X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FAnalysis.pm;h=fda51376077c9e67781e1c5e134c203704fd706a;hb=679f17e1a60a81370df8cbb49b94a2b5d19e3a98;hp=049211b72b2d412cb00fd2c55ee293ced07c4df4;hpb=bebec0e9db285a2161d08667ba98fa3bcb91d9e4;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Analysis.pm b/lib/Text/Tradition/Analysis.pm index 049211b..fda5137 100644 --- a/lib/Text/Tradition/Analysis.pm +++ b/lib/Text/Tradition/Analysis.pm @@ -3,7 +3,11 @@ package Text::Tradition::Analysis; use strict; use warnings; use Benchmark; +use Encode qw/ encode_utf8 /; use Exporter 'import'; +use Graph; +use JSON qw/ encode_json decode_json /; +use LWP::UserAgent; use Text::Tradition; use Text::Tradition::Stemma; @@ -37,15 +41,24 @@ between readings are actually kept. =head1 SUBROUTINES -=head2 run_analysis( $tradition, $stemma_id, @merge_relationship_types ) +=head2 run_analysis( $tradition, %opts ) -Runs the analysis described in analyze_variant_location on every location -in the collation of the given tradition, against the stemma specified in -$stemma_id. If $stemma_id is not specified, it defaults to 0 (referencing -the first stemma saved for the tradition.) +Runs the analysis described in analyze_variant_location on every location in the +collation of the given tradition, with the given options. These include: -The optional @merge_relationship_types contains a list of relationship types -to treat as equivalent for the analysis. +=over 4 + +=item * stemma_id - Specify which of the tradition's stemmata to use. Default +is 0 (i.e. the first). + +=item * ranks - Specify a list of location ranks to analyze; exclude the rest. + +=item * merge_types - Specify a list of relationship types, where related readings +should be treated as identical for the purposes of analysis. + +=item * exclude_type1 - Exclude those ranks whose groupings have only type-1 variants. + +=back =begin testing @@ -59,48 +72,120 @@ my $tradition = Text::Tradition->new( 'input' => 'TEI', my $s = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' ); is( ref( $s ), 'Text::Tradition::Stemma', "Added stemma to tradition" ); +my %expected_genealogical = ( + 1 => 0, + 2 => 1, + 3 => 0, + 5 => 0, + 7 => 0, + 8 => 0, + 10 => 0, + 13 => 1, + 33 => 0, + 34 => 0, + 37 => 0, + 60 => 0, + 81 => 1, + 84 => 0, + 87 => 0, + 101 => 0, + 102 => 0, + 122 => 1, + 157 => 0, + 166 => 1, + 169 => 1, + 200 => 0, + 216 => 1, + 217 => 1, + 219 => 1, + 241 => 1, + 242 => 1, + 243 => 1, +); + my $data = run_analysis( $tradition ); -# TODO Check genealogical count -is( $data->{'genealogical_count'}, 13, "Got right genealogical count" ); -is( $data->{'conflict_count'}, 16, "Got right conflict count" ); -is( $data->{'variant_count'}, 28, "Got right total variant number" ); +foreach my $row ( @{$data->{'variants'}} ) { + # Account for rows that used to be "not useful" + unless( exists $expected_genealogical{$row->{'id'}} ) { + $expected_genealogical{$row->{'id'}} = 1; + } + is( $row->{'genealogical'}, $expected_genealogical{$row->{'id'}}, + "Got correct genealogical flag for row " . $row->{'id'} ); +} +is( $data->{'variant_count'}, 58, "Got right total variant number" ); +# TODO Make something meaningful of conflict count, maybe test other bits =end testing =cut sub run_analysis { - my( $tradition, $stemma_id, @collapse ) = @_; - $stemma_id = 0 unless $stemma_id; - - # Run the variant analysis on every rank in the graph that doesn't - # have a common reading. Return the results. - my @variants; # holds results from analyze_variant_location - my $genealogical; # counter of 'genealogical' variants - my $conflicts; # counter of conflicting readings - - # Find and mark 'common' ranks for exclusion. - my %common_rank; - foreach my $rdg ( $tradition->collation->common_readings ) { - $common_rank{$rdg->rank} = 1; + my( $tradition, %opts ) = @_; + my $c = $tradition->collation; + + my $stemma_id = $opts{'stemma_id'} || 0; + my @ranks = ref( $opts{'ranks'} ) eq 'ARRAY' ? @{$opts{'ranks'}} : (); + my @collapse = ref( $opts{'merge_types'} ) eq 'ARRAY' ? @{$opts{'merge_types'}} : (); + + # Get the stemma + my $stemma = $tradition->stemma( $stemma_id ); + + # Figure out which witnesses we are working with + my @lacunose = $stemma->hypotheticals; + my @tradition_wits = map { $_->sigil } $tradition->witnesses; + map { push( @tradition_wits, $_->sigil.$c->ac_label ) if $_->is_layered } + $tradition->witnesses; + push( @lacunose, _symmdiff( [ $stemma->witnesses ], \@tradition_wits ) ); + + # Find and mark 'common' ranks for exclusion, unless they were + # explicitly specified. + unless( @ranks ) { + my %common_rank; + foreach my $rdg ( $c->common_readings ) { + $common_rank{$rdg->rank} = 1; + } + @ranks = grep { !$common_rank{$_} } ( 1 .. $c->end->rank-1 ); } - foreach my $rank ( 1 .. $tradition->collation->end->rank-1 ) { - next if $common_rank{$rank}; - my $variant_row = analyze_variant_location( - $tradition, $rank, $stemma_id, @collapse ); - next unless $variant_row; - push( @variants, $variant_row ); - $genealogical++ if $variant_row->{'genealogical'}; - $conflicts += grep { $_->{'conflict'} } @{$variant_row->{'readings'}}; + # Group the variants to send to the solver + my @groups; + my %lacunae; + foreach my $rank ( @ranks ) { + my $missing = [ @lacunose ]; + my $rankgroup = group_variants( $tradition, $rank, $missing, \@collapse ); + if( $opts{'exclude_type1'} ) { + # Check to see whether this is a "useful" group. + my( $rdgs, $grps ) = _useful_variant( $rankgroup ); + next unless @$rdgs; + } + push( @groups, $rankgroup ); + $lacunae{$rank} = $missing; } + $DB::single = 1; + # Parse the answer + my $answer = solve_variants( $stemma, @groups ); + + # Do further analysis on the answer + my $conflict_count = 0; + foreach my $idx ( 0 .. $#ranks ) { + my $location = $answer->{'variants'}->[$idx]; + # Add the rank back in + $location->{'id'} = $ranks[$idx]; + # Add the lacunae back in + $location->{'missing'} = $lacunae{$ranks[$idx]}; + # Run the extra analysis we need. + analyze_location( $tradition, $stemma->graph, $location ); + # Add the reading text back in + foreach my $rdghash ( @{$location->{'readings'}} ) { + $conflict_count++ + if exists $rdghash->{'conflict'} && $rdghash->{'conflict'}; + my $rdg = $c->reading( $rdghash->{'readingid'} ); + $rdghash->{'text'} = $rdg ? $rdg->text : $rdghash->{'readingid'}; + } + } + $answer->{'conflict_count'} = $conflict_count; - return { - 'variants' => \@variants, - 'variant_count' => scalar @variants, # TODO redundant - 'conflict_count' => $conflicts, - 'genealogical_count' => $genealogical, - }; + return $answer; } =head2 group_variants( $tradition, $rank, $lacunose, @merge_relationship_types ) @@ -110,8 +195,8 @@ relationships in @merge_relationship_types as equivalent. $lacunose should be a reference to an array, to which the sigla of lacunose witnesses at this rank will be appended. -Returns two ordered lists $readings, $groups, where $readings->[$n] is attested -by the witnesses listed in $groups->[$n]. +Returns a hash $group_readings where $rdg is attested by the witnesses listed +in $group_readings->{$rdg}. =cut @@ -119,36 +204,38 @@ by the witnesses listed in $groups->[$n]. sub group_variants { my( $tradition, $rank, $lacunose, $collapse ) = @_; my $c = $tradition->collation; + my $aclabel = $c->ac_label; # Get the alignment table readings my %readings_at_rank; my @gap_wits; - foreach my $tablewit ( @{$tradition->collation->alignment_table->{'alignment'}} ) { + foreach my $tablewit ( @{$c->alignment_table->{'alignment'}} ) { my $rdg = $tablewit->{'tokens'}->[$rank-1]; + my $wit = $tablewit->{'witness'}; if( $rdg && $rdg->{'t'}->is_lacuna ) { - _add_to_witlist( $tablewit->{'witness'}, $lacunose, - $tradition->collation->ac_label ); + _add_to_witlist( $wit, $lacunose, $aclabel ); } elsif( $rdg ) { $readings_at_rank{$rdg->{'t'}->text} = $rdg->{'t'}; } else { - _add_to_witlist( $tablewit->{'witness'}, \@gap_wits, - $tradition->collation->ac_label ); + _add_to_witlist( $wit, \@gap_wits, $aclabel ); } } # Group the readings, collapsing groups by relationship if needed my %grouped_readings; - foreach my $rdg ( sort { $b->witnesses <=> $a->witnesses } values %readings_at_rank ) { + foreach my $rdg ( sort { $b->witnesses <=> $a->witnesses } + values %readings_at_rank ) { # Skip readings that have been collapsed into others. - next if exists $grouped_readings{$rdg->text} && !$grouped_readings{$rdg->text}; + next if exists $grouped_readings{$rdg->id} && !$grouped_readings{$rdg->id}; my @wits = $rdg->witnesses; if( $collapse ) { my $filter = sub { my $r = $_[0]; grep { $_ eq $r->type } @$collapse; }; foreach my $other ( $rdg->related_readings( $filter ) ) { - push( @wits, $other->witnesses ); - $grouped_readings{$other->text} = 0; + my @otherwits = $other->witnesses; + push( @wits, @otherwits ); + $grouped_readings{$other->id} = 0; } } - $grouped_readings{$rdg->text} = \@wits; + $grouped_readings{$rdg->id} = \@wits; } $grouped_readings{'(omitted)'} = \@gap_wits if @gap_wits; # Get rid of our collapsed readings @@ -159,68 +246,316 @@ sub group_variants { return \%grouped_readings; } -=head2 analyze_variant_location( $tradition, $rank, $stemma_id, @merge_relationship_types ) +=head2 solve_variants( $graph, @groups ) + +Sends the set of groups to the external graph solver service and returns +a cleaned-up answer, adding the rank IDs back where they belong. + +The JSON has the form + { "graph": [ stemmagraph DOT string without newlines ], + "groupings": [ array of arrays of groups, one per rank ] } + +The answer has the form + { "variants" => [ array of variant location structures ], + "variant_count" => total, + "conflict_count" => number of conflicts detected, + "genealogical_count" => number of solutions found } + +=cut + +sub solve_variants { + my( $stemma, @groups ) = @_; + + # Make the json with stemma + groups + my $groupings = []; + foreach my $ghash ( @groups ) { + my @grouping; + foreach my $k ( keys %$ghash ) { + push( @grouping, $ghash->{$k} ); + } + push( @$groupings, \@grouping ); + } + ## Witness map is a HACK to get around limitations in node names from IDP + my $witness_map = {}; + my $json = encode_json( _safe_wit_strings( $stemma, $groupings, $witness_map ) ); + + # Send it off and get the result + my $solver_url = 'http://byzantini.st/cgi-bin/graphcalc.cgi'; + my $ua = LWP::UserAgent->new(); + my $resp = $ua->post( $solver_url, 'Content-Type' => 'application/json', + 'Content' => $json ); + + my $answer; + my $used_idp; + if( $resp->is_success ) { + $answer = _desanitize_names( decode_json( $resp->content ), $witness_map ); + $used_idp = 1; + } else { + # Fall back to the old method. + warn "IDP solver returned " . $resp->status_line . " / " . $resp->content + . "; falling back to perl method"; + $answer = perl_solver( $stemma, @$groupings ); + } + + # Fold the result back into what we know about the groups. + my $variants = []; + my $genealogical = 0; + foreach my $idx ( 0 .. $#groups ) { + my( $calc_groups, $result ) = @{$answer->[$idx]}; + if( $result ) { + $genealogical++; + # Prune the calculated groups, in case the IDP solver failed to. + if( $used_idp ) { + my @pruned_groups; + foreach my $cg ( @$calc_groups ) { + my @pg = _prune_group( $cg, $stemma ); + push( @pruned_groups, \@pg ); + } + $calc_groups = \@pruned_groups; + } + } + my $input_group = $groups[$idx]; + foreach my $k ( sort keys %$input_group ) { + my $cg = shift @$calc_groups; + $input_group->{$k} = $cg; + } + my $vstruct = { + 'genealogical' => $result, + 'readings' => [], + }; + foreach my $k ( keys %$input_group ) { + push( @{$vstruct->{'readings'}}, + { 'readingid' => $k, 'group' => $input_group->{$k}} ); + } + push( @$variants, $vstruct ); + } + + return { 'variants' => $variants, + 