1 package Text::Tradition::Collation;
4 use Encode qw( decode_utf8 );
8 use IPC::Run qw( run binary );
10 use Text::Tradition::Collation::Reading;
11 use Text::Tradition::Collation::RelationshipStore;
12 use Text::Tradition::Error;
13 use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx );
19 default => sub { Graph->new() },
27 isa => 'Text::Tradition::Collation::RelationshipStore',
29 relationships => 'relationships',
30 related_readings => 'related_readings',
31 get_relationship => 'get_relationship',
32 del_relationship => 'del_relationship',
33 equivalence => 'equivalence',
34 equivalence_graph => 'equivalence_graph',
36 writer => '_set_relations',
41 isa => 'Text::Tradition',
42 writer => '_set_tradition',
47 isa => 'HashRef[Text::Tradition::Collation::Reading]',
51 _add_reading => 'set',
52 del_reading => 'delete',
53 has_reading => 'exists',
56 default => sub { {} },
59 has 'wit_list_separator' => (
68 default => 'base text',
91 isa => 'Text::Tradition::Collation::Reading',
92 writer => '_set_start',
98 isa => 'Text::Tradition::Collation::Reading',
103 has 'cached_table' => (
106 predicate => 'has_cached_table',
107 clearer => 'wipe_table',
110 has '_graphcalc_done' => (
118 Text::Tradition::Collation - a software model for a text collation
123 my $t = Text::Tradition->new(
124 'name' => 'this is a text',
126 'file' => '/path/to/tei_parallel_seg_file.xml' );
128 my $c = $t->collation;
129 my @readings = $c->readings;
130 my @paths = $c->paths;
131 my @relationships = $c->relationships;
133 my $svg_variant_graph = $t->collation->as_svg();
137 Text::Tradition is a library for representation and analysis of collated
138 texts, particularly medieval ones. The Collation is the central feature of
139 a Tradition, where the text, its sequence of readings, and its relationships
140 between readings are actually kept.
146 The constructor. Takes a hash or hashref of the following arguments:
150 =item * tradition - The Text::Tradition object to which the collation
153 =item * linear - Whether the collation should be linear; that is, whether
154 transposed readings should be treated as two linked readings rather than one,
155 and therefore whether the collation graph is acyclic. Defaults to true.
157 =item * baselabel - The default label for the path taken by a base text
158 (if any). Defaults to 'base text'.
160 =item * wit_list_separator - The string to join a list of witnesses for
161 purposes of making labels in display graphs. Defaults to ', '.
163 =item * ac_label - The extra label to tack onto a witness sigil when
164 representing another layer of path for the given witness - that is, when
165 a text has more than one possible reading due to scribal corrections or
166 the like. Defaults to ' (a.c.)'.
168 =item * wordsep - The string used to separate words in the original text.
179 =head2 wit_list_separator
187 Simple accessors for collation attributes.
191 The meta-reading at the start of every witness path.
195 The meta-reading at the end of every witness path.
199 Returns all Reading objects in the graph.
201 =head2 reading( $id )
203 Returns the Reading object corresponding to the given ID.
205 =head2 add_reading( $reading_args )
207 Adds a new reading object to the collation.
208 See L<Text::Tradition::Collation::Reading> for the available arguments.
210 =head2 del_reading( $object_or_id )
212 Removes the given reading from the collation, implicitly removing its
213 paths and relationships.
215 =head2 merge_readings( $main, $second, $concatenate, $with_str )
217 Merges the $second reading into the $main one. If $concatenate is true, then
218 the merged node will carry the text of both readings, concatenated with either
219 $with_str (if specified) or a sensible default (the empty string if the
220 appropriate 'join_*' flag is set on either reading, or else $self->wordsep.)
222 The first two arguments may be either readings or reading IDs.
224 =head2 has_reading( $id )
226 Predicate to see whether a given reading ID is in the graph.
228 =head2 reading_witnesses( $object_or_id )
230 Returns a list of sigils whose witnesses contain the reading.
234 Returns all reading paths within the document - that is, all edges in the
235 collation graph. Each path is an arrayref of [ $source, $target ] reading IDs.
237 =head2 add_path( $source, $target, $sigil )
239 Links the given readings in the collation in sequence, under the given witness
240 sigil. The readings may be specified by object or ID.
242 =head2 del_path( $source, $target, $sigil )
244 Links the given readings in the collation in sequence, under the given witness
245 sigil. The readings may be specified by object or ID.
247 =head2 has_path( $source, $target );
249 Returns true if the two readings are linked in sequence in any witness.
250 The readings may be specified by object or ID.
254 Returns all Relationship objects in the collation.
256 =head2 add_relationship( $reading, $other_reading, $options )
258 Adds a new relationship of the type given in $options between the two readings,
259 which may be specified by object or ID. Returns a value of ( $status, @vectors)
260 where $status is true on success, and @vectors is a list of relationship edges
261 that were ultimately added.
262 See L<Text::Tradition::Collation::Relationship> for the available options.
268 $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) );
269 $self->_set_start( $self->add_reading(
270 { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) );
271 $self->_set_end( $self->add_reading(
272 { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) );
275 ### Reading construct/destruct functions
278 my( $self, $reading ) = @_;
279 unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) {
280 my %args = %$reading;
281 if( $args{'init'} ) {
282 # If we are initializing an empty collation, don't assume that we
283 # have set a tradition.
284 delete $args{'init'};
285 } elsif( $self->tradition->has_language && !exists $args{'language'} ) {
286 $args{'language'} = $self->tradition->language;
288 $reading = Text::Tradition::Collation::Reading->new(
289 'collation' => $self,
292 # First check to see if a reading with this ID exists.
293 if( $self->reading( $reading->id ) ) {
294 throw( "Collation already has a reading with id " . $reading->id );
296 $self->_graphcalc_done(0);
297 $self->_add_reading( $reading->id => $reading );
298 # Once the reading has been added, put it in both graphs.
299 $self->sequence->add_vertex( $reading->id );
300 $self->relations->add_reading( $reading->id );
304 around del_reading => sub {
309 if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) {
312 # Remove the reading from the graphs.
