1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Data;
11 use Text::Tradition::Collation::Reading;
12 use Text::Tradition::Collation::RelationshipStore;
13 use Text::Tradition::Error;
14 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
16 use XML::LibXML::XPathContext;
20 isa => 'Text::Tradition::Collation::Data',
58 isa => 'Text::Tradition',
59 writer => '_set_tradition',
65 Text::Tradition::Collation - a software model for a text collation
70 my $t = Text::Tradition->new(
71 'name' => 'this is a text',
73 'file' => '/path/to/tei_parallel_seg_file.xml' );
75 my $c = $t->collation;
76 my @readings = $c->readings;
77 my @paths = $c->paths;
78 my @relationships = $c->relationships;
80 my $svg_variant_graph = $t->collation->as_svg();
84 Text::Tradition is a library for representation and analysis of collated
85 texts, particularly medieval ones. The Collation is the central feature of
86 a Tradition, where the text, its sequence of readings, and its relationships
87 between readings are actually kept.
93 The constructor. Takes a hash or hashref of the following arguments:
97 =item * tradition - The Text::Tradition object to which the collation
100 =item * linear - Whether the collation should be linear; that is, whether
101 transposed readings should be treated as two linked readings rather than one,
102 and therefore whether the collation graph is acyclic. Defaults to true.
104 =item * baselabel - The default label for the path taken by a base text
105 (if any). Defaults to 'base text'.
107 =item * wit_list_separator - The string to join a list of witnesses for
108 purposes of making labels in display graphs. Defaults to ', '.
110 =item * ac_label - The extra label to tack onto a witness sigil when
111 representing another layer of path for the given witness - that is, when
112 a text has more than one possible reading due to scribal corrections or
113 the like. Defaults to ' (a.c.)'.
115 =item * wordsep - The string used to separate words in the original text.
126 =head2 wit_list_separator
134 Simple accessors for collation attributes.
138 The meta-reading at the start of every witness path.
142 The meta-reading at the end of every witness path.
146 Returns all Reading objects in the graph.
148 =head2 reading( $id )
150 Returns the Reading object corresponding to the given ID.
152 =head2 add_reading( $reading_args )
154 Adds a new reading object to the collation.
155 See L<Text::Tradition::Collation::Reading> for the available arguments.
157 =head2 del_reading( $object_or_id )
159 Removes the given reading from the collation, implicitly removing its
160 paths and relationships.
162 =head2 merge_readings( $main, $second, $concatenate, $with_str )
164 Merges the $second reading into the $main one. If $concatenate is true, then
165 the merged node will carry the text of both readings, concatenated with either
166 $with_str (if specified) or a sensible default (the empty string if the
167 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
169 The first two arguments may be either readings or reading IDs.
171 =head2 has_reading( $id )
173 Predicate to see whether a given reading ID is in the graph.
175 =head2 reading_witnesses( $object_or_id )
177 Returns a list of sigils whose witnesses contain the reading.
181 Returns all reading paths within the document - that is, all edges in the
182 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
184 =head2 add_path( $source, $target, $sigil )
186 Links the given readings in the collation in sequence, under the given witness
187 sigil. The readings may be specified by object or ID.
189 =head2 del_path( $source, $target, $sigil )
191 Links the given readings in the collation in sequence, under the given witness
192 sigil. The readings may be specified by object or ID.
194 =head2 has_path( $source, $target );
196 Returns true if the two readings are linked in sequence in any witness.
197 The readings may be specified by object or ID.
201 Returns all Relationship objects in the collation.
203 =head2 add_relationship( $reading, $other_reading, $options )
205 Adds a new relationship of the type given in $options between the two readings,
206 which may be specified by object or ID. Returns a value of ( $status, @vectors)
207 where $status is true on success, and @vectors is a list of relationship edges
208 that were ultimately added.
209 See L<Text::Tradition::Collation::Relationship> for the available options.
214 my ( $class, @args ) = @_;
215 my %args = @args == 1 ? %{ $args[0] } : @args;
216 # TODO determine these from the Moose::Meta object
217 my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel
218 linear wordsep start end cached_table _graphcalc_done);
220 for my $attr (@delegate_attrs) {
221 $data_args{$attr} = delete $args{$attr} if exists $args{$attr};
223 $args{_data} = Text::Tradition::Collation::Data->new(%data_args);
229 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
230 $self->_set_start( $self->add_reading(
231 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
232 $self->_set_end( $self->add_reading(
233 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
236 sub register_relationship_type {
238 my %args = @_ == 1 ? %{$_[0]} : @_;
239 if( $self->relations->has_type( $args{name} ) ) {
240 throw( 'Relationship type ' . $args{name} . ' already registered' );
242 $self->relations->add_type( %args );
245 ### Reading construct/destruct functions
248 my( $self, $reading ) = @_;
249 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
250 my %args = %$reading;
251 if( $args{'init'} ) {
252 # If we are initializing an empty collation, don't assume that we
253 # have set a tradition.
254 delete $args{'init'};
255 } elsif( $self->tradition->can('language') && $self->tradition->has_language
256 && !exists $args{'language'} ) {
257 $args{'language'} = $self->tradition->language;
259 $reading = Text::Tradition::Collation::Reading->new(
260 'collation' => $self,
263 # First check to see if a reading with this ID exists.
264 if( $self->reading( $reading->id ) ) {
265 throw( "Collation already has a reading with id " . $reading->id );
267 $self->_graphcalc_done(0);
268 $self->_add_reading( $reading->id => $reading );
269 # Once the reading has been added, put it in both graphs.
270 $self->sequence->add_vertex( $reading->id );
271 $self->relations->add_reading( $reading->id );
275 around del_reading => sub {
280 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
283 # Remove the reading from the graphs.
