X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=5b3cd96c2c168abd51cba4998350d1153f7a75f7;hb=9fef629bd3a741a6d74d130f10056898d504fb47;hp=4686d5e1ed813c725316d8ce109bb1f60591a08b;hpb=6a222840d60d8ef64c47e5ce0c4f033db1f72e2b;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index 4686d5e..5b3cd96 100644 --- a/lib/Text/Tradition/Collation.pm +++ b/lib/Text/Tradition/Collation.pm @@ -1,60 +1,62 @@ package Text::Tradition::Collation; -use Graph::Easy; +use feature 'say'; +use Encode qw( decode_utf8 ); +use File::Temp; +use File::Which; +use Graph; use IPC::Run qw( run binary ); +use Text::CSV; use Text::Tradition::Collation::Reading; +use Text::Tradition::Collation::RelationshipStore; +use Text::Tradition::Error; +use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx ); +use XML::LibXML; +use XML::LibXML::XPathContext; use Moose; -has 'graph' => ( +has 'sequence' => ( is => 'ro', - isa => 'Graph::Easy', + isa => 'Graph', + default => sub { Graph->new() }, handles => { - add_reading => 'add_node', - del_reading => 'del_node', - add_path => 'add_edge', - del_path => 'del_edge', - reading => 'node', - path => 'edge', - readings => 'nodes', - paths => 'edges', + paths => 'edges', }, - default => sub { Graph::Easy->new( undirected => 0 ) }, ); - + +has 'relations' => ( + is => 'ro', + isa => 'Text::Tradition::Collation::RelationshipStore', + handles => { + relationships => 'relationships', + related_readings => 'related_readings', + get_relationship => 'get_relationship', + del_relationship => 'del_relationship', + equivalence => 'equivalence', + equivalence_graph => 'equivalence_graph', + }, + writer => '_set_relations', + ); has 'tradition' => ( - is => 'rw', - isa => 'Text::Tradition', - ); - -has 'svg' => ( - is => 'ro', - isa => 'Str', - writer => '_save_svg', - predicate => 'has_svg', - ); - -has 'graphviz' => ( - is => 'ro', - isa => 'Str', - writer => '_save_graphviz', - predicate => 'has_graphviz', - ); - -has 'graphml' => ( is => 'ro', - isa => 'XML::LibXML::Document', - writer => '_save_graphml', - predicate => 'has_graphml', + isa => 'Text::Tradition', + writer => '_set_tradition', + weak_ref => 1, ); -# Keeps track of the lemmas within the collation. At most one lemma -# per position in the graph. -has 'lemmata' => ( - is => 'ro', - isa => 'HashRef[Maybe[Str]]', +has 'readings' => ( + isa => 'HashRef[Text::Tradition::Collation::Reading]', + traits => ['Hash'], + handles => { + reading => 'get', + _add_reading => 'set', + del_reading => 'delete', + has_reading => 'exists', + readings => 'values', + }, default => sub { {} }, - ); + ); has 'wit_list_separator' => ( is => 'rw', @@ -68,125 +70,955 @@ has 'baselabel' => ( default => 'base text', ); -has 'collapsed' => ( - is => 'rw', - isa => 'Bool', - ); +has 'linear' => ( + is => 'rw', + isa => 'Bool', + default => 1, + ); + +has 'ac_label' => ( + is => 'rw', + isa => 'Str', + default => ' (a.c.)', + ); + +has 'wordsep' => ( + is => 'rw', + isa => 'Str', + default => ' ', + ); + +has 'start' => ( + is => 'ro', + isa => 'Text::Tradition::Collation::Reading', + writer => '_set_start', + weak_ref => 1, + ); + +has 'end' => ( + is => 'ro', + isa => 'Text::Tradition::Collation::Reading', + writer => '_set_end', + weak_ref => 1, + ); + +has 'cached_svg' => ( + is => 'rw', + isa => 'Str', + predicate => 'has_cached_svg', + clearer => 'wipe_svg', + ); + +has 'cached_table' => ( + is => 'rw', + isa => 'HashRef', + predicate => 'has_cached_table', + clearer => 'wipe_table', + ); + +has '_graphcalc_done' => ( + is => 'rw', + isa => 'Bool', + default => undef, + ); + +=head1 NAME + +Text::Tradition::Collation - a software model for a text collation + +=head1 SYNOPSIS + + use Text::Tradition; + my $t = Text::Tradition->new( + 'name' => 'this is a text', + 'input' => 'TEI', + 'file' => '/path/to/tei_parallel_seg_file.xml' ); + + my $c = $t->collation; + my @readings = $c->readings; + my @paths = $c->paths; + my @relationships = $c->relationships; + + my $svg_variant_graph = $t->collation->as_svg(); + +=head1 DESCRIPTION +Text::Tradition is a library for representation and analysis of collated +texts, particularly medieval ones. The Collation is the central feature of +a Tradition, where the text, its sequence of readings, and its relationships +between readings are actually kept. -# The collation can be created two ways: -# 1. Collate a set of witnesses (with CollateX I guess) and process -# the results as in 2. -# 2. Read a pre-prepared collation in one of a variety of formats, -# and make the graph from that. +=head1 CONSTRUCTOR -# The graph itself will (for now) be immutable, and the positions -# within the graph will also be immutable. We need to calculate those -# positions upon graph construction. The equivalences between graph -# nodes will be mutable, entirely determined by the user (or possibly -# by some semantic pre-processing provided by the user.) So the -# constructor should just make an empty equivalences object. The -# constructor will also need to make the witness objects, if we didn't -# come through option 1. +=head2 new -sub BUILD { - my( $self, $args ) = @_; - $self->graph->use_class('node', 'Text::Tradition::Collation::Reading'); +The constructor. Takes a hash or hashref of the following arguments: - # Pass through any graph-specific options. - my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse'; - $self->graph->set_attribute( 'node', 'shape', $shape ); -} +=over + +=item * tradition - The Text::Tradition object to which the collation +belongs. Required. + +=item * linear - Whether the collation should be linear; that is, whether +transposed readings should be treated as two linked readings rather than one, +and therefore whether the collation graph is acyclic. Defaults to true. + +=item * baselabel - The default label for the path taken by a base text +(if any). Defaults to 'base text'. + +=item * wit_list_separator - The string to join a list of witnesses for +purposes of making labels in display graphs. Defaults to ', '. + +=item * ac_label - The extra label to tack onto a witness sigil when +representing another layer of path for the given witness - that is, when +a text has more than one possible reading due to scribal corrections or +the like. Defaults to ' (a.c.)'. + +=item * wordsep - The string used to separate words in the original text. +Defaults to ' '. + +=back + +=head1 ACCESSORS + +=head2 tradition + +=head2 linear + +=head2 wit_list_separator + +=head2 baselabel -# Wrapper around add_path +=head2 ac_label -around add_path => sub { - my $orig = shift; +=head2 wordsep + +Simple accessors for collation attributes. + +=head2 start + +The meta-reading at the start of every witness path. + +=head2 end + +The meta-reading at the end of every witness path. + +=head2 readings + +Returns all Reading objects in the graph. + +=head2 reading( $id ) + +Returns the Reading object corresponding to the given ID. + +=head2 add_reading( $reading_args ) + +Adds a new reading object to the collation. +See L for the available arguments. + +=head2 del_reading( $object_or_id ) + +Removes the given reading from the collation, implicitly removing its +paths and relationships. + +=head2 merge_readings( $main, $second, $concatenate, $with_str ) + +Merges the $second reading into the $main one. If $concatenate is true, then +the merged node will carry the text of both readings, concatenated with either +$with_str (if specified) or a sensible default (the empty string if the +appropriate 'join_*' flag is set on either reading, or else $self->wordsep.) + +The first two arguments may be either readings or reading IDs. + +=head2 has_reading( $id ) + +Predicate to see whether a given reading ID is in the graph. + +=head2 reading_witnesses( $object_or_id ) + +Returns a list of sigils whose witnesses contain the reading. + +=head2 paths + +Returns all reading paths within the document - that is, all edges in the +collation graph. Each path is an arrayref of [ $source, $target ] reading IDs. + +=head2 add_path( $source, $target, $sigil ) + +Links the given readings in the collation in sequence, under the given witness +sigil. The readings may be specified by object or ID. + +=head2 del_path( $source, $target, $sigil ) + +Links the given readings in the collation in sequence, under the given witness +sigil. The readings may be specified by object or ID. + +=head2 has_path( $source, $target ); + +Returns true if the two readings are linked in sequence in any witness. +The readings may be specified by object or ID. + +=head2 relationships + +Returns all Relationship objects in the collation. + +=head2 add_relationship( $reading, $other_reading, $options ) + +Adds a new relationship of the type given in $options between the two readings, +which may be specified by object or ID. Returns a value of ( $status, @vectors) +where $status is true on success, and @vectors is a list of relationship edges +that were ultimately added. +See L for the available options. + +=cut + +sub BUILD { my $self = shift; + $self->_set_relations( Text::Tradition::Collation::RelationshipStore->new( 'collation' => $self ) ); + $self->_set_start( $self->add_reading( + { 'collation' => $self, 'is_start' => 1, 'init' => 1 } ) ); + $self->_set_end( $self->add_reading( + { 'collation' => $self, 'is_end' => 1, 'init' => 1 } ) ); +} - # Make sure there are three arguments - unless( @_ == 3 ) { - warn "Call add_path with args source, target, witness"; - return; - } - # Make sure the proposed path does not yet exist - my( $source, $target, $wit ) = @_; - $source = $self->reading( $source ) - unless ref( $source ) eq 'Text::Tradition::Collation::Reading'; - $target = $self->reading( $target ) - unless ref( $target ) eq 'Text::Tradition::Collation::Reading'; - foreach my $path ( $source->edges_to( $target ) ) { - if( $path->label eq $wit ) { - return; +### Reading construct/destruct functions + +sub add_reading { + my( $self, $reading ) = @_; + unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) { + my %args = %$reading; + if( $args{'init'} ) { + # If