 # GraphML output version.
1031 my %graph_data_keys;
1033 my %graph_attributes = ( 'version' => 'string' );
1034 # Graph attributes include those of Tradition and those of Collation.
1036 my $tmeta = $self->tradition->meta;
1037 my $cmeta = $self->meta;
1038 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1039 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1040 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1041 next if $attr->name =~ /^_/;
1042 next if $skipsave{$attr->name};
1043 next unless $save_types{$attr->type_constraint->name};
1044 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1046 # Extra custom keys for complex objects that should be saved in some form.
1047 # The subroutine should return a string, or undef/empty.
1048 $graph_attributes{'stemmata'} = sub {
1050 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1051 $self->tradition->stemmata;
1052 join( "\n", @stemstrs );
1054 $graph_attributes{'user'} = sub {
1055 $self->tradition->user ? $self->tradition->user->id : undef
1058 foreach my $datum ( sort keys %graph_attributes ) {
1059 $graph_data_keys{$datum} = 'dg'.$gdi++;
1060 my $key = $root->addNewChild( $graphml_ns, 'key' );
1061 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1062 : $graph_attributes{$datum};
1063 $key->setAttribute( 'attr.name', $datum );
1064 $key->setAttribute( 'attr.type', $dtype );
1065 $key->setAttribute( 'for', 'graph' );
1066 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1069 # Add the data keys for reading nodes
1070 my %reading_attributes;
1071 my $rmeta = Text::Tradition::Collation::Reading->meta;
1072 foreach my $attr( $rmeta->get_all_attributes ) {
1073 next if $attr->name =~ /^_/;
1074 next if $skipsave{$attr->name};
1075 next unless $save_types{$attr->type_constraint->name};
1076 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1078 # Extra custom key for the reading morphology
1079 $reading_attributes{'lexemes'} = 'string';
1083 foreach my $datum ( sort keys %reading_attributes ) {
1084 $node_data_keys{$datum} = 'dn'.$ndi++;
1085 my $key = $root->addNewChild( $graphml_ns, 'key' );
1086 $key->setAttribute( 'attr.name', $datum );
1087 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1088 $key->setAttribute( 'for', 'node' );
1089 $key->setAttribute( 'id', $node_data_keys{$datum} );
1092 # Add the data keys for edges, that is, paths and relationships. Path
1093 # data does not come from a Moose class so is here manually.
1096 my %edge_attributes = (
1097 witness => 'string', # ID/label for a path
1098 extra => 'boolean', # Path key
1100 my @path_attributes = keys %edge_attributes; # track our manual additions
1101 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1102 foreach my $attr( $pmeta->get_all_attributes ) {
1103 next if $attr->name =~ /^_/;
1104 next if $skipsave{$attr->name};
1105 next unless $save_types{$attr->type_constraint->name};
1106 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1108 foreach my $datum ( sort keys %edge_attributes ) {
1109 $edge_data_keys{$datum} = 'de'.$edi++;
1110 my $key = $root->addNewChild( $graphml_ns, 'key' );
1111 $key->setAttribute( 'attr.name', $datum );
1112 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1113 $key->setAttribute( 'for', 'edge' );
1114 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1117 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1118 my $xmlidname = $self->tradition->name;
1119 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1120 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1121 $xmlidname = '_'.$xmlidname;
1123 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1124 $sgraph->setAttribute( 'edgedefault', 'directed' );
1125 $sgraph->setAttribute( 'id', $xmlidname );
1126 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1127 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1128 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1129 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1130 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1132 # Tradition/collation attribute data
1133 foreach my $datum ( keys %graph_attributes ) {
1135 if( $datum eq 'version' ) {
1137 } elsif( ref( $graph_attributes{$datum} ) ) {
1138 my $sub = $graph_attributes{$datum};
1140 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1141 $value = $self->tradition->$datum;
1143 $value = $self->$datum;
1145 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1150 # Add our readings to the graph
1151 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1152 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1153 ( $n->rank < $start->rank || $n->rank > $end->rank );
1154 $use_readings{$n->id} = 1;
1155 # Add to the main graph
1156 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1157 my $node_xmlid = 'n' . $node_ctr++;
1158 $node_hash{ $n->id } = $node_xmlid;
1159 $node_el->setAttribute( 'id', $node_xmlid );
1160 foreach my $d ( keys %reading_attributes ) {
1162 # Custom serialization
1163 if( $d eq 'lexemes' ) {
1164 # If nval is a true value, we have lexemes so we need to
1165 # serialize them. Otherwise set nval to undef so that the
1166 # key is excluded from this reading.
