X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=fb1b7cb5535feadfb38e1b7b27b9e0b64b37057f;hb=e4b73942dd7d58e6e3cb7a42b1f348b56021192e;hp=3f62b11c6f2ff9a982571de8cdafabb4bc53d209;hpb=826d8773c081da08f139d1e48fa5ca1abb725f8e;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index 3f62b11..fb1b7cb 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', ); @@ -271,8 +274,10 @@ 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 @@ -281,6 +286,13 @@ 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 ); @@ -363,11 +375,23 @@ 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( $deleted ) ) { @@ -381,12 +405,10 @@ sub merge_readings { @wits{keys %$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 ); my $joinstr = $combine_char; unless( defined $joinstr ) { $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior; @@ -428,10 +450,13 @@ 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 $self = shift; @@ -453,6 +478,7 @@ sub del_path { } unless( keys %{$self->sequence->get_edge_attributes( $source, $target )} ) { $self->sequence->delete_edge( $source, $target ); + $self->relations->delete_equivalence_edge( $source, $target ); } } @@ -550,7 +576,8 @@ 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 || $opts->{'nocalc'} ); + $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 +659,15 @@ 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 +689,7 @@ 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 ) ); } @@ -711,18 +739,18 @@ 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"; @@ -806,6 +834,7 @@ sub _path_display_label { # 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 ); @@ -876,6 +905,13 @@ 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 @@ -919,6 +955,7 @@ sub as_graphml { 'Str' => 'string', 'Int' => 'int', 'Bool' => 'boolean', + 'ReadingID' => 'string', 'RelationshipType' => 'string', 'RelationshipScope' => 'string', ); @@ -945,6 +982,8 @@ 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++; @@ -964,6 +1003,9 @@ 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 ) { @@ -1000,21 +1042,31 @@ 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 { @@ -1037,9 +1089,19 @@ sub as_graphml { $node_el->setAttribute( 'id', $node_xmlid ); foreach my $d ( keys %reading_attributes ) { my $nval = $n->$d; - if( $rankoffset && $d eq 'rank' ) { + # 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 + 1 : $nval - $rankoffset; + $nval = $n eq $self->end ? $end->rank - $rankoffset + 1 + : $nval - $rankoffset; } _add_graphml_data( $node_el, $node_data_keys{$d}, $nval ) if defined $nval; @@ -1052,8 +1114,10 @@ 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] }, @@ -1078,6 +1142,10 @@ 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, @@ -1370,11 +1438,10 @@ 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; @@ -1435,69 +1502,6 @@ sub make_witness_path { $wit->clear_uncorrected_path; } -=head2 equivalence_graph( \%readingmap, $startrank, $endrank, @testvector ) - -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.) - -It is also possible to pass in a test relationship in @testvector, and get -the resulting equivalence graph before the relationship has been made. - -=cut - -sub equivalence_graph { - my( $self, $map, $start, $end, @newvector ) = @_; - $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 = $self->related_readings( $r->id, 'colocated' ); - push( @rels, $r->id ); - # Make an equivalence vertex - my $rn = 'equivalence_' . $rel_ctr++; - $eqgraph->add_vertex( $rn ); - # Note which readings belong to this vertex. - push( @rels, $r->id ); - foreach( @rels ) { - $map->{$_} = $rn; - } - } - - # 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 ); - } - - # Collapse the vertices in @newvector if applicable. - if( @newvector ) { - my( $eqs, $eqt ) = map { $map->{$_} } @newvector; - $DB::single = 1 unless $eqs && $eqt; - unless( $eqs eq $eqt ) { - # Combine the vertices. - map { $eqgraph->add_edge( $eqs, $_ ) } $eqgraph->successors( $eqt ); - map { $eqgraph->add_edge( $_, $eqs ) } $eqgraph->predecessors( $eqt ); - $eqgraph->delete_vertex( $eqt ); - } - } - return $eqgraph; -} - =head2 calculate_ranks Calculate the reading ranks (that is, their aligned positions relative @@ -1533,28 +1537,19 @@ 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}; - 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 ); - } + + # 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 $node_ranks->{$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?" ); @@ -1576,41 +1571,6 @@ 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; @@ -1633,8 +1593,17 @@ 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"; + #print STDERR "Combining readings at same rank: $key\n"; $changed = 1; $self->merge_readings( $unique_rank_rdg{$key}, $rdg ); # TODO see if this now makes a common point. @@ -1762,12 +1731,12 @@ 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 @@ -1806,7 +1775,7 @@ sub _common_in_path { if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r1' ) { # print STDERR "Path candidate $p from $lc\n"; push( @candidates, $p ); - } else { + } elsif( !$all_seen{$p->id} ) { $all_seen{$p->id} = 'r1'; push( @new_lc1, $p ); } @@ -1817,7 +1786,7 @@ sub _common_in_path { if( $all_seen{$p->id} && $all_seen{$p->id} ne 'r2' ) { # print STDERR "Path candidate $p from $lc\n"; push( @candidates, $p ); - } else { + } elsif( !$all_seen{$p->id} ) { $all_seen{$p->id} = 'r2'; push( @new_lc2, $p ); }