we are initializing an empty collation, don't assume that we + # have set a tradition. + delete $args{'init'}; + } elsif( $self->tradition->has_language && !exists $args{'language'} ) { + $args{'language'} = $self->tradition->language; + } + $reading = Text::Tradition::Collation::Reading->new( + 'collation' => $self, + %args ); } - } - # Do the deed - $self->$orig( @_ ); + # First check to see if a reading with this ID exists. + if( $self->reading( $reading->id ) ) { + throw( "Collation already has a reading with id " . $reading->id ); + } + $self->_graphcalc_done(0); + $self->_add_reading( $reading->id => $reading ); + # Once the reading has been added, put it in both graphs. + $self->sequence->add_vertex( $reading->id ); + $self->relations->add_reading( $reading->id ); + return $reading; }; -# Wrapper around merge_nodes +around del_reading => sub { + my $orig = shift; + my $self = shift; + my $arg = shift; + + if( ref( $arg ) eq 'Text::Tradition::Collation::Reading' ) { + $arg = $arg->id; + } + # Remove the reading from the graphs. + $self->_graphcalc_done(0); + $self->_clear_cache; # Explicitly clear caches to GC the reading + $self->sequence->delete_vertex( $arg ); + $self->relations->delete_reading( $arg ); + + # Carry on. + $self->$orig( $arg ); +}; + +=begin testing + +use Text::Tradition; + +my $cxfile = 't/data/Collatex-16.xml'; +my $t = Text::Tradition->new( + 'name' => 'inline', + 'input' => 'CollateX', + 'file' => $cxfile, + ); +my $c = $t->collation; + +my $rno = scalar $c->readings; +# Split n21 for testing purposes +my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } ); +my $old_r = $c->reading( 'n21' ); +$old_r->alter_text( 'to' ); +$c->del_path( 'n20', 'n21', 'A' ); +$c->add_path( 'n20', 'n21p0', 'A' ); +$c->add_path( 'n21p0', 'n21', 'A' ); +$c->flatten_ranks(); +ok( $c->reading( 'n21p0' ), "New reading exists" ); +is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" ); + +# Combine n3 and n4 ( with his ) +$c->merge_readings( 'n3', 'n4', 1 ); +ok( !$c->reading('n4'), "Reading n4 is gone" ); +is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" ); + +# Collapse n9 and n10 ( rood / root ) +$c->merge_readings( 'n9', 'n10' ); +ok( !$c->reading('n10'), "Reading n10 is gone" ); +is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" ); + +# Combine n21 and n21p0 +my $remaining = $c->reading('n21'); +$remaining ||= $c->reading('n22'); # one of these should still exist +$c->merge_readings( 'n21p0', $remaining, 1 ); +ok( !$c->reading('n21'), "Reading $remaining is gone" ); +is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" ); + +=end testing + +=cut sub merge_readings { - my $self = shift; - my $first_node = shift; - my $second_node = shift; - $first_node->merge_from( $second_node ); - unshift( @_, $first_node, $second_node ); - return $self->graph->merge_nodes( @_ ); + my $self = shift; + + # Sanity check + my( $kept_obj, $del_obj, $combine, $combine_char ) = $self->_objectify_args( @_ ); + my $mergemeta = $kept_obj->is_meta; + throw( "Cannot merge meta and non-meta reading" ) + unless ( $mergemeta && $del_obj->is_meta ) + || ( !$mergemeta && !$del_obj->is_meta ); + if( $mergemeta ) { + throw( "Cannot merge with start or end node" ) + if( $kept_obj eq $self->start || $kept_obj eq $self->end + || $del_obj eq $self->start || $del_obj eq $self->end ); + } + # We only need the IDs for adding paths to the graph, not the reading + # objects themselves. + my $kept = $kept_obj->id; + my $deleted = $del_obj->id; + $self->_graphcalc_done(0); + + # The kept reading should inherit the paths and the relationships + # of the deleted reading. + foreach my $path ( $self->sequence->edges_at( $deleted ) ) { + my @vector = ( $kept ); + push( @vector, $path->[1] ) if $path->[0] eq $deleted; + unshift( @vector, $path->[0] ) if $path->[1] eq $deleted; + next if $vector[0] eq $vector[1]; # Don't add a self loop + my %wits = %{$self->sequence->get_edge_attributes( @$path )}; + $self->sequence->add_edge( @vector ); + my $fwits = $self->sequence->get_edge_attributes( @vector ); + @wits{keys %$fwits} = values %$fwits; + $self->sequence->set_edge_attributes( @vector, \%wits ); + } + $self->relations->merge_readings( $kept, $deleted, $combine ); + + # Do the deletion deed. + if( $combine ) { + # Combine the text of the readings + my $joinstr = $combine_char; + unless( defined $joinstr ) { + $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior; + $joinstr = $self->wordsep unless defined $joinstr; + } + $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) ); + # Change this reading to a joining one if necessary + $kept_obj->_set_join_next( $del_obj->join_next ); + $kept_obj->normal_form( + join( $joinstr, $kept_obj->normal_form, $del_obj->normal_form ) ); + # Combine the lexemes present in the readings + if( $kept_obj->has_lexemes && $del_obj->has_lexemes ) { + $kept_obj->add_lexeme( $del_obj->lexemes ); + } + } + $self->del_reading( $deleted ); } -=head2 Output method(s) +=head2 compress_readings -=over +Where possible in the graph, compresses plain sequences of readings into a +single reading. The sequences must consist of readings with no +relationships to other readings, with only a single witness path between +them and no other witness paths from either that would skip the other. The +readings must also not be marked as nonsense or bad grammar. + +WARNING: This operation cannot be undone. + +=cut + +sub compress_readings { + my $self = shift; + # Anywhere in the graph that there is a reading that joins only to a single + # successor, and neither of these have any relationships, just join the two + # readings. + my %gobbled; + foreach my $rdg ( sort { $a->rank <=> $b->rank } $self->readings ) { + next if $rdg->is_meta; + next if $gobbled{$rdg->id}; + next if $rdg->grammar_invalid || $rdg->is_nonsense; + next if $rdg->related_readings(); + my %seen; + while( $self->sequence->successors( $rdg ) == 1 ) { + my( $next ) = $self->reading( $self->sequence->successors( $rdg ) ); + throw( "Infinite loop" ) if $seen{$next->id}; + $seen{$next->id} = 1; + last if $self->sequence->predecessors( $next ) > 1; + last if $next->is_meta; + last if $next->grammar_invalid || $next->is_nonsense; + last if $next->related_readings(); + say "Joining readings $rdg and $next"; + $self->merge_readings( $rdg, $next, 1 ); + } + } + # Make sure we haven't screwed anything up + foreach my $wit ( $self->tradition->witnesses ) { + my $pathtext = $self->path_text( $wit->sigil ); + my $origtext = join( ' ', @{$wit->text} ); + throw( "Text differs for witness " . $wit->sigil ) + unless $pathtext eq $origtext; + if( $wit->is_layered ) { + $pathtext = $self->path_text( $wit->sigil.$self->ac_label ); + $origtext = join( ' ', @{$wit->layertext} ); + throw( "Ante-corr text differs for witness " . $wit->sigil ) + unless $pathtext eq $origtext; + } + } + + $self->relations->rebuild_equivalence(); + $self->calculate_ranks(); +} + +# Helper function for manipulating the graph. +sub _stringify_args { + my( $self, $first, $second, @args ) = @_; + $first = $first->id + if ref( $first ) eq 'Text::Tradition::Collation::Reading'; + $second = $second->id + if ref( $second ) eq 'Text::Tradition::Collation::Reading'; + return( $first, $second, @args ); +} + +# Helper function for manipulating the graph. +sub _objectify_args { + my( $self, $first, $second, $arg ) = @_; + $first = $self->reading( $first ) + unless ref( $first ) eq 'Text::Tradition::Collation::Reading'; + $second = $self->reading( $second ) + unless ref( $second ) eq 'Text::Tradition::Collation::Reading'; + return( $first, $second, $arg ); +} +### Path logic + +sub add_path { + my $self = shift; + + # We only need the IDs for adding paths to the graph, not the reading + # objects themselves. + my( $source, $target, $wit ) = $self->_stringify_args( @_ ); + + $self->_graphcalc_done(0); + # Connect the readings + unless( $self->sequence->has_edge( $source, $target ) ) { + $self->sequence->add_edge( $source, $target ); + $self->relations->add_equivalence_edge( $source, $target ); + } + # Note the witness in question + $self->sequence->set_edge_attribute( $source, $target, $wit, 1 ); +} + +sub del_path { + my $self = shift; + my @args; + if( ref( $_[0] ) eq 'ARRAY' ) { + my $e = shift @_; + @args = ( @$e, @_ ); + } else { + @args = @_; + } + + # We only need the IDs for adding paths to the graph, not the reading + # objects themselves. + my( $source, $target, $wit ) = $self->_stringify_args( @args ); + + $self->_graphcalc_done(0); + if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) { + $self->sequence->delete_edge_attribute( $source, $target, $wit ); + } + unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) { + $self->sequence->delete_edge( $source, $target ); + $self->relations->delete_equivalence_edge( $source, $target ); + } +} + + +# Extra graph-alike utility +sub has_path { + my $self = shift; + my( $source, $target, $wit ) = $self->_stringify_args( @_ ); + return undef unless $self->sequence->has_edge( $source, $target ); + return $self->sequence->has_edge_attribute( $source, $target, $wit ); +} + +=head2 clear_witness( @sigil_list ) + +Clear the given witnesses out of the collation entirely, removing references +to them in paths, and removing readings that belong only to them. Should only +be called via $tradition->del_witness. + +=cut + +sub clear_witness { + my( $self, @sigils ) = @_; -=item B + $self->_graphcalc_done(0); + # Clear the witness(es) out of the paths + foreach my $e ( $self->paths ) { + foreach my $sig ( @sigils ) { + $self->del_path( $e, $sig ); + } + } + + # Clear out the newly unused readings + foreach my $r ( $self->readings ) { + unless( $self->reading_witnesses( $r ) ) { + $self->del_reading( $r ); + } + } +} + +sub add_relationship { + my $self = shift; + my( $source, $target, $opts ) = $self->_stringify_args( @_ ); + my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts ); + $self->_graphcalc_done(0); + return @vectors; +} + +around qw/ get_relationship del_relationship / => sub { + my $orig = shift; + my $self = shift; + my @args = @_; + if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) { + @args = @{$_[0]}; + } + my( $source, $target ) = $self->_stringify_args( @args ); + $self->$orig( $source, $target ); +}; + +=head2 reading_witnesses( $reading ) -print $graph->as_svg( $recalculate ); +Return a list of sigils corresponding to the witnesses in which the reading appears. -Returns an SVG string that represents the graph. Uses GraphViz to do -this, because Graph::Easy doesn\'t cope well with long graphs. Unless -$recalculate is passed (and is a true value), the method will return a -cached copy of the SVG after the first call to the method. +=cut + +sub reading_witnesses { + my( $self, $reading ) = @_; + # We need only check either the incoming or the outgoing edges; I have + # arbitrarily chosen "incoming". Thus, special-case the start node. + if( $reading eq $self->start ) { + return map { $_->sigil } $self->tradition->witnesses; + } + my %all_witnesses; + foreach my $e ( $self->sequence->edges_to( $reading ) ) { + my $wits = $self->sequence->get_edge_attributes( @$e ); + @all_witnesses{ keys %$wits } = 1; + } + my $acstr = $self->ac_label; + foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) { + delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit}; + } + return keys %all_witnesses; +} + +=head1 OUTPUT METHODS + +=head2 as_svg( \%options ) + +Returns an SVG string that represents the graph, via as_dot and graphviz. +See as_dot for a list of options. Must have GraphViz (dot) installed to run. =cut sub as_svg { - my( $self, $recalc ) = @_; - return $self->svg if $self->has_svg; - - $self->collapse_graph_edges(); - $self->_save_graphviz( $self->graph->as_graphviz() ) - unless( $self->has_graphviz && !$recalc ); + my( $self, $opts ) = @_; + throw( "Need GraphViz installed to output SVG" ) + unless File::Which::which( 'dot' ); + my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'}; + $self->calculate_ranks() + unless( $self->_graphcalc_done || $opts->{'nocalc'} || !$self->linear ); + if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) { + my @cmd = qw/dot -Tsvg/; + my( $svg, $err ); + my $dotfile = File::Temp->new(); + ## USE FOR DEBUGGING + # $dotfile->unlink_on_destroy(0); + binmode $dotfile, ':utf8'; + print $dotfile $self->as_dot( $opts ); + push( @cmd, $dotfile->filename ); + run( \@cmd, ">", binary(), \$svg ); + $svg = decode_utf8( $svg ); + $self->cached_svg( $svg ) unless $want_subgraph; + return $svg; + } else { + return $self->cached_svg; + } +} + + +=head2 as_dot( \%options ) + +Returns a string that is the collation graph expressed in dot +(i.e. GraphViz) format. Options include: + +=over 4 + +=item * from + +=item * to + +=item * color_common + +=back + +=cut + +sub as_dot { + my( $self, $opts ) = @_; + my $startrank = $opts->{'from'} if $opts; + my $endrank = $opts->{'to'} if $opts; + my $color_common = $opts->{'color_common'} if $opts; + my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank + && $self->end->rank > 100; + $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs + + # Check the arguments + if( $startrank ) { + return if $endrank && $startrank > $endrank; + return if $startrank > $self->end->rank; + } + if( defined $endrank ) { + return if $endrank < 0; + $endrank = undef if $endrank == $self->end->rank; + } + + my $graph_name = $self->tradition->name; + $graph_name =~ s/[^\w\s]//g; + $graph_name = join( '_', split( /\s+/, $graph_name ) ); + + my %graph_attrs = ( + 'rankdir' => 'LR', + 'bgcolor' => 'none', + ); + my %node_attrs = ( + 'fontsize' => 14, + 'fillcolor' => 'white', + 'style' => 'filled', + 'shape' => 'ellipse' + ); + my %edge_attrs = ( + 'arrowhead' => 'open', + 'color' => '#000000', + 'fontcolor' => '#000000', + ); + + my $dot = sprintf( "digraph %s {\n", $graph_name ); + $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n"; + $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n"; + + # Output substitute start/end readings if necessary + if( $startrank ) { + $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n"; + } + if( $endrank ) { + $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n"; + } + if( $STRAIGHTENHACK ) { + ## HACK part 1 + my $startlabel = $startrank ? '__SUBSTART__' : '__START__'; + $dot .= "\tsubgraph { rank=same \"$startlabel\" \"#SILENT#\" }\n"; + $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];" + } + my %used; # Keep track of the readings that actually appear in the graph + # Sort the readings by rank if we have ranks; this speeds layout. + my @all_readings = $self->end->has_rank + ? sort { $a->rank <=> $b->rank } $self->readings + : $self->readings; + # TODO Refrain from outputting lacuna nodes - just grey out the edges. + foreach my $reading ( @all_readings ) { + # Only output readings within our rank range. + next if $startrank && $reading->rank < $startrank; + next if $endrank && $reading->rank > $endrank; + $used{$reading->id} = 1; + # Need not output nodes without separate labels + next if $reading->id eq $reading->text; + my $rattrs; + my $label = $reading->text; + $label .= '-' if $reading->join_next; + $label = "-$label" if $reading->join_prior; + $label =~ s/\"/\\\"/g; + $rattrs->{'label'} = $label; + $rattrs->{'id'} = $reading->id; + $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common; + $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) ); + } - my @cmd = qw/dot -Tsvg/; - my( $svg, $err ); - my $in = $self->graphviz; - run( \@cmd, \$in, ">", binary(), \$svg ); - $self->{'svg'} = $svg; - $self->expand_graph_edges(); - return $svg; + # Add the real edges. Need to weight one edge per rank jump, in a + # continuous line. + # my $weighted = $self->_add_edge_weights; + my @edges = $self->paths; + my( %substart, %subend ); + foreach my $edge ( @edges ) { + # Do we need to output this edge? + if( $used{$edge->[0]} && $used{$edge->[1]} ) { + my $label = $self->_path_display_label( $self->path_witnesses( $edge ) ); + my $variables = { %edge_attrs, 'label' => $label }; + + # Account for the rank gap if necessary + my $rank0 = $self->reading( $edge->[0] )->rank + if $self->reading( $edge->[0] )->has_rank; + my $rank1 = $self->reading( $edge->[1] )->rank + if $self->reading( $edge->[1] )->has_rank; + if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) { + $variables->{'minlen'} = $rank1 - $rank0; + } + + # Add the calculated edge weights + # if( exists $weighted->{$edge->[0]} + # && $weighted->{$edge->[0]} eq $edge->[1] ) { + # # $variables->{'color'} = 'red'; + # $variables->{'weight'} = 3.0; + # } + + # EXPERIMENTAL: make edge width reflect no. of witnesses + my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2; + $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit + + my $varopts = _dot_attr_string( $variables ); + $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n", + $edge->[0], $edge->[1], $varopts ); + } elsif( $used{$edge->[0]} ) { + $subend{$edge->[0]} = $edge->[1]; + } elsif( $used{$edge->[1]} ) { + $substart{$edge->[1]} = $edge->[0]; + } + } + # Add substitute start and end edges if necessary + foreach my $node ( keys %substart ) { + my $witstr = $self->_path_display_label ( $self->path_witnesses( $substart{$node}, $node ) ); + my $variables = { %edge_attrs, 'label' => $witstr }; + my $nrdg = $self->reading( $node ); + if( $nrdg->has_rank && $nrdg->rank > $startrank ) { + # Substart is actually one lower than $startrank + $variables->{'minlen'} = $nrdg->rank - ( $startrank - 1 ); + } + my $varopts = _dot_attr_string( $variables ); + $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n"; + } + foreach my $node ( keys %subend ) { + my $witstr = $self->_path_display_label ( $self->path_witnesses( $node, $subend{$node} ) ); + my $variables = { %edge_attrs, 'label' => $witstr }; + my $varopts = _dot_attr_string( $variables ); + $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n"; + } + # HACK part 2 + if( $STRAIGHTENHACK ) { + my $endlabel = $endrank ? '__SUBEND__' : '__END__'; + $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; + } + + $dot .= "}\n"; + return $dot; } -=item B +sub _dot_attr_string { + my( $hash ) = @_; + my @attrs; + foreach my $k ( sort keys %$hash ) { + my $v = $hash->{$k}; + push( @attrs, $k.'="'.$v.'"' ); + } + return( '[ ' . join( ', ', @attrs ) . ' ]' ); +} + +sub _add_edge_weights { + my $self = shift; + # Walk the graph from START to END, choosing the successor node with + # the largest number of witness paths each time. + my $weighted = {}; + my $curr = $self->start->id; + my $ranked = $self->end->has_rank; + while( $curr ne $self->end->id ) { + my $rank = $ranked ? $self->reading( $curr )->rank : 0; + my @succ = sort { $self->path_witnesses( $curr, $a ) + <=> $self->path_witnesses( $curr, $b ) } + $self->sequence->successors( $curr ); + my $next = pop @succ; + my $nextrank = $ranked ? $self->reading( $next )->rank : 0; + # Try to avoid lacunae in the weighted path. + while( @succ && + ( $self->reading( $next )->is_lacuna || + $nextrank - $rank > 1 ) ){ + $next = pop @succ; + } + $weighted->{$curr} = $next; + $curr = $next; + } + return $weighted; +} -print $graph->as_graphml( $recalculate ) +=head2 path_witnesses( $edge ) -Returns a GraphML representation of the collation graph, with -transposition information and position information. Unless -$recalculate is passed (and is a true value), the method will return a -cached copy of the SVG after the first call to the method. +Returns the list of sigils whose witnesses are associated with the given edge. +The edge can be passed as either an array or an arrayref of ( $source, $target ). =cut -sub as_graphml { - my( $self, $recalc ) = @_; - return $self->graphml if $self->has_graphml; +sub path_witnesses { + my( $self, @edge ) = @_; + # If edge is an arrayref, cope. + if( @edge == 1 && ref( $edge[0] ) eq 'ARRAY' ) { + my $e = shift @edge; + @edge = @$e; + } + my @wits = keys %{$self->sequence->get_edge_attributes( @edge )}; + return @wits; +} + +# Helper function. Make a display label for the given witnesses, showing a.c. +# witnesses only where the main witness is not also in the list. +sub _path_display_label { + my $self = shift; + my %wits; + map { $wits{$_} = 1 } @_; + + # If an a.c. wit is listed, remove it if the main wit is also listed. + # Otherwise keep it for explicit listing. + my $aclabel = $self->ac_label; + my @disp_ac; + foreach my $w ( sort keys %wits ) { + if( $w =~ /^(.*)\Q$aclabel\E$/ ) { + if( exists $wits{$1} ) { + delete $wits{$w}; + } else { + push( @disp_ac, $w ); + } + } + } + + # See if we are in a majority situation. + my $maj = scalar( $self->tradition->witnesses ) * 0.6; + $maj = $maj > 5 ? $maj : 5; + if( scalar keys %wits > $maj ) { + unshift( @disp_ac, 'majority' ); + return join( ', ', @disp_ac ); + } else { + return join( ', ', sort keys %wits ); + } +} + +=head2 readings_at_rank( $rank ) + +Returns a list of readings at a given rank, taken from the alignment table. + +=cut + +sub readings_at_rank { + my( $self, $rank ) = @_; + my $table = $self->alignment_table; + # Table rank is real rank - 1. + my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}}; + my %readings; + foreach my $e ( @elements ) { + next unless ref( $e ) eq 'HASH'; + next unless exists $e->{'t'}; + $readings{$e->{'t'}->id} = $e->{'t'}; + } + return values %readings; +} + +=head2 as_graphml + +Returns a GraphML representation of the collation. The GraphML will contain +two graphs. The first expresses the attributes of the readings and the witness +paths that link them; the second expresses the relationships that link the +readings. This is the native transfer format for a tradition. + +=begin testing + +use Text::Tradition; + +my $READINGS = 311; +my $PATHS = 361; + +my $datafile = 't/data/florilegium_tei_ps.xml'; +my $tradition = Text::Tradition->new( 'input' => 'TEI', + 'name' => 'test0', + 'file' => $datafile, + 'linear' => 1 ); + +ok( $tradition, "Got a tradition object" ); +is( scalar $tradition->witnesses, 13, "Found all witnesses" ); +ok( $tradition->collation, "Tradition has a collation" ); + +my $c = $tradition->collation; +is( scalar $c->readings, $READINGS, "Collation has all readings" ); +is( scalar $c->paths, $PATHS, "Collation has all paths" ); +is( scalar $c->relationships, 0, "Collation has all relationships" ); + +# Add a few relationships +$c->add_relationship( 'w123', 'w125', { 'type' => 'collated' } ); +$c->add_relationship( 'w193', 'w196', { 'type' => 'collated' } ); +$c->add_relationship( 'w257', 'w262', { 'type' => 'transposition' } ); + +# Now write it to GraphML and parse it again. + +my $graphml = $c->as_graphml; +my $st = Text::Tradition->new( 'input' => 'Self', 'string' => $graphml ); +is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all readings" ); +is( scalar $st->collation->paths, $PATHS, "Reparsed collation has all paths" ); +is( scalar $st->collation->relationships, 3, "Reparsed collation has new relationships" ); + +# Now add a stemma, write to GraphML, and look at the output. +my $stemma = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' ); +is( ref( $stemma ), 'Text::Tradition::Stemma', "Parsed dotfile into stemma" ); +is( $tradition->stemmata, 1, "Tradition now has the stemma" ); +$graphml = $c->as_graphml; +like( $graphml, qr/digraph/, "Digraph declaration exists in GraphML" ); + +# Now add a user, write to GraphML, and look at the output. +unlike( $graphml, qr/testuser/, "Test user name does not exist in GraphML yet" ); +my $testuser = Text::Tradition::User->new( + id => 'testuser', password => 'testpass' ); +is( ref( $testuser ), 'Text::Tradition::User', "Created test user object" ); +$testuser->add_tradition( $tradition ); +is( $tradition->user->id, $testuser->id, "Tradition assigned to test user" ); +$graphml = $c->as_graphml; +like( $graphml, qr/testuser/, "Test user name now exists in GraphML" ); + +=end testing + +=cut +sub as_graphml { + my( $self, $options ) = @_; + $self->calculate_ranks unless $self->_graphcalc_done; + + my $start = $options->{'from'} + ? $self->reading( $options->{'from'} ) : $self->start; + my $end = $options->{'to'} + ? $self->reading( $options->{'to'} ) : $self->end; + if( $start->has_rank && $end->has_rank && $end->rank < $start->rank ) { + throw( 'Start node must be before end node' ); + } + # The readings need to be ranked for this to work. + $start = $self->start unless $start->has_rank; + $end = $self->end unless $end->has_rank; + my $rankoffset = 0; + unless( $start eq $self->start ) { + $rankoffset = $start->rank - 1; + } + my %use_readings; + # Some namespaces my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns'; my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance'; my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' . - 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; + 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; # Create the document and root node my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" ); @@ -194,227 +1026,379 @@ sub as_graphml { $graphml->setDocumentElement( $root ); $root->setNamespace( $xsi_ns, 'xsi', 0 ); $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema ); - - # Add the data keys for nodes - my @node_data = ( 'name', 'token', 'identical', 'position' ); - foreach my $ndi ( 0 .. $#node_data ) { - my $key = $root->addNewChild( $graphml_ns, 'key' ); - $key->setAttribute( 'attr.name', $node_data[$ndi] ); - $key->setAttribute( 'attr.type', 'string' ); - $key->setAttribute( 'for', 'node' ); - $key->setAttribute( 'id', 'd'.$ndi ); + + # List of attribute types to save on our objects and their corresponding + # GraphML types + my %save_types = ( + 'Str' => 'string', + 'Int' => 'int', + 'Bool' => 'boolean', + 'ReadingID' => 'string', + 'RelationshipType' => 'string', + 'RelationshipScope' => 'string', + ); + + # List of attribute names *not* to save on our objects. + # We will also not save any attribute beginning with _. + my %skipsave; + map { $skipsave{$_} = 1 } qw/ cached_svg /; + + # Add the data keys for the graph. Include an extra key 'version' for the + # GraphML output version. + my %graph_data_keys; + my $gdi = 0; + my %graph_attributes = ( 'version' => 'string' ); + # Graph attributes include those of Tradition and those of Collation. + my %gattr_from; + my $tmeta = $self->tradition->meta; + my $cmeta = $self->meta; + map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes; + map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes; + foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; + } + # Extra custom keys for complex objects that should be saved in some form. + # The subroutine should return a string, or undef/empty. + $graph_attributes{'stemmata'} = sub { + my @stemstrs; + map { push( @stemstrs, $_->editable( {linesep => ''} ) ) } + $self->tradition->stemmata; + join( "\n", @stemstrs ); + }; + $graph_attributes{'user'} = sub { + $self->tradition->user ? $self->tradition->user->id : undef + }; + + foreach my $datum ( sort keys %graph_attributes ) { + $graph_data_keys{$datum} = 'dg'.$gdi++; + my $key = $root->addNewChild( $graphml_ns, 'key' ); + my $dtype = ref( $graph_attributes{$datum} ) ? 'string' + : $graph_attributes{$datum}; + $key->setAttribute( 'attr.name', $datum ); + $key->setAttribute( 'attr.type', $dtype ); + $key->setAttribute( 'for', 'graph' ); + $key->setAttribute( 'id', $graph_data_keys{$datum} ); } - # Add the data keys for edges - my %wit_hash; - my $wit_ctr = 0; - foreach my $wit ( @{$self->tradition->witnesses} ) { - my $wit_key = 'w' . $wit_ctr++; - $wit_hash{$wit} = $wit_key; - my $key = $root->addNewChild( $graphml_ns, 'key' ); - $key->setAttribute( 'attr.name', $wit ); - $key->setAttribute( 'attr.type', 'string' ); - $key->setAttribute( 'for', 'edge' ); - $key->setAttribute( 'id', $wit_key ); + # Add the data keys for reading nodes + my %reading_attributes; + my $rmeta = Text::Tradition::Collation::Reading->meta; + foreach my $attr( $rmeta->get_all_attributes ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; + } + # Extra custom key for the reading morphology + $reading_attributes{'lexemes'} = 'string'; + + my %node_data_keys; + my $ndi = 0; + foreach my $datum ( sort keys %reading_attributes ) { + $node_data_keys{$datum} = 'dn'.$ndi++; + my $key = $root->addNewChild( $graphml_ns, 'key' ); + $key->setAttribute( 'attr.name', $datum ); + $key->setAttribute( 'attr.type', $reading_attributes{$datum} ); + $key->setAttribute( 'for', 'node' ); + $key->setAttribute( 'id', $node_data_keys{$datum} ); } - # Add the graph, its nodes, and its edges - $self->collapse_graph_edges(); - my $graph = $root->addNewChild( $graphml_ns, 'graph' ); - $graph->setAttribute( 'edgedefault', 'directed' ); - $graph->setAttribute( 'id', 'g0' ); # TODO make this meaningful - $graph->setAttribute( 'parse.edgeids', 'canonical' ); - $graph->setAttribute( 'parse.edges', $self->edges() ); - $graph->setAttribute( 'parse.nodeids', 'canonical' ); - $graph->setAttribute( 'parse.nodes', $self->nodes() ); - $graph->setAttribute( 'parse.order', 'nodesfirst' ); - - my $node_ctr = 0; - my %node_hash; - foreach my $n ( $self->readings ) { - my %this_node_data = (); - foreach my $ndi ( 0 .. $#node_data ) { - my $value; - $this_node_data{'d'.$ndi} = $n->name if $node_data[$ndi] eq 'name'; - $this_node_data{'d'.$ndi} = $n->label - if $node_data[$ndi] eq 'token'; - $this_node_data{'d'.$ndi} = $n->primary->name if $n->has_primary; - $this_node_data{'d'.$ndi} = - $self->{'positions'}->node_position( $n ) - if $node_data[$ndi] eq 'position'; - } - my $node_el = $graph->addNewChild( $graphml_ns, 'node' ); - my $node_xmlid = 'n' . $node_ctr++; - $node_hash{ $n->name } = $node_xmlid; - $node_el->setAttribute( 'id', $node_xmlid ); - - foreach my $dk ( keys %this_node_data ) { - my $d_el = $node_el->addNewChild( $graphml_ns, 'data' ); - $d_el->setAttribute( 'key', $dk ); - $d_el->appendTextChild( $this_node_data{$dk} ); + # Add the data keys for edges, that is, paths and relationships. Path + # data does not come from a Moose class so is here manually. + my $edi = 0; + my %edge_data_keys; + my %edge_attributes = ( + witness => 'string', # ID/label for a path + extra => 'boolean', # Path key + ); + my @path_attributes = keys %edge_attributes; # track our manual additions + my $pmeta = Text::Tradition::Collation::Relationship->meta; + foreach my $attr( $pmeta->get_all_attributes ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; } + foreach my $datum ( sort keys %edge_attributes ) { + $edge_data_keys{$datum} = 'de'.