Commit | Line | Data |
d71100ed |
1 | package Text::Tradition::Analysis; |
2 | |
3 | use strict; |
4 | use warnings; |
5 | use Text::Tradition; |
6 | use Text::Tradition::Stemma; |
7 | |
8 | sub new { |
9 | my( $class, $args ) = @_; |
10 | my $self = {}; |
11 | # Our object needs to have a stemma graph and a variant table. |
12 | my( $svg, $variants ) = run_analysis( $args->{'file'}, $args->{'stemmadot'} ); |
13 | $self->{'svg'} = $svg; |
14 | $self->{'variants'} = $variants; |
15 | |
16 | bless( $self, $class ); |
17 | return $self; |
18 | } |
19 | |
20 | sub run_analysis { |
21 | my( $file, $stemmadot ) = @_; |
22 | # What we will return |
23 | my $svg; |
24 | my $variants = []; |
25 | |
3d79e248 |
26 | # Read in the file and stemma |
d71100ed |
27 | my $tradition = Text::Tradition->new( |
3d79e248 |
28