1167 $nval = $nval ? $n->_serialize_lexemes : undef;
1168 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1171 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1172 # Adjust the ranks within the subgraph.
1173 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1174 : $nval - $rankoffset;
1176 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1181 # Add the path edges to the sequence graph
1183 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1184 # We add an edge in the graphml for every witness in $e.
1185 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1186 my @edge_wits = sort $self->path_witnesses( $e );
1187 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1188 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1189 # Skip any path from start to end; that witness is not in the subgraph.
1190 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1191 foreach my $wit ( @edge_wits ) {
1192 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1193 $node_hash{ $e->[0] },
1194 $node_hash{ $e->[1] } );
1195 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1196 $edge_el->setAttribute( 'source', $from );
1197 $edge_el->setAttribute( 'target', $to );
1198 $edge_el->setAttribute( 'id', $id );
1200 # It's a witness path, so add the witness
1202 my $key = $edge_data_keys{'witness'};
1203 # Is this an ante-corr witness?
1204 my $aclabel = $self->ac_label;
1205 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1206 # Keep the base witness
1208 # ...and record that this is an 'extra' reading path
1209 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1211 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1215 # Report the actual number of nodes and edges that went in
1216 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1217 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1219 # Add the relationship graph to the XML
1220 map { delete $edge_data_keys{$_} } @path_attributes;
1221 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1222 $node_data_keys{'id'}, \%edge_data_keys );
1224 # Save and return the thing
1225 my $result = decode_utf8( $graphml->toString(1) );
1229 sub _add_graphml_data {
1230 my( $el, $key, $value ) = @_;
1231 return unless defined $value;
1232 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1233 $data_el->setAttribute( 'key', $key );
1234 $data_el->appendText( $value );
1239 Returns a CSV alignment table representation of the collation graph, one
1240 row per witness (or witness uncorrected.)
1246 my $table = $self->alignment_table;
1247 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1249 # Make the header row
1250 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1251 push( @result, decode_utf8( $csv->string ) );
1252 # Make the rest of the rows
1253 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1254 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1255 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1256 $csv->combine( @row );
1257 push( @result, decode_utf8( $csv->string ) );
1259 return join( "\n", @result );
1262 =head2 alignment_table( $use_refs, $include_witnesses )
1264 Return a reference to an alignment table, in a slightly enhanced CollateX
1265 format which looks like this:
1267 $table = { alignment => [ { witness => "SIGIL",
1268 tokens => [ { t => "TEXT" }, ... ] },
1269 { witness => "SIG2",
1270 tokens => [ { t => "TEXT" }, ... ] },
1272 length => TEXTLEN };
1274 If $use_refs is set to 1, the reading object is returned in the table
1275 instead of READINGTEXT; if not, the text of the reading is returned.
1277 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1278 keys have a true hash value will be included.
1282 sub alignment_table {
1284 $self->calculate_ranks() unless $self->_graphcalc_done;
1285 return $self->cached_table if $self->has_cached_table;
1287 # Make sure we can do this
1288 throw( "Need a linear graph in order to make an alignment table" )
1289 unless $self->linear;
1290 $self->calculate_ranks unless $self->end->has_rank;
1292 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1293 my @all_pos = ( 1 .. $self->end->rank - 1 );
1294 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1295 # say STDERR "Making witness row(s) for " . $wit->sigil;
1296 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1297 my @row = _make_witness_row( \@wit_path, \@all_pos );
1298 push( @{$table->{'alignment'}},
1299 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1300 if( $wit->is_layered ) {
1301 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1302 $wit->sigil.$self->ac_label );
1303 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1304 push( @{$table->{'alignment'}},
1305 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1308 $self->cached_table( $table );
1312 sub _make_witness_row {
1313 my( $path, $positions ) = @_;
1315 map { $char_hash{$_} = undef } @$positions;
1317 foreach my $rdg ( @$path ) {
1318 my $rtext = $rdg->text;
1319 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1320 say STDERR "rank " . $rdg->rank if $debug;
1321 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1322 $char_hash{$rdg->rank} = { 't' => $rdg };
1324 my @row = map { $char_hash{$_} } @$positions;
1325 # Fill in lacuna markers for undef spots in the row
1326 my $last_el = shift @row;
1327 my @filled_row = ( $last_el );
1328 foreach my $el ( @row ) {
1329 # If we are using node reference, make the lacuna node appear many times
1330 # in the table. If not, use the lacuna tag.