'variant_count' => scalar @$variants, + 'genealogical_count' => $genealogical }; +} + +#### HACKERY to cope with IDP's limited idea of what a node name looks like ### + +sub _safe_wit_strings { + my( $stemma, $groupings, $witness_map ) = @_; + my $safegraph = Graph->new(); + # Convert the graph to a safe representation and store the conversion. + foreach my $n ( $stemma->graph->vertices ) { + my $sn = _safe_witstr( $n ); + warn "Ambiguous stringification $sn for $n and " . $witness_map->{$sn} + if exists $witness_map->{$sn}; + $witness_map->{$sn} = $n; + $safegraph->add_vertex( $sn ); + $safegraph->set_vertex_attributes( $sn, + $stemma->graph->get_vertex_attributes( $n ) ); + } + foreach my $e ( $stemma->graph->edges ) { + my @safe_e = ( _safe_witstr( $e->[0] ), _safe_witstr( $e->[1] ) ); + $safegraph->add_edge( @safe_e ); + } + my $safe_stemma = Text::Tradition::Stemma->new( + 'collation' => $stemma->collation, 'graph' => $safegraph ); + + # Now convert the witness groupings to a safe representation. + my $safe_groupings = []; + foreach my $grouping ( @$groupings ) { + my $safe_grouping = []; + foreach my $group ( @$grouping ) { + my $safe_group = []; + foreach my $n ( @$group ) { + my $sn = _safe_witstr( $n ); + warn "Ambiguous stringification $sn for $n and " . $witness_map->{$sn} + if exists $witness_map->{$sn} && $witness_map->{$sn} ne $n; + $witness_map->{$sn} = $n; + push( @$safe_group, $sn ); + } + push( @$safe_grouping, $safe_group ); + } + push( @$safe_groupings, $safe_grouping ); + } + + # Return it all in the struct we expect. We have stored the reductions + # in the $witness_map that we were passed. + return { 'graph' => $safe_stemma->editable( ' ' ), 'groupings' => $safe_groupings }; +} + +sub _safe_witstr { + my $witstr = shift; + $witstr =~ s/\s+/_/g; + $witstr =~ s/[^\w\d-]//g; + return $witstr; +} + +sub _desanitize_names { + my( $jsonstruct, $witness_map ) = @_; + my $result = []; + foreach my $grouping ( @$jsonstruct ) { + my $real_grouping = []; + foreach my $element ( @$grouping ) { + if( ref( $element ) eq 'ARRAY' ) { + # it's the groupset. + my $real_groupset = []; + foreach my $group ( @$element ) { + my $real_group = []; + foreach my $n ( @$group ) { + my $rn = $witness_map->{$n}; + push( @$real_group, $rn ); + } + push( @$real_groupset, $real_group ); + } + push( @$real_grouping, $real_groupset ); + } else { + # It is the boolean, not actually a group. + push( @$real_grouping, $element ); + } + } + push( @$result, $real_grouping ); + } + return $result; +} + +### END HACKERY ### + +=head2 analyze_location ( $tradition, $graph, $location_hash ) + +Given the tradition, its stemma graph, and the solution from the graph solver, +work out the rest of the information we want. For each reading we need missing, +conflict, reading_parents, independent_occurrence, followed, not_followed, and follow_unknown. Alters the location_hash in place. + +=cut + +sub analyze_location { + my ( $tradition, $graph, $variant_row ) = @_; + + # Make a hash of all known node memberships, and make the subgraphs. + my $contig = {}; + my $reading_roots = {}; + my $subgraph = {}; + foreach my $rdghash ( @{$variant_row->{'readings'}} ) { + my $rid = $rdghash->{'readingid'}; + map { $contig->{$_} = $rid } @{$rdghash->{'group'}}; + + # Make the subgraph. + my $part = $graph->copy; + my %these_vertices; + map { $these_vertices{$_} = 1 } @{$rdghash->{'group'}}; + $part->delete_vertices( grep { !