$edi++; + my $key = $root->addNewChild( $graphml_ns, 'key' ); + $key->setAttribute( 'attr.name', $datum ); + $key->setAttribute( 'attr.type', $edge_attributes{$datum} ); + $key->setAttribute( 'for', 'edge' ); + $key->setAttribute( 'id', $edge_data_keys{$datum} ); } - foreach my $e ( $self->paths() ) { - my( $name, $from, $to ) = ( $e->name, - $node_hash{ $e->from()->name() }, - $node_hash{ $e->to()->name() } ); - my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' ); - $edge_el->setAttribute( 'source', $from ); - $edge_el->setAttribute( 'target', $to ); - $edge_el->setAttribute( 'id', $name ); - # TODO Got to add the witnesses + # Add the collation graph itself. First, sanitize the name to a valid XML ID. + my $xmlidname = $self->tradition->name; + $xmlidname =~ s/(?!$xml10_namechar_rx)./_/g; + if( $xmlidname !~ /^$xml10_namestartchar_rx/ ) { + $xmlidname = '_'.$xmlidname; } + my $sgraph = $root->addNewChild( $graphml_ns, 'graph' ); + $sgraph->setAttribute( 'edgedefault', 'directed' ); + $sgraph->setAttribute( 'id', $xmlidname ); + $sgraph->setAttribute( 'parse.edgeids', 'canonical' ); + $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later + $sgraph->setAttribute( 'parse.nodeids', 'canonical' ); + $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later + $sgraph->setAttribute( 'parse.order', 'nodesfirst' ); + + # Tradition/collation attribute data + foreach my $datum ( keys %graph_attributes ) { + my $value; + if( $datum eq 'version' ) { + $value = '3.2'; + } elsif( ref( $graph_attributes{$datum} ) ) { + my $sub = $graph_attributes{$datum}; + $value = &$sub(); + } elsif( $gattr_from{$datum} eq 'Tradition' ) { + $value = $self->tradition->$datum; + } else { + $value = $self->$datum; + } + _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value ); + } - # Return the thing - $self->_save_graphml( $graphml ); - $self->expand_graph_edges(); - return $graphml; -} + my $node_ctr = 0; + my %node_hash; + # Add our readings to the graph + foreach my $n ( sort { $a->id cmp $b->id } $self->readings ) { + next if $n->has_rank && $n ne $self->start && $n ne $self->end && + ( $n->rank < $start->rank || $n->rank > $end->rank ); + $use_readings{$n->id} = 1; + # Add to the main graph + my $node_el = $sgraph->addNewChild( $graphml_ns, 'node' ); + my $node_xmlid = 'n' . $node_ctr++; + $node_hash{ $n->id } = $node_xmlid; + $node_el->setAttribute( 'id', $node_xmlid ); + foreach my $d ( keys %reading_attributes ) { + my $nval = $n->$d; + # Custom serialization + if( $d eq 'lexemes' ) { + # If nval is a true value, we have lexemes so we need to + # serialize them. Otherwise set nval to undef so that the + # key is excluded from this reading. + $nval = $nval ? $n->_serialize_lexemes : undef; + } elsif( $d eq 'normal_form' && $n->normal_form eq $n->text ) { + $nval = undef; + } + if( $rankoffset && $d eq 'rank' && $n ne $self->start ) { + # Adjust the ranks within the subgraph. + $nval = $n eq $self->end ? $end->rank - $rankoffset + 1 + : $nval - $rankoffset; + } + _add_graphml_data( $node_el, $node_data_keys{$d}, $nval ) + if defined $nval; + } + } -sub collapse_graph_edges { - my $self = shift; - # Our collation graph has an edge per witness. This is great for - # calculation purposes, but terrible for display. Thus we want to - # display only one edge between any two nodes. - - return if $self->collapsed; - - print STDERR "Collapsing path edges in graph...\n"; - - # Don't list out every witness if we have more than half to list. - my $majority = int( scalar( @{$self->tradition->witnesses} ) / 2 ) + 1; - foreach my $node( $self->readings ) { - my $newlabels = {}; - # We will visit each node, so we only look ahead. - foreach my $edge ( $node->outgoing() ) { - add_hash_entry( $newlabels, $edge->to->name, $edge->name ); - $self->del_path( $edge ); - } - - foreach my $newdest ( keys %$newlabels ) { - my $label; - my @compressed_wits = (); - if( @{$newlabels->{$newdest}} < $majority ) { - $label = join( ', ', @{$newlabels->{$newdest}} ); - } else { - ## TODO FIX THIS HACK - my @aclabels; - foreach my $wit ( @{$newlabels->{$newdest}} ) { - if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ ) { - push( @aclabels, $wit ); - } else { - push( @compressed_wits, $wit ); - } + # Add the path edges to the sequence graph + my $edge_ctr = 0; + foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) { + # We add an edge in the graphml for every witness in $e. + next unless( $use_readings{$e->[0]} || $use_readings{$e->[1]} ); + my @edge_wits = sort $self->path_witnesses( $e ); + $e->[0] = $self->start->id unless $use_readings{$e->[0]}; + $e->[1] = $self->end->id unless $use_readings{$e->[1]}; + # Skip any path from start to end; that witness is not in the subgraph. + next if ( $e->[0] eq $self->start->id && $e->[1] eq $self->end->id ); + foreach my $wit ( @edge_wits ) { + my( $id, $from, $to ) = ( 'e'.$edge_ctr++, + $node_hash{ $e->[0] }, + $node_hash{ $e->[1] } ); + my $edge_el = $sgraph->addNewChild( $graphml_ns, 'edge' ); + $edge_el->setAttribute( 'source', $from ); + $edge_el->setAttribute( 'target', $to ); + $edge_el->setAttribute( 'id', $id ); + + # It's a witness path, so add the witness + my $base = $wit; + my $key = $edge_data_keys{'witness'}; + # Is this an ante-corr witness? + my $aclabel = $self->ac_label; + if( $wit =~ /^(.*)\Q$aclabel\E$/ ) { + # Keep the base witness + $base = $1; + # ...and record that this is an 'extra' reading path + _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel ); + } + _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base ); } - $label = join( ', ', 'majority', @aclabels ); - } - - my $newedge = - $self->add_path( $node, $self->reading( $newdest ), $label ); - if( @compressed_wits ) { - ## TODO fix this hack too. - $newedge->set_attribute( 'class', - join( '|', @compressed_wits ) ); - } } - } + + # Report the actual number of nodes and edges that went in + $sgraph->setAttribute( 'parse.edges', $edge_ctr ); + $sgraph->setAttribute( 'parse.nodes', $node_ctr ); + + # Add the relationship graph to the XML + map { delete $edge_data_keys{$_} } @path_attributes; + $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash, + $node_data_keys{'id'}, \%edge_data_keys ); + + # Save and return the thing + my $result = decode_utf8( $graphml->toString(1) ); + return $result; +} - $self->collapsed( 1 ); +sub _add_graphml_data { + my( $el, $key, $value ) = @_; + return unless defined $value; + my $data_el = $el->addNewChild( $el->namespaceURI, 'data' ); + $data_el->setAttribute( 'key', $key ); + $data_el->appendText( $value ); } -sub expand_graph_edges { - my $self = shift; - # Our collation graph has only one edge between any two nodes. - # This is great for display, but not so great for analysis. - # Expand this so that each witness has its own edge between any - # two reading nodes. - return unless $self->collapsed; - - print STDERR "Expanding path edges in graph...\n"; +=head2 as_csv - foreach my $edge( $self->paths ) { - my $from = $edge->from; - my $to = $edge->to; - my @wits = split( /, /, $edge->label ); - if( grep { $_ eq 'majority' } @wits ) { - push( @wits, split( /\|/, $edge->get_attribute( 'class' ) ) ); - } - $self->del_path( $edge ); - foreach ( @wits ) { - $self->add_path( $from, $to, $_ ); - } +Returns a CSV alignment table representation of the collation graph, one +row per witness (or witness uncorrected.) + +=cut + +sub as_csv { + my( $self ) = @_; + my $table = $self->alignment_table; + my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } ); + my @result; + # Make the header row + $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} ); + push( @result, decode_utf8( $csv->string ) ); + # Make the rest of the rows + foreach my $idx ( 0 .. $table->{'length'} - 1 ) { + my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}}; + my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs; + $csv->combine( @row ); + push( @result, decode_utf8( $csv->string ) ); } - $self->collapsed( 0 ); + return join( "\n", @result ); } -=back - -=head2 Navigation methods +=head2 alignment_table( $use_refs, $include_witnesses ) -=over +Return a reference to an alignment table, in a slightly enhanced CollateX +format which looks like this: -=item B + $table = { alignment => [ { witness => "SIGIL", + tokens => [ { t => "TEXT" }, ... ] }, + { witness => "SIG2", + tokens => [ { t => "TEXT" }, ... ] }, + ... ], + length => TEXTLEN }; -my $beginning = $collation->start(); +If $use_refs is set to 1, the reading object is returned in the table +instead of READINGTEXT; if not, the text of the reading is returned. -Returns the beginning of the collation, a meta-reading with label '#START#'. +If $include_witnesses is set to a hashref, only the witnesses whose sigil +keys have a true hash value will be included. =cut -sub start { - # Return the beginning reading of the graph. - my $self = shift; - my( $new_start ) = @_; - if( $new_start ) { - $self->del_reading( '#START#' ); - $self->graph->rename_node( $new_start, '#START#' ); +sub alignment_table { + my( $self ) = @_; + $self->calculate_ranks() unless $self->_graphcalc_done; + return $self->cached_table if $self->has_cached_table; + + # Make sure we can do this + throw( "Need a linear graph in order to make an alignment table" ) + unless $self->linear; + $self->calculate_ranks unless $self->end->has_rank; + + my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 }; + my @all_pos = ( 1 .. $self->end->rank - 1 ); + foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) { + # say STDERR "Making witness row(s) for " . $wit->sigil; + my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); + my @row = _make_witness_row( \@wit_path, \@all_pos ); + push( @{$table->{'alignment'}}, + { 'witness' => $wit->sigil, 'tokens' => \@row } ); + if( $wit->is_layered ) { + my @wit_ac_path = $self->reading_sequence( $self->start, $self->end, + $wit->sigil.