1331 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1334 push( @filled_row, $el );
1340 =head1 NAVIGATION METHODS
1342 =head2 reading_sequence( $first, $last, $sigil, $backup )
1344 Returns the ordered list of readings, starting with $first and ending
1345 with $last, for the witness given in $sigil. If a $backup sigil is
1346 specified (e.g. when walking a layered witness), it will be used wherever
1347 no $sigil path exists. If there is a base text reading, that will be
1348 used wherever no path exists for $sigil or $backup.
1352 # TODO Think about returning some lazy-eval iterator.
1353 # TODO Get rid of backup; we should know from what witness is whether we need it.
1355 sub reading_sequence {
1356 my( $self, $start, $end, $witness ) = @_;
1358 $witness = $self->baselabel unless $witness;
1359 my @readings = ( $start );
1362 while( $n && $n->id ne $end->id ) {
1363 if( exists( $seen{$n->id} ) ) {
1364 throw( "Detected loop for $witness at " . $n->id );
1368 my $next = $self->next_reading( $n, $witness );
1370 throw( "Did not find any path for $witness from reading " . $n->id );
1372 push( @readings, $next );
1375 # Check that the last reading is our end reading.
1376 my $last = $readings[$#readings];
1377 throw( "Last reading found from " . $start->text .
1378 " for witness $witness is not the end!" ) # TODO do we get this far?
1379 unless $last->id eq $end->id;
1384 =head2 next_reading( $reading, $sigil );
1386 Returns the reading that follows the given reading along the given witness
1392 # Return the successor via the corresponding path.
1394 my $answer = $self->_find_linked_reading( 'next', @_ );
1395 return undef unless $answer;
1396 return $self->reading( $answer );
1399 =head2 prior_reading( $reading, $sigil )
1401 Returns the reading that precedes the given reading along the given witness
1407 # Return the predecessor via the corresponding path.
1409 my $answer = $self->_find_linked_reading( 'prior', @_ );
1410 return $self->reading( $answer );
1413 sub _find_linked_reading {
1414 my( $self, $direction, $node, $path ) = @_;
1416 # Get a backup if we are dealing with a layered witness
1418 my $aclabel = $self->ac_label;
1419 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1423 my @linked_paths = $direction eq 'next'
1424 ? $self->sequence->edges_from( $node )
1425 : $self->sequence->edges_to( $node );
1426 return undef unless scalar( @linked_paths );
1428 # We have to find the linked path that contains all of the
1429 # witnesses supplied in $path.
1430 my( @path_wits, @alt_path_wits );
1431 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1432 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1435 foreach my $le ( @linked_paths ) {
1436 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1439 my @le_wits = sort $self->path_witnesses( $le );
1440 if( _is_within( \@path_wits, \@le_wits ) ) {
1441 # This is the right path.
1442 return $direction eq 'next' ? $le->[1] : $le->[0];
1443 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1447 # Got this far? Return the alternate path if it exists.
1448 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1451 # Got this far? Return the base path if it exists.
1452 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1455 # Got this far? We have no appropriate path.
1456 warn "Could not find $direction node from " . $node->id
1457 . " along path $path";
1463 my( $set1, $set2 ) = @_;
1464 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1465 foreach my $el ( @$set1 ) {
1466 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1471 # Return the string that joins together a list of witnesses for
1472 # display on a single path.
1473 sub _witnesses_of_label {
1474 my( $self, $label ) = @_;
1475 my $regex = $self->wit_list_separator;
1476 my @answer = split( /\Q$regex\E/, $label );
1480 =head2 common_readings
1482 Returns the list of common readings in the graph (i.e. those readings that are
1483 shared by all non-lacunose witnesses.)
1487 sub common_readings {
1489 my @common = grep { $_->is_common } $self->readings;
1493 =head2 path_text( $sigil, [, $start, $end ] )
1495 Returns the text of a witness (plus its backup, if we are using a layer)
1496 as stored in the collation. The text is returned as a string, where the
1497 individual readings are joined with spaces and the meta-readings (e.g.
1498 lacunae) are omitted. Optional specification of $start and $end allows
1499 the generation of a subset of the witness text.
1504 my( $self, $wit, $start, $end ) = @_;
1505 $start = $self->start unless $start;
1506 $end = $self->end unless $end;
1507 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1510 foreach my $r ( @path ) {
1511 unless ( $r->join_prior || !$last || $last->join_next ) {
1514 $pathtext .= $r->text;
1520 =head1 INITIALIZATION METHODS
1522 These are mostly for use by parsers.
1524 =head2 make_witness_path( $witness )
1526 Link the array of readings contained in $witness->path (and in
1527 $witness->uncorrected_path if it exists) into collation paths.
