X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?p=scpubgit%2Fstemmatology.git;a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=dda84852ee08ae2914484ba243577d9c99c9c2cd;hp=68b2adf5997badcdbbd2e3c48b421913c43dd150;hb=428bcf0bc79f77a7857b21ef881708faa792e33a;hpb=f2fec47dc44816775c844e8b50cef03b12b69dae diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index 68b2adf..dda8485 100644 --- a/lib/Text/Tradition/Collation.pm +++ b/lib/Text/Tradition/Collation.pm @@ -1,16 +1,16 @@ package Text::Tradition::Collation; +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_XS; +use Text::CSV; use Text::Tradition::Collation::Reading; use Text::Tradition::Collation::RelationshipStore; use Text::Tradition::Error; -use XML::LibXML; -use XML::LibXML::XPathContext; +use XML::Easy::Syntax qw( $xml10_namestartchar_rx $xml10_namechar_rx ); use Moose; has 'sequence' => ( @@ -30,6 +30,8 @@ has 'relations' => ( related_readings => 'related_readings', get_relationship => 'get_relationship', del_relationship => 'del_relationship', + equivalence => 'equivalence', + equivalence_graph => 'equivalence_graph', }, writer => '_set_relations', ); @@ -37,6 +39,7 @@ has 'relations' => ( has 'tradition' => ( is => 'ro', isa => 'Text::Tradition', + writer => '_set_tradition', weak_ref => 1, ); @@ -270,8 +273,10 @@ See L for the available options. 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 } ) ); - $self->_set_end( $self->add_reading( { 'collation' => $self, 'is_end' => 1 } ) ); + $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 } ) ); } ### Reading construct/destruct functions @@ -280,6 +285,13 @@ 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 ); @@ -362,11 +374,23 @@ is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" ); sub merge_readings { 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, $deleted, $combine, $combine_char ) = $self->_stringify_args( @_ ); + 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( $deleted ) ) { @@ -380,22 +404,91 @@ sub merge_readings { @wits{keys %$fwits} = values %$fwits; $self->sequence->set_edge_attributes( @vector, \%wits ); } - $self->relations->merge_readings( $kept, $deleted, $combine_char ); + $self->relations->merge_readings( $kept, $deleted, $combine ); # Do the deletion deed. if( $combine ) { - my $kept_obj = $self->reading( $kept ); - my $del_obj = $self->reading( $deleted ); + # 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 compress_readings + +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 ) { + # While we are here, get rid of any extra wordforms from a disambiguated + # reading. + if( $rdg->disambiguated ) { + foreach my $lex ( $rdg->lexemes ) { + $lex->clear_matching_forms(); + $lex->add_matching_form( $lex->form ); + } + } + # Now look for readings that can be joined to their successors. + 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 { @@ -427,10 +520,13 @@ sub add_path { $self->_graphcalc_done(0); # Connect the readings - $self->sequence->add_edge( $source, $target ); + 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 $self = shift; @@ -452,6 +548,7 @@ sub del_path { } unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) { $self->sequence->delete_edge( $source, $target ); + $self->relations->delete_equivalence_edge( $source, $target ); } } @@ -494,8 +591,7 @@ sub clear_witness { sub add_relationship { my $self = shift; my( $source, $target, $opts ) = $self->_stringify_args( @_ ); - my( @vectors ) = $self->relations->add_relationship( $source, - $self->reading( $source ), $target, $self->reading( $target ), $opts ); + my( @vectors ) = $self->relations->add_relationship( $source, $target, $opts ); $self->_graphcalc_done(0); return @vectors; } @@ -550,7 +646,8 @@ sub as_svg { 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; + $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 ); @@ -632,15 +729,15 @@ sub as_dot { # Output substitute start/end readings if necessary if( $startrank ) { - $dot .= "\t\"#SUBSTART#\" [ label=\"...\" ];\n"; + $dot .= "\t\"__SUBSTART__\" [ label=\"...\",id=\"__START__\" ];\n"; } if( $endrank ) { - $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n"; + $dot .= "\t\"__SUBEND__\" [ label=\"...\",id=\"__END__\" ];\n"; } if( $STRAIGHTENHACK ) { ## HACK part 1 - my $startlabel = $startrank ? 'SUBSTART' : 'START'; - $dot .= "\tsubgraph { rank=same \"#$startlabel#\" \"#SILENT#\" }\n"; + 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 @@ -662,6 +759,7 @@ sub as_dot { $label = "-$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 ) ); } @@ -701,28 +799,33 @@ sub as_dot { $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n", $edge->[0], $edge->[1], $varopts ); } elsif( $used{$edge->[0]} ) { - $subend{$edge->[0]} = 1; + $subend{$edge->[0]} = $edge->[1]; } elsif( $used{$edge->[1]} ) { - $substart{$edge->[1]} = 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->reading_witnesses( $self->reading( $node ) ) ); + 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;"; + $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;\n"; } foreach my $node ( keys %subend ) { - my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) ); + 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;"; + $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;\n"; } # HACK part 2 if( $STRAIGHTENHACK ) { - my $endlabel = $endrank ? 