313 $self->_graphcalc_done(0);
314 $self->_clear_cache; # Explicitly clear caches to GC the reading
315 $self->sequence->delete_vertex( $arg );
316 $self->relations->delete_reading( $arg );
319 $self->$orig( $arg );
326 my $cxfile = 't/data/Collatex-16.xml';
327 my $t = Text::Tradition->new(
329 'input' => 'CollateX',
332 my $c = $t->collation;
334 my $rno = scalar $c->readings;
335 # Split n21 for testing purposes
336 my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } );
337 my $old_r = $c->reading( 'n21' );
338 $old_r->alter_text( 'to' );
339 $c->del_path( 'n20', 'n21', 'A' );
340 $c->add_path( 'n20', 'n21p0', 'A' );
341 $c->add_path( 'n21p0', 'n21', 'A' );
343 ok( $c->reading( 'n21p0' ), "New reading exists" );
344 is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" );
346 # Combine n3 and n4 ( with his )
347 $c->merge_readings( 'n3', 'n4', 1 );
348 ok( !$c->reading('n4'), "Reading n4 is gone" );
349 is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" );
351 # Collapse n9 and n10 ( rood / root )
352 $c->merge_readings( 'n9', 'n10' );
353 ok( !$c->reading('n10'), "Reading n10 is gone" );
354 is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" );
356 # Combine n21 and n21p0
357 my $remaining = $c->reading('n21');
358 $remaining ||= $c->reading('n22'); # one of these should still exist
359 $c->merge_readings( 'n21p0', $remaining, 1 );
360 ok( !$c->reading('n21'), "Reading $remaining is gone" );
361 is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" );
371 my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ );
372 my $mergemeta = $kept_obj->is_meta;
373 throw( "Cannot merge meta and non-meta reading" )
374 unless ( $mergemeta && $del_obj->is_meta )
375 || ( !$mergemeta && !$del_obj->is_meta );
377 throw( "Cannot merge with start or end node" )
378 if( $kept_obj eq $self->start || $kept_obj eq $self->end
379 || $del_obj eq $self->start || $del_obj eq $self->end );
381 # We only need the IDs for adding paths to the graph, not the reading
382 # objects themselves.
383 my $kept = $kept_obj->id;
384 my $deleted = $del_obj->id;
385 $self->_graphcalc_done(0);
387 # The kept reading should inherit the paths and the relationships
388 # of the deleted reading.
389 foreach my $path ( $self->sequence->edges_at( $deleted ) ) {
390 my @vector = ( $kept );
391 push( @vector, $path->[1] ) if $path->[0] eq $deleted;
392 unshift( @vector, $path->[0] ) if $path->[1] eq $deleted;
393 next if $vector[0] eq $vector[1]; # Don't add a self loop
394 my %wits = %{$self->sequence->get_edge_attributes( @$path )};
395 $self->sequence->add_edge( @vector );
396 my $fwits = $self->sequence->get_edge_attributes( @vector );
397 @wits{keys %$fwits} = values %$fwits;
398 $self->sequence->set_edge_attributes( @vector, \%wits );
400 $self->relations->merge_readings( $kept, $deleted, $combine );
402 # Do the deletion deed.
404 # Combine the text of the readings
405 my $joinstr = $combine_char;
406 unless( defined $joinstr ) {
407 $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior;
408 $joinstr = $self->wordsep unless defined $joinstr;
410 $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) );
411 # Change this reading to a joining one if necessary
412 $kept_obj->_set_join_next( $del_obj->join_next );
413 $kept_obj->normal_form(
414 join( $joinstr, $kept_obj->normal_form, $del_obj->normal_form ) );
415 # Combine the lexemes present in the readings
416 if( $kept_obj->has_lexemes && $del_obj->has_lexemes ) {
417 $kept_obj->add_lexeme( $del_obj->lexemes );
420 $self->del_reading( $deleted );
423 =head2 compress_readings
425 Where possible in the graph, compresses plain sequences of readings into a
426 single reading. The sequences must consist of readings with no
427 relationships to other readings, with only a single witness path between
428 them and no other witness paths from either that would skip the other. The
429 readings must also not be marked as nonsense or bad grammar.
431 WARNING: This operation cannot be undone.
435 sub compress_readings {
437 # Anywhere in the graph that there is a reading that joins only to a single
438 # successor, and neither of these have any relationships, just join the two
441 foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) {
442 # While we are here, get rid of any extra wordforms from a disambiguated
444 if( $rdg->disambiguated ) {
445 foreach my $lex ( $rdg->lexemes ) {
446 $lex->clear_matching_forms();
447 $lex->add_matching_form( $lex->form );
450 # Now look for readings that can be joined to their successors.
451 next if $rdg->is_meta;
452 next if $gobbled{$rdg->id};
453 next if $rdg->grammar_invalid || $rdg->is_nonsense;
454 next if $rdg->related_readings();
456 while( $self->sequence->successors( $rdg ) == 1 ) {
457 my( $next ) = $self->reading( $self->sequence->successors( $rdg ) );
458 throw( "Infinite loop" ) if $seen{$next->id};
459 $seen{$next->id} = 1;
460 last if $self->sequence->predecessors( $next ) > 1;
461 last if $next->is_meta;
462 last if $next->grammar_invalid || $next->is_nonsense;
463 last if $next->related_readings();
464 say "Joining readings $rdg and $next";
465 $self->merge_readings( $rdg, $next, 1 );
468 # Make sure we haven't screwed anything up
469 foreach my $wit ( $self->tradition->witnesses ) {
470 my $pathtext = $self->path_text( $wit->sigil );
471 my $origtext = join( ' ', @{$wit->text} );
472 throw( "Text differs for witness " . $wit->sigil )
473 unless $pathtext eq $origtext;
474 if( $wit->is_layered ) {
475 $pathtext = $self->path_text( $wit->sigil.$self->ac_label );
476 $origtext = join( ' ', @{$wit->layertext} );
477 throw( "Ante-corr text differs for witness " . $wit->sigil )
478 unless $pathtext eq $origtext;
482 $self->relations->rebuild_equivalence();
483 $self->calculate_ranks();
486 # Helper function for manipulating the graph.
487 sub _stringify_args {
488 my( $self, $first, $second, @args ) = @_;
490 if ref( $first ) eq 'Text::Tradition::Collation::Reading';
491 $second = $second->id
492 if ref( $second ) eq 'Text::Tradition::Collation::Reading';
493 return( $first, $second, @args );
496 # Helper function for manipulating the graph.
497 sub _objectify_args {
498 my( $self, $first, $second, $arg ) = @_;
499 $first = $self->reading( $first )
500 unless ref( $first ) eq 'Text::Tradition::Collation::Reading';
501 $second = $self->reading( $second )
502 unless ref( $second ) eq 'Text::Tradition::Collation::Reading';
503 return( $first, $second, $arg );
510 # We only need the IDs for adding paths to the graph, not the reading
511 # objects themselves.