1528 Clear out the arrays when finished.
1530 =head2 make_witness_paths
1532 Call make_witness_path for all witnesses in the tradition.
1536 # For use when a collation is constructed from a base text and an apparatus.
1537 # We have the sequences of readings and just need to add path edges.
1538 # When we are done, clear out the witness path attributes, as they are no
1540 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1542 sub make_witness_paths {
1544 foreach my $wit ( $self->tradition->witnesses ) {
1545 # say STDERR "Making path for " . $wit->sigil;
1546 $self->make_witness_path( $wit );
1550 sub make_witness_path {
1551 my( $self, $wit ) = @_;
1552 my @chain = @{$wit->path};
1553 my $sig = $wit->sigil;
1554 # Add start and end if necessary
1555 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1556 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1557 foreach my $idx ( 0 .. $#chain-1 ) {
1558 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1560 if( $wit->is_layered ) {
1561 @chain = @{$wit->uncorrected_path};
1562 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1563 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1564 foreach my $idx( 0 .. $#chain-1 ) {
1565 my $source = $chain[$idx];
1566 my $target = $chain[$idx+1];
1567 $self->add_path( $source, $target, $sig.$self->ac_label )
1568 unless $self->has_path( $source, $target, $sig );
1572 $wit->clear_uncorrected_path;
1575 =head2 calculate_ranks
1577 Calculate the reading ranks (that is, their aligned positions relative
1578 to each other) for the graph. This can only be called on linear collations.
1582 use Text::Tradition;
1584 my $cxfile = 't/data/Collatex-16.xml';
1585 my $t = Text::Tradition->new(
1587 'input' => 'CollateX',
1590 my $c = $t->collation;
1593 my $table = $c->alignment_table;
1594 ok( $c->has_cached_table, "Alignment table was cached" );
1595 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1596 $c->calculate_ranks;
1597 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1598 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1599 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1605 sub calculate_ranks {
1607 # Save the existing ranks, in case we need to invalidate the cached SVG.
1609 map { $existing_ranks{$_} = $_->rank } $self->readings;
1611 # Do the rankings based on the relationship equivalence graph, starting
1612 # with the start node.
1613 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1615 # Transfer our rankings from the topological graph to the real one.
1616 foreach my $r ( $self->readings ) {
1617 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1618 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1620 # Die. Find the last rank we calculated.
1621 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1622 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1624 my $last = pop @all_defined;
1625 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1628 # Do we need to invalidate the cached data?
1629 if( $self->has_cached_table ) {
1630 foreach my $r ( $self->readings ) {
1631 next if defined( $existing_ranks{$r} )
1632 && $existing_ranks{$r} == $r->rank;
1633 # Something has changed, so clear the cache
1634 $self->_clear_cache;
1635 # ...and recalculate the common readings.
1636 $self->calculate_common_readings();
1640 # The graph calculation information is now up to date.
1641 $self->_graphcalc_done(1);
1646 $self->wipe_table if $self->has_cached_table;
1650 =head2 flatten_ranks
1652 A convenience method for parsing collation data. Searches the graph for readings
1653 with the same text at the same rank, and merges any that are found.
1659 my %unique_rank_rdg;
1661 foreach my $rdg ( $self->readings ) {
1662 next unless $rdg->has_rank;
1663 my $key = $rdg->rank . "||" . $rdg->text;
1664 if( exists $unique_rank_rdg{$key} ) {
1665 # Make sure they don't have different grammatical forms
1666 my $ur = $unique_rank_rdg{$key};
1667 if( $rdg->disambiguated && $ur->disambiguated ) {
1668 my $rform = join( '//', map { $_->form->to_string } $rdg->lexemes );
1669 my $uform = join( '//', map { $_->form->to_string } $ur->lexemes );
1670 next unless $rform eq $uform;
1671 } elsif( $rdg->disambiguated xor $ur->disambiguated ) {
1675 #say STDERR "Combining readings at same rank: $key";
1677 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1678 # TODO see if this now makes a common point.
1680 $unique_rank_rdg{$key} = $rdg;
1683 # If we merged readings, the ranks are still fine but the alignment
1684 # table is wrong. Wipe it.
