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 has_reading( $id )
164 Predicate to see whether a given reading ID is in the graph.
166 =head2 reading_witnesses( $object_or_id )
168 Returns a list of sigils whose witnesses contain the reading.
172 Returns all reading paths within the document - that is, all edges in the
173 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
175 =head2 add_path( $source, $target, $sigil )
177 Links the given readings in the collation in sequence, under the given witness
178 sigil. The readings may be specified by object or ID.
180 =head2 del_path( $source, $target, $sigil )
182 Links the given readings in the collation in sequence, under the given witness
183 sigil. The readings may be specified by object or ID.
185 =head2 has_path( $source, $target );
187 Returns true if the two readings are linked in sequence in any witness.
188 The readings may be specified by object or ID.
192 Returns all Relationship objects in the collation.
194 =head2 add_relationship( $reading, $other_reading, $options )
196 Adds a new relationship of the type given in $options between the two readings,
197 which may be specified by object or ID. Returns a value of ( $status, @vectors)
198 where $status is true on success, and @vectors is a list of relationship edges
199 that were ultimately added.
200 See L<Text::Tradition::Collation::Relationship> for the available options.
205 my ( $class, @args ) = @_;
206 my %args = @args == 1 ? %{ $args[0] } : @args;
207 # TODO determine these from the Moose::Meta object
208 my @delegate_attrs = qw(sequence relations readings wit_list_separator baselabel
209 linear wordsep start end cached_table _graphcalc_done);
211 for my $attr (@delegate_attrs) {
212 $data_args{$attr} = delete $args{$attr} if exists $args{$attr};
214 $args{_data} = Text::Tradition::Collation::Data->new(%data_args);
220 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
221 $self->_set_start( $self->add_reading(
222 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
223 $self->_set_end( $self->add_reading(
224 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
227 sub register_relationship_type {
229 my %args = @_ == 1 ? %{$_[0]} : @_;
230 if( $self->relations->has_type( $args{name} ) ) {
231 throw( 'Relationship type ' . $args{name} . ' already registered' );
233 $self->relations->add_type( %args );
236 ### Reading construct/destruct functions
239 my( $self, $reading ) = @_;
240 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
241 my %args = %$reading;
242 if( $args{'init'} ) {
243 # If we are initializing an empty collation, don't assume that we
244 # have set a tradition.
245 delete $args{'init'};
246 } elsif( $self->tradition->can('language') && $self->tradition->has_language
247 && !exists $args{'language'} ) {
248 $args{'language'} = $self->tradition->language;
250 $reading = Text::Tradition::Collation::Reading->new(
251 'collation' => $self,
254 # First check to see if a reading with this ID exists.
255 if( $self->reading( $reading->id ) ) {
256 throw( "Collation already has a reading with id " . $reading->id );
258 $self->_graphcalc_done(0);
259 $self->_add_reading( $reading->id => $reading );
260 # Once the reading has been added, put it in both graphs.
261 $self->sequence->add_vertex( $reading->id );
262 $self->relations->add_reading( $reading->id );
266 around del_reading => sub {
271 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
274 # Remove the reading from the graphs.
275 $self->_graphcalc_done(0);
276 $self->_clear_cache; # Explicitly clear caches to GC the reading
277 $self->sequence->delete_vertex( $arg );
278 $self->relations->delete_reading( $arg );
281 $self->$orig( $arg );
284 =head2 merge_readings( $main, $second, $concatenate, $with_str )
286 Merges the $second reading into the $main one. If $concatenate is true, then
287 the merged node will carry the text of both readings, concatenated with either
288 $with_str (if specified) or a sensible default (the empty string if the
289 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
291 The first two arguments may be either readings or reading IDs.
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 ('unto') 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' );
313 $c->add_relationship( 'n21', 'n22', { type => 'collated', scope => 'local' } );
315 ok( $c->reading( 'n21p0' ), "New reading exists" );
316 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
318 # Combine n3 and n4 ( with his )
319 $c->merge_readings( 'n3', 'n4', 1 );
320 ok( !$c->reading('n4'), "Reading n4 is gone" );
321 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
323 # Collapse n9 and n10 ( rood / root )
324 $c->merge_readings( 'n9', 'n10' );
325 ok( !$c->reading('n10'), "Reading n10 is gone" );
326 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
328 # Combine n21 and n21p0
329 my $remaining = $c->reading('n21');
330 $remaining ||= $c->reading('n22'); # one of these should still exist
331 $c->merge_readings( 'n21p0', $remaining, 1 );
332 ok( !$c->reading('n21'), "Reading $remaining is gone" );
333 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
343 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
344 my $mergemeta = $kept_obj->is_meta;
345 throw( "Cannot merge meta and non-meta reading" )
346 unless ( $mergemeta && $del_obj->is_meta )
347 || ( !$mergemeta && !$del_obj->is_meta );
349 throw( "Cannot merge with start or end node" )
350 if( $kept_obj eq $self->start || $kept_obj eq $self->end
351 || $del_obj eq $self->start || $del_obj eq $self->end );
352 throw( "Cannot combine text of meta readings" ) if $combine;
354 # We only need the IDs for adding paths to the graph, not the reading
355 # objects themselves.
356 my $kept = $kept_obj->id;
357 my $deleted = $del_obj->id;
358 $self->_graphcalc_done(0);
360 # The kept reading should inherit the paths and the relationships
361 # of the deleted reading.
362 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
363 my @vector = ( $kept );
364 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
365 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
366 next if $vector[0] eq $vector[1]; # Don't add a self loop
367 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
368 $self->sequence->add_edge( @vector );
369 my $fwits = $self->sequence->get_edge_attributes( @vector );
370 @wits{keys %$fwits} = values %$fwits;
371 $self->sequence->set_edge_attributes( @vector, \%wits );
373 $self->relations->merge_readings( $kept, $deleted, $combine );
375 # Do the deletion deed.
377 # Combine the text of the readings
378 my $joinstr = $combine_char;
379 unless( defined $joinstr ) {
380 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
381 $joinstr = $self->wordsep unless defined $joinstr;
383 $kept_obj->_combine( $del_obj, $joinstr );
385 $self->del_reading( $deleted );
388 =head2 compress_readings
390 Where possible in the graph, compresses plain sequences of readings into a
391 single reading. The sequences must consist of readings with no
392 relationships to other readings, with only a single witness path between
393 them and no other witness paths from either that would skip the other. The
394 readings must also not be marked as nonsense or bad grammar.
396 WARNING: This operation cannot be undone.
400 sub compress_readings {
402 # Anywhere in the graph that there is a reading that joins only to a single
403 # successor, and neither of these have any relationships, just join the two
405 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
406 # Now look for readings that can be joined to their successors.