512 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
514 $self->_graphcalc_done(0);
515 # Connect the readings
516 unless( $self->sequence->has_edge( $source, $target ) ) {
517 $self->sequence->add_edge( $source, $target );
518 $self->relations->add_equivalence_edge( $source, $target );
520 # Note the witness in question
521 $self->sequence->set_edge_attribute( $source, $target, $wit, 1 );
527 if( ref( $_[0] ) eq 'ARRAY' ) {
534 # We only need the IDs for adding paths to the graph, not the reading
535 # objects themselves.
536 my( $source, $target, $wit ) = $self->_stringify_args( @args );
538 $self->_graphcalc_done(0);
539 if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) {
540 $self->sequence->delete_edge_attribute( $source, $target, $wit );
542 unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) {
543 $self->sequence->delete_edge( $source, $target );
544 $self->relations->delete_equivalence_edge( $source, $target );
549 # Extra graph-alike utility
552 my( $source, $target, $wit ) = $self->_stringify_args( @_ );
553 return undef unless $self->sequence->has_edge( $source, $target );
554 return $self->sequence->has_edge_attribute( $source, $target, $wit );
557 =head2 clear_witness( @sigil_list )
559 Clear the given witnesses out of the collation entirely, removing references
560 to them in paths, and removing readings that belong only to them. Should only
561 be called via $tradition->del_witness.
566 my( $self, @sigils ) = @_;
568 $self->_graphcalc_done(0);
569 # Clear the witness(es) out of the paths
570 foreach my $e ( $self->paths ) {
571 foreach my $sig ( @sigils ) {
572 $self->del_path( $e, $sig );
576 # Clear out the newly unused readings
577 foreach my $r ( $self->readings ) {
578 unless( $self->reading_witnesses( $r ) ) {
579 $self->del_reading( $r );
584 sub add_relationship {
586 my( $source, $target, $opts ) = $self->_stringify_args( @_ );
587 my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts );
588 $self->_graphcalc_done(0);
592 around qw/ get_relationship del_relationship / => sub {
596 if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) {
599 my( $source, $target ) = $self->_stringify_args( @args );
600 $self->$orig( $source, $target );
603 =head2 reading_witnesses( $reading )
605 Return a list of sigils corresponding to the witnesses in which the reading appears.
609 sub reading_witnesses {
610 my( $self, $reading ) = @_;
611 # We need only check either the incoming or the outgoing edges; I have
612 # arbitrarily chosen "incoming". Thus, special-case the start node.
613 if( $reading eq $self->start ) {
614 return map { $_->sigil } $self->tradition->witnesses;
617 foreach my $e ( $self->sequence->edges_to( $reading ) ) {
618 my $wits = $self->sequence->get_edge_attributes( @$e );
619 @all_witnesses{ keys %$wits } = 1;
621 my $acstr = $self->ac_label;
622 foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) {
623 delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit};
625 return keys %all_witnesses;
628 =head1 OUTPUT METHODS
630 =head2 as_svg( \%options )
632 Returns an SVG string that represents the graph, via as_dot and graphviz.
633 See as_dot for a list of options. Must have GraphViz (dot) installed to run.
638 my( $self, $opts ) = @_;
639 throw( "Need GraphViz installed to output SVG" )
640 unless File::Which::which( 'dot' );
641 my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'};
642 $self->calculate_ranks()
643 unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear );
644 my @cmd = qw/dot -Tsvg/;
646 my $dotfile = File::Temp->new();
648 # $dotfile->unlink_on_destroy(0);
649 binmode $dotfile, ':utf8';
650 print $dotfile $self->as_dot( $opts );
651 push( @cmd, $dotfile->filename );
652 run( \@cmd, ">", binary(), \$svg );
653 $svg = decode_utf8( $svg );
658 =head2 as_dot( \%options )
660 Returns a string that is the collation graph expressed in dot
661 (i.e. GraphViz) format. Options include:
676 my( $self, $opts ) = @_;
677 my $startrank = $opts->{'from'} if $opts;
678 my $endrank = $opts->{'to'} if $opts;
679 my $color_common = $opts->{'color_common'} if $opts;
680 my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank
681 && $self->end->rank > 100;
682 $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs
684 # Check the arguments
686 return if $endrank && $startrank > $endrank;
687 return if $startrank > $self->end->rank;
689 if( defined $endrank ) {
690 return if $endrank < 0;
691 $endrank = undef if $endrank == $self->end->rank;
694 my $graph_name = $self->tradition->name;
695 $graph_name =~ s/[^\w\s]//g;
696 $graph_name = join( '_', split( /\s+/, $graph_name ) );
704 'fillcolor' => 'white',
709 'arrowhead' => 'open',
710 'color' => '#000000',
711 'fontcolor' => '#000000',
714 my $dot = sprintf( "digraph %s {\n", $graph_name );
715 $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
716 $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
718 # Output substitute start/end readings if necessary
720 $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n";
723 $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n";
725 if( $STRAIGHTENHACK ) {
727 my $startlabel = $startrank ? '__SUBSTART__' : '__START__';
728 $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n";
729 $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];"
731 my %used; # Keep track of the readings that actually appear in the graph
732 # Sort the readings by rank if we have ranks; this speeds layout.
733 my @all_readings = $self->end->has_rank
734 ? sort { $a->rank <=> $b->rank } $self->readings
736 # TODO Refrain from outputting lacuna nodes - just grey out the edges.
737 foreach my $reading ( @all_readings ) {
738 # Only output readings within our rank range.
739 next if $startrank && $reading->rank < $startrank;
740 next if $endrank && $reading->rank > $endrank;
741 $used{$reading->id} = 1;
742 # Need not output nodes without separate labels
743 next if $reading->id eq $reading->text;
745 my $label = $reading->text;
746 $label .= '-' if $reading->join_next;
747 $label = "-$label" if $reading->join_prior;
748 $label =~ s/\"/\\\"/g;
749 $rattrs->{'label'} = $label;
750 $rattrs->{'id'} = $reading->id;
751 $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common;
752 $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
755 # Add the real edges. Need to weight one edge per rank jump, in a
757 # my $weighted = $self->_add_edge_weights;
758 my @edges = $self->paths;
759 my( %substart, %subend );
760 foreach my $edge ( @edges ) {
761 # Do we need to output this edge?