284 $self->_graphcalc_done(0);
285 $self->_clear_cache; # Explicitly clear caches to GC the reading
286 $self->sequence->delete_vertex( $arg );
287 $self->relations->delete_reading( $arg );
290 $self->$orig( $arg );
297 my $cxfile = 't/data/Collatex-16.xml';
298 my $t = Text::Tradition->new(
300 'input' => 'CollateX',
303 my $c = $t->collation;
305 my $rno = scalar $c->readings;
306 # Split n21 for testing purposes
307 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
308 my $old_r = $c->reading( 'n21' );
309 $old_r->alter_text( 'to' );
310 $c->del_path( 'n20', 'n21', 'A' );
311 $c->add_path( 'n20', 'n21p0', 'A' );
312 $c->add_path( 'n21p0', 'n21', 'A' );
314 ok( $c->reading( 'n21p0' ), "New reading exists" );
315 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
317 # Combine n3 and n4 ( with his )
318 $c->merge_readings( 'n3', 'n4', 1 );
319 ok( !$c->reading('n4'), "Reading n4 is gone" );
320 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
322 # Collapse n9 and n10 ( rood / root )
323 $c->merge_readings( 'n9', 'n10' );
324 ok( !$c->reading('n10'), "Reading n10 is gone" );
325 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
327 # Combine n21 and n21p0
328 my $remaining = $c->reading('n21');
329 $remaining ||= $c->reading('n22'); # one of these should still exist
330 $c->merge_readings( 'n21p0', $remaining, 1 );
331 ok( !$c->reading('n21'), "Reading $remaining is gone" );
332 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
342 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
343 my $mergemeta = $kept_obj->is_meta;
344 throw( "Cannot merge meta and non-meta reading" )
345 unless ( $mergemeta && $del_obj->is_meta )
346 || ( !$mergemeta && !$del_obj->is_meta );
348 throw( "Cannot merge with start or end node" )
349 if( $kept_obj eq $self->start || $kept_obj eq $self->end
350 || $del_obj eq $self->start || $del_obj eq $self->end );
351 throw( "Cannot combine text of meta readings" ) if $combine;
353 # We only need the IDs for adding paths to the graph, not the reading
354 # objects themselves.
355 my $kept = $kept_obj->id;
356 my $deleted = $del_obj->id;
357 $self->_graphcalc_done(0);
359 # The kept reading should inherit the paths and the relationships
360 # of the deleted reading.
361 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
362 my @vector = ( $kept );
363 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
364 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
365 next if $vector[0] eq $vector[1]; # Don't add a self loop
366 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
367 $self->sequence->add_edge( @vector );
368 my $fwits = $self->sequence->get_edge_attributes( @vector );
369 @wits{keys %$fwits} = values %$fwits;
370 $self->sequence->set_edge_attributes( @vector, \%wits );
372 $self->relations->merge_readings( $kept, $deleted, $combine );
374 # Do the deletion deed.
376 # Combine the text of the readings
377 my $joinstr = $combine_char;
378 unless( defined $joinstr ) {
379 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
380 $joinstr = $self->wordsep unless defined $joinstr;
382 $kept_obj->_combine( $del_obj, $joinstr );
384 $self->del_reading( $deleted );
387 =head2 compress_readings
389 Where possible in the graph, compresses plain sequences of readings into a
390 single reading. The sequences must consist of readings with no
391 relationships to other readings, with only a single witness path between
392 them and no other witness paths from either that would skip the other. The
393 readings must also not be marked as nonsense or bad grammar.
395 WARNING: This operation cannot be undone.
399 sub compress_readings {
401 # Anywhere in the graph that there is a reading that joins only to a single
402 # successor, and neither of these have any relationships, just join the two
404 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
405 # Now look for readings that can be joined to their successors.
406 next unless $rdg->is_combinable;
408 while( $self->sequence->successors( $rdg ) == 1 ) {
409 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
410 throw( "Infinite loop" ) if $seen{$next->id};
411 $seen{$next->id} = 1;
412 last if $self->sequence->predecessors( $next ) > 1;
413 last unless $next->is_combinable;
414 say "Joining readings $rdg and $next";
415 $self->merge_readings( $rdg, $next, 1 );
418 # Make sure we haven't screwed anything up
419 foreach my $wit ( $self->tradition->witnesses ) {
420 my $pathtext = $self->path_text( $wit->sigil );
421 my $origtext = join( ' ', @{$wit->text} );
422 throw( "Text differs for witness " . $wit->sigil )
423 unless $pathtext eq $origtext;
424 if( $wit->is_layered ) {
425 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
426 $origtext = join( ' ', @{$wit->layertext} );
427 throw( "Ante-corr text differs for witness " . $wit->sigil )
428 unless $pathtext eq $origtext;
432 $self->relations->rebuild_equivalence();
433 $self->calculate_ranks();
436 # Helper function for manipulating the graph.
437 sub _stringify_args {
438 my( $self, $first, $second, @args ) = @_;
440 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
441 $second = $second->id
442 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
443 return( $first, $second, @args );
446 # Helper function for manipulating the graph.
447 sub _objectify_args {
448 my( $self, $first, $second, $arg ) = @_;
449 $first = $self->reading( $first )
450 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
451 $second = $self->reading( $second )
452 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
453 return( $first, $second, $arg );
456 =head2 duplicate_reading( $reading, @witlist )
458 Split the given reading into two, so that the new reading is in the path for
459 the witnesses given in @witlist.
465 my $st = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/collatecorr.xml' );
466 is( ref( $st ), 'Text::Tradition', "Got a tradition from test file" );
467 ok( $st->has_witness('Ba96'), "Tradition has the affected witness" );
469 my $sc = $st->collation;
471 ok( $sc->reading('n131'), "Tradition has the affected reading" );
472 is( scalar( $sc->readings ), $numr, "There are $numr readings in the graph" );
473 is( $sc->end->rank, 14, "There are fourteen ranks in the graph" );
475 # Detach the erroneously collated reading
476 $sc->duplicate_reading( 'n131', 'Ba96' );
477 ok( $sc->reading('n131_0'), "Detached the bad collation with a new reading" );
478 is( scalar( $sc->readings ), $numr + 1, "A reading was added to the graph" );
479 is( $sc->end->rank, 10, "There are now only ten ranks in the graph" );
481 # Check that the bad transposition is gone
482 is( $sc->get_relationship( 'n130', 'n135' ), undef, "Bad transposition relationship is gone" );
485 ok( $sc->add_relationship( 'n124', 'n131_0', { type => 'collated', scope => 'local' } ),
486 "Collated the readings correctly" );
487 $sc->calculate_ranks();
488 $sc->flatten_ranks();
489 is( $sc->end->rank, 11, "The ranks shifted appropriately" );
490 is( scalar( $sc->readings ), $numr - 3, "Now we are collated correctly" );
496 sub duplicate_reading {
497 my( $self, $r, @wits ) = @_;
498 # Add the new reading, duplicating $r.
499 unless( ref( $r ) eq 'Text::Tradition::Collation::Reading' ) {
500 $r = $self->reading( $r );
502 throw( "Cannot duplicate a meta-reading" )
505 # Get all the reading attributes and duplicate them.
506 my $rmeta = Text::Tradition::Collation::Reading->meta;
508 foreach my $attr( $rmeta->get_all_attributes ) {
509 next if $attr->name =~ /^_/;
510 my $acc = $attr->get_read_method;
511 if( !$acc && $attr->has_applied_traits ) {
512 my $tr = $attr->applied_traits;
513 if( $tr->[0] =~ /::(Array|Hash)$/ ) {
515 my %methods = reverse %{$attr->handles};
516 $acc = $methods{elements};
517 $args{$attr->name} = $which eq 'Array'
518 ? [ $r->$acc ] : { $r->$acc };
521 $args{$attr->name} = $r->$acc if $acc;
524 # By definition the new reading will no longer be common.
525 $args{is_common} = 0;
526 # The new reading also needs its own ID.
527 $args{id} = $self->_generate_dup_id( $r->id );
529 # Try to make the new reading.
530 my $newr = $self->add_reading( \%args );
531 # The old reading is also no longer common.
534 # For each of the witnesses, dissociate from the old reading and
535 # associate with the new.
536 foreach my $wit ( @wits ) {
537 my $prior = $self->prior_reading( $r, $wit );
538 my $next = $self->next_reading( $r, $wit );
539 $self->del_path( $prior, $r, $wit );
540 $self->add_path( $prior, $newr, $wit );
541 $self->del_path( $r, $next, $wit );
542 $self->add_path( $newr, $next, $wit );
545 # Hash the reading ranks and find the closest common successor to our
549 if( $self->end->has_rank ) {
550 $succ = $self->common_successor( $r, $newr );
551 foreach my $rdg ( $self->readings ) {
552 $rrk{$rdg->id} = $rdg->rank;
556 # Rebuild the equivalence graph and calculate the new ranks
557 $self->relations->rebuild_equivalence();
558 $self->calculate_ranks();
560 # Check for invalid non-colocated relationships among changed-rank readings
561 # from where the ranks start changing up to $succ
562 if( $self->end->has_rank ) {
563 my $lastrank = $succ->rank;
564 foreach my $rdg ( $self->readings ) {
565 next if $rdg->rank > $lastrank;
566 next if $rdg->rank == $rrk{$rdg->id};
567 my @noncolo = $rdg->related_readings( sub { !$_[0]->colocated } );
568 next unless @noncolo;
569 foreach my $nc ( @noncolo ) {
570 $self->relations->verify_or_delete( $rdg, $nc );
576 sub _generate_dup_id {
577 my( $self, $rid ) = @_;
582 if( $self->has_reading( $newid ) ) {
595 # We only need the IDs for adding paths to the graph, not the reading
596 # objects themselves.
597 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
599 $self->_graphcalc_done(0);
600 # Connect the readings
601 unless( $self->sequence->has_edge( $source, $target ) ) {
602 $self->sequence->add_edge( $source, $target );
603 $self->relations->add_equivalence_edge( $source, $target );
605 # Note the witness in question
606 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
612 if( ref( $_[0] ) eq 'ARRAY' ) {
619 # We only need the IDs for removing paths from the graph, not the reading
620 # objects themselves.
621 my( $source, $target, $wit ) = $self->_stringify_args( @args );
623 $self->_graphcalc_done(0);
624 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
625 $self->sequence->delete_edge_attribute( $source, $target, $wit );
627 unless( $self->sequence->has_edge_attributes( $source, $target ) ) {
628 $self->sequence->delete_edge( $source, $target );
629 $self->relations->delete_equivalence_edge( $source, $target );
634 # Extra graph-alike utility
637 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
638 return undef unless $self->sequence->has_edge( $source, $target );
639 return $self->sequence->has_edge_attribute( $source, $target, $wit );
642 =head2 clear_witness( @sigil_list )
644 Clear the given witnesses out of the collation entirely, removing references
645 to them in paths, and removing readings that belong only to them. Should only
646 be called via $tradition->del_witness.
651 my( $self, @sigils ) = @_;
653 $self->_graphcalc_done(0);
654 # Clear the witness(es) out of the paths
655 foreach my $e ( $self->paths ) {
656 foreach my $sig ( @sigils ) {
657 $self->del_path( $e, $sig );
661 # Clear out the newly unused readings
662 foreach my $r ( $self->readings ) {
663 unless( $self->reading_witnesses( $r ) ) {
664 $self->del_reading( $r );
669 sub add_relationship {
671 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
672 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
673 foreach my $v ( @vectors ) {
674 next unless $self->get_relationship( $v )->colocated;
675 if( $self->reading( $v->[0] )->has_rank && $self->reading( $v->[1] )->has_rank
676 && $self->reading( $v->[0] )->rank ne $self->reading( $v->[1] )->rank ) {
677 $self->_graphcalc_done(0);
685 around qw/ get_relationship del_relationship / => sub {
689 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
692 my( $source, $target ) = $self->_stringify_args( @args );
693 $self->$orig( $source, $target );
696 =head2 reading_witnesses( $reading )
698 Return a list of sigils corresponding to the witnesses in which the reading appears.
702 sub reading_witnesses {
703 my( $self, $reading ) = @_;
704 # We need only check either the incoming or the outgoing edges; I have
705 # arbitrarily chosen "incoming". Thus, special-case the start node.
706 if( $reading eq $self->start ) {
707 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
710 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
711 my $wits = $self->sequence->get_edge_attributes( @$e );
712 @all_witnesses{ keys %$wits } = 1;
714 my $acstr = $self->ac_label;
715 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
716 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
718 return keys %all_witnesses;
721 =head1 OUTPUT METHODS
723 =head2 as_svg( \%options )
725 Returns an SVG string that represents the graph, via as_dot and graphviz.
726 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
731 my( $self, $opts ) = @_;
732 throw( "Need GraphViz installed to output SVG" )
733 unless File::Which::which( 'dot' );
734 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
735 $self->calculate_ranks()
736 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
737 my @cmd = qw/dot -Tsvg/;
739 my $dotfile = File::Temp->new();
741 # $dotfile->unlink_on_destroy(0);
742 binmode $dotfile, ':utf8';
743 print $dotfile $self->as_dot( $opts );
744 push( @cmd, $dotfile->filename );
745 run( \@cmd, ">", binary(), \$svg );
746 $svg = decode_utf8( $svg );
751 =head2 as_dot( \%options )
753 Returns a string that is the collation graph expressed in dot
754 (i.e. GraphViz) format. Options include:
769 my( $self, $opts ) = @_;
770 my $startrank = $opts->{'from'} if $opts;
771 my $endrank = $opts->{'to'} if $opts;
772 my $color_common = $opts->{'color_common'} if $opts;
773 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
774 && $self->end->rank > 100;
775 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
777 # Check the arguments
779 return if $endrank && $startrank > $endrank;
780 return if $startrank > $self->end->rank;
782 if( defined $endrank ) {
783 return if $endrank < 0;
784 $endrank = undef if $endrank == $self->end->rank;
787 my $graph_name = $self->tradition->name;
788 $graph_name =~ s/[^\w\s]//g;
789 $graph_name = join( '_', split( /\s+/, $graph_name ) );
797 'fillcolor' => 'white',
802 'arrowhead' => 'open',
803 'color' => '#000000',
804 'fontcolor' => '#000000',
807 my $dot = sprintf( "digraph %s {\n", $graph_name );
808 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
809 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
811 # Output substitute start/end readings if necessary
813 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
816 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
818 if( $STRAIGHTENHACK ) {
820 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
821 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
822 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
824 my %used; # Keep track of the readings that actually appear in the graph
825 # Sort the readings by rank if we have ranks; this speeds layout.
826 my @all_readings = $self->end->has_rank
827 ? sort { $a->rank <=> $b->rank } $self->readings
829 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
830 foreach my $reading ( @all_readings ) {
831 # Only output readings within our rank range.
832 next if $startrank && $reading->rank < $startrank;
833 next if $endrank && $reading->rank > $endrank;
834 $used{$reading->id} = 1;
835 # Need not output nodes without separate labels
836 next if $reading->id eq $reading->text;
838 my $label = $reading->text;
839 $label .= '-' if $reading->join_next;
840 $label = "-$label" if $reading->join_prior;
841 $label =~ s/\"/\\\"/g;
842 $rattrs->{'label'} = $label;
843 $rattrs->{'id'} = $reading->id;
844 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
845 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
848 # Add the real edges. Need to weight one edge per rank jump, in a
850 # my $weighted = $self->_add_edge_weights;
851 my @edges = $self->paths;
852 my( %substart, %subend );
853 foreach my $edge ( @edges ) {
854 # Do we need to output this edge?
855 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
856 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
857 my $variables = { %edge_attrs, 'label' => $label };
859 # Account for the rank gap if necessary
860 my $rank0 = $self->reading( $edge->[0] )->rank
861 if $self->reading( $edge->[0] )->has_rank;
862 my $rank1 = $self->reading( $edge->[1] )->rank
863 if $self->reading( $edge->[1] )->has_rank;
864 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
865 $variables->{'minlen'} = $rank1 - $rank0;
868 # Add the calculated edge weights
869 # if( exists $weighted->{$edge->[0]}
870 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
871 # # $variables->{'color'} = 'red';
872 # $variables->{'weight'} = 3.0;
875 # EXPERIMENTAL: make edge width reflect no. of witnesses
876 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
877 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
879 my $varopts = _dot_attr_string( $variables );
880 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
881 $edge->[0], $edge->[1], $varopts );
882 } elsif( $used{$edge->[0]} ) {
883 $subend{$edge->[0]} = $edge->[1];
884 } elsif( $used{$edge->[1]} ) {
885 $substart{$edge->[1]} = $edge->[0];
889 # If we are asked to, add relationship links
890 if( exists $opts->{show_relations} ) {
891 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
892 if( $filter eq 'transposition' ) {
893 $filter =~ qr/^transposition$/;
895 foreach my $redge ( $self->relationships ) {
896 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
897 if( $filter ne 'all' ) {
898 my $rel = $self->get_relationship( $redge );
899 next unless $rel->type =~ /$filter/;
903 constraint => 'false',
904 label => uc( substr( $rel->type, 0, 4 ) ),
907 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
908 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
914 # Add substitute start and end edges if necessary
915 foreach my $node ( keys %substart ) {
916 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
917 my $variables = { %edge_attrs, 'label' => $witstr };
918 my $nrdg = $self->reading( $node );
919 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
920 # Substart is actually one lower than $startrank
921 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
923 my $varopts = _dot_attr_string( $variables );
924 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
926 foreach my $node ( keys %subend ) {
927 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
928 my $variables = { %edge_attrs, 'label' => $witstr };
929 my $varopts = _dot_attr_string( $variables );
930 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
933 if( $STRAIGHTENHACK ) {
934 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
935 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
942 sub _dot_attr_string {
945 foreach my $k ( sort keys %$hash ) {
947 push( @attrs, $k.'="'.$v.'"' );
949 return( '[ ' . join( ', ', @attrs ) . ' ]' );
952 sub _add_edge_weights {
954 # Walk the graph from START to END, choosing the successor node with
955 # the largest number of witness paths each time.
957 my $curr = $self->start->id;
958 my $ranked = $self->end->has_rank;
959 while( $curr ne $self->end->id ) {
960 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
961 my @succ = sort { $self->path_witnesses( $curr, $a )
962 <=> $self->path_witnesses( $curr, $b ) }
963 $self->sequence->successors( $curr );
964 my $next = pop @succ;
965 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
966 # Try to avoid lacunae in the weighted path.
968 ( $self->reading( $next )->is_lacuna ||
969 $nextrank - $rank > 1 ) ){
972 $weighted->{$curr} = $next;
978 =head2 path_witnesses( $edge )
980 Returns the list of sigils whose witnesses are associated with the given edge.
981 The edge can be passed as either an array or an arrayref of ( $source, $target ).
986 my( $self, @edge ) = @_;
987 # If edge is an arrayref, cope.
988 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
992 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
996 # Helper function. Make a display label for the given witnesses, showing a.c.
997 # witnesses only where the main witness is not also in the list.
998 sub _path_display_label {
1001 map { $wits{$_} = 1 } @_;
1003 # If an a.c. wit is listed, remove it if the main wit is also listed.
1004 # Otherwise keep it for explicit listing.
1005 my $aclabel = $self->ac_label;
1007 foreach my $w ( sort keys %wits ) {
1008 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
1009 if( exists $wits{$1} ) {
1012 push( @disp_ac, $w );
1017 # See if we are in a majority situation.