$these_vertices{$_} } $part->vertices ); + $subgraph->{$rid} = $part; + # Get the reading roots. + map { $reading_roots->{$_} = $rid } $part->predecessorless_vertices; + } + + # Now that we have all the node group memberships, calculate followed/ + # non-followed/unknown values for each reading. Also figure out the + # reading's evident parent(s). + foreach my $rdghash ( @{$variant_row->{'readings'}} ) { + # Group string key - TODO do we need this? + my $gst = wit_stringify( $rdghash->{'group'} ); + my $rid = $rdghash->{'readingid'}; + # Get the subgraph + my $part = $subgraph->{$rid}; + + # Start figuring things out. + my @roots = $part->predecessorless_vertices; + $rdghash->{'independent_occurrence'} = scalar @roots; + $rdghash->{'followed'} = scalar( $part->vertices ) - scalar( @roots ); + # Find the parent readings, if any, of this reading. + my %rdgparents; + foreach my $wit ( @roots ) { + # Look in the main stemma to find this witness's extant or known-reading + # immediate ancestor(s), and look up the reading that each ancestor olds. + my @check = $graph->predecessors( $wit ); + while( @check ) { + my @next; + foreach my $wparent( @check ) { + my $preading = $contig->{$wparent}; + if( $preading ) { + $rdgparents{$preading} = 1; + } else { + push( @next, $graph->predecessors( $wparent ) ); + } + } + @check = @next; + } + } + $rdghash->{'reading_parents'} = [ keys %rdgparents ]; + + # Find the number of times this reading was altered, and the number of + # times we're not sure. + my( %nofollow, %unknownfollow ); + foreach my $wit ( $part->vertices ) { + foreach my $wchild ( $graph->successors( $wit ) ) { + next if $part->has_vertex( $wchild ); + if( $reading_roots->{$wchild} && $contig->{$wchild} ) { + # It definitely changed here. + $nofollow{$wchild} = 1; + } elsif( !($contig->{$wchild}) ) { + # The child is a hypothetical node not definitely in + # any group. Answer is unknown. + $unknownfollow{$wchild} = 1; + } # else it's a non-root node in a known group, and therefore + # is presumed to have its reading from its group, not this link. + } + } + $rdghash->{'not_followed'} = keys %nofollow; + $rdghash->{'follow_unknown'} = keys %unknownfollow; + + # Now say whether this reading represents a conflict. + unless( $variant_row->{'genealogical'} ) { + $rdghash->{'conflict'} = @roots != 1; + } + } +} + + +=head2 perl_solver( $tradition, $rank, $stemma_id, @merge_relationship_types ) + +** NOTE ** This method should hopefully not be called - it is not guaranteed +to be correct. Serves as a backup for the real solver. Runs an analysis of the given tradition, at the location given in $rank, against the graph of the stemma specified in $stemma_id. The argument @merge_relationship_types is an optional list of relationship types for which readings so related should be treated as equivalent. -Returns a data structure as follows: +Returns a nested array data structure as follows: - { 'id' => $rank, - 'genealogical' => boolean, - 'readings => [ { text => $reading_text, - group => [ witnesses ], - conflict => [ conflicting ], - missing => [ excluded ] }, ... ] - } -where 'conflicting' is the list of witnesses whose readings conflict with -this group, and 'excluded' is the list of witnesses either not present in -the stemma or lacunose at this location. + [ [ group_list, is_genealogical ], [ group_list, is_genealogical ] ... ] + +where the group list is the array of arrays passed in for each element of @groups, +possibly with the addition of hypothetical readings. + =cut -sub analyze_variant_location { - my( $tradition, $rank, $sid, @collapse ) = @_; - # Get the readings in this tradition at this rank - my @rank_rdgs = grep { $_->rank == $rank } $tradition->collation->readings; - # Get the applicable stemma - my $undirected; # TODO Allow undirected distance tree analysis too - my $stemma = $tradition->stemma( $sid ); +sub perl_solver { + my( $stemma, @groups ) = @_; my $graph = $stemma->graph; - # Figure out which witnesses we are working with - my @lacunose = $stemma->hypotheticals; - push( @lacunose, _symmdiff( [ $stemma->witnesses ], - [ map { $_->sigil } $tradition->witnesses ] ) ); - - # Now group the readings - my( $readings, $groups ) = _useful_variant( - group_variants( $tradition, $rank, \@lacunose, \@collapse ), - $graph, $tradition->collation->ac_label ); - return unless scalar @$readings; - my $group_readings = {}; - # Lookup