762 if( $used{$edge->[0]} && $used{$edge->[1]} ) {
763 my $label = $self->_path_display_label( $self->path_witnesses( $edge ) );
764 my $variables = { %edge_attrs, 'label' => $label };
766 # Account for the rank gap if necessary
767 my $rank0 = $self->reading( $edge->[0] )->rank
768 if $self->reading( $edge->[0] )->has_rank;
769 my $rank1 = $self->reading( $edge->[1] )->rank
770 if $self->reading( $edge->[1] )->has_rank;
771 if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) {
772 $variables->{'minlen'} = $rank1 - $rank0;
775 # Add the calculated edge weights
776 # if( exists $weighted->{$edge->[0]}
777 # && $weighted->{$edge->[0]} eq $edge->[1] ) {
778 # # $variables->{'color'} = 'red';
779 # $variables->{'weight'} = 3.0;
782 # EXPERIMENTAL: make edge width reflect no. of witnesses
783 my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
784 $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
786 my $varopts = _dot_attr_string( $variables );
787 $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
788 $edge->[0], $edge->[1], $varopts );
789 } elsif( $used{$edge->[0]} ) {
790 $subend{$edge->[0]} = $edge->[1];
791 } elsif( $used{$edge->[1]} ) {
792 $substart{$edge->[1]} = $edge->[0];
795 # Add substitute start and end edges if necessary
796 foreach my $node ( keys %substart ) {
797 my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) );
798 my $variables = { %edge_attrs, 'label' => $witstr };
799 my $nrdg = $self->reading( $node );
800 if( $nrdg->has_rank && $nrdg->rank > $startrank ) {
801 # Substart is actually one lower than $startrank
802 $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 );
804 my $varopts = _dot_attr_string( $variables );
805 $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n";
807 foreach my $node ( keys %subend ) {
808 my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) );
809 my $variables = { %edge_attrs, 'label' => $witstr };
810 my $varopts = _dot_attr_string( $variables );
811 $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n";
814 if( $STRAIGHTENHACK ) {
815 my $endlabel = $endrank ? '__SUBEND__' : '__END__';
816 $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n";
823 sub _dot_attr_string {
826 foreach my $k ( sort keys %$hash ) {
828 push( @attrs, $k.'="'.$v.'"' );
830 return( '[ ' . join( ', ', @attrs ) . ' ]' );
833 sub _add_edge_weights {
835 # Walk the graph from START to END, choosing the successor node with
836 # the largest number of witness paths each time.
838 my $curr = $self->start->id;
839 my $ranked = $self->end->has_rank;
840 while( $curr ne $self->end->id ) {
841 my $rank = $ranked ? $self->reading( $curr )->rank : 0;
842 my @succ = sort { $self->path_witnesses( $curr, $a )
843 <=> $self->path_witnesses( $curr, $b ) }
844 $self->sequence->successors( $curr );
845 my $next = pop @succ;
846 my $nextrank = $ranked ? $self->reading( $next )->rank : 0;
847 # Try to avoid lacunae in the weighted path.
849 ( $self->reading( $next )->is_lacuna ||
850 $nextrank - $rank > 1 ) ){
853 $weighted->{$curr} = $next;
859 =head2 path_witnesses( $edge )
861 Returns the list of sigils whose witnesses are associated with the given edge.
862 The edge can be passed as either an array or an arrayref of ( $source, $target ).
867 my( $self, @edge ) = @_;
868 # If edge is an arrayref, cope.
869 if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) {
873 my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
877 # Helper function. Make a display label for the given witnesses, showing a.c.
878 # witnesses only where the main witness is not also in the list.
879 sub _path_display_label {
882 map { $wits{$_} = 1 } @_;
884 # If an a.c. wit is listed, remove it if the main wit is also listed.
885 # Otherwise keep it for explicit listing.
886 my $aclabel = $self->ac_label;
888 foreach my $w ( sort keys %wits ) {
889 if( $w =~ /^(.*)\Q$aclabel\E$/ ) {
890 if( exists $wits{$1} ) {
893 push( @disp_ac, $w );
898 # See if we are in a majority situation.
899 my $maj = scalar( $self->tradition->witnesses ) * 0.6;
900 $maj = $maj > 5 ? $maj : 5;
901 if( scalar keys %wits > $maj ) {
902 unshift( @disp_ac, 'majority' );
903 return join( ', ', @disp_ac );
905 return join( ', ', sort keys %wits );
909 =head2 readings_at_rank( $rank )
911 Returns a list of readings at a given rank, taken from the alignment table.
915 sub readings_at_rank {
916 my( $self, $rank ) = @_;
917 my $table = $self->alignment_table;
918 # Table rank is real rank - 1.
919 my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}};
921 foreach my $e ( @elements ) {
922 next unless ref( $e ) eq 'HASH';
923 next unless exists $e->{'t'};
924 $readings{$e->{'t'}->id} = $e->{'t'};
926 return values %readings;
931 Returns a GraphML representation of the collation. The GraphML will contain
932 two graphs. The first expresses the attributes of the readings and the witness
933 paths that link them; the second expresses the relationships that link the
934 readings. This is the native transfer format for a tradition.
943 my $datafile = 't/data/florilegium_tei_ps.xml';
944 my $tradition = Text::Tradition->new( 'input' => 'TEI',
949 ok( $tradition, "Got a tradition object" );
950 is( scalar $tradition->witnesses, 13, "Found all witnesses" );
951 ok( $tradition->collation, "Tradition has a collation" );
953 my $c = $tradition->collation;
954 is( scalar $c->readings, $READINGS, "Collation has all readings" );
955 is( scalar $c->paths, $PATHS, "Collation has all paths" );
956 is( scalar $c->relationships, 0, "Collation has all relationships" );
958 # Add a few relationships
959 $c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } );
960 $c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } );
961 $c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } );
963 # Now write it to GraphML and parse it again.
965 my $graphml = $c->as_graphml;
966 my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml );
967 is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" );
968 is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" );
969 is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" );
971 # Now add a stemma, write to GraphML, and look at the output.
972 my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' );
973 is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" );
974 is( $tradition->stemmata, 1, "Tradition now has the stemma" );
975 $graphml = $c->as_graphml;
976 like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" );
978 # Now add a user, write to GraphML, and look at the output.
979 unlike( $graphml, qr/testuser/, "Test user name does not exist in GraphML yet" );
980 my $testuser = Text::Tradition::User->new(
981 id => 'testuser', password => 'testpass' );
982 is( ref( $testuser ), 'Text::Tradition::User', "Created test user object" );
983 $testuser->add_tradition( $tradition );
984 is( $tradition->user->id, $testuser->id, "Tradition assigned to test user" );
985 $graphml = $c->as_graphml;
986 like( $graphml, qr/testuser/, "Test user name now exists in GraphML" );
993 my( $self, $options ) = @_;
994 $self->calculate_ranks unless $self->_graphcalc_done;
996 my $start = $options->{'from'}
997 ? $self->reading( $options->{'from'} ) : $self->start;
998 my $end = $options->{'to'}
999 ? $self->reading( $options->{'to'} ) : $self->end;
1000 if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) {
1001 throw( 'Start node must be before end node' );
1003 # The readings need to be ranked for this to work.