1685 $self->wipe_table() if $changed;
1689 =head2 calculate_common_readings
1691 Goes through the graph identifying the readings that appear in every witness
1692 (apart from those with lacunae at that spot.) Marks them as common and returns
1697 use Text::Tradition;
1699 my $cxfile = 't/data/Collatex-16.xml';
1700 my $t = Text::Tradition->new(
1702 'input' => 'CollateX',
1705 my $c = $t->collation;
1707 my @common = $c->calculate_common_readings();
1708 is( scalar @common, 8, "Found correct number of common readings" );
1709 my @marked = sort $c->common_readings();
1710 is( scalar @common, 8, "All common readings got marked as such" );
1711 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1712 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1718 sub calculate_common_readings {
1721 map { $_->is_common( 0 ) } $self->readings;
1722 # Implicitly calls calculate_ranks
1723 my $table = $self->alignment_table;
1724 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1725 my @row = map { $_->{'tokens'}->[$idx]
1726 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1727 @{$table->{'alignment'}};
1729 foreach my $r ( @row ) {
1731 $hash{$r->id} = $r unless $r->is_meta;
1733 $hash{'UNDEF'} = $r;
1736 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1737 my( $r ) = values %hash;
1739 push( @common, $r );
1745 =head2 text_from_paths
1747 Calculate the text array for all witnesses from the path, for later consistency
1748 checking. Only to be used if there is no non-graph-based way to know the
1753 sub text_from_paths {
1755 foreach my $wit ( $self->tradition->witnesses ) {
1756 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1758 foreach my $r ( @readings ) {
1759 next if $r->is_meta;
1760 push( @text, $r->text );
1762 $wit->text( \@text );
1763 if( $wit->is_layered ) {
1764 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1765 $wit->sigil.$self->ac_label );
1767 foreach my $r ( @ucrdgs ) {
1768 next if $r->is_meta;
1769 push( @uctext, $r->text );
1771 $wit->layertext( \@uctext );
1776 =head1 UTILITY FUNCTIONS
1778 =head2 common_predecessor( $reading_a, $reading_b )
1780 Find the last reading that occurs in sequence before both the given readings.
1781 At the very least this should be $self->start.
1783 =head2 common_successor( $reading_a, $reading_b )
1785 Find the first reading that occurs in sequence after both the given readings.
1786 At the very least this should be $self->end.
1790 use Text::Tradition;
1792 my $cxfile = 't/data/Collatex-16.xml';
1793 my $t = Text::Tradition->new(
1795 'input' => 'CollateX',
1798 my $c = $t->collation;
1800 is( $c->common_predecessor( 'n24', 'n23' )->id,
1801 'n20', "Found correct common predecessor" );
1802 is( $c->common_successor( 'n24', 'n23' )->id,
1803 '__END__', "Found correct common successor" );
1805 is( $c->common_predecessor( 'n19', 'n17' )->id,
1806 'n16', "Found correct common predecessor for readings on same path" );
1807 is( $c->common_successor( 'n21', 'n10' )->id,
1808 '__END__', "Found correct common successor for readings on same path" );
1814 ## Return the closest reading that is a predecessor of both the given readings.
1815 sub common_predecessor {
1817 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1818 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1821 sub common_successor {
1823 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1824 return $self->_common_in_path( $r1, $r2, 'successors' );
1828 # TODO think about how to do this without ranks...
1829 sub _common_in_path {
1830 my( $self, $r1, $r2, $dir ) = @_;
1831 my $iter = $self->end->rank;
1833 my @last_r1 = ( $r1 );
1834 my @last_r2 = ( $r2 );
1835 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1837 # say STDERR "Finding common $dir for $r1, $r2";
1838 while( !@candidates ) {
1839 last unless $iter--; # Avoid looping infinitely
1840 # Iterate separately down the graph from r1 and r2
1841 my( @new_lc1, @new_lc2 );
1842 foreach my $lc ( @last_r1 ) {
1843 foreach my $p ( $lc->$dir ) {
1844 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1845 # say STDERR "Path candidate $p from $lc";
1846 push( @candidates, $p );
1847 } elsif( !$all_seen{$p->id} ) {
1848 $all_seen{$p->id} = 'r1';
1849 push( @new_lc1, $p );
1853 foreach my $lc ( @last_r2 ) {
1854 foreach my $p ( $lc->$dir ) {
1855 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1856 # say STDERR "Path candidate $p from $lc";
1857 push( @candidates, $p );
1858 } elsif( !$all_seen{$p->id} ) {
1859 $all_seen{$p->id} = 'r2';
1860 push( @new_lc2, $p );
1864 @last_r1 = @new_lc1;
1865 @last_r2 = @new_lc2;
1867 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1868 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1872 Text::Tradition::Error->throw(
1873 'ident' => 'Collation error',
1879 __PACKAGE__->meta->make_immutable;
1883 This package is free software and is provided "as is" without express
1884 or implied warranty. You can redistribute it and/or modify it under
1885 the same terms as Perl itself.
1889 Tara L Andrews E<lt>aurum@cpan.orgE<gt>