table group string -> readings - foreach my $x ( 0 .. $#$groups ) { - $group_readings->{wit_stringify( $groups->[$x] )} = $readings->[$x]; + my @answer; + foreach my $g ( @groups ) { + push( @answer, _solve_variant_location( $graph, $g ) ); } + return \@answer; +} +sub _solve_variant_location { + my( $graph, $groups ) = @_; # Now do the work. my $contig = {}; my $subgraph = {}; my $is_conflicted; my $conflict = {}; - my %reading_roots; - my $variant_row = { 'id' => $rank, 'readings' => [] }; + # Mark each ms as in its own group, first. foreach my $g ( @$groups ) { my $gst = wit_stringify( $g ); map { $contig->{$_} = $gst } @$g; } + # Now for each unmarked node in the graph, initialize an array # for possible group memberships. We will use this later to # resolve potential conflicts. - $DB::single = 1 if $rank == 636; map { $contig->{$_} = [] unless $contig->{$_} } $graph->vertices; foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { my $gst = wit_stringify( $g ); # This is the group name @@ -231,42 +566,34 @@ sub analyze_variant_location { grep { !ref( $contig->{$_} ) && $contig->{$_} ne $gst } $graph->vertices ); # Now look to see if our group is connected. - if( $undirected ) { # For use with distance trees etc. - # Find all vertices reachable from the first (arbitrary) group - # member. If we are genealogical this should include them all. - my $reachable = {}; - map { $reachable->{$_} = 1 } $part->all_reachable( $g->[0] ); - # TODO This is a terrible way to do distance trees, since all - # non-leaf nodes are included in every graph part now. We may - # have to go back to SPDP. - } else { - if( @$g > 1 ) { - # We have to take directionality into account. - # How many root nodes do we have? - my @roots = grep { ref( $contig->{$_} ) || $contig->{$_} eq $gst } - $part->predecessorless_vertices; - # Assuming that @$g > 1, find the first root node that has at - # least one successor belonging to our group. If this reading - # is genealogical, there should be only one, but we will check - # that implicitly later. - foreach my $root ( @roots ) { - # Prune the tree to get rid of extraneous hypotheticals. - $root = _prune_subtree( $part, $root, $contig ); - next unless $root; - # Save this root for our group. - push( @group_roots, $root ); - # Get all the successor nodes of our root. - } - } else { - # Dispense with the trivial case of one reading. - @group_roots = @$g; - _prune_subtree( $part, @group_roots, $contig ); - } + if( @$g > 1 ) { + # We have to take directionality into account. + # How many root nodes do we have? + my @roots = grep { ref( $contig->{$_} ) || $contig->{$_} eq $gst } + $part->predecessorless_vertices; + # Assuming that @$g > 1, find the first root node that has at + # least one successor belonging to our group. If this reading + # is genealogical, there should be only one, but we will check + # that implicitly later. + foreach my $root ( @roots ) { + # Prune the tree to get rid of extraneous hypotheticals. + $root = _prune_subtree( $part, $root, $contig ); + next unless $root; + # Save this root for our group. + push( @group_roots, $root ); + # Get all the successor nodes of our root. + } + } else { + # Dispense with the trivial case of one reading. + my $wit = $g->[0]; + @group_roots = ( $wit ); + foreach my $v ( $part->vertices ) { + $part->delete_vertex( $v ) unless $v eq $wit; + } } - map { $reading_roots{$_} = 1 } @group_roots; if( @group_roots > 1 ) { - $conflict->{$group_readings->{$gst}} = 1; + $conflict->{$gst} = 1; $is_conflicted = 1; } # Paint the 'hypotheticals' with our group. @@ -279,13 +606,8 @@ sub analyze_variant_location { } - # Start to write the reading, and save the group subgraph. - my $reading = { 'text' => $group_readings->{$gst}, - 'missing' => wit_stringify( \@lacunose ), - 'group' => $gst }; # This will change if we find no conflict # Save the relevant