1004 $start = $self->start unless $start->has_rank;
1005 $end = $self->end unless $end->has_rank;
1007 unless( $start eq $self->start ) {
1008 $rankoffset = $start->rank - 1;
1013 my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
1014 my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
1015 my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
1016 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';
1018 # Create the document and root node
1019 require XML::LibXML;
1020 my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
1021 my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
1022 $graphml->setDocumentElement( $root );
1023 $root->setNamespace( $xsi_ns, 'xsi', 0 );
1024 $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
1026 # List of attribute types to save on our objects and their corresponding
1031 'Bool' => 'boolean',
1032 'ReadingID' => 'string',
1033 'RelationshipType' => 'string',
1034 'RelationshipScope' => 'string',
1037 # Add the data keys for the graph. Include an extra key 'version' for the
1038 # GraphML output version.
1039 my %graph_data_keys;
1041 my %graph_attributes = ( 'version' => 'string' );
1042 # Graph attributes include those of Tradition and those of Collation.
1044 my $tmeta = $self->tradition->meta;
1045 my $cmeta = $self->meta;
1046 map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes;
1047 map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes;
1048 foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) {
1049 next if $attr->name =~ /^_/;
1050 next unless $save_types{$attr->type_constraint->name};
1051 $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1053 # Extra custom keys for complex objects that should be saved in some form.
1054 # The subroutine should return a string, or undef/empty.
1055 $graph_attributes{'stemmata'} = sub {
1057 map { push( @stemstrs, $_->editable( {linesep => ''} ) ) }
1058 $self->tradition->stemmata;
1059 join( "\n", @stemstrs );
1061 $graph_attributes{'user'} = sub {
1062 $self->tradition->user ? $self->tradition->user->id : undef
1065 foreach my $datum ( sort keys %graph_attributes ) {
1066 $graph_data_keys{$datum} = 'dg'.$gdi++;
1067 my $key = $root->addNewChild( $graphml_ns, 'key' );
1068 my $dtype = ref( $graph_attributes{$datum} ) ? 'string'
1069 : $graph_attributes{$datum};
1070 $key->setAttribute( 'attr.name', $datum );
1071 $key->setAttribute( 'attr.type', $dtype );
1072 $key->setAttribute( 'for', 'graph' );
1073 $key->setAttribute( 'id', $graph_data_keys{$datum} );
1076 # Add the data keys for reading nodes
1077 my %reading_attributes;
1078 my $rmeta = Text::Tradition::Collation::Reading->meta;
1079 foreach my $attr( $rmeta->get_all_attributes ) {
1080 next if $attr->name =~ /^_/;
1081 next unless $save_types{$attr->type_constraint->name};
1082 $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1084 # Extra custom key for the reading morphology
1085 $reading_attributes{'lexemes'} = 'string';
1089 foreach my $datum ( sort keys %reading_attributes ) {
1090 $node_data_keys{$datum} = 'dn'.$ndi++;
1091 my $key = $root->addNewChild( $graphml_ns, 'key' );
1092 $key->setAttribute( 'attr.name', $datum );
1093 $key->setAttribute( 'attr.type', $reading_attributes{$datum} );
1094 $key->setAttribute( 'for', 'node' );
1095 $key->setAttribute( 'id', $node_data_keys{$datum} );
1098 # Add the data keys for edges, that is, paths and relationships. Path
1099 # data does not come from a Moose class so is here manually.
1102 my %edge_attributes = (
1103 witness => 'string', # ID/label for a path
1104 extra => 'boolean', # Path key
1106 my @path_attributes = keys %edge_attributes; # track our manual additions
1107 my $pmeta = Text::Tradition::Collation::Relationship->meta;
1108 foreach my $attr( $pmeta->get_all_attributes ) {
1109 next if $attr->name =~ /^_/;
1110 next unless $save_types{$attr->type_constraint->name};
1111 $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name};
1113 foreach my $datum ( sort keys %edge_attributes ) {
1114 $edge_data_keys{$datum} = 'de'.$edi++;
1115 my $key = $root->addNewChild( $graphml_ns, 'key' );
1116 $key->setAttribute( 'attr.name', $datum );
1117 $key->setAttribute( 'attr.type', $edge_attributes{$datum} );
1118 $key->setAttribute( 'for', 'edge' );
1119 $key->setAttribute( 'id', $edge_data_keys{$datum} );
1122 # Add the collation graph itself. First, sanitize the name to a valid XML ID.
1123 my $xmlidname = $self->tradition->name;
1124 $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g;
1125 if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) {
1126 $xmlidname = '_'.$xmlidname;
1128 my $sgraph = $root->addNewChild( $graphml_ns, 'graph' );
1129 $sgraph->setAttribute( 'edgedefault', 'directed' );
1130 $sgraph->setAttribute( 'id', $xmlidname );
1131 $sgraph->setAttribute( 'parse.edgeids', 'canonical' );
1132 $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later
1133 $sgraph->setAttribute( 'parse.nodeids', 'canonical' );
1134 $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later
1135 $sgraph->setAttribute( 'parse.order', 'nodesfirst' );
1137 # Tradition/collation attribute data
1138 foreach my $datum ( keys %graph_attributes ) {
1140 if( $datum eq 'version' ) {
1142 } elsif( ref( $graph_attributes{$datum} ) ) {
1143 my $sub = $graph_attributes{$datum};
1145 } elsif( $gattr_from{$datum} eq 'Tradition' ) {
1146 $value = $self->tradition->$datum;
1148 $value = $self->$datum;
1150 _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value );
1155 # Add our readings to the graph
1156 foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) {
1157 next if $n->has_rank && $n ne $self->start && $n ne $self->end &&
1158 ( $n->rank < $start->rank || $n->rank > $end->rank );
1159 $use_readings{$n->id} = 1;
1160 # Add to the main graph
1161 my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' );
1162 my $node_xmlid = 'n' . $node_ctr++;
1163 $node_hash{ $n->id } = $node_xmlid;
1164 $node_el->setAttribute( 'id', $node_xmlid );
1165 foreach my $d ( keys %reading_attributes ) {
1167 # Custom serialization
1168 if( $d eq 'lexemes' ) {
1169 # If nval is a true value, we have lexemes so we need to
1170 # serialize them. Otherwise set nval to undef so that the
1171 # key is excluded from this reading.
1172 $nval = $nval ? $n->_serialize_lexemes : undef;
1173 } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) {
1176 if( $rankoffset && $d eq 'rank' && $n ne $self->start ) {
1177 # Adjust the ranks within the subgraph.
1178 $nval = $n eq $self->end ? $end->rank - $rankoffset + 1
1179 : $nval - $rankoffset;
1181 _add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
1186 # Add the path edges to the sequence graph
1188 foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
1189 # We add an edge in the graphml for every witness in $e.
1190 next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} );
1191 my @edge_wits = sort $self->path_witnesses( $e );
1192 $e->[0] = $self->start->id unless $use_readings{$e->[0]};
1193 $e->[1] = $self->end->id unless $use_readings{$e->[1]};
1194 # Skip any path from start to end; that witness is not in the subgraph.
1195 next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id );
1196 foreach my $wit ( @edge_wits ) {
1197 my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
1198 $node_hash{ $e->[0] },
1199 $node_hash{ $e->[1] } );
1200 my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' );
1201 $edge_el->setAttribute( 'source', $from );
1202 $edge_el->setAttribute( 'target', $to );
1203 $edge_el->setAttribute( 'id', $id );
1205 # It's a witness path, so add the witness
1207 my $key = $edge_data_keys{'witness'};
1208 # Is this an ante-corr witness?
1209 my $aclabel = $self->ac_label;
1210 if( $wit =~ /^(.*)\Q$aclabel\E$/ ) {
1211 # Keep the base witness
1213 # ...and record that this is an 'extra' reading path
1214 _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel );
1216 _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base );
1220 # Report the actual number of nodes and edges that went in
1221 $sgraph->setAttribute( 'parse.edges', $edge_ctr );
1222 $sgraph->setAttribute( 'parse.nodes', $node_ctr );
1224 # Add the relationship graph to the XML
1225 map { delete $edge_data_keys{$_} } @path_attributes;
1226 $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash,
1227 $node_data_keys{'id'}, \%edge_data_keys );
1229 # Save and return the thing
1230 my $result = decode_utf8( $graphml->toString(1) );
1234 sub _add_graphml_data {
1235 my( $el, $key, $value ) = @_;
1236 return unless defined $value;
1237 my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
1238 $data_el->setAttribute( 'key', $key );
1239 $data_el->appendText( $value );
1244 Returns a CSV alignment table representation of the collation graph, one
1245 row per witness (or witness uncorrected.)
1251 my $table = $self->alignment_table;
1252 my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } );
1254 # Make the header row
1255 $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} );
1256 push( @result, decode_utf8( $csv->string ) );
1257 # Make the rest of the rows
1258 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1259 my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}};
1260 my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs;
1261 $csv->combine( @row );
1262 push( @result, decode_utf8( $csv->string ) );
1264 return join( "\n", @result );
1267 =head2 alignment_table( $use_refs, $include_witnesses )
1269 Return a reference to an alignment table, in a slightly enhanced CollateX
1270 format which looks like this:
1272 $table = { alignment => [ { witness => "SIGIL",
1273 tokens => [ { t => "TEXT" }, ... ] },
1274 { witness => "SIG2",
1275 tokens => [ { t => "TEXT" }, ... ] },
1277 length => TEXTLEN };
1279 If $use_refs is set to 1, the reading object is returned in the table
1280 instead of READINGTEXT; if not, the text of the reading is returned.
1282 If $include_witnesses is set to a hashref, only the witnesses whose sigil
1283 keys have a true hash value will be included.
1287 sub alignment_table {
1289 $self->calculate_ranks() unless $self->_graphcalc_done;
1290 return $self->cached_table if $self->has_cached_table;
1292 # Make sure we can do this
1293 throw( "Need a linear graph in order to make an alignment table" )
1294 unless $self->linear;
1295 $self->calculate_ranks unless $self->end->has_rank;
1297 my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
1298 my @all_pos = ( 1 .. $self->end->rank - 1 );
1299 foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) {
1300 # say STDERR "Making witness row(s) for " . $wit->sigil;
1301 my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1302 my @row = _make_witness_row( \@wit_path, \@all_pos );
1303 push( @{$table->{'alignment'}},
1304 { 'witness' => $wit->sigil, 'tokens' => \@row } );
1305 if( $wit->is_layered ) {
1306 my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
1307 $wit->sigil.$self->ac_label );
1308 my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos );
1309 push( @{$table->{'alignment'}},
1310 { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
1313 $self->cached_table( $table );
1317 sub _make_witness_row {
1318 my( $path, $positions ) = @_;
1320 map { $char_hash{$_} = undef } @$positions;
1322 foreach my $rdg ( @$path ) {
1323 my $rtext = $rdg->text;
1324 $rtext = '#LACUNA#' if $rdg->is_lacuna;
1325 say STDERR "rank " . $rdg->rank if $debug;
1326 # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank;
1327 $char_hash{$rdg->rank} = { 't' => $rdg };
1329 my @row = map { $char_hash{$_} } @$positions;
1330 # Fill in lacuna markers for undef spots in the row
1331 my $last_el = shift @row;
1332 my @filled_row = ( $last_el );
1333 foreach my $el ( @row ) {
1334 # If we are using node reference, make the lacuna node appear many times
1335 # in the table. If not, use the lacuna tag.
1336 if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) {
1339 push( @filled_row, $el );
1345 =head1 NAVIGATION METHODS
1347 =head2 reading_sequence( $first, $last, $sigil, $backup )
1349 Returns the ordered list of readings, starting with $first and ending
1350 with $last, for the witness given in $sigil. If a $backup sigil is
1351 specified (e.g. when walking a layered witness), it will be used wherever
1352 no $sigil path exists. If there is a base text reading, that will be
1353 used wherever no path exists for $sigil or $backup.
1357 # TODO Think about returning some lazy-eval iterator.
1358 # TODO Get rid of backup; we should know from what witness is whether we need it.
1360 sub reading_sequence {
1361 my( $self, $start, $end, $witness ) = @_;
1363 $witness = $self->baselabel unless $witness;
1364 my @readings = ( $start );
1367 while( $n && $n->id ne $end->id ) {
1368 if( exists( $seen{$n->id} ) ) {
1369 throw( "Detected loop for $witness at " . $n->id );
1373 my $next = $self->next_reading( $n, $witness );
1375 throw( "Did not find any path for $witness from reading " . $n->id );
1377 push( @readings, $next );
1380 # Check that the last reading is our end reading.
1381 my $last = $readings[$#readings];
1382 throw( "Last reading found from " . $start->text .
1383 " for witness $witness is not the end!" ) # TODO do we get this far?
1384 unless $last->id eq $end->id;
1389 =head2 next_reading( $reading, $sigil );
1391 Returns the reading that follows the given reading along the given witness
1397 # Return the successor via the corresponding path.
1399 my $answer = $self->_find_linked_reading( 'next', @_ );
1400 return undef unless $answer;
1401 return $self->reading( $answer );
1404 =head2 prior_reading( $reading, $sigil )
1406 Returns the reading that precedes the given reading along the given witness
1412 # Return the predecessor via the corresponding path.