'SUBEND' : 'END'; - $dot .= "\t\"#$endlabel#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; + my $endlabel = $endrank ? '__SUBEND__' : '__END__'; + $dot .= "\t\"$endlabel\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; } $dot .= "}\n"; @@ -783,15 +886,35 @@ sub path_witnesses { 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 = sort @_; + 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; - if( scalar @wits > $maj ) { - # TODO break out a.c. wits - return 'majority'; + $maj = $maj > 5 ? $maj : 5; + if( scalar keys %wits > $maj ) { + unshift( @disp_ac, 'majority' ); + return join( ', ', @disp_ac ); } else { - return join( ', ', @wits ); + return join( ', ', sort keys %wits ); } } @@ -857,14 +980,47 @@ is( scalar $st->collation->readings, $READINGS, "Reparsed collation has all read 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 ) = @_; + 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'; @@ -872,6 +1028,7 @@ sub as_graphml { 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; # Create the document and root node + require XML::LibXML; my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" ); my $root = $graphml->createElementNS( $graphml_ns, 'graphml' ); $graphml->setDocumentElement( $root ); @@ -884,6 +1041,7 @@ sub as_graphml { 'Str' => 'string', 'Int' => 'int', 'Bool' => 'boolean', + 'ReadingID' => 'string', 'RelationshipType' => 'string', 'RelationshipScope' => 'string', ); @@ -910,12 +1068,25 @@ sub as_graphml { 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', $graph_attributes{$datum} ); + $key->setAttribute( 'attr.type', $dtype ); $key->setAttribute( 'for', 'graph' ); $key->setAttribute( 'id', $graph_data_keys{$datum} ); } @@ -929,6 +1100,9 @@ sub as_graphml { 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 ) { @@ -965,21 +1139,29 @@ sub as_graphml { $key->setAttribute( 'id', $edge_data_keys{$datum} ); } - # Add the collation graph itself + # 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', $self->tradition->name ); + $sgraph->setAttribute( 'id', $xmlidname ); $sgraph->setAttribute( 'parse.edgeids', 'canonical' ); - $sgraph->setAttribute( 'parse.edges', scalar($self->paths) ); + $sgraph->setAttribute( 'parse.edges', 0 ); # fill in later $sgraph->setAttribute( 'parse.nodeids', 'canonical' ); - $sgraph->setAttribute( 'parse.nodes', scalar($self->readings) ); + $sgraph->setAttribute( 'parse.nodes', 0 ); # fill in later $sgraph->setAttribute( 'parse.order', 'nodesfirst' ); - # Collation attribute data + # Tradition/collation attribute data foreach my $datum ( keys %graph_attributes ) { my $value; if( $datum eq 'version' ) { - $value = '3.1'; + $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 { @@ -992,6 +1174,9 @@ sub as_graphml { 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++; @@ -999,6 +1184,20 @@ sub as_graphml { $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; } @@ -1008,7 +1207,13 @@ sub as_graphml { 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. - foreach my $wit ( sort $self->path_witnesses( $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] } ); @@ -1032,6 +1237,10 @@ sub as_graphml { } } + # 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, @@ -1060,7 +1269,7 @@ row per witness (or witness uncorrected.) sub as_csv { my( $self ) = @_; my $table = $self->alignment_table; - my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } ); + my $csv = Text::CSV->new( { binary => 1, quote_null => 0 } ); my @result; # Make the header row $csv->combine( map { $_->{'witness'} } @{$table->{'alignment'}} ); @@ -1108,7 +1317,7 @@ sub alignment_table { 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 ) { - # print STDERR "Making witness row(s) for " . $wit->sigil . "\n"; + # 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'}}, @@ -1133,8 +1342,8 @@ sub _make_witness_row { foreach my $rdg ( @$path ) { my $rtext = $rdg->text; $rtext = '#LACUNA#' if $rdg->is_lacuna; - print STDERR "rank " . $rdg->rank . "\n" if $debug; - # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank; + 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; @@ -1306,7 +1515,7 @@ sub common_readings { return @common; } -=head2 path_text( $sigil, $mainsigil [, $start, $end ] ) +=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 @@ -1324,11 +1533,10 @@ sub path_text { my $pathtext = ''; my $last; foreach my $r ( @path ) { - if( $r->join_prior || !$last || $last->join_next ) { - $pathtext .= $r->text; - } else { - $pathtext .= ' ' . $r->text; - } + unless ( $r->join_prior || !$last || $last->join_next ) { + $pathtext .= ' '; + } + $pathtext .