407 next unless $rdg->is_combinable;
409 while( $self->sequence->successors( $rdg ) == 1 ) {
410 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
411 throw( "Infinite loop" ) if $seen{$next->id};
412 $seen{$next->id} = 1;
413 last if $self->sequence->predecessors( $next ) > 1;
414 last unless $next->is_combinable;
415 say "Joining readings $rdg and $next";
416 $self->merge_readings( $rdg, $next, 1 );
419 # Make sure we haven't screwed anything up
420 foreach my $wit ( $self->tradition->witnesses ) {
421 my $pathtext = $self->path_text( $wit->sigil );
422 my $origtext = join( ' ', @{$wit->text} );
423 throw( "Text differs for witness " . $wit->sigil )
424 unless $pathtext eq $origtext;
425 if( $wit->is_layered ) {
426 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
427 $origtext = join( ' ', @{$wit->layertext} );
428 throw( "Ante-corr text differs for witness " . $wit->sigil )
429 unless $pathtext eq $origtext;
433 $self->relations->rebuild_equivalence();
434 $self->calculate_ranks();
437 # Helper function for manipulating the graph.
438 sub _stringify_args {
439 my( $self, $first, $second, @args ) = @_;
441 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
442 $second = $second->id
443 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
444 return( $first, $second, @args );
447 # Helper function for manipulating the graph.
448 sub _objectify_args {
449 my( $self, $first, $second, $arg ) = @_;
450 $first = $self->reading( $first )
451 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
452 $second = $self->reading( $second )
453 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
454 return( $first, $second, $arg );
457 =head2 duplicate_reading( $reading, @witlist )
459 Split the given reading into two, so that the new reading is in the path for
460 the witnesses given in @witlist. If the result is that certain non-colocated
461 relationships (e.g. transpositions) are no longer valid, these will be removed.
462 Returns the newly-created reading.
466 use Test::More::UTF8;
470 my $st = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/collatecorr.xml' );
471 is( ref( $st ), 'Text::Tradition', "Got a tradition from test file" );
472 ok( $st->has_witness('Ba96'), "Tradition has the affected witness" );
474 my $sc = $st->collation;
476 ok( $sc->reading('n131'), "Tradition has the affected reading" );
477 is( scalar( $sc->readings ), $numr, "There are $numr readings in the graph" );
478 is( $sc->end->rank, 14, "There are fourteen ranks in the graph" );
480 # Detach the erroneously collated reading
481 my( $newr, @del_rdgs ) = $sc->duplicate_reading( 'n131', 'Ba96' );
482 ok( $newr, "New reading was created" );
483 ok( $sc->reading('n131_0'), "Detached the bad collation with a new reading" );
484 is( scalar( $sc->readings ), $numr + 1, "A reading was added to the graph" );
485 is( $sc->end->rank, 10, "There are now only ten ranks in the graph" );
486 my $csucc = $sc->common_successor( 'n131', 'n131_0' );
487 is( $csucc->id, 'n136', "Found correct common successor to duped reading" );
489 # Check that the bad transposition is gone
490 is( scalar @del_rdgs, 1, "Deleted reading was returned by API call" );
491 is( $sc->get_relationship( 'n130', 'n135' ), undef, "Bad transposition relationship is gone" );
493 # The collation should not be fixed
494 my @pairs = $sc->identical_readings();
495 is( scalar @pairs, 0, "Not re-collated yet" );
497 ok( $sc->merge_readings( 'n124', 'n131_0' ), "Collated the readings correctly" );
498 @pairs = $sc->identical_readings( start => 'n124', end => $csucc->id );
499 is( scalar @pairs, 3, "Found three more identical readings" );
500 is( $sc->end->rank, 11, "The ranks shifted appropriately" );
501 $sc->flatten_ranks();
502 is( scalar( $sc->readings ), $numr - 3, "Now we are collated correctly" );
504 # Check that we can't "duplicate" a reading with no wits or with all wits
506 my( $badr, @del_rdgs ) = $sc->duplicate_reading( 'n124' );
507 ok( 0, "Reading duplication without witnesses throws an error" );
508 } catch( Text::Tradition::Error $e ) {
509 like( $e->message, qr/Must specify one or more witnesses/,
510 "Reading duplication without witnesses throws the expected error" );
512 ok( 0, "Reading duplication without witnesses threw the wrong error" );
516 my( $badr, @del_rdgs ) = $sc->duplicate_reading( 'n124', 'Ba96', 'Mü11475' );
517 ok( 0, "Reading duplication with all witnesses throws an error" );
518 } catch( Text::Tradition::Error $e ) {
519 like( $e->message, qr/Cannot join all witnesses/,
520 "Reading duplication with all witnesses throws the expected error" );
522 ok( 0, "Reading duplication with all witnesses threw the wrong error" );
529 sub duplicate_reading {
530 my( $self, $r, @wits ) = @_;
531 # Check that we are not doing anything unwise.
532 throw( "Must specify one or more witnesses for the duplicated reading" )
534 unless( ref( $r ) eq 'Text::Tradition::Collation::Reading' ) {
535 $r = $self->reading( $r );
537 throw( "Cannot duplicate a meta-reading" )
539 my $ordered_req_wits = join( ',', sort @wits );
540 my $ordered_rdg_wits = join( ',', $r->witnesses );
541 throw( "Cannot join all witnesses to the new reading" )
542 if $ordered_req_wits eq $ordered_rdg_wits;
544 # Get all the reading attributes and duplicate them.
545 my $rmeta = Text::Tradition::Collation::Reading->meta;
547 foreach my $attr( $rmeta->get_all_attributes ) {
548 next if $attr->name =~ /^_/;
549 my $acc = $attr->get_read_method;
550 if( !$acc && $attr->has_applied_traits ) {
551 my $tr = $attr->applied_traits;
552 if( $tr->[0] =~ /::(Array|Hash)$/ ) {
554 my %methods = reverse %{$attr->handles};
555 $acc = $methods{elements};
556 $args{$attr->name} = $which eq 'Array'
557 ? [ $r->$acc ] : { $r->$acc };
560 $args{$attr->name} = $r->$acc if $acc;
563 # By definition the new reading will no longer be common.
564 $args{is_common} = 0;
565 # The new reading also needs its own ID.
566 $args{id} = $self->_generate_dup_id( $r->id );
568 # Try to make the new reading.
569 my $newr = $self->add_reading( \%args );
570 # The old reading is also no longer common.
573 # For each of the witnesses, dissociate from the old reading and
574 # associate with the new.
575 foreach my $wit ( @wits ) {
576 my $prior = $self->prior_reading( $r, $wit );
577 my $next = $self->next_reading( $r, $wit );
578 $self->del_path( $prior, $r, $wit );
579 $self->add_path( $prior, $newr, $wit );
580 $self->del_path( $r, $next, $wit );
581 $self->add_path( $newr, $next, $wit );
584 # If the graph is ranked, we need to look for relationships that are now
585 # invalid (i.e. 'non-colocation' types that might now be colocated) and
586 # remove them. If not, we can skip it.
589 my @deleted_relations;
590 if( $self->end->has_rank ) {
591 # Find the point where we can stop checking
592 $succ = $self->common_successor( $r, $newr );
594 # Hash the existing ranks
595 foreach my $rdg ( $self->readings ) {
596 $rrk{$rdg->id} = $rdg->rank;
598 # Calculate the new ranks
599 $self->calculate_ranks();
601 # Check for invalid non-colocated relationships among changed-rank readings
602 # from where the ranks start changing up to $succ
603 my $lastrank = $succ->rank;
604 foreach my $rdg ( $self->readings ) {
605 next if $rdg->rank > $lastrank;
606 next if $rdg->rank == $rrk{$rdg->id};
607 my @noncolo = $rdg->related_readings( sub { !$_[0]->colocated } );
608 next unless @noncolo;
609 foreach my $nc ( @noncolo ) {
610 unless( $self->relations->verify_or_delete( $rdg, $nc ) ) {
611 push( @deleted_relations, [ $rdg->id, $nc->id ] );
616 return ( $newr, @deleted_relations );
619 sub _generate_dup_id {
620 my( $self, $rid ) = @_;
625 if( $self->has_reading( $newid ) ) {
638 # We only need the IDs for adding paths to the graph, not the reading
639 # objects themselves.
640 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
642 $self->_graphcalc_done(0);
643 # Connect the readings
644 unless( $self->sequence->has_edge( $source, $target ) ) {
645 $self->sequence->add_edge( $source, $target );
646 $self->relations->add_equivalence_edge( $source, $target );
648 # Note the witness in question
649 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
655 if( ref( $_[0] ) eq 'ARRAY' ) {
662 # We only need the IDs for removing paths from the graph, not the reading
663 # objects themselves.
664 my( $source, $target, $wit ) = $self->_stringify_args( @args );
666 $self->_graphcalc_done(0);
667 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
668 $self->sequence->delete_edge_attribute( $source, $target, $wit );
670 unless( $self->sequence->has_edge_attributes( $source, $target ) ) {
671 $self->sequence->delete_edge( $source, $target );
672 $self->relations->delete_equivalence_edge( $source, $target );
677 # Extra graph-alike utility
680 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
681 return undef unless $self->sequence->has_edge( $source, $target );
682 return $self->sequence->has_edge_attribute( $source, $target, $wit );
685 =head2 clear_witness( @sigil_list )
687 Clear the given witnesses out of the collation entirely, removing references
688 to them in paths, and removing readings that belong only to them. Should only
689 be called via $tradition->del_witness.
694 my( $self, @sigils ) = @_;
696 $self->_graphcalc_done(0);
697 # Clear the witness(es) out of the paths
698 foreach my $e ( $self->paths ) {
699 foreach my $sig ( @sigils ) {
700 $self->del_path( $e, $sig );
704 # Clear out the newly unused readings
705 foreach my $r ( $self->readings ) {
706 unless( $self->reading_witnesses( $r ) ) {
707 $self->del_reading( $r );
712 sub add_relationship {
714 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
715 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
716 foreach my $v ( @vectors ) {
717 next unless $self->get_relationship( $v )->colocated;
718 if( $self->reading( $v->[0] )->has_rank && $self->reading( $v->[1] )->has_rank
719 && $self->reading( $v->[0] )->rank ne $self->reading( $v->[1] )->rank ) {
720 $self->_graphcalc_done(0);
728 around qw/ get_relationship del_relationship / => sub {
732 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
735 my @stringargs = $self->_stringify_args( @args );
736 $self->$orig( @stringargs );
739 =head2 reading_witnesses( $reading )
741 Return a list of sigils corresponding to the witnesses in which the reading appears.
745 sub reading_witnesses {
746 my( $self, $reading ) = @_;
747 # We need only check either the incoming or the outgoing edges; I have
748 # arbitrarily chosen "incoming". Thus, special-case the start node.
749 if( $reading eq $self->start ) {
750 return map { $_->sigil } grep { $_->is_collated } $self->tradition->witnesses;
753 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
754 my $wits = $self->sequence->get_edge_attributes( @$e );
755 @all_witnesses{ keys %$wits } = 1;
757 my $acstr = $self->ac_label;
758 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
759 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
761 return keys %all_witnesses;
764 =head1 OUTPUT METHODS
766 =head2 as_svg( \%options )
768 Returns an SVG string that represents the graph, via as_dot and graphviz.
769 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
774 my( $self, $opts ) = @_;
775 throw( "Need GraphViz installed to output SVG" )
776 unless File::Which::which( 'dot' );
777 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
778 $self->calculate_ranks()
779 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
780 my @cmd = qw/dot -Tsvg/;
782 my $dotfile = File::Temp->new();
784 # $dotfile->unlink_on_destroy(0);
785 binmode $dotfile, ':utf8';
786 print $dotfile $self->as_dot( $opts );
787 push( @cmd, $dotfile->filename );
788 run( \@cmd, ">", binary(), \$svg );
789 $svg = decode_utf8( $svg );
794 =head2 as_dot( \%options )
796 Returns a string that is the collation graph expressed in dot
797 (i.e. GraphViz) format. Options include:
812 my( $self, $opts ) = @_;
813 my $startrank = $opts->{'from'} if $opts;
814 my $endrank = $opts->{'to'} if $opts;
815 my $color_common = $opts->{'color_common'} if $opts;
816 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
817 && $self->end->rank > 100;
818 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
820 # Check the arguments
822 return if $endrank && $startrank > $endrank;
823 return if $startrank > $self->end->rank;
825 if( defined $endrank ) {
826 return if $endrank < 0;
827 $endrank = undef if $endrank == $self->end->rank;
830 my $graph_name = $self->tradition->name;
831 $graph_name =~ s/[^\w\s]//g;
832 $graph_name = join( '_', split( /\s+/, $graph_name ) );
840 'fillcolor' => 'white',
845 'arrowhead' => 'open',
846 'color' => '#000000',
847 'fontcolor' => '#000000',
850 my $dot = sprintf( "digraph %s {\n", $graph_name );
851 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
852 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
854 # Output substitute start/end readings if necessary
856 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
859 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
861 if( $STRAIGHTENHACK ) {
863 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
864 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
865 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
867 my %used; # Keep track of the readings that actually appear in the graph
868 # Sort the readings by rank if we have ranks; this speeds layout.
869 my @all_readings = $self->end->has_rank
870 ? sort { $a->rank <=> $b->rank } $self->readings
872 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
873 foreach my $reading ( @all_readings ) {
874 # Only output readings within our rank range.
875 next if $startrank && $reading->rank < $startrank;
876 next if $endrank && $reading->rank > $endrank;
877 $used{$reading->id} = 1;
878 # Need not output nodes without separate labels
879 next if $reading->id eq $reading->text;
881 my $label = $reading->text;
882 $label .= '-' if $reading->join_next;
883 $label = "-$label" if $reading->join_prior;
884 $label =~ s/\"/\\\"/g;
885 $rattrs->{'label'} = $label;
886 $rattrs->{'id'} = $reading->id;
887 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
888 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
891 # Add the real edges. Need to weight one edge per rank jump, in a
893 # my $weighted = $self->_add_edge_weights;
894 my @edges = $self->paths;
895 my( %substart, %subend );
896 foreach my $edge ( @edges ) {
897 # Do we need to output this edge?