subgraph. $subgraph->{$gst} = $part; - push( @{$variant_row->{'readings'}}, $reading ); } # For each of our hypothetical readings, flatten its 'contig' array if @@ -335,7 +657,7 @@ sub analyze_variant_location { # $h is needed by this group. if( exists( $still_contig->{$h} ) ) { # Conflict! - $conflict->{$group_readings->{$gst}} = 1; + $conflict->{$gst} = 1; $still_contig->{$h} = ''; } else { $still_contig->{$h} = $gst; @@ -356,7 +678,7 @@ sub analyze_variant_location { # We need $h. if( exists $still_contig->{$h} ) { # Conflict! - $conflict->{$group_readings->{$gst}} = 1; + $conflict->{$gst} = 1; $still_contig->{$h} = ''; } else { $still_contig->{$h} = $gst; @@ -376,61 +698,35 @@ sub analyze_variant_location { } } # end if @resolve - # Now that we have all the node group memberships, calculate followed/ - # non-followed/unknown values for each reading. Also figure out the - # reading's evident parent(s). - foreach my $rdghash ( @{$variant_row->{'readings'}} ) { - my $gst = $rdghash->{'group'}; - my $part = $subgraph->{$gst}; - my @roots = $part->predecessorless_vertices; - $rdghash->{'independent_occurrence'} = scalar @roots; - $rdghash->{'followed'} = scalar( $part->vertices ) - scalar( @roots ); - # Find the parent readings, if any, of this reading. - my @rdgparents; - foreach my $wit ( @roots ) { - # Look in the main stemma to find this witness's parent(s), and look - # up the reading that the parent holds. - foreach my $wparent( $graph->predecessors( $wit ) ) { - my $pgroup = $contig->{$wparent}; - if( $pgroup ) { - push( @rdgparents, $group_readings->{$pgroup} ); - } - } - } - $rdghash->{'reading_parents'} = \@rdgparents; - - # Find the number of times this reading was altered, and the number of - # times we're not sure. - my( %nofollow, %unknownfollow ); - foreach my $wit ( $part->vertices ) { - foreach my $wchild ( $graph->successors( $wit ) ) { - next if $part->has_vertex( $wchild ); - if( $reading_roots{$wchild} && $contig->{$wchild} ) { - # It definitely changed here. - $nofollow{$wchild} = 1; - } elsif( !($contig->{$wchild}) ) { - # The child is a hypothetical node not definitely in - # any group. Answer is unknown. - $unknownfollow{$wchild} = 1; - } # else it's a non-root node in a known group, and therefore - # is presumed to have its reading from its group, not this link. - } - } - $rdghash->{'not_followed'} = keys %nofollow; - $rdghash->{'follow_unknown'} = keys %unknownfollow; + my $is_genealogical = keys %$conflict ? JSON::false : JSON::true; + my $variant_row = [ [], $is_genealogical ]; + # Fill in the groupings from $contig. + foreach my $g ( @$groups ) { + my $gst = wit_stringify( $g ); + my @realgroup = grep { $contig->{$_} eq $gst } keys %$contig; + push( @{$variant_row->[0]}, \@realgroup ); } - - # Now write the group and conflict information into the respective rows. - foreach my $rdghash ( @{$variant_row->{'readings'}} ) { - $rdghash->{'conflict'} = $conflict->{$rdghash->{'text'}}; - my @members = grep { $contig->{$_} eq $rdghash->{'group'} } keys %$contig; - $rdghash->{'group'} = wit_stringify( \@members ); - } - - $variant_row->{'genealogical'} = !( keys %$conflict ); return $variant_row; } +sub _prune_group { + my( $group, $stemma ) = @_; + # Get these into a form prune_subtree will recognize. Make a "contighash" + my $hypohash = {}; + map { $hypohash->{$_} = 1 } @$group; + # ...with reference values for hypotheticals. + map { $hypohash->{$_} = [] } $stemma->hypotheticals; + # Make our subgraph + my $subgraph = $stemma->graph->copy; + map { $subgraph->delete_vertex( $_ ) unless exists $hypohash->{$_} } + $subgraph->vertices; + # ...and find the root. + my( $root ) = $subgraph->predecessorless_vertices; + # Now prune and return the remaining vertices. + _prune_subtree( $subgraph, $root, $hypohash ); + return $subgraph->vertices; +} + sub _prune_subtree { my( $tree, $root, $contighash ) = @_; # First, delete hypothetical leaves / orphans until there are none left.