$self->ac_label ); + my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos ); + push( @{$table->{'alignment'}}, + { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } ); + } } - return $self->reading('#START#'); + $self->cached_table( $table ); + return $table; } -=item B +sub _make_witness_row { + my( $path, $positions ) = @_; + my %char_hash; + map { $char_hash{$_} = undef } @$positions; + my $debug = 0; + foreach my $rdg ( @$path ) { + my $rtext = $rdg->text; + $rtext = '#LACUNA#' if $rdg->is_lacuna; + say STDERR "rank " . $rdg->rank if $debug; + # say STDERR "No rank for " . $rdg->id unless defined $rdg->rank; + $char_hash{$rdg->rank} = { 't' => $rdg }; + } + my @row = map { $char_hash{$_} } @$positions; + # Fill in lacuna markers for undef spots in the row + my $last_el = shift @row; + my @filled_row = ( $last_el ); + foreach my $el ( @row ) { + # If we are using node reference, make the lacuna node appear many times + # in the table. If not, use the lacuna tag. + if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) { + $el = $last_el; + } + push( @filled_row, $el ); + $last_el = $el; + } + return @filled_row; +} -my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] ); +=head1 NAVIGATION METHODS + +=head2 reading_sequence( $first, $last, $sigil, $backup ) Returns the ordered list of readings, starting with $first and ending -with $last, along the given witness path. If no path is specified, -assume that the path is that of the base text (if any.) +with $last, for the witness given in $sigil. If a $backup sigil is +specified (e.g. when walking a layered witness), it will be used wherever +no $sigil path exists. If there is a base text reading, that will be +used wherever no path exists for $sigil or $backup. =cut +# TODO Think about returning some lazy-eval iterator. +# TODO Get rid of backup; we should know from what witness is whether we need it. + sub reading_sequence { - my( $self, $start, $end, $witness, $backup ) = @_; + my( $self, $start, $end, $witness ) = @_; $witness = $self->baselabel unless $witness; my @readings = ( $start ); my %seen; my $n = $start; - while( $n && $n ne $end ) { - if( exists( $seen{$n->name()} ) ) { - warn "Detected loop at " . $n->name(); - last; - } - $seen{$n->name()} = 1; - - my $next = $self->next_reading( $n, $witness, $backup ); - warn "Did not find any path for $witness from reading " . $n->name - unless $next; - push( @readings, $next ); - $n = $next; + while( $n && $n->id ne $end->id ) { + if( exists( $seen{$n->id} ) ) { + throw( "Detected loop for $witness at " . $n->id ); + } + $seen{$n->id} = 1; + + my $next = $self->next_reading( $n, $witness ); + unless( $next ) { + throw( "Did not find any path for $witness from reading " . $n->id ); + } + push( @readings, $next ); + $n = $next; } # Check that the last reading is our end reading. my $last = $readings[$#readings]; - warn "Last reading found from " . $start->label() . - " for witness $witness is not the end!" - unless $last eq $end; + throw( "Last reading found from " . $start->text . + " for witness $witness is not the end!" ) # TODO do we get this far? + unless $last->id eq $end->id; return @readings; } -=item B - -my $next_reading = $graph->next_reading( $reading, $witpath ); +=head2 next_reading( $reading, $sigil ); Returns the reading that follows the given reading along the given witness path. @@ -424,12 +1408,12 @@ path. sub next_reading { # Return the successor via the corresponding path. my $self = shift; - return $self->_find_linked_reading( 'next', @_ ); + my $answer = $self->_find_linked_reading( 'next', @_ ); + return undef unless $answer; + return $self->reading( $answer ); } -=item B - -my $prior_reading = $graph->prior_reading( $reading, $witpath ); +=head2 prior_reading( $reading, $sigil ) Returns the reading that precedes the given reading along the given witness path. @@ -439,46 +1423,55 @@ path. sub prior_reading { # Return the predecessor via the corresponding path. my $self = shift; - return $self->_find_linked_reading( 'prior', @_ ); + my $answer = $self->_find_linked_reading( 'prior', @_ ); + return $self->reading( $answer ); } sub _find_linked_reading { - my( $self, $direction, $node, $path, $alt_path ) = @_; + my( $self, $direction, $node, $path ) = @_; + + # Get a backup if we are dealing with a layered witness + my $alt_path; + my $aclabel = $self->ac_label; + if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) { + $alt_path = $1; + } + my @linked_paths = $direction eq 'next' - ? $node->outgoing() : $node->incoming(); + ? $self->sequence->edges_from( $node ) + : $self->sequence->edges_to( $node ); return undef unless scalar( @linked_paths ); # We have to find the linked path that contains all of the # witnesses supplied in $path. my( @path_wits, @alt_path_wits ); - @path_wits = $self->witnesses_of_label( $path ) if $path; - @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path; + @path_wits = sort( $self->_witnesses_of_label( $path ) ) if $path; + @alt_path_wits = sort( $self->_witnesses_of_label( $alt_path ) ) if $alt_path; my $base_le; my $alt_le; foreach my $le ( @linked_paths ) { - if( $le->name eq $self->baselabel ) { - $base_le = $le; - } else { - my @le_wits = $self->witnesses_of_label( $le->name ); - if( _is_within( \@path_wits, \@le_wits ) ) { - # This is the right path. - return $direction eq 'next' ? $le->to() : $le->from(); - } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) { - $alt_le = $le; - } - } + if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) { + $base_le = $le; + } + my @le_wits = sort $self->path_witnesses( $le ); + if( _is_within( \@path_wits, \@le_wits ) ) { + # This is the right path. + return $direction eq 'next' ? $le->[1] : $le->[0]; + } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) { + $alt_le = $le; + } } # Got this far? Return the alternate path if it exists. - return $direction eq 'next' ? $alt_le->to() : $alt_le->from() - if $alt_le; + return $direction eq 'next' ? $alt_le->[1] : $alt_le->[0] + if $alt_le; # Got this far? Return the base path if it exists. - return $direction eq 'next' ? $base_le->to() : $base_le->from() - if $base_le; + return $direction eq 'next' ? $base_le->[1] : $base_le->[0] + if $base_le; # Got this far? We have no appropriate path. - warn "Could not find $direction node from " . $node->label - . " along path $path"; + warn "Could not find $direction node from " . $node->id + . " along path $path"; return undef; } @@ -487,419 +1480,428 @@ sub _is_within { my( $set1, $set2 ) = @_; my $ret = @$set1; # will be 0, i.e. false, if set1 is empty foreach my $el ( @$set1 ) { - $ret = 0 unless grep { /^\Q$el\E$/ } @$set2; + $ret = 0 unless grep { /^\Q$el\E$/ } @$set2; } return $ret; } +# Return the string that joins together a list of witnesses for +# display on a single path. +sub _witnesses_of_label { + my( $self, $label ) = @_; + my $regex = $self->wit_list_separator; + my @answer = split( /\Q$regex\E/, $label ); + return @answer; +} -## INITIALIZATION METHODS - for use by parsers -# Walk the paths for each witness in the graph, and return the nodes -# that the graph has in common. If $using_base is true, some -# different logic is needed. - -sub walk_witness_paths { - my( $self, $end ) = @_; - # For each witness, walk the path through the graph. - # Then we need to find the common nodes. - # TODO This method is going to fall down if we have a very gappy - # text in the collation. - my $paths = {}; - my @common_readings; - foreach my $wit ( @{$self->tradition->witnesses} ) { - my $curr_reading = $self->start; - my @wit_path = $self->reading_sequence( $self->start, $end, - $wit->sigil ); - $wit->path( \@wit_path ); - - # Detect the common readings. - @common_readings = _find_common( \@common_readings, \@wit_path ); - } +=head2 common_readings - # Mark all the nodes as either common or not. - foreach my $cn ( @common_readings ) { - print STDERR "Setting " . $cn->name . " / " . $cn->label - . " as common node\n"; - $cn->make_common; - } - foreach my $n ( $self->readings() ) { - $n->make_variant unless $n->is_common; - } - # Return an array of the common nodes in order. - return @common_readings; +Returns the list of common readings in the graph (i.e. those readings that are +shared by all non-lacunose witnesses.) + +=cut + +sub common_readings { + my $self = shift; + my @common = grep { $_->is_common } $self->readings; + return @common; } -sub _find_common { - my( $common_readings, $new_path ) = @_; - my @cr; - if( @$common_readings ) { - foreach my $n ( @$new_path ) { - push( @cr, $n ) if grep { $_ eq $n } @$common_readings; +=head2 path_text( $sigil, [, $start, $end ] ) + +Returns the text of a witness (plus its backup, if we are using a layer) +as stored in the collation. The text is returned as a string, where the +individual readings are joined with spaces and the meta-readings (e.g. +lacunae) are omitted. Optional specification of $start and $end allows +the generation of a subset of the witness text. + +=cut + +sub path_text { + my( $self, $wit, $start, $end ) = @_; + $start = $self->start unless $start; + $end = $self->end unless $end; + my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit ); + my $pathtext = ''; + my $last; + foreach my $r ( @path ) { + unless ( $r->join_prior || !$last || $last->join_next ) { + $pathtext .= ' '; + } + $pathtext .