= $r->text; $last = $r; } return $pathtext; @@ -1359,7 +1567,7 @@ Call make_witness_path for all witnesses in the tradition. sub make_witness_paths { my( $self ) = @_; foreach my $wit ( $self->tradition->witnesses ) { - # print STDERR "Making path for " . $wit->sigil . "\n"; + # say STDERR "Making path for " . $wit->sigil; $self->make_witness_path( $wit ); } } @@ -1368,11 +1576,16 @@ sub make_witness_path { my( $self, $wit ) = @_; my @chain = @{$wit->path}; 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 ); } 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]; @@ -1418,58 +1631,20 @@ sub calculate_ranks { my $self = shift; # Save the existing ranks, in case we need to invalidate the cached SVG. my %existing_ranks; - # Walk a version of the graph where every node linked by a relationship - # edge is fundamentally the same node, and do a topological ranking on - # the nodes in this graph. - my $topo_graph = Graph->new(); - my %rel_containers; - my $rel_ctr = 0; - # Add the nodes - foreach my $r ( $self->readings ) { - next if exists $rel_containers{$r->id}; - my @rels = $r->related_readings( 'colocated' ); - if( @rels ) { - # Make a relationship container. - push( @rels, $r ); - my $rn = 'rel_container_' . $rel_ctr++; - $topo_graph->add_vertex( $rn ); - foreach( @rels ) { - $rel_containers{$_->id} = $rn; - } - } else { - # Add a new node to mirror the old node. - $rel_containers{$r->id} = $r->id; - $topo_graph->add_vertex( $r->id ); - } - } + 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(); - # Add the edges. - foreach my $r ( $self->readings ) { - $existing_ranks{$r} = $r->rank; - foreach my $n ( $self->sequence->successors( $r->id ) ) { - my( $tfrom, $tto ) = ( $rel_containers{$r->id}, - $rel_containers{$n} ); - # $DB::single = 1 unless $tfrom && $tto; - $topo_graph->add_edge( $tfrom, $tto ); - } - } - - # Now do the rankings, starting with the start node. - my $topo_start = $rel_containers{$self->start->id}; - my $node_ranks = { $topo_start => 0 }; - my @curr_origin = ( $topo_start ); - # A little iterative function. - while( @curr_origin ) { - @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin ); - } # Transfer our rankings from the topological graph to the real one. foreach my $r ( $self->readings ) { - if( defined $node_ranks->{$rel_containers{$r->id}} ) { - $r->rank( $node_ranks->{$rel_containers{$r->id}} ); + 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 { $node_ranks->{$rel_containers{$a->id}} - <=> $node_ranks->{$rel_containers{$b->id}} } + 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?" ); @@ -1491,41 +1666,6 @@ sub calculate_ranks { $self->_graphcalc_done(1); } -sub _assign_rank { - my( $graph, $node_ranks, @current_nodes ) = @_; - # Look at each of the children of @current_nodes. If all the child's - # parents have a rank, assign it the highest rank + 1 and add it to - # @next_nodes. Otherwise skip it; we will return when the highest-ranked - # parent gets a rank. - my @next_nodes; - foreach my $c ( @current_nodes ) { - warn "Current reading $c has no rank!" - unless exists $node_ranks->{$c}; - # print STDERR "Looking at child of node $c, rank " - # . $node_ranks->{$c} . "\n"; - foreach my $child ( $graph->successors( $c ) ) { - next if exists $node_ranks->{$child}; - my $highest_rank = -1; - my $skip = 0; - foreach my $parent ( $graph->predecessors( $child ) ) { - if( exists $node_ranks->{$parent} ) { - $highest_rank = $node_ranks->{$parent} - if $highest_rank <= $node_ranks->{$parent}; - } else { - $skip = 1; - last; - } - } - next if $skip; - my $c_rank = $highest_rank + 1; - # print STDERR "Assigning rank $c_rank to node $child \n"; - $node_ranks->{$child} = $c_rank; - push( @next_nodes, $child ); - } - } - return @next_nodes; -} - sub _clear_cache { my $self = shift; $self->wipe_svg if $self->has_cached_svg; @@ -1548,8 +1688,17 @@ sub flatten_ranks { 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! - # print STDERR "Combining readings at same rank: $key\n"; + #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. @@ -1655,10 +1804,12 @@ sub text_from_paths { =head2 common_predecessor( $reading_a, $reading_b ) Find the last reading that occurs in sequence before both the given readings. +At the very least this should be $self->start. =head2 common_successor( $reading_a, $reading_b ) Find the first reading that occurs in sequence after both the given readings. +At the very least this should be $self->end. =begin testing @@ -1675,12 +1826,12 @@ 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" ); + '__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__', "Found correct common successor for readings on same path" ); =end testing @@ -1699,26 +1850,45 @@ sub common_successor { return $self->_common_in_path( $r1, $r2, 'successors' ); } + +# TODO think about how to do this without ranks... sub _common_in_path { my( $self, $r1, $r2, $dir ) = @_; - my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank; - $iter = $self->end->rank - $iter if $dir eq 'successors'; + my $iter = $self->end->rank; my @candidates; - my @last_checked = ( $r1, $r2 ); + 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 ) { - my @new_lc; - foreach my $lc ( @last_checked ) { + 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} ) { + if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) { + # say STDERR "Path candidate $p from $lc"; push( @candidates, $p ); - } else { - $all_seen{$p->id} = 1; - push( @new_lc, $p ); + } elsif( !$all_seen{$p->id} ) { + $all_seen{$p->id} = 'r2'; + push( @new_lc2, $p ); } } } - @last_checked = @new_lc; + @last_r1 = @new_lc1; + @last_r2 = @new_lc2; } my @answer = sort { $a->rank <=> $b->rank } @candidates; return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );