X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=3c65b4ce3f372acc2e57a2cb6447b642712de93c;hb=10943ab0b79fbd489f6beb3b81a13ed8cbcfafcf;hp=f0901a655dbd8cd8ef4f0e69b7d6b9c837a8ba24;hpb=ac4d7ac24adc55494d609a4c0396a84fef26d8b3;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index f0901a6..3c65b4c 100644 --- a/lib/Text/Tradition/Collation.pm +++ b/lib/Text/Tradition/Collation.pm @@ -9,6 +9,7 @@ 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; @@ -30,6 +31,8 @@ has 'relations' => ( related_readings => 'related_readings', get_relationship => 'get_relationship', del_relationship => 'del_relationship', + equivalence => 'equivalence', + equivalence_graph => 'equivalence_graph', }, writer => '_set_relations', ); @@ -281,6 +284,9 @@ sub add_reading { my( $self, $reading ) = @_; unless( ref( $reading ) eq 'Text::Tradition::Collation::Reading' ) { my %args = %$reading; + if( $self->tradition->has_language && !exists $args{'language'} ) { + $args{'language'} = $self->tradition->language; + } $reading = Text::Tradition::Collation::Reading->new( 'collation' => $self, %args ); @@ -367,7 +373,7 @@ sub merge_readings { # objects themselves. my( $kept, $deleted, $combine, $combine_char ) = $self->_stringify_args( @_ ); $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( $deleted ) ) { @@ -381,7 +387,7 @@ sub merge_readings { @wits{keys %$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 ) { @@ -428,10 +434,13 @@ 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 $self = shift; @@ -453,6 +462,7 @@ sub del_path { } unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) { $self->sequence->delete_edge( $source, $target ); + $self->relations->delete_equivalence_edge( $source, $target ); } } @@ -495,8 +505,7 @@ 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; } @@ -633,15 +642,15 @@ 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 @@ -663,6 +672,7 @@ 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 ) ); } @@ -712,18 +722,18 @@ sub as_dot { my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) ); my $variables = { %edge_attrs, 'label' => $witstr }; my $varopts = _dot_attr_string( $variables ); - $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;"; + $dot .= "\t\"__SUBSTART__\" -> \"$node\" $varopts;"; } foreach my $node ( keys %subend ) { my $witstr = $self->_path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) ); my $variables = { %edge_attrs, 'label' => $witstr }; my $varopts = _dot_attr_string( $variables ); - $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;"; + $dot .= "\t\"$node\" -> \"__SUBEND__\" $varopts;"; } # 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"; @@ -877,6 +887,13 @@ 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 parse again. +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" ); + =end testing =cut @@ -895,6 +912,10 @@ sub as_graphml { # 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 @@ -916,6 +937,7 @@ sub as_graphml { 'Str' => 'string', 'Int' => 'int', 'Bool' => 'boolean', + 'ReadingID' => 'string', 'RelationshipType' => 'string', 'RelationshipScope' => 'string', ); @@ -942,6 +964,8 @@ sub as_graphml { next unless $save_types{$attr->type_constraint->name}; $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; } + # Extra custom key for the tradition stemma(ta) + $graph_attributes{'stemmata'} = 'string'; foreach my $datum ( sort keys %graph_attributes ) { $graph_data_keys{$datum} = 'dg'.$gdi++; @@ -961,6 +985,9 @@ 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 ) { @@ -997,21 +1024,31 @@ 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( $datum eq 'stemmata' ) { + my @stemstrs; + map { push( @stemstrs, $_->editable( {linesep => ''} ) ) } + $self->tradition->stemmata; + $value = join( "\n", @stemstrs ); } elsif( $gattr_from{$datum} eq 'Tradition' ) { $value = $self->tradition->$datum; } else { @@ -1034,6 +1071,18 @@ 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; + } + 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; } @@ -1045,8 +1094,10 @@ sub as_graphml { # 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 unless $use_readings{$e->[0]}; - $e->[1] = $self->end unless $use_readings{$e->[1]}; + $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] }, @@ -1071,6 +1122,10 @@ 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, @@ -1363,11 +1418,10 @@ 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; @@ -1428,60 +1482,6 @@ sub make_witness_path { $wit->clear_uncorrected_path; } -=head2 equivalence_graph( \%readingmap, $startrank, $endrank ) - -Returns an equivalence graph of the collation, in which all readings -related via a 'colocated' relationship are transformed into a single -vertex. Can be used to determine the validity of a new relationship. The -mapping between equivalence vertices and reading IDs will be stored in the -hash whose reference is passed as readingmap. For a subset of the graph, -pass in a start and/or an ending rank (this only works if L -has been called at least once.) - -=cut - -sub equivalence_graph { - my( $self, $map, $start, $end ) = @_; - $start = undef unless $self->end->has_rank; - $end = undef unless $self->end->has_rank; - - my $eqgraph = Graph->new(); - my $rel_ctr = 0; - # Add the nodes - foreach my $r ( $self->readings ) { - unless( $r eq $self->start || $r eq $self->end ) { - next if $start && $r->rank < $start; - next if $end && $r->rank > $end; - } - next if exists $map->{$r->id}; - my @rels = $r->related_readings( 'colocated' ); - if( @rels ) { - # Make an equivalence vertex - my $rn = 'equivalence_' . $rel_ctr++; - $eqgraph->add_vertex( $rn ); - # Note which readings belong to this vertex. - push( @rels, $r ); - foreach( @rels ) { - $map->{$_->id} = $rn; - } - } else { - # Add a new node to mirror the old node. - $map->{$r->id} = $r->id; - $eqgraph->add_vertex( $r->id ); - } - } - - # Add the edges. - foreach my $p ( $self->paths ) { - my $efrom = exists $map->{$p->[0]} ? $map->{$p->[0]} - : $map->{$self->start->id}; - my $eto = exists $map->{$p->[1]} ? $map->{$p->[1]} - : $map->{$self->end->id}; - $eqgraph->add_edge( $efrom, $eto ); - } - return $eqgraph; -} - =head2 calculate_ranks Calculate the reading ranks (that is, their aligned positions relative @@ -1517,28 +1517,25 @@ sub calculate_ranks { # Save the existing ranks, in case we need to invalidate the cached SVG. my %existing_ranks; map { $existing_ranks{$_} = $_->rank } $self->readings; - # 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 %rel_containers; - my $topo_graph = $self->equivalence_graph( \%rel_containers ); - - # Now do the rankings, starting with the start node. - my $topo_start = $rel_containers{$self->start->id}; + + # Do the rankings based on the relationship equivalence graph, starting + # with the start node. + my $topo_start = $self->equivalence( $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 ); + @curr_origin = _assign_rank( $self->equivalence_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}}||-1 ) - <=> ( $node_ranks->{$rel_containers{$b->id}}||-1 ) } + 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?" ); @@ -1724,10 +1721,12 @@ 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 @@ -1744,12 +1743,12 @@ 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 @@ -1768,26 +1767,45 @@ 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; + # print STDERR "Finding common $dir for $r1, $r2\n"; 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' ) { + # print STDERR "Path candidate $p from $lc\n"; + 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' ) { + # print STDERR "Path candidate $p from $lc\n"; 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 );