898 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
899 my $label = $self->_path_display_label( $opts,
900 $self->path_witnesses( $edge ) );
901 my $variables = { %edge_attrs, 'label' => $label };
903 # Account for the rank gap if necessary
904 my $rank0 = $self->reading( $edge->[0] )->rank
905 if $self->reading( $edge->[0] )->has_rank;
906 my $rank1 = $self->reading( $edge->[1] )->rank
907 if $self->reading( $edge->[1] )->has_rank;
908 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
909 $variables->{'minlen'} = $rank1 - $rank0;
912 # Add the calculated edge weights
913 # if( exists $weighted->{$edge->[0]}
914 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
915 # # $variables->{'color'} = 'red';
916 # $variables->{'weight'} = 3.0;
919 # EXPERIMENTAL: make edge width reflect no. of witnesses
920 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
921 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
923 my $varopts = _dot_attr_string( $variables );
924 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
925 $edge->[0], $edge->[1], $varopts );
926 } elsif( $used{$edge->[0]} ) {
927 $subend{$edge->[0]} = $edge->[1];
928 } elsif( $used{$edge->[1]} ) {
929 $substart{$edge->[1]} = $edge->[0];
933 # If we are asked to, add relationship links
934 if( exists $opts->{show_relations} ) {
935 my $filter = $opts->{show_relations}; # can be 'transposition' or 'all'
936 if( $filter eq 'transposition' ) {
937 $filter =~ qr/^transposition$/;
939 foreach my $redge ( $self->relationships ) {
940 if( $used{$redge->[0]} && $used{$redge->[1]} ) {
941 if( $filter ne 'all' ) {
942 my $rel = $self->get_relationship( $redge );
943 next unless $rel->type =~ /$filter/;
947 constraint => 'false',
948 label => uc( substr( $rel->type, 0, 4 ) ),
951 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
952 $redge->[0], $redge->[1], _dot_attr_string( $variables ) );
958 # Add substitute start and end edges if necessary
959 foreach my $node ( keys %substart ) {
960 my $witstr = $self->_path_display_label( $opts,
961 $self->path_witnesses( $substart{$node}, $node ) );
962 my $variables = { %edge_attrs, 'label' => $witstr };
963 my $nrdg = $self->reading( $node );
964 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
965 # Substart is actually one lower than $startrank
966 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
968 my $varopts = _dot_attr_string( $variables );
969 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
971 foreach my $node ( keys %subend ) {
972 my $witstr = $self->_path_display_label( $opts,
973 $self->path_witnesses( $node, $subend{$node} ) );
974 my $variables = { %edge_attrs, 'label' => $witstr };
975 my $varopts = _dot_attr_string( $variables );
976 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
979 if( $STRAIGHTENHACK ) {
980 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
981 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
988 sub _dot_attr_string {
991 foreach my $k ( sort keys %$hash ) {
993 push( @attrs, $k.'="'.$v.'"' );
995 return( '[ ' . join( ', ', @attrs ) . ' ]' );
998 sub _add_edge_weights {
1000 # Walk the graph from START to END, choosing the successor node with
1001 # the largest number of witness paths each time.
1003 my $curr = $self->start->id;
1004 my $ranked = $self->end->has_rank;
1005 while( $curr ne $self->end->id ) {
1006 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
1007 my @succ = sort { $self->path_witnesses( $curr, $a )
1008 <=> $self->path_witnesses( $curr, $b ) }
1009 $self->sequence->successors( $curr );
1010 my $next = pop @succ;
1011 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
1012 # Try to avoid lacunae in the weighted path.
1014 ( $self->reading( $next )->is_lacuna ||
1015 $nextrank - $rank > 1 ) ){
1018 $weighted->{$curr} = $next;
1024 =head2 path_witnesses( $edge )
1026 Returns the list of sigils whose witnesses are associated with the given edge.
1027 The edge can be passed as either an array or an arrayref of ( $source, $target ).
1031 sub path_witnesses {
1032 my( $self, @edge ) = @_;
1033 # If edge is an arrayref, cope.
1034 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
1035 my $e = shift @edge;
1038 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
1042 # Helper function. Make a display label for the given witnesses, showing a.c.
1043 # witnesses only where the main witness is not also in the list.
1044 sub _path_display_label {
1048 map { $wits{$_} = 1 } @_;
1050 # If an a.c. wit is listed, remove it if the main wit is also listed.
1051 # Otherwise keep it for explicit listing.
1052 my $aclabel = $self->ac_label;
1054 foreach my $w ( sort keys %wits ) {
1055 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
1056 if( exists $wits{$1} ) {
1059 push( @disp_ac, $w );
1064 if( $opts->{'explicit_wits'} ) {
1065 return join( ', ', sort keys %wits );
1067 # See if we are in a majority situation.
1068 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
1069 $maj = $maj > 5 ? $maj : 5;
1070 if( scalar keys %wits > $maj ) {
1071 unshift( @disp_ac, 'majority' );
1072 return join( ', ', @disp_ac );
1074 return join( ', ', sort keys %wits );
1079 =head2 readings_at_rank( $rank )
1081 Returns a list of readings at a given rank, taken from the alignment table.
1085 sub readings_at_rank {
1086 my( $self, $rank ) = @_;
1087 my $table = $self->alignment_table;
1088 # Table rank is real rank - 1.
1089 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
1091 foreach my $e ( @elements ) {
1092 next unless ref( $e ) eq 'HASH';
1093 next unless exists $e->{'t'};
1094 $readings{$e->{'t'}->id} = $e->{'t'};
1096 return values %readings;
1101 Returns a GraphML representation of the collation. The GraphML will contain
1102 two graphs. The first expresses the attributes of the readings and the witness
1103 paths that link them; the second expresses the relationships that link the
1104 readings. This is the native transfer format for a tradition.
1108 use Text::Tradition;
1114 my $datafile = 't/data/florilegium_tei_ps.xml';
1115 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1117 'file' => $datafile,
1120 ok( $tradition, "Got a tradition object" );
1121 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
1122 ok( $tradition->collation, "Tradition has a collation" );
1124 my $c = $tradition->collation;
1125 is( scalar $c->readings, $READINGS, "Collation has all readings" );
1126 is( scalar $c->paths, $PATHS, "Collation has all paths" );
1127 is( scalar $c->relationships, 0, "Collation has all relationships" );
1129 # Add a few relationships
1130 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
1131 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
1132 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
1134 # Now write it to GraphML and parse it again.
1136 my $graphml = $c->as_graphml;
1137 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
1138 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
1139 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
1140 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
1142 # Now add a stemma, write to GraphML, and look at the output.
1144 skip "Analysis module not present", 3 unless $tradition->can( 'add_stemma' );
1145 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
1146 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
1147 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
1148 $graphml = $c->as_graphml;
1149 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
1156 ## TODO MOVE this to Tradition.pm and modularize it better
1158 my( $self, $options ) = @_;
1159 $self->calculate_ranks unless $self->_graphcalc_done;
1161 my $start = $options->{'from'}
1162 ? $self->reading( $options->{'from'} ) : $self->start;
1163 my $end = $options->{'to'}
1164 ? $self->reading( $options->{'to'} ) : $self->end;
1165 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1166 throw( 'Start node must be before end node' );
1168 # The readings need to be ranked for this to work.
1169 $start = $self->start unless $start->has_rank;
1170 $end = $self->end unless $end->has_rank;
1172 unless( $start eq $self->start ) {
1173 $rankoffset = $start->rank - 1;
1178 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1179 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1180 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1181 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1183 # Create the document and root node
1184 require XML::LibXML;
1185 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1186 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1187 $graphml->setDocumentElement( $root );
1188 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1189 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1191 # List of attribute types to save on our objects and their corresponding
1196 'Bool' => 'boolean',
1197 'ReadingID' => 'string',
1198 'RelationshipType' => 'string',
1199 'RelationshipScope' => 'string',
1202 # Add the data keys for the graph. Include an extra key 'version' for the
1203 # GraphML output version.