1018 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
1019 $maj = $maj > 5 ? $maj : 5;
1020 if( scalar keys %wits > $maj ) {
1021 unshift( @disp_ac, 'majority' );
1022 return join( ', ', @disp_ac );
1024 return join( ', ', sort keys %wits );
1028 =head2 readings_at_rank( $rank )
1030 Returns a list of readings at a given rank, taken from the alignment table.
1034 sub readings_at_rank {
1035 my( $self, $rank ) = @_;
1036 my $table = $self->alignment_table;
1037 # Table rank is real rank - 1.
1038 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
1040 foreach my $e ( @elements ) {
1041 next unless ref( $e ) eq 'HASH';
1042 next unless exists $e->{'t'};
1043 $readings{$e->{'t'}->id} = $e->{'t'};
1045 return values %readings;
1050 Returns a GraphML representation of the collation. The GraphML will contain
1051 two graphs. The first expresses the attributes of the readings and the witness
1052 paths that link them; the second expresses the relationships that link the
1053 readings. This is the native transfer format for a tradition.
1057 use Text::Tradition;
1063 my $datafile = 't/data/florilegium_tei_ps.xml';
1064 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1066 'file' => $datafile,
1069 ok( $tradition, "Got a tradition object" );
1070 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
1071 ok( $tradition->collation, "Tradition has a collation" );
1073 my $c = $tradition->collation;
1074 is( scalar $c->readings, $READINGS, "Collation has all readings" );
1075 is( scalar $c->paths, $PATHS, "Collation has all paths" );
1076 is( scalar $c->relationships, 0, "Collation has all relationships" );
1078 # Add a few relationships
1079 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
1080 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
1081 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
1083 # Now write it to GraphML and parse it again.
1085 my $graphml = $c->as_graphml;
1086 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
1087 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
1088 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
1089 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
1091 # Now add a stemma, write to GraphML, and look at the output.
1093 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
1094 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
1095 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
1096 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
1097 $graphml = $c->as_graphml;
1098 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
1105 ## TODO MOVE this to Tradition.pm and modularize it better
1107 my( $self, $options ) = @_;
1108 $self->calculate_ranks unless $self->_graphcalc_done;
1110 my $start = $options->{'from'}
1111 ? $self->reading( $options->{'from'} ) : $self->start;
1112 my $end = $options->{'to'}
1113 ? $self->reading( $options->{'to'} ) : $self->end;
1114 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1115 throw( 'Start node must be before end node' );
1117 # The readings need to be ranked for this to work.
1118 $start = $self->start unless $start->has_rank;
1119 $end = $self->end unless $end->has_rank;
1121 unless( $start eq $self->start ) {
1122 $rankoffset = $start->rank - 1;
1127 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1128 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1129 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1130 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1132 # Create the document and root node
1133 require XML::LibXML;
1134 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1135 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1136 $graphml->setDocumentElement( $root );
1137 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1138 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1140 # List of attribute types to save on our objects and their corresponding
1145 'Bool' => 'boolean',
1146 'ReadingID' => 'string',
1147 'RelationshipType' => 'string',
1148 'RelationshipScope' => 'string',
1151 # Add the data keys for the graph. Include an extra key 'version' for the
1152 # GraphML output version.
1153 my %graph_data_keys;
1155 my %graph_attributes = ( 'version' => 'string' );
1156 # Graph attributes include those of Tradition and those of Collation.
1158 # TODO Use meta introspection method from duplicate_reading to do this
1159 # instead of naming custom keys.
1160 my $tmeta = $self->tradition->meta;
1161 my $cmeta = $self->meta;
1162 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1163 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1164 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1165 next if $attr->name =~ /^_/;
1166 next unless $save_types{$attr->type_constraint->name};
1167 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1169 # Extra custom keys for complex objects that should be saved in some form.
1170 # The subroutine should return a string, or undef/empty.
1171 if( $tmeta->has_method('stemmata') ) {
1172 $graph_attributes{'stemmata'} = sub {
1174 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1175 $self->tradition->stemmata;
1176 join( "\n", @stemstrs );
1180 if( $tmeta->has_method('user') ) {
1181 $graph_attributes{'user'} = sub {
1182 $self->tradition->user ? $self->tradition->user->id : undef
1186 foreach my $datum ( sort keys %graph_attributes ) {
1187 $graph_data_keys{$datum} = 'dg'.$gdi++;
1188 my $key = $root->addNewChild( $graphml_ns, 'key' );
1189 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1190 : $graph_attributes{$datum};
1191 $key->setAttribute( 'attr.name', $datum );
1192 $key->setAttribute( 'attr.type', $dtype );
1193 $key->setAttribute( 'for', 'graph' );
1194 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1197 # Add the data keys for reading nodes
1198 my %reading_attributes;
1199 my $rmeta = Text::Tradition::Collation::Reading->meta;
1200 foreach my $attr( $rmeta->get_all_attributes ) {
1201 next if $attr->name =~ /^_/;
1202 next unless $save_types{$attr->type_constraint->name};
1203 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1205 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1206 # Extra custom key for the reading morphology
1207 $reading_attributes{'lexemes'} = 'string';
1212 foreach my $datum ( sort keys %reading_attributes ) {
1213 $node_data_keys{$datum} = 'dn'.$ndi++;
1214 my $key = $root->addNewChild( $graphml_ns, 'key' );
1215 $key->setAttribute( 'attr.name', $datum );
1216 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1217 $key->setAttribute( 'for', 'node' );
1218 $key->setAttribute( 'id', $node_data_keys{$datum} );
1221 # Add the data keys for edges, that is, paths and relationships. Path
1222 # data does not come from a Moose class so is here manually.
1225 my %edge_attributes = (
1226 witness => 'string', # ID/label for a path
1227 extra => 'boolean', # Path key
1229 my @path_attributes = keys %edge_attributes; # track our manual additions
1230 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1231 foreach my $attr( $pmeta->get_all_attributes ) {
1232 next if $attr->name =~ /^_/;
1233 next unless $save_types{$attr->type_constraint->name};
1234 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1236 foreach my $datum ( sort keys %edge_attributes ) {
1237 $edge_data_keys{$datum} = 'de'.$edi++;
1238 my $key = $root->addNewChild( $graphml_ns, 'key' );
1239 $key->setAttribute( 'attr.name', $datum );
1240 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1241 $key->setAttribute( 'for', 'edge' );
1242 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1245 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1246 my $xmlidname = $self->tradition->name;
1247 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1248 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1249 $xmlidname = '_'.$xmlidname;
1251 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1252 $sgraph->setAttribute( 'edgedefault', 'directed' );
1253 $sgraph->setAttribute( 'id', $xmlidname );
1254 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1255 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1256 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1257 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1258 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1260 # Tradition/collation attribute data
1261 foreach my $datum ( keys %graph_attributes ) {
1263 if( $datum eq 'version' ) {
1265 } elsif( ref( $graph_attributes{$datum} ) ) {
1266 my $sub = $graph_attributes{$datum};
1268 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1269 $value = $self->tradition->$datum;
1271 $value = $self->$datum;
1273 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1278 # Add our readings to the graph
1279 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1280 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1281 ( $n->rank < $start->rank || $n->rank > $end->rank );
1282 $use_readings{$n->id} = 1;
1283 # Add to the main graph
1284 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1285 my $node_xmlid = 'n' . $node_ctr++;
1286 $node_hash{ $n->id } = $node_xmlid;
1287 $node_el->setAttribute( 'id', $node_xmlid );
1288 foreach my $d ( keys %reading_attributes ) {
1290 # Custom serialization
1291 if( $d eq 'lexemes' ) {
1292 # If nval is a true value, we have lexemes so we need to
1293 # serialize them. Otherwise set nval to undef so that the
1294 # key is excluded from this reading.
1295 $nval = $nval ? $n->_serialize_lexemes : undef;
1296 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1299 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1300 # Adjust the ranks within the subgraph.
1301 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1302 : $nval - $rankoffset;
1304 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1309 # Add the path edges to the sequence graph
1311 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1312 # We add an edge in the graphml for every witness in $e.
1313 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1314 my @edge_wits = sort $self->path_witnesses( $e );
1315 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1316 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1317 # Skip any path from start to end; that witness is not in the subgraph.
1318 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1319 foreach my $wit ( @edge_wits ) {
1320 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1321 $node_hash{ $e->[0] },
1322 $node_hash{ $e->[1] } );
1323 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1324 $edge_el->setAttribute( 'source', $from );
1325 $edge_el->setAttribute( 'target', $to );
1326 $edge_el->setAttribute( 'id', $id );
1328 # It's a witness path, so add the witness
1330 my $key = $edge_data_keys{'witness'};
1331 # Is this an ante-corr witness?
1332 my $aclabel = $self->ac_label;
1333 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1334 # Keep the base witness
1336 # ...and record that this is an 'extra' reading path
1337 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1339 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1343 # Report the actual number of nodes and edges that went in
1344 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1345 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1347 # Add the relationship graph to the XML
1348 map { delete $edge_data_keys{$_} } @path_attributes;
1349 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1350 $node_data_keys{'id'}, \%edge_data_keys );
1352 # Save and return the thing
1353 my $result = decode_utf8( $graphml->toString(1) );
1357 sub _add_graphml_data {
1358 my( $el, $key, $value ) = @_;
1359 return unless defined $value;
1360 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1361 $data_el->setAttribute( 'key', $key );
1362 $data_el->appendText( $value );
1367 Returns a CSV alignment table representation of the collation graph, one
1368 row per witness (or witness uncorrected.)
1374 my $table = $self->alignment_table;
1375 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1377 # Make the header row
1378 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1379 push( @result, decode_utf8( $csv->string ) );
1380 # Make the rest of the rows
1381 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1382 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1383 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1384 $csv->combine( @row );
1385 push( @result, decode_utf8( $csv->string ) );
1387 return join( "\n", @result );
1390 =head2 alignment_table
1392 Return a reference to an alignment table, in a slightly enhanced CollateX
1393 format which looks like this:
1395 $table = { alignment => [ { witness => "SIGIL",
1396 tokens => [ { t => "TEXT" }, ... ] },
1397 { witness => "SIG2",
1398 tokens => [ { t => "TEXT" }, ... ] },
1400 length => TEXTLEN };
1404 sub alignment_table {
1406 return $self->cached_table if $self->has_cached_table;
1408 # Make sure we can do this
1409 throw( "Need a linear graph in order to make an alignment table" )
1410 unless $self->linear;
1411 $self->calculate_ranks()
1412 unless $self->_graphcalc_done && $self->end->has_rank;
1414 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1415 my @all_pos = ( 1 .. $self->end->rank - 1 );
1416 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1417 # say STDERR "Making witness row(s) for " . $wit->sigil;
1418 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1419 my @row = _make_witness_row( \@wit_path, \@all_pos );
1420 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1421 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1422 push( @{$table->{'alignment'}}, $witobj );
1423 if( $wit->is_layered ) {
1424 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1425 $wit->sigil.$self->ac_label );
1426 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1427 my $witacobj = { 'witness' => $wit->sigil.$self->ac_label,
1428 'tokens' => \@ac_row };
1429 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1430 push( @{$table->{'alignment'}}, $witacobj );
1433 $self->cached_table( $table );
1437 sub _make_witness_row {
1438 my( $path, $positions ) = @_;
1440 map { $char_hash{$_} = undef } @$positions;
1442 foreach my $rdg ( @$path ) {
1443 say STDERR "rank " . $rdg->rank if $debug;
1444 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1445 $char_hash{$rdg->rank} = { 't' => $rdg };
1447 my @row = map { $char_hash{$_} } @$positions;
1448 # Fill in lacuna markers for undef spots in the row
1449 my $last_el = shift @row;
1450 my @filled_row = ( $last_el );
1451 foreach my $el ( @row ) {
1452 # If we are using node reference, make the lacuna node appear many times
1453 # in the table. If not, use the lacuna tag.
1454 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1457 push( @filled_row, $el );
1464 =head1 NAVIGATION METHODS
1466 =head2 reading_sequence( $first, $last, $sigil, $backup )
1468 Returns the ordered list of readings, starting with $first and ending
1469 with $last, for the witness given in $sigil. If a $backup sigil is
1470 specified (e.g. when walking a layered witness), it will be used wherever
1471 no $sigil path exists. If there is a base text reading, that will be
1472 used wherever no path exists for $sigil or $backup.
1476 # TODO Think about returning some lazy-eval iterator.
1477 # TODO Get rid of backup; we should know from what witness is whether we need it.
1479 sub reading_sequence {
1480 my( $self, $start, $end, $witness ) = @_;
1482 $witness = $self->baselabel unless $witness;
1483 my @readings = ( $start );
1486 while( $n && $n->id ne $end->id ) {
1487 if( exists( $seen{$n->id} ) ) {
1488 throw( "Detected loop for $witness at " . $n->id );
1492 my $next = $self->next_reading( $n, $witness );
1494 throw( "Did not find any path for $witness from reading " . $n->id );
1496 push( @readings, $next );
1499 # Check that the last reading is our end reading.
1500 my $last = $readings[$#readings];
1501 throw( "Last reading found from " . $start->text .
1502 " for witness $witness is not the end!" ) # TODO do we get this far?
1503 unless $last->id eq $end->id;
1508 =head2 next_reading( $reading, $sigil );
1510 Returns the reading that follows the given reading along the given witness
1516 # Return the successor via the corresponding path.
1518 my $answer = $self->_find_linked_reading( 'next', @_ );
1519 return undef unless $answer;
1520 return $self->reading( $answer );
1523 =head2 prior_reading( $reading, $sigil )
1525 Returns the reading that precedes the given reading along the given witness
1531 # Return the predecessor via the corresponding path.
1533 my $answer = $self->_find_linked_reading( 'prior', @_ );
1534 return $self->reading( $answer );
1537 sub _find_linked_reading {
1538 my( $self, $direction, $node, $path ) = @_;
1540 # Get a backup if we are dealing with a layered witness
1542 my $aclabel = $self->ac_label;
1543 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1547 my @linked_paths = $direction eq 'next'
1548 ? $self->sequence->edges_from( $node )
1549 : $self->sequence->edges_to( $node );
1550 return undef unless scalar( @linked_paths );
1552 # We have to find the linked path that contains all of the
1553 # witnesses supplied in $path.
1554 my( @path_wits, @alt_path_wits );
1555 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1556 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1559 foreach my $le ( @linked_paths ) {
1560 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1563 my @le_wits = sort $self->path_witnesses( $le );
1564 if( _is_within( \@path_wits, \@le_wits ) ) {
1565 # This is the right path.
1566 return $direction eq 'next' ? $le->[1] : $le->[0];
1567 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1571 # Got this far? Return the alternate path if it exists.
1572 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1575 # Got this far? Return the base path if it exists.
1576 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1579 # Got this far? We have no appropriate path.
1580 warn "Could not find $direction node from " . $node->id
1581 . " along path $path";
1587 my( $set1, $set2 ) = @_;
1588 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1589 foreach my $el ( @$set1 ) {
1590 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1595 # Return the string that joins together a list of witnesses for
1596 # display on a single path.
1597 sub _witnesses_of_label {
1598 my( $self, $label ) = @_;
1599 my $regex = $self->wit_list_separator;
1600 my @answer = split( /\Q$regex\E/, $label );
1604 =head2 common_readings
1606 Returns the list of common readings in the graph (i.e. those readings that are
1607 shared by all non-lacunose witnesses.)
1611 sub common_readings {
1613 my @common = grep { $_->is_common } $self->readings;
1617 =head2 path_text( $sigil, [, $start, $end ] )
1619 Returns the text of a witness (plus its backup, if we are using a layer)
1620 as stored in the collation. The text is returned as a string, where the
1621 individual readings are joined with spaces and the meta-readings (e.g.
1622 lacunae) are omitted. Optional specification of $start and $end allows
1623 the generation of a subset of the witness text.
1628 my( $self, $wit, $start, $end ) = @_;
1629 $start = $self->start unless $start;
1630 $end = $self->end unless $end;
1631 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1634 foreach my $r ( @path ) {
1635 unless ( $r->join_prior || !$last || $last->join_next ) {
1638 $pathtext .= $r->text;
1644 =head1 INITIALIZATION METHODS
1646 These are mostly for use by parsers.
1648 =head2 make_witness_path( $witness )
1650 Link the array of readings contained in $witness->path (and in
1651 $witness->uncorrected_path if it exists) into collation paths.
1652 Clear out the arrays when finished.
1654 =head2 make_witness_paths
1656 Call make_witness_path for all witnesses in the tradition.
1660 # For use when a collation is constructed from a base text and an apparatus.
1661 # We have the sequences of readings and just need to add path edges.
1662 # When we are done, clear out the witness path attributes, as they are no
1664 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1666 sub make_witness_paths {
1668 foreach my $wit ( $self->tradition->witnesses ) {
1669 # say STDERR "Making path for " . $wit->sigil;
1670 $self->make_witness_path( $wit );
1674 sub make_witness_path {
1675 my( $self, $wit ) = @_;
1676 my @chain = @{$wit->path};
1677 my $sig = $wit->sigil;
1678 # Add start and end if necessary
1679 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1680 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1681 foreach my $idx ( 0 .. $#chain-1 ) {
1682 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1684 if( $wit->is_layered ) {
1685 @chain = @{$wit->uncorrected_path};
1686 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1687 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1688 foreach my $idx( 0 .. $#chain-1 ) {
1689 my $source = $chain[$idx];
1690 my $target = $chain[$idx+1];
1691 $self->add_path( $source, $target, $sig.$self->ac_label )
1692 unless $self->has_path( $source, $target, $sig );
1696 $wit->clear_uncorrected_path;
1699 =head2 calculate_ranks
1701 Calculate the reading ranks (that is, their aligned positions relative
1702 to each other) for the graph. This can only be called on linear collations.
1706 use Text::Tradition;
1708 my $cxfile = 't/data/Collatex-16.xml';
1709 my $t = Text::Tradition->new(
1711 'input' => 'CollateX',
1714 my $c = $t->collation;
1717 my $table = $c->alignment_table;
1718 ok( $c->has_cached_table, "Alignment table was cached" );
1719 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1720 $c->calculate_ranks;
1721 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1722 $c->add_relationship( 'n13', 'n23', { type => 'repetition' } );
1723 is( $c->alignment_table, $table, "Alignment table unchanged after non-colo relationship add" );
1724 $c->add_relationship( 'n24', 'n23', { type => 'spelling' } );
1725 isnt( $c->alignment_table, $table, "Alignment table changed after colo relationship add" );
1731 sub calculate_ranks {
1733 # Save the existing ranks, in case we need to invalidate the cached SVG.
1735 map { $existing_ranks{$_} = $_->rank } $self->readings;
1737 # Do the rankings based on the relationship equivalence graph, starting
1738 # with the start node.
1739 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1741 # Transfer our rankings from the topological graph to the real one.
1742 foreach my $r ( $self->readings ) {
1743 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1744 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1746 # Die. Find the last rank we calculated.
1747 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1748 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1750 my $last = pop @all_defined;
1751 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1754 # Do we need to invalidate the cached data?
1755 if( $self->has_cached_table ) {
1756 foreach my $r ( $self->readings ) {
1757 next if defined( $existing_ranks{$r} )
1758 && $existing_ranks{$r} == $r->rank;
1759 # Something has changed, so clear the cache
1760 $self->_clear_cache;
1761 # ...and recalculate the common readings.
1762 $self->calculate_common_readings();
1766 # The graph calculation information is now up to date.
1767 $self->_graphcalc_done(1);
1772 $self->wipe_table if $self->has_cached_table;
1776 =head2 flatten_ranks
1778 A convenience method for parsing collation data. Searches the graph for readings
1779 with the same text at the same rank, and merges any that are found.
1785 my %unique_rank_rdg;
1787 foreach my $rdg ( $self->readings ) {
1788 next unless $rdg->has_rank;
1789 my $key = $rdg->rank . "||" . $rdg->text;
1790 if( exists $unique_rank_rdg{$key} ) {
1791 # Make sure they don't have different grammatical forms
1792 my $ur = $unique_rank_rdg{$key};
1793 if( $rdg->is_identical( $ur ) ) {
1795 #say STDERR "Combining readings at same rank: $key";
1797 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1798 # TODO see if this now makes a common point.
1801 $unique_rank_rdg{$key} = $rdg;
1804 # If we merged readings, the ranks are still fine but the alignment
1805 # table is wrong. Wipe it.
1806 $self->wipe_table() if $changed;
1810 =head2 calculate_common_readings
1812 Goes through the graph identifying the readings that appear in every witness
1813 (apart from those with lacunae at that spot.) Marks them as common and returns
1818 use Text::Tradition;
1820 my $cxfile = 't/data/Collatex-16.xml';
1821 my $t = Text::Tradition->new(
1823 'input' => 'CollateX',
1826 my $c = $t->collation;
1828 my @common = $c->calculate_common_readings();
1829 is( scalar @common, 8, "Found correct number of common readings" );
1830 my @marked = sort $c->common_readings();
1831 is( scalar @common, 8, "All common readings got marked as such" );
1832 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1833 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1839 sub calculate_common_readings {
1842 map { $_->is_common( 0 ) } $self->readings;
1843 # Implicitly calls calculate_ranks
1844 my $table = $self->alignment_table;
1845 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1846 my @row = map { $_->{'tokens'}->[$idx]
1847 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1848 @{$table->{'alignment'}};
1850 foreach my $r ( @row ) {
1852 $hash{$r->id} = $r unless $r->is_meta;
1854 $hash{'UNDEF'} = $r;
1857 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1858 my( $r ) = values %hash;
1860 push( @common, $r );
1866 =head2 text_from_paths
1868 Calculate the text array for all witnesses from the path, for later consistency
1869 checking. Only to be used if there is no non-graph-based way to know the
1874 sub text_from_paths {
1876 foreach my $wit ( $self->tradition->witnesses ) {
1877 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1879 foreach my $r ( @readings ) {
1880 next if $r->is_meta;
1881 push( @text, $r->text );
1883 $wit->text( \@text );
1884 if( $wit->is_layered ) {
1885 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1886 $wit->sigil.$self->ac_label );
1888 foreach my $r ( @ucrdgs ) {
1889 next if $r->is_meta;
1890 push( @uctext, $r->text );
1892 $wit->layertext( \@uctext );
1897 =head1 UTILITY FUNCTIONS
1899 =head2 common_predecessor( $reading_a, $reading_b )
1901 Find the last reading that occurs in sequence before both the given readings.
1902 At the very least this should be $self->start.
1904 =head2 common_successor( $reading_a, $reading_b )
1906 Find the first reading that occurs in sequence after both the given readings.
1907 At the very least this should be $self->end.
1911 use Text::Tradition;
1913 my $cxfile = 't/data/Collatex-16.xml';
1914 my $t = Text::Tradition->new(
1916 'input' => 'CollateX',
1919 my $c = $t->collation;
1921 is( $c->common_predecessor( 'n24', 'n23' )->id,
1922 'n20', "Found correct common predecessor" );
1923 is( $c->common_successor( 'n24', 'n23' )->id,
1924 '__END__', "Found correct common successor" );
1926 is( $c->common_predecessor( 'n19', 'n17' )->id,
1927 'n16', "Found correct common predecessor for readings on same path" );
1928 is( $c->common_successor( 'n21', 'n10' )->id,
1929 '__END__', "Found correct common successor for readings on same path" );
1935 ## Return the closest reading that is a predecessor of both the given readings.
1936 sub common_predecessor {
1938 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1939 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1942 sub common_successor {
1944 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1945 return $self->_common_in_path( $r1, $r2, 'successors' );
1949 # TODO think about how to do this without ranks...
1950 sub _common_in_path {
1951 my( $self, $r1, $r2, $dir ) = @_;
1952 my $iter = $self->end->rank;
1954 my @last_r1 = ( $r1 );
1955 my @last_r2 = ( $r2 );
1956 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1958 # say STDERR "Finding common $dir for $r1, $r2";
1959 while( !@candidates ) {
1960 last unless $iter--; # Avoid looping infinitely
1961 # Iterate separately down the graph from r1 and r2
1962 my( @new_lc1, @new_lc2 );
1963 foreach my $lc ( @last_r1 ) {
1964 foreach my $p ( $lc->$dir ) {
1965 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1966 # say STDERR "Path candidate $p from $lc";
1967 push( @candidates, $p );
1968 } elsif( !$all_seen{$p->id} ) {
1969 $all_seen{$p->id} = 'r1';
1970 push( @new_lc1, $p );
1974 foreach my $lc ( @last_r2 ) {
1975 foreach my $p ( $lc->$dir ) {
1976 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1977 # say STDERR "Path candidate $p from $lc";
1978 push( @candidates, $p );
1979 } elsif( !$all_seen{$p->id} ) {
1980 $all_seen{$p->id} = 'r2';
1981 push( @new_lc2, $p );
1985 @last_r1 = @new_lc1;
1986 @last_r2 = @new_lc2;
1988 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1989 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1993 Text::Tradition::Error->throw(
1994 'ident' => 'Collation error',
2000 __PACKAGE__->meta->make_immutable;
2006 =item * Rework XML serialization in a more modular way
2012 This package is free software and is provided "as is" without express
2013 or implied warranty. You can redistribute it and/or modify it under
2014 the same terms as Perl itself.
2018 Tara L Andrews E<lt>aurum@cpan.orgE<gt>