1414 my $answer = $self->_find_linked_reading( 'prior', @_ );
1415 return $self->reading( $answer );
1418 sub _find_linked_reading {
1419 my( $self, $direction, $node, $path ) = @_;
1421 # Get a backup if we are dealing with a layered witness
1423 my $aclabel = $self->ac_label;
1424 if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
1428 my @linked_paths = $direction eq 'next'
1429 ? $self->sequence->edges_from( $node )
1430 : $self->sequence->edges_to( $node );
1431 return undef unless scalar( @linked_paths );
1433 # We have to find the linked path that contains all of the
1434 # witnesses supplied in $path.
1435 my( @path_wits, @alt_path_wits );
1436 @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path;
1437 @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path;
1440 foreach my $le ( @linked_paths ) {
1441 if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
1444 my @le_wits = sort $self->path_witnesses( $le );
1445 if( _is_within( \@path_wits, \@le_wits ) ) {
1446 # This is the right path.
1447 return $direction eq 'next' ? $le->[1] : $le->[0];
1448 } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
1452 # Got this far? Return the alternate path if it exists.
1453 return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0]
1456 # Got this far? Return the base path if it exists.
1457 return $direction eq 'next' ? $base_le->[1] : $base_le->[0]
1460 # Got this far? We have no appropriate path.
1461 warn "Could not find $direction node from " . $node->id
1462 . " along path $path";
1468 my( $set1, $set2 ) = @_;
1469 my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
1470 foreach my $el ( @$set1 ) {
1471 $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
1476 # Return the string that joins together a list of witnesses for
1477 # display on a single path.
1478 sub _witnesses_of_label {
1479 my( $self, $label ) = @_;
1480 my $regex = $self->wit_list_separator;
1481 my @answer = split( /\Q$regex\E/, $label );
1485 =head2 common_readings
1487 Returns the list of common readings in the graph (i.e. those readings that are
1488 shared by all non-lacunose witnesses.)
1492 sub common_readings {
1494 my @common = grep { $_->is_common } $self->readings;
1498 =head2 path_text( $sigil, [, $start, $end ] )
1500 Returns the text of a witness (plus its backup, if we are using a layer)
1501 as stored in the collation. The text is returned as a string, where the
1502 individual readings are joined with spaces and the meta-readings (e.g.
1503 lacunae) are omitted. Optional specification of $start and $end allows
1504 the generation of a subset of the witness text.
1509 my( $self, $wit, $start, $end ) = @_;
1510 $start = $self->start unless $start;
1511 $end = $self->end unless $end;
1512 my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
1515 foreach my $r ( @path ) {
1516 unless ( $r->join_prior || !$last || $last->join_next ) {
1519 $pathtext .= $r->text;
1525 =head1 INITIALIZATION METHODS
1527 These are mostly for use by parsers.
1529 =head2 make_witness_path( $witness )
1531 Link the array of readings contained in $witness->path (and in
1532 $witness->uncorrected_path if it exists) into collation paths.
1533 Clear out the arrays when finished.
1535 =head2 make_witness_paths
1537 Call make_witness_path for all witnesses in the tradition.
1541 # For use when a collation is constructed from a base text and an apparatus.
1542 # We have the sequences of readings and just need to add path edges.
1543 # When we are done, clear out the witness path attributes, as they are no
1545 # TODO Find a way to replace the witness path attributes with encapsulated functions?
1547 sub make_witness_paths {
1549 foreach my $wit ( $self->tradition->witnesses ) {
1550 # say STDERR "Making path for " . $wit->sigil;
1551 $self->make_witness_path( $wit );
1555 sub make_witness_path {
1556 my( $self, $wit ) = @_;
1557 my @chain = @{$wit->path};
1558 my $sig = $wit->sigil;
1559 # Add start and end if necessary
1560 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1561 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1562 foreach my $idx ( 0 .. $#chain-1 ) {
1563 $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
1565 if( $wit->is_layered ) {
1566 @chain = @{$wit->uncorrected_path};
1567 unshift( @chain, $self->start ) unless $chain[0] eq $self->start;
1568 push( @chain, $self->end ) unless $chain[-1] eq $self->end;
1569 foreach my $idx( 0 .. $#chain-1 ) {
1570 my $source = $chain[$idx];
1571 my $target = $chain[$idx+1];
1572 $self->add_path( $source, $target, $sig.$self->ac_label )
1573 unless $self->has_path( $source, $target, $sig );
1577 $wit->clear_uncorrected_path;
1580 =head2 calculate_ranks
1582 Calculate the reading ranks (that is, their aligned positions relative
1583 to each other) for the graph. This can only be called on linear collations.
1587 use Text::Tradition;
1589 my $cxfile = 't/data/Collatex-16.xml';
1590 my $t = Text::Tradition->new(
1592 'input' => 'CollateX',
1595 my $c = $t->collation;
1598 my $table = $c->alignment_table;
1599 ok( $c->has_cached_table, "Alignment table was cached" );
1600 is( $c->alignment_table, $table, "Cached table returned upon second call" );
1601 $c->calculate_ranks;
1602 is( $c->alignment_table, $table, "Cached table retained with no rank change" );
1603 $c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } );
1604 isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" );
1610 sub calculate_ranks {
1612 # Save the existing ranks, in case we need to invalidate the cached SVG.
1614 map { $existing_ranks{$_} = $_->rank } $self->readings;
1616 # Do the rankings based on the relationship equivalence graph, starting
1617 # with the start node.
1618 my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks();
1620 # Transfer our rankings from the topological graph to the real one.
1621 foreach my $r ( $self->readings ) {
1622 if( defined $node_ranks->{$self->equivalence( $r->id )} ) {
1623 $r->rank( $node_ranks->{$self->equivalence( $r->id )} );
1625 # Die. Find the last rank we calculated.
1626 my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 )
1627 <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) }
1629 my $last = pop @all_defined;
1630 throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
1633 # Do we need to invalidate the cached data?
1634 if( $self->has_cached_table ) {
1635 foreach my $r ( $self->readings ) {
1636 next if defined( $existing_ranks{$r} )
1637 && $existing_ranks{$r} == $r->rank;
1638 # Something has changed, so clear the cache
1639 $self->_clear_cache;
1640 # ...and recalculate the common readings.
1641 $self->calculate_common_readings();
1645 # The graph calculation information is now up to date.
1646 $self->_graphcalc_done(1);
1651 $self->wipe_table if $self->has_cached_table;
1655 =head2 flatten_ranks
1657 A convenience method for parsing collation data. Searches the graph for readings
1658 with the same text at the same rank, and merges any that are found.