1204 my %graph_data_keys;
1206 my %graph_attributes = ( 'version' => 'string' );
1207 # Graph attributes include those of Tradition and those of Collation.
1209 # TODO Use meta introspection method from duplicate_reading to do this
1210 # instead of naming custom keys.
1211 my $tmeta = $self->tradition->meta;
1212 my $cmeta = $self->meta;
1213 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1214 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1215 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1216 next if $attr->name =~ /^_/;
1217 next unless $save_types{$attr->type_constraint->name};
1218 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1220 # Extra custom keys for complex objects that should be saved in some form.
1221 # The subroutine should return a string, or undef/empty.
1222 if( $tmeta->has_method('stemmata') ) {
1223 $graph_attributes{'stemmata'} = sub {
1225 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1226 $self->tradition->stemmata;
1227 join( "\n", @stemstrs );
1231 if( $tmeta->has_method('user') ) {
1232 $graph_attributes{'user'} = sub {
1233 $self->tradition->user ? $self->tradition->user->id : undef
1237 foreach my $datum ( sort keys %graph_attributes ) {
1238 $graph_data_keys{$datum} = 'dg'.$gdi++;
1239 my $key = $root->addNewChild( $graphml_ns, 'key' );
1240 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1241 : $graph_attributes{$datum};
1242 $key->setAttribute( 'attr.name', $datum );
1243 $key->setAttribute( 'attr.type', $dtype );
1244 $key->setAttribute( 'for', 'graph' );
1245 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1248 # Add the data keys for reading nodes
1249 my %reading_attributes;
1250 my $rmeta = Text::Tradition::Collation::Reading->meta;
1251 foreach my $attr( $rmeta->get_all_attributes ) {
1252 next if $attr->name =~ /^_/;
1253 next unless $save_types{$attr->type_constraint->name};
1254 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1256 if( $self->start->does('Text::Tradition::Morphology' ) ) {
1257 # Extra custom key for the reading morphology
1258 $reading_attributes{'lexemes'} = 'string';
1263 foreach my $datum ( sort keys %reading_attributes ) {
1264 $node_data_keys{$datum} = 'dn'.$ndi++;
1265 my $key = $root->addNewChild( $graphml_ns, 'key' );
1266 $key->setAttribute( 'attr.name', $datum );
1267 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1268 $key->setAttribute( 'for', 'node' );
1269 $key->setAttribute( 'id', $node_data_keys{$datum} );
1272 # Add the data keys for edges, that is, paths and relationships. Path
1273 # data does not come from a Moose class so is here manually.
1276 my %edge_attributes = (
1277 witness => 'string', # ID/label for a path
1278 extra => 'boolean', # Path key
1280 my @path_attributes = keys %edge_attributes; # track our manual additions
1281 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1282 foreach my $attr( $pmeta->get_all_attributes ) {
1283 next if $attr->name =~ /^_/;
1284 next unless $save_types{$attr->type_constraint->name};
1285 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1287 foreach my $datum ( sort keys %edge_attributes ) {
1288 $edge_data_keys{$datum} = 'de'.$edi++;
1289 my $key = $root->addNewChild( $graphml_ns, 'key' );
1290 $key->setAttribute( 'attr.name', $datum );
1291 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1292 $key->setAttribute( 'for', 'edge' );
1293 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1296 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1297 my $xmlidname = $self->tradition->name;
1298 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1299 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1300 $xmlidname = '_'.$xmlidname;
1302 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1303 $sgraph->setAttribute( 'edgedefault', 'directed' );
1304 $sgraph->setAttribute( 'id', $xmlidname );
1305 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1306 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1307 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1308 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1309 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1311 # Tradition/collation attribute data
1312 foreach my $datum ( keys %graph_attributes ) {
1314 if( $datum eq 'version' ) {
1316 } elsif( ref( $graph_attributes{$datum} ) ) {
1317 my $sub = $graph_attributes{$datum};
1319 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1320 $value = $self->tradition->$datum;
1322 $value = $self->$datum;
1324 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1329 # Add our readings to the graph
1330 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1331 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1332 ( $n->rank < $start->rank || $n->rank > $end->rank );
1333 $use_readings{$n->id} = 1;
1334 # Add to the main graph
1335 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1336 my $node_xmlid = 'n' . $node_ctr++;
1337 $node_hash{ $n->id } = $node_xmlid;
1338 $node_el->setAttribute( 'id', $node_xmlid );
1339 foreach my $d ( keys %reading_attributes ) {
1341 # Custom serialization
1342 if( $d eq 'lexemes' ) {
1343 # If nval is a true value, we have lexemes so we need to
1344 # serialize them. Otherwise set nval to undef so that the
1345 # key is excluded from this reading.
1346 $nval = $nval ? $n->_serialize_lexemes : undef;
1347 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1350 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1351 # Adjust the ranks within the subgraph.
1352 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1353 : $nval - $rankoffset;
1355 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1360 # Add the path edges to the sequence graph
1362 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1363 # We add an edge in the graphml for every witness in $e.
1364 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1365 my @edge_wits = sort $self->path_witnesses( $e );
1366 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1367 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1368 # Skip any path from start to end; that witness is not in the subgraph.
1369 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1370 foreach my $wit ( @edge_wits ) {
1371 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1372 $node_hash{ $e->[0] },
1373 $node_hash{ $e->[1] } );
1374 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1375 $edge_el->setAttribute( 'source', $from );
1376 $edge_el->setAttribute( 'target', $to );
1377 $edge_el->setAttribute( 'id', $id );
1379 # It's a witness path, so add the witness
1381 my $key = $edge_data_keys{'witness'};
1382 # Is this an ante-corr witness?
1383 my $aclabel = $self->ac_label;
1384 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1385 # Keep the base witness
1387 # ...and record that this is an 'extra' reading path
1388 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1390 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1394 # Report the actual number of nodes and edges that went in
1395 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1396 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1398 # Add the relationship graph to the XML
1399 map { delete $edge_data_keys{$_} } @path_attributes;
1400 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1401 $node_data_keys{'id'}, \%edge_data_keys );
1403 # Save and return the thing
1404 my $result = decode_utf8( $graphml->toString(1) );
1408 sub _add_graphml_data {
1409 my( $el, $key, $value ) = @_;
1410 return unless defined $value;
1411 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1412 $data_el->setAttribute( 'key', $key );
1413 $data_el->appendText( $value );
1418 Returns a CSV alignment table representation of the collation graph, one
1419 row per witness (or witness uncorrected.)
1423 Returns a tab-separated alignment table representation of the collation graph,
1424 one row per witness (or witness uncorrected.)
1428 use Text::Tradition;
1436 my $datafile = 't/data/florilegium_tei_ps.xml';
1437 my $tradition = Text::Tradition->new( 'input' => 'TEI',
1439 'file' => $datafile,
1442 my $c = $tradition->collation;
1443 # Export the thing to CSV
1444 my $csvstr = $c->as_csv();
1446 my $csv = Text::CSV->new({ sep_char => ',', binary => 1 });
1447 my @lines = split(/\n/, $csvstr );
1448 ok( $csv->parse( $lines[0] ), "Successfully parsed first line of CSV" );
1449 is( scalar( $csv->fields ), $WITS + $WITAC, "CSV has correct number of witness columns" );
1450 my @q_ac = grep { $_ eq 'Q'.$c->ac_label } $csv->fields;
1451 ok( @q_ac, "Found a layered witness" );
1453 my $t2 = Text::Tradition->new( input => 'Tabular',
1457 is( scalar $t2->collation->readings, $READINGS, "Reparsed CSV collation has all readings" );
1458 is( scalar $t2->collation->paths, $PATHS, "Reparsed CSV collation has all paths" );
1460 # Now do it with TSV
1461 my $tsvstr = $c->as_tsv();
1462 my $t3 = Text::Tradition->new( input => 'Tabular',
1466 is( scalar $t3->collation->readings, $READINGS, "Reparsed TSV collation has all readings" );
1467 is( scalar $t3->collation->paths, $PATHS, "Reparsed TSV collation has all paths" );
1469 my $table = $c->alignment_table;
1470 my $noaccsv = $c->as_csv({ noac => 1 });
1471 my @noaclines = split(/\n/, $noaccsv );
1472 ok( $csv->parse( $noaclines[0] ), "Successfully parsed first line of no-ac CSV" );
1473 is( scalar( $csv->fields ), $WITS, "CSV has correct number of witness columns" );
1474 is( $c->alignment_table, $table, "Request for CSV did not alter the alignment table" );
1476 my $safecsv = $c->as_csv({ safe_ac => 1});
1477 my @safelines = split(/\n/, $safecsv );
1478 ok( $csv->parse( $safelines[0] ), "Successfully parsed first line of safe CSV" );
1479 is( scalar( $csv->fields ), $WITS + $WITAC, "CSV has correct number of witness columns" );
1480 @q_ac = grep { $_ eq 'Q__L' } $csv->fields;
1481 ok( @q_ac, "Found a sanitized layered witness" );
1482 is( $c->alignment_table, $table, "Request for CSV did not alter the alignment table" );
1489 my( $self, $opts ) = @_;
1490 my $table = $self->alignment_table( $opts );
1491 my $csv_options = { binary => 1, quote_null => 0 };
1492 $csv_options->{'sep_char'} = $opts->{fieldsep};
1493 if( $opts->{fieldsep} eq "\t" ) {
1494 # If it is really tab separated, nothing is an escape char.
1495 $csv_options->{'quote_char'} = undef;
1496 $csv_options->{'escape_char'} = '';
1498 my $csv = Text::CSV->new( $csv_options );
1500 # Make the header row
1501 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1502 push( @result, $csv->string );
1503 # Make the rest of the rows
1504 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1505 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1506 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1507 $csv->combine( @row );
1508 push( @result, $csv->string );
1510 return join( "\n", @result );
1515 my $opts = shift || {};
1516 $opts->{fieldsep} = ',';
1517 return $self->_tabular( $opts );
1522 my $opts = shift || {};
1523 $opts->{fieldsep} = "\t";
1524 return $self->_tabular( $opts );
1527 =head2 alignment_table
1529 Return a reference to an alignment table, in a slightly enhanced CollateX
1530 format which looks like this:
1532 $table = { alignment => [ { witness => "SIGIL",
1533 tokens => [ { t => "TEXT" }, ... ] },
1534 { witness => "SIG2",
1535 tokens => [ { t => "TEXT" }, ... ] },
1537 length => TEXTLEN };
1541 sub alignment_table {
1542 my( $self, $opts ) = @_;
1543 if( $self->has_cached_table ) {
1544 return $self->cached_table
1545 unless $opts->{noac} || $opts->{safe_ac};
1548 # Make sure we can do this
1549 throw( "Need a linear graph in order to make an alignment table" )
1550 unless $self->linear;
1551 $self->calculate_ranks()
1552 unless $self->_graphcalc_done && $self->end->has_rank;
1554 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1555 my @all_pos = ( 1 .. $self->end->rank - 1 );
1556 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1557 # say STDERR "Making witness row(s) for " . $wit->sigil;
1558 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1559 my @row = _make_witness_row( \@wit_path, \@all_pos );
1560 my $witobj = { 'witness' => $wit->sigil, 'tokens' => \@row };
1561 $witobj->{'identifier'} = $wit->identifier if $wit->identifier;
1562 push( @{$table->{'alignment'}}, $witobj );
1563 if( $wit->is_layered && !$opts->{noac} ) {
1564 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1565 $wit->sigil.$self->ac_label );
1566 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1567 my $witlabel = $opts->{safe_ac}
1568 ? $wit->sigil . '__L' : $wit->sigil.$self->ac_label;
1569 my $witacobj = { 'witness' => $witlabel,
1570 'tokens' => \@ac_row };
1571 $witacobj->{'identifier'} = $wit->identifier if $wit->identifier;
1572 push( @{$table->{'alignment'}}, $witacobj );
1575 unless( $opts->{noac} || $opts->{safe_ac} ) {
1576 $self->cached_table( $table );
1581 sub _make_witness_row {
1582 my( $path, $positions ) = @_;
1584 map { $char_hash{$_} = undef } @$positions;
1586 foreach my $rdg ( @$path ) {
1587 say STDERR "rank " . $rdg->rank if $debug;
1588 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1589 $char_hash{$rdg->rank} = { 't' => $rdg };
1591 my @row = map { $char_hash{$_} } @$positions;
1592 # Fill in lacuna markers for undef spots in the row
1593 my $last_el = shift @row;
1594 my @filled_row = ( $last_el );
1595 foreach my $el ( @row ) {
1596 # If we are using node reference, make the lacuna node appear many times
1597 # in the table. If not, use the lacuna tag.
1598 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1601 push( @filled_row, $el );
1608 =head1 NAVIGATION METHODS
1610 =head2 reading_sequence( $first, $last, $sigil, $backup )
1612 Returns the ordered list of readings, starting with $first and ending
1613 with $last, for the witness given in $sigil. If a $backup sigil is
1614 specified (e.g. when walking a layered witness), it will be used wherever
1615 no $sigil path exists. If there is a base text reading, that will be
1616 used wherever no path exists for $sigil or $backup.
1620 # TODO Think about returning some lazy-eval iterator.
1621 # TODO Get rid of backup; we should know from what witness is whether we need it.
1623 sub reading_sequence {
1624 my( $self, $start, $end, $witness ) = @_;
1626 $witness = $self->baselabel unless $witness;
1627 my @readings = ( $start );
1630 while( $n && $n->id ne $end->id ) {
1631 if( exists( $seen{$n->id} ) ) {
1632 throw( "Detected loop for $witness at " . $n->id );
1636 my $next = $self->next_reading( $n, $witness );
1638 throw( "Did not find any path for $witness from reading " . $n->id );
1640 push( @readings, $next );
1643 # Check that the last reading is our end reading.
1644 my $last = $readings[$#readings];
1645 throw( "Last reading found from " . $start->text .
1646 " for witness $witness is not the end!" ) # TODO do we get this far?
1647 unless $last->id eq $end->id;
1652 =head2 next_reading( $reading, $sigil );
1654 Returns the reading that follows the given reading along the given witness
1660 # Return the successor via the corresponding path.
1662 my $answer = $self->_find_linked_reading( 'next', @_ );
1663 return undef unless $answer;
1664 return $self->reading( $answer );
1667 =head2 prior_reading( $reading, $sigil )
1669 Returns the reading that precedes the given reading along the given witness
1675 # Return the predecessor via the corresponding path.
1677 my $answer = $self->_find_linked_reading( 'prior', @_ );
1678 return $self->reading( $answer );
1681 sub _find_linked_reading {
1682 my( $self, $direction, $node, $path ) = @_;
1684 # Get a backup if we are dealing with a layered witness
1686 my $aclabel = $self->ac_label;
1687 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1691 my @linked_paths = $direction eq 'next'
1692 ? $self->sequence->edges_from( $node )
1693 : $self->sequence->edges_to( $node );
1694 return undef unless scalar( @linked_paths );
1696 # We have to find the linked path that contains all of the
1697 # witnesses supplied in $path.
1698 my( @path_wits, @alt_path_wits );
1699 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1700 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1703 foreach my $le ( @linked_paths ) {
1704 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1707 my @le_wits = sort $self->path_witnesses( $le );
1708 if( _is_within( \@path_wits, \@le_wits ) ) {
1709 # This is the right path.
1710 return $direction eq 'next' ? $le->[1] : $le->[0];
1711 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1715 # Got this far? Return the alternate path if it exists.
1716 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1719 # Got this far? Return the base path if it exists.
1720 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1723 # Got this far? We have no appropriate path.
1724 warn "Could not find $direction node from " . $node->id
1725 . " along path $path";
1731 my( $set1, $set2 ) = @_;
1732 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1733 foreach my $el ( @$set1 ) {
1734 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1739 # Return the string that joins together a list of witnesses for
1740 # display on a single path.
1741 sub _witnesses_of_label {
1742 my( $self, $label ) = @_;
1743 my $regex = $self->wit_list_separator;
1744 my @answer = split( /\Q$regex\E/, $label );
1748 =head2 common_readings
1750 Returns the list of common readings in the graph (i.e. those readings that are
1751 shared by all non-lacunose witnesses.)
1755 sub common_readings {
1757 my @common = grep { $_->is_common } $self->readings;
1761 =head2 path_text( $sigil, [, $start, $end ] )
1763 Returns the text of a witness (plus its backup, if we are using a layer)
1764 as stored in the collation. The text is returned as a string, where the
1765 individual readings are joined with spaces and the meta-readings (e.g.
1766 lacunae) are omitted. Optional specification of $start and $end allows
1767 the generation of a subset of the witness text.
1772 my( $self, $wit, $start, $end ) = @_;
1773 $start = $self->start unless $start;
1774 $end = $self->end unless $end;
1775 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1778 foreach my $r ( @path ) {
1779 unless ( $r->join_prior || !$last || $last->join_next ) {
1782 $pathtext .= $r->text;
1788 =head1 INITIALIZATION METHODS
1790 These are mostly for use by parsers.
1792 =head2 make_witness_path( $witness )
1794 Link the array of readings contained in $witness->path (and in
1795 $witness->uncorrected_path if it exists) into collation paths.
1796 Clear out the arrays when finished.
1798 =head2 make_witness_paths
1800 Call make_witness_path for all witnesses in the tradition.
1804 # For use when a collation is constructed from a base text and an apparatus.
1805 # We have the sequences of readings and just need to add path edges.
1806 # When we are done, clear out the witness path attributes, as they are no
1808 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1810 sub make_witness_paths {
1812 foreach my $wit ( $self->tradition->witnesses ) {
1813 # say STDERR "Making path for " . $wit->sigil;
1814 $self->make_witness_path( $wit );
1818 sub make_witness_path {
1819 my( $self, $wit ) = @_;
1820 my @chain = @{$wit->path};
1821 my $sig = $wit->sigil;
1822 # Add start and end if necessary
1823 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1824 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1825 foreach my $idx ( 0 .. $#chain-1 ) {
1826 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1828 if( $wit->is_layered ) {
1829 @chain = @{$wit->uncorrected_path};
1830 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1831 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1832 foreach my $idx( 0 .. $#chain-1 ) {
1833 my $source = $chain[$idx];
1834 my $target = $chain[$idx+1];
1835 $self->add_path( $source, $target, $sig.$self->ac_label )
1836 unless $self->has_path( $source, $target, $sig );
1840 $wit->clear_uncorrected_path;
1843 =head2 calculate_ranks
1845 Calculate the reading ranks (that is, their aligned positions relative
1846 to each other) for the graph. This can only be called on linear collations.
1850 use Text::Tradition;
1852 my $cxfile = 't/data/Collatex-16.xml';
1853 my $t = Text::Tradition->new(
1855 'input' => 'CollateX',
1858 my $c = $t->collation;
1861 my $table = $c->alignment_table;
1862 ok( $c->has_cached_table, "Alignment table was cached" );
1863 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1864 $c->calculate_ranks;
1865 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1866 $c->add_relationship( 'n13', 'n23', { type => 'repetition' } );
1867 is( $c->alignment_table, $table, "Alignment table unchanged after non-colo relationship add" );
1868 $c->add_relationship( 'n24', 'n23', { type => 'spelling' } );
1869 isnt( $c->alignment_table, $table, "Alignment table changed after colo relationship add" );
1875 sub calculate_ranks {
1877 # Save the existing ranks, in case we need to invalidate the cached SVG.
1879 map { $existing_ranks{$_} = $_->rank } $self->readings;
1881 # Do the rankings based on the relationship equivalence graph, starting
1882 # with the start node.
1883 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1885 # Transfer our rankings from the topological graph to the real one.
1886 foreach my $r ( $self->readings ) {
1887 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1888 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1890 # Die. Find the last rank we calculated.
1891 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1892 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1894 my $last = pop @all_defined;
1895 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1898 # Do we need to invalidate the cached data?
1899 if( $self->has_cached_table ) {
1900 foreach my $r ( $self->readings ) {
1901 next if defined( $existing_ranks{$r} )
1902 && $existing_ranks{$r} == $r->rank;
1903 # Something has changed, so clear the cache
1904 $self->_clear_cache;
1905 # ...and recalculate the common readings.
1906 $self->calculate_common_readings();
1910 # The graph calculation information is now up to date.
1911 $self->_graphcalc_done(1);
1916 $self->wipe_table if $self->has_cached_table;
1920 =head2 flatten_ranks
1922 A convenience method for parsing collation data. Searches the graph for readings
1923 with the same text at the same rank, and merges any that are found.
1928 my ( $self, %args ) = shift;
1929 my %unique_rank_rdg;
1931 foreach my $p ( $self->identical_readings( %args ) ) {
1932 # say STDERR "Combining readings at same rank: @$p";
1934 $self->merge_readings( @$p );
1935 # TODO see if this now makes a common point.
1937 # If we merged readings, the ranks are still fine but the alignment
1938 # table is wrong. Wipe it.