= $r->text; + $last = $r; } - } else { - push( @cr, @$new_path ); - } - return @cr; + return $pathtext; } -sub _remove_common { - my( $common_readings, $divergence ) = @_; - my @cr; - my %diverged; - map { $diverged{$_->name} = 1 } @$divergence; - foreach( @$common_readings ) { - push( @cr, $_ ) unless $diverged{$_->name}; - } - return @cr; -} +=head1 INITIALIZATION METHODS + +These are mostly for use by parsers. + +=head2 make_witness_path( $witness ) + +Link the array of readings contained in $witness->path (and in +$witness->uncorrected_path if it exists) into collation paths. +Clear out the arrays when finished. + +=head2 make_witness_paths + +Call make_witness_path for all witnesses in the tradition. +=cut -# An alternative to walk_witness_paths, for use when a collation is -# constructed from a base text and an apparatus. We have the -# sequences of readings and just need to add path edges. +# For use when a collation is constructed from a base text and an apparatus. +# We have the sequences of readings and just need to add path edges. +# When we are done, clear out the witness path attributes, as they are no +# longer needed. +# TODO Find a way to replace the witness path attributes with encapsulated functions? sub make_witness_paths { my( $self ) = @_; - - my @common_readings; - foreach my $wit ( @{$self->tradition->witnesses} ) { - $self->make_witness_path( $wit ); - @common_readings = _find_common( \@common_readings, $wit->path ); - - # If we have pre-corrected readings, we need to add paths - # for those as well. - if( $wit->has_ante_corr ) { - my @path = @{$wit->path}; - foreach my $ac ( @{$wit->ante_corr} ) { - # my( $offset, $length, $items ) = @$ac; - # Figure out where the path needs to start and - # end its divergence. - my $start = $ac->[0] - 1; - my $end = $ac->[0] + $ac->[1]; - my @chain; - push( @chain, $path[$start] ); - push( @chain, @{$ac->[2]} ); - push( @chain, $path[$end] ); - $self->make_path_uncorrection( $wit->sigil, @chain ); - } - @common_readings = _find_common( \@common_readings, - $wit->uncorrected_path ); - } + foreach my $wit ( $self->tradition->witnesses ) { + # say STDERR "Making path for " . $wit->sigil; + $self->make_witness_path( $wit ); } - return @common_readings; } sub make_witness_path { my( $self, $wit ) = @_; my @chain = @{$wit->path}; - $self->connect_readings_for_witness( $wit->sigil, @chain ); -} - -sub make_path_uncorrection { - my( $self, $sig, @chain ) = @_; - $sig .= ' (a.c.)'; - $self->connect_readings_for_witness( $sig, @chain ); -} - -sub connect_readings_for_witness { - my( $self, $sig, @chain ) = @_; + my $sig = $wit->sigil; + # Add start and end if necessary + unshift( @chain, $self->start ) unless $chain[0] eq $self->start; + push( @chain, $self->end ) unless $chain[-1] eq $self->end; foreach my $idx ( 0 .. $#chain-1 ) { - $self->add_path( $chain[$idx], $chain[$idx+1], $sig ); + $self->add_path( $chain[$idx], $chain[$idx+1], $sig ); + } + if( $wit->is_layered ) { + @chain = @{$wit->uncorrected_path}; + unshift( @chain, $self->start ) unless $chain[0] eq $self->start; + push( @chain, $self->end ) unless $chain[-1] eq $self->end; + foreach my $idx( 0 .. $#chain-1 ) { + my $source = $chain[$idx]; + my $target = $chain[$idx+1]; + $self->add_path( $source, $target, $sig.$self->ac_label ) + unless $self->has_path( $source, $target, $sig ); + } } + $wit->clear_path; + $wit->clear_uncorrected_path; } +=head2 calculate_ranks -sub common_readings { - my $self = shift; - my @common = grep { $_->is_common } $self->readings(); - return sort { _cmp_position( $a->position, $b->position ) } @common; -} - -# Calculate the relative positions of nodes in the graph, if they -# were not given to us. -sub calculate_positions { - my( $self, @ordered_common ) = @_; - - # We have to calculate the position identifiers for each word, - # keyed on the common nodes. This will be 'fun'. The end result - # is a hash per witness, whose key is the word node and whose - # value is its position in the text. Common nodes are always N,1 - # so have identical positions in each text. - - my $node_pos = {}; - foreach my $wit ( @{$self->tradition->witnesses} ) { - print STDERR "Calculating positions in " . $wit->sigil . "\n"; - _update_positions_from_path( $wit->path, @ordered_common ); - _update_positions_from_path( $wit->uncorrected_path, @ordered_common ) - if $wit->has_ante_corr; - } - - # DEBUG - foreach my $r ( $self->readings() ) { - print STDERR "Reading " . $r->name . "/" . $r->label . " has no position\n" - unless( $r->has_position ); - } +Calculate the reading ranks (that is, their aligned positions relative +to each other) for the graph. This can only be called on linear collations. - $self->init_lemmata(); -} +=begin testing -sub _update_positions_from_path { - my( $path, @ordered_common ) = @_; +use Text::Tradition; - # First we walk the given path, making a matrix for the witness - # that corresponds to its eventual position identifier. Common - # nodes always start a new row, and are thus always in the first - # column. - my $wit_matrix = []; - my $cn = 0; # We should hit the common readings in order. - my $row = []; - foreach my $wn ( @{$path} ) { - if( $wn eq $ordered_common[$cn] ) { - # Set up to look for the next common node, and - # start a new row of words. - $cn++; - push( @$wit_matrix, $row ) if scalar( @$row ); - $row = []; - } - push( @$row, $wn ); +my $cxfile = 't/data/Collatex-16.xml'; +my $t = Text::Tradition->new( + 'name' => 'inline', + 'input' => 'CollateX', + 'file' => $cxfile, + ); +my $c = $t->collation; + +# Make an svg +my $table = $c->alignment_table; +ok( $c->has_cached_table, "Alignment table was cached" ); +is( $c->alignment_table, $table, "Cached table returned upon second call" ); +$c->calculate_ranks; +is( $c->alignment_table, $table, "Cached table retained with no rank change" ); +$c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } ); +isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" ); + +=end testing + +=cut + +sub calculate_ranks { + my $self = shift; + # Save the existing ranks, in case we need to invalidate the cached SVG. + my %existing_ranks; + map { $existing_ranks{$_} = $_->rank } $self->readings; + + # Do the rankings based on the relationship equivalence graph, starting + # with the start node. + my ( $node_ranks, $rank_nodes ) = $self->relations->equivalence_ranks(); + + # Transfer our rankings from the topological graph to the real one. + foreach my $r ( $self->readings ) { + if( defined $node_ranks->{$self->equivalence( $r->id )} ) { + $r->rank( $node_ranks->{$self->equivalence( $r->id )} ); + } else { + # Die. Find the last rank we calculated. + my @all_defined = sort { ( $node_ranks->{$self->equivalence( $a->id )}||-1 ) + <=> ( $node_ranks->{$self->equivalence( $b->id )}||-1 ) } + $self->readings; + my $last = pop @all_defined; + throw( "Ranks not calculated after $last - do you have a cycle in the graph?" ); + } } - push( @$wit_matrix, $row ); # Push the last row onto the matrix - - # Now we have a matrix per witness, so that each row in the - # matrix begins with a common node, and continues with all the - # variant words that appear in the witness. We turn this into - # real positions in row,cell format. But we need some - # trickery in order to make sure that each node gets assigned - # to only one position. - - foreach my $li ( 1..scalar(@$wit_matrix) ) { - foreach my $di ( 1..scalar(@{$wit_matrix->[$li-1]}) ) { - my $reading = $wit_matrix->[$li-1]->[$di-1]; - my $position = "$li,$di"; - $DB::single = 1 unless ref( $reading ) eq 'Text::Tradition::Collation::Reading'; - - # If we have seen this node before, we need to compare - # its position with what went before. - unless( $reading->has_position && - _cmp_position( $position, $reading->position ) < 1 ) { - # The new position ID replaces the old one. - $reading->position( $position ); - } # otherwise, the old position needs to stay. - } + # Do we need to invalidate the cached data? + if( $self->has_cached_svg || $self->has_cached_table ) { + foreach my $r ( $self->readings ) { + next if defined( $existing_ranks{$r} ) + && $existing_ranks{$r} == $r->rank; + # Something has changed, so clear the cache + $self->_clear_cache; + # ...and recalculate the common readings. + $self->calculate_common_readings(); + last; + } } + # The graph calculation information is now up to date. + $self->_graphcalc_done(1); } -sub _cmp_position { - my( $a, $b ) = @_; - if ( $a && $b ) { - my @pos_a = split(/,/, $a ); - my @pos_b = split(/,/, $b ); - - my $big_cmp = $pos_a[0] <=> $pos_b[0]; - return $big_cmp if $big_cmp; - # else - return $pos_a[1] <=> $pos_b[1]; - } elsif ( $b ) { # a is undefined - return -1; - } elsif ( $a ) { # b is undefined - return 1; - } - return 0; # they are both undefined -} +sub _clear_cache { + my $self = shift; + $self->wipe_svg if $self->has_cached_svg; + $self->wipe_table if $self->has_cached_table; +} -sub all_positions { - my $self = shift; - my %positions = (); - map { $positions{$_->position} = 1 } $self->readings; - my @answer = sort { _cmp_position( $a, $b ) } keys( %positions ); - return @answer; -} -sub readings_at_position { - my( $self, $pos ) = @_; - my @answer = grep { $_->position eq $pos } $self->readings; - return @answer; -} +=head2 flatten_ranks -## Lemmatizer functions +A convenience method for parsing collation data. Searches the graph for readings +with the same text at the same rank, and merges any that are found. -sub init_lemmata { - my $self = shift; - - foreach my $position ( $self->all_positions ) { - $self->lemmata->{$position} = undef; - } +=cut - foreach my $cr ( $self->common_readings ) { - $self->lemmata->{$cr->position} = $cr->name; +sub flatten_ranks { + my $self = shift; + my %unique_rank_rdg; + my $changed; + foreach my $rdg ( $self->readings ) { + next unless $rdg->has_rank; + my $key = $rdg->rank . "||" . $rdg->text; + if( exists $unique_rank_rdg{$key} ) { + # Make sure they don't have different grammatical forms + my $ur = $unique_rank_rdg{$key}; + if( $rdg->disambiguated && $ur->disambiguated ) { + my $rform = join( '//', map { $_->form->to_string } $rdg->lexemes ); + my $uform = join( '//', map { $_->form->to_string } $ur->lexemes ); + next unless $rform eq $uform; + } elsif( $rdg->disambiguated xor $ur->disambiguated ) { + next; + } + # Combine! + #say STDERR "Combining readings at same rank: $key"; + $changed = 1; + $self->merge_readings( $unique_rank_rdg{$key}, $rdg ); + # TODO see if this now makes a common point. + } else { + $unique_rank_rdg{$key} = $rdg; + } } + # If we merged readings, the ranks are still fine but the alignment + # table is wrong. Wipe it. + $self->wipe_table() if $changed; } - -=item B + -my @state = $graph->lemma_readings( @readings_delemmatized ); +=head2 calculate_common_readings -Takes a list of readings that have just been delemmatized, and returns -a set of tuples of the form ['reading', 'state'] that indicates what -changes need to be made to the graph. +Goes through the graph identifying the readings that appear in every witness +(apart from those with lacunae at that spot.) Marks them as common and returns +the list. -=over +=begin testing + +use Text::Tradition; -=item * +my $cxfile = 't/data/Collatex-16.xml'; +my $t = Text::Tradition->new( + 'name' => 'inline', + 'input' => 'CollateX', + 'file' => $cxfile, + ); +my $c = $t->collation; -A state of 1 means 'lemmatize this reading' +my @common = $c->calculate_common_readings(); +is( scalar @common, 8, "Found correct number of common readings" ); +my @marked = sort $c->common_readings(); +is( scalar @common, 8, "All common readings got marked as such" ); +my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /; +is_deeply( \@marked, \@expected, "Found correct list of common readings" ); -=item * +=end testing -A state of 0 means 'delemmatize this reading' +=cut -=item * +sub calculate_common_readings { + my $self = shift; + my @common; + map { $_->is_common( 0 ) } $self->readings; + # Implicitly calls calculate_ranks + my $table = $self->alignment_table; + foreach my $idx ( 0 .. $table->{'length'} - 1 ) { + my @row = map { $_->{'tokens'}->[$idx] + ? $_->{'tokens'}->[$idx]->{'t'} : '' } + @{$table->{'alignment'}}; + my %hash; + foreach my $r ( @row ) { + if( $r ) { + $hash{$r->id} = $r unless $r->is_meta; + } else { + $hash{'UNDEF'} = $r; + } + } + if( keys %hash == 1 && !exists $hash{'UNDEF'} ) { + my( $r ) = values %hash; + $r->is_common( 1 ); + push( @common, $r ); + } + } + return @common; +} -A state of undef means 'an ellipsis belongs in the text here because -no decision has been made / an earlier decision was backed out' +=head2 text_from_paths -=back +Calculate the text array for all witnesses from the path, for later consistency +checking. Only to be used if there is no non-graph-based way to know the +original texts. =cut -sub lemma_readings { - my( $self, @toggled_off_nodes ) = @_; - - # First get the positions of those nodes which have been - # toggled off. - my $positions_off = {}; - map { $positions_off->{ $_->position } = $_->name } @toggled_off_nodes; - - # Now for each position, we have to see if a node is on, and we - # have to see if a node has been turned off. - my @answer; - foreach my $pos ( $self->all_positions() ) { - # Find the state of this position. If there is an active node, - # its name will be the state; otherwise the state will be 0 - # (nothing at this position) or undef (ellipsis at this position) - my $active = $self->lemmata->{$pos}; - - # Is there a formerly active node that was toggled off? - if( exists( $positions_off->{$pos} ) ) { - my $off_node = $positions_off->{$pos}; - if( $active && $active ne $off_node) { - push( @answer, [ $off_node, 0 ], [ $active, 1 ] ); - } else { - push( @answer, [ $off_node, $active ] ); - } - - # No formerly active node, so we just see if there is a currently - # active one. - } elsif( $active ) { - # Push the active node, whatever it is. - push( @answer, [ $active, 1 ] ); - } else { - # Push the state that is there. Arbitrarily use the first node - # at that position. - my @pos_nodes = $self->readings_at_position( $pos ); - push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] ); - } - } - - return @answer; +sub text_from_paths { + my $self = shift; + foreach my $wit ( $self->tradition->witnesses ) { + my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); + my @text; + foreach my $r ( @readings ) { + next if $r->is_meta; + push( @text, $r->text ); + } + $wit->text( \@text ); + if( $wit->is_layered ) { + my @ucrdgs = $self->reading_sequence( $self->start, $self->end, + $wit->sigil.$self->ac_label ); + my @uctext; + foreach my $r ( @ucrdgs ) { + next if $r->is_meta; + push( @uctext, $r->text ); + } + $wit->layertext( \@uctext ); + } + } } -=item B +=head1 UTILITY FUNCTIONS -my @readings_delemmatized = $graph->toggle_reading( $reading_name ); +=head2 common_predecessor( $reading_a, $reading_b ) -Takes a reading node name, and either lemmatizes or de-lemmatizes -it. Returns a list of all readings that are de-lemmatized as a result -of the toggle. +Find the last reading that occurs in sequence before both the given readings. +At the very least this should be $self->start. -=cut +=head2 common_successor( $reading_a, $reading_b ) -sub toggle_reading { - my( $self, $rname ) = @_; - - return unless $rname; - my $reading = $self->reading( $rname ); - if( !$reading || $reading->is_common() ) { - # Do nothing, it's a common node. - return; - } +Find the first reading that occurs in sequence after both the given readings. +At the very least this should be $self->end. - my $pos = $reading->position; - my $old_state = $self->lemmata->{$pos}; - my @readings_off; - if( $old_state && $old_state eq $rname ) { - # Turn off the node. We turn on no others by default. - push( @readings_off, $reading ); - } else { - # Turn on the node. - $self->lemmata->{$pos} = $rname; - # Any other 'on' readings in the same position should be off. - push( @readings_off, $self->same_position_as( $reading ) ); - # Any node that is an identical transposed one should be off. - push( @readings_off, $reading->identical_readings ); - } - @readings_off = unique_list( @readings_off ); - - # Turn off the readings that need to be turned off. - my @readings_delemmatized; - foreach my $n ( @readings_off ) { - my $state = $self->lemmata->{$n->position}; - if( $state && $state eq $n->name ) { - # this reading is still on, so turn it off - push( @readings_delemmatized, $n ); - my $new_state = undef; - if( $n eq $reading ) { - # This is the reading that was clicked, so if there are no - # other readings there, turn off the position. In all other - # cases, restore the ellipsis. - my @other_n = $self->same_position_as( $n ); - $new_state = 0 unless @other_n; - } - $self->lemmata->{$n->position} = $new_state; - } elsif( $old_state && $old_state eq $n->name ) { - # another reading has already been turned on here - push( @readings_delemmatized, $n ); - } # else some other reading was on anyway, so pass. - } - return @readings_delemmatized; -} +=begin testing -sub same_position_as { - my( $self, $reading ) = @_; - my $pos = $reading->position; - my @same = grep { $_ ne $reading } $self->readings_at_position( $reading->position ); - return @same; +use Text::Tradition; + +my $cxfile = 't/data/Collatex-16.xml'; +my $t = Text::Tradition->new( + 'name' => 'inline', + 'input' => 'CollateX', + 'file' => $cxfile, + ); +my $c = $t->collation; + +is( $c->common_predecessor( 'n24', 'n23' )->id, + 'n20', "Found correct common predecessor" ); +is( $c->common_successor( 'n24', 'n23' )->id, + '__END__', "Found correct common successor" ); + +is( $c->common_predecessor( 'n19', 'n17' )->id, + 'n16', "Found correct common predecessor for readings on same path" ); +is( $c->common_successor( 'n21', 'n10' )->id, + '__END__', "Found correct common successor for readings on same path" ); + +=end testing + +=cut + +## Return the closest reading that is a predecessor of both the given readings. +sub common_predecessor { + my $self = shift; + my( $r1, $r2 ) = $self->_objectify_args( @_ ); + return $self->_common_in_path( $r1, $r2, 'predecessors' ); } -# Return the string that joins together a list of witnesses for -# display on a single path. -sub path_label { - my $self = shift; - return join( $self->wit_list_separator, @_ ); +sub common_successor { + my $self = shift; + my( $r1, $r2 ) = $self->_objectify_args( @_ ); + return $self->_common_in_path( $r1, $r2, 'successors' ); } -sub witnesses_of_label { - my( $self, $label ) = @_; - my $regex = $self->wit_list_separator; - my @answer = split( /\Q$regex\E/, $label ); - return @answer; -} -sub unique_list { - my( @list ) = @_; - my %h; - map { $h{$_->name} = $_ } @list; - return values( %h ); +# TODO think about how to do this without ranks... +sub _common_in_path { + my( $self, $r1, $r2, $dir ) = @_; + my $iter = $self->end->rank; + my @candidates; + my @last_r1 = ( $r1 ); + my @last_r2 = ( $r2 ); + # my %all_seen = ( $r1 => 'r1', $r2 => 'r2' ); + my %all_seen; + # say STDERR "Finding common $dir for $r1, $r2"; + while( !@candidates ) { + last unless $iter--; # Avoid looping infinitely + # Iterate separately down the graph from r1 and r2 + my( @new_lc1, @new_lc2 ); + foreach my $lc ( @last_r1 ) { + foreach my $p ( $lc->$dir ) { + if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) { + # say STDERR "Path candidate $p from $lc"; + push( @candidates, $p ); + } elsif( !$all_seen{$p->id} ) { + $all_seen{$p->id} = 'r1'; + push( @new_lc1, $p ); + } + } + } + foreach my $lc ( @last_r2 ) { + foreach my $p ( $lc->$dir ) { + if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) { + # say STDERR "Path candidate $p from $lc"; + push( @candidates, $p ); + } elsif( !$all_seen{$p->id} ) { + $all_seen{$p->id} = 'r2'; + push( @new_lc2, $p ); + } + } + } + @last_r1 = @new_lc1; + @last_r2 = @new_lc2; + } + my @answer = sort { $a->rank <=> $b->rank } @candidates; + return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer ); } -sub add_hash_entry { - my( $hash, $key, $entry ) = @_; - if( exists $hash->{$key} ) { - push( @{$hash->{$key}}, $entry ); - } else { - $hash->{$key} = [ $entry ]; - } +sub throw { + Text::Tradition::Error->throw( + 'ident' => 'Collation error', + 'message' => $_[0], + ); } no Moose; __PACKAGE__->meta->make_immutable; + +=head1 LICENSE + +This package is free software and is provided "as is" without express +or implied warranty. You can redistribute it and/or modify it under +the same terms as Perl itself. + +=head1 AUTHOR + +Tara L Andrews Eaurum@cpan.orgE