1664 my %unique_rank_rdg;
1666 foreach my $rdg ( $self->readings ) {
1667 next unless $rdg->has_rank;
1668 my $key = $rdg->rank . "||" . $rdg->text;
1669 if( exists $unique_rank_rdg{$key} ) {
1670 # Make sure they don't have different grammatical forms
1671 my $ur = $unique_rank_rdg{$key};
1672 if( $rdg->disambiguated && $ur->disambiguated ) {
1673 my $rform = join( '//', map { $_->form->to_string } $rdg->lexemes );
1674 my $uform = join( '//', map { $_->form->to_string } $ur->lexemes );
1675 next unless $rform eq $uform;
1676 } elsif( $rdg->disambiguated xor $ur->disambiguated ) {
1680 #say STDERR "Combining readings at same rank: $key";
1682 $self->merge_readings( $unique_rank_rdg{$key}, $rdg );
1683 # TODO see if this now makes a common point.
1685 $unique_rank_rdg{$key} = $rdg;
1688 # If we merged readings, the ranks are still fine but the alignment
1689 # table is wrong. Wipe it.
1690 $self->wipe_table() if $changed;
1694 =head2 calculate_common_readings
1696 Goes through the graph identifying the readings that appear in every witness
1697 (apart from those with lacunae at that spot.) Marks them as common and returns
1702 use Text::Tradition;
1704 my $cxfile = 't/data/Collatex-16.xml';
1705 my $t = Text::Tradition->new(
1707 'input' => 'CollateX',
1710 my $c = $t->collation;
1712 my @common = $c->calculate_common_readings();
1713 is( scalar @common, 8, "Found correct number of common readings" );
1714 my @marked = sort $c->common_readings();
1715 is( scalar @common, 8, "All common readings got marked as such" );
1716 my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /;
1717 is_deeply( \@marked, \@expected, "Found correct list of common readings" );
1723 sub calculate_common_readings {
1726 map { $_->is_common( 0 ) } $self->readings;
1727 # Implicitly calls calculate_ranks
1728 my $table = $self->alignment_table;
1729 foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
1730 my @row = map { $_->{'tokens'}->[$idx]
1731 ? $_->{'tokens'}->[$idx]->{'t'} : '' }
1732 @{$table->{'alignment'}};
1734 foreach my $r ( @row ) {
1736 $hash{$r->id} = $r unless $r->is_meta;
1738 $hash{'UNDEF'} = $r;
1741 if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
1742 my( $r ) = values %hash;
1744 push( @common, $r );
1750 =head2 text_from_paths
1752 Calculate the text array for all witnesses from the path, for later consistency
1753 checking. Only to be used if there is no non-graph-based way to know the
1758 sub text_from_paths {
1760 foreach my $wit ( $self->tradition->witnesses ) {
1761 my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil );
1763 foreach my $r ( @readings ) {
1764 next if $r->is_meta;
1765 push( @text, $r->text );
1767 $wit->text( \@text );
1768 if( $wit->is_layered ) {
1769 my @ucrdgs = $self->reading_sequence( $self->start, $self->end,
1770 $wit->sigil.$self->ac_label );
1772 foreach my $r ( @ucrdgs ) {
1773 next if $r->is_meta;
1774 push( @uctext, $r->text );
1776 $wit->layertext( \@uctext );
1781 =head1 UTILITY FUNCTIONS
1783 =head2 common_predecessor( $reading_a, $reading_b )
1785 Find the last reading that occurs in sequence before both the given readings.
1786 At the very least this should be $self->start.
1788 =head2 common_successor( $reading_a, $reading_b )
1790 Find the first reading that occurs in sequence after both the given readings.
1791 At the very least this should be $self->end.
1795 use Text::Tradition;
1797 my $cxfile = 't/data/Collatex-16.xml';
1798 my $t = Text::Tradition->new(
1800 'input' => 'CollateX',
1803 my $c = $t->collation;
1805 is( $c->common_predecessor( 'n24', 'n23' )->id,
1806 'n20', "Found correct common predecessor" );
1807 is( $c->common_successor( 'n24', 'n23' )->id,
1808 '__END__', "Found correct common successor" );
1810 is( $c->common_predecessor( 'n19', 'n17' )->id,
1811 'n16', "Found correct common predecessor for readings on same path" );
1812 is( $c->common_successor( 'n21', 'n10' )->id,
1813 '__END__', "Found correct common successor for readings on same path" );
1819 ## Return the closest reading that is a predecessor of both the given readings.
1820 sub common_predecessor {
1822 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1823 return $self->_common_in_path( $r1, $r2, 'predecessors' );
1826 sub common_successor {
1828 my( $r1, $r2 ) = $self->_objectify_args( @_ );
1829 return $self->_common_in_path( $r1, $r2, 'successors' );
1833 # TODO think about how to do this without ranks...
1834 sub _common_in_path {
1835 my( $self, $r1, $r2, $dir ) = @_;
1836 my $iter = $self->end->rank;
1838 my @last_r1 = ( $r1 );
1839 my @last_r2 = ( $r2 );
1840 # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' );
1842 # say STDERR "Finding common $dir for $r1, $r2";
1843 while( !@candidates ) {
1844 last unless $iter--; # Avoid looping infinitely
1845 # Iterate separately down the graph from r1 and r2
1846 my( @new_lc1, @new_lc2 );
1847 foreach my $lc ( @last_r1 ) {
1848 foreach my $p ( $lc->$dir ) {
1849 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) {
1850 # say STDERR "Path candidate $p from $lc";
1851 push( @candidates, $p );
1852 } elsif( !$all_seen{$p->id} ) {
1853 $all_seen{$p->id} = 'r1';
1854 push( @new_lc1, $p );
1858 foreach my $lc ( @last_r2 ) {
1859 foreach my $p ( $lc->$dir ) {
1860 if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) {
1861 # say STDERR "Path candidate $p from $lc";
1862 push( @candidates, $p );
1863 } elsif( !$all_seen{$p->id} ) {
1864 $all_seen{$p->id} = 'r2';
1865 push( @new_lc2, $p );
1869 @last_r1 = @new_lc1;
1870 @last_r2 = @new_lc2;
1872 my @answer = sort { $a->rank <=> $b->rank } @candidates;
1873 return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
1877 Text::Tradition::Error->throw(
1878 'ident' => 'Collation error',
1884 __PACKAGE__->meta->make_immutable;
1888 This package is free software and is provided "as is" without express
1889 or implied warranty. You can redistribute it and/or modify it under
1890 the same terms as Perl itself.
1894 Tara L Andrews E<lt>aurum@cpan.orgE<gt>