1939 $self->wipe_table() if $changed;
1942 =head2 identical_readings
1943 =head2 identical_readings( start => $startnode, end => $endnode )
1944 =head2 identical_readings( startrank => $startrank, endrank => $endrank )
1946 Goes through the graph identifying all pairs of readings that appear to be
1947 identical, and therefore able to be merged into a single reading. Returns the
1948 relevant identical pairs. Can be restricted to run over only a part of the
1949 graph, specified either by node or by rank.
1953 sub identical_readings {
1954 my ( $self, %args ) = @_;
1955 # Find where we should start and end.
1956 my $startrank = $args{startrank} || 0;
1957 if( $args{start} ) {
1958 throw( "Starting reading has no rank" ) unless $self->reading( $args{start} )
1959 && $self->reading( $args{start} )->has_rank;
1960 $startrank = $self->reading( $args{start} )->rank;
1962 my $endrank = $args{endrank} || $self->end->rank;
1964 throw( "Ending reading has no rank" ) unless $self->reading( $args{end} )
1965 && $self->reading( $args{end} )->has_rank;
1966 $endrank = $self->reading( $args{end} )->rank;
1969 # Make sure the ranks are correct.
1970 unless( $self->_graphcalc_done ) {
1971 $self->calculate_ranks;
1973 # Go through the readings looking for duplicates.
1974 my %unique_rank_rdg;
1976 foreach my $rdg ( $self->readings ) {
1977 next unless $rdg->has_rank;
1978 my $rk = $rdg->rank;
1979 next if $rk > $endrank || $rk < $startrank;
1980 my $key = $rk . "||" . $rdg->text;
1981 if( exists $unique_rank_rdg{$key} ) {
1982 # Make sure they don't have different grammatical forms
1983 my $ur = $unique_rank_rdg{$key};
1984 if( $rdg->is_identical( $ur ) ) {
1985 push( @pairs, [ $ur, $rdg ] );
1988 $unique_rank_rdg{$key} = $rdg;
1996 =head2 calculate_common_readings
1998 Goes through the graph identifying the readings that appear in every witness
1999 (apart from those with lacunae at that spot.) Marks them as common and returns
2004 use Text::Tradition;
2006 my $cxfile = 't/data/Collatex-16.xml';
2007 my $t = Text::Tradition->new(
2009 'input' => 'CollateX',
2012 my $c = $t->collation;
2014 my @common = $c->calculate_common_readings();
2015 is( scalar @common, 8, "Found correct number of common readings" );
2016 my @marked = sort $c->common_readings();
2017 is( scalar @common, 8, "All common readings got marked as such" );
2018 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
2019 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
2025 sub calculate_common_readings {
2028 map { $_->is_common( 0 ) } $self->readings;
2029 # Implicitly calls calculate_ranks
2030 my $table = $self->alignment_table;
2031 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
2032 my @row = map { $_->{'tokens'}->[$idx]
2033 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
2034 @{$table->{'alignment'}};
2036 foreach my $r ( @row ) {
2038 $hash{$r->id} = $r unless $r->is_meta;
2040 $hash{'UNDEF'} = $r;
2043 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
2044 my( $r ) = values %hash;
2046 push( @common, $r );
2052 =head2 text_from_paths
2054 Calculate the text array for all witnesses from the path, for later consistency
2055 checking. Only to be used if there is no non-graph-based way to know the
2060 sub text_from_paths {
2062 foreach my $wit ( $self->tradition->witnesses ) {
2063 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
2065 foreach my $r ( @readings ) {
2066 next if $r->is_meta;
2067 push( @text, $r->text );
2069 $wit->text( \@text );
2070 if( $wit->is_layered ) {
2071 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
2072 $wit->sigil.$self->ac_label );
2074 foreach my $r ( @ucrdgs ) {
2075 next if $r->is_meta;
2076 push( @uctext, $r->text );
2078 $wit->layertext( \@uctext );
2083 =head1 UTILITY FUNCTIONS
2085 =head2 common_predecessor( $reading_a, $reading_b )
2087 Find the last reading that occurs in sequence before both the given readings.
2088 At the very least this should be $self->start.
2090 =head2 common_successor( $reading_a, $reading_b )
2092 Find the first reading that occurs in sequence after both the given readings.
2093 At the very least this should be $self->end.
2097 use Text::Tradition;
2099 my $cxfile = 't/data/Collatex-16.xml';
2100 my $t = Text::Tradition->new(
2102 'input' => 'CollateX',
2105 my $c = $t->collation;
2107 is( $c->common_predecessor( 'n24', 'n23' )->id,
2108 'n20', "Found correct common predecessor" );
2109 is( $c->common_successor( 'n24', 'n23' )->id,
2110 '__END__', "Found correct common successor" );
2112 is( $c->common_predecessor( 'n19', 'n17' )->id,
2113 'n16', "Found correct common predecessor for readings on same path" );
2114 is( $c->common_successor( 'n21', 'n10' )->id,
2115 '__END__', "Found correct common successor for readings on same path" );
2121 ## Return the closest reading that is a predecessor of both the given readings.
2122 sub common_predecessor {
2124 my( $r1, $r2 ) = $self->_objectify_args( @_ );
2125 return $self->_common_in_path( $r1, $r2, 'predecessors' );
2128 sub common_successor {
2130 my( $r1, $r2 ) = $self->_objectify_args( @_ );
2131 return $self->_common_in_path( $r1, $r2, 'successors' );
2135 # TODO think about how to do this without ranks...
2136 sub _common_in_path {
2137 my( $self, $r1, $r2, $dir ) = @_;
2138 my $iter = $self->end->rank;
2140 my @last_r1 = ( $r1 );
2141 my @last_r2 = ( $r2 );
2142 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
2144 # say STDERR "Finding common $dir for $r1, $r2";
2145 while( !@candidates ) {
2146 last unless $iter--; # Avoid looping infinitely
2147 # Iterate separately down the graph from r1 and r2
2148 my( @new_lc1, @new_lc2 );
2149 foreach my $lc ( @last_r1 ) {
2150 foreach my $p ( $lc->$dir ) {
2151 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
2152 # say STDERR "Path candidate $p from $lc";
2153 push( @candidates, $p );
2154 } elsif( !$all_seen{$p->id} ) {
2155 $all_seen{$p->id} = 'r1';
2156 push( @new_lc1, $p );
2160 foreach my $lc ( @last_r2 ) {
2161 foreach my $p ( $lc->$dir ) {
2162 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
2163 # say STDERR "Path candidate $p from $lc";
2164 push( @candidates, $p );
2165 } elsif( !$all_seen{$p->id} ) {
2166 $all_seen{$p->id} = 'r2';
2167 push( @new_lc2, $p );
2171 @last_r1 = @new_lc1;
2172 @last_r2 = @new_lc2;
2174 my @answer = sort { $a->rank <=> $b->rank } @candidates;
2175 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
2179 Text::Tradition::Error->throw(
2180 'ident' => 'Collation error',
2186 __PACKAGE__->meta->make_immutable;
2192 =item * Rework XML serialization in a more modular way
2198 This package is free software and is provided "as is" without express
2199 or implied warranty. You can redistribute it and/or modify it under
2200 the same terms as Perl itself.
2204 Tara L Andrews E<lt>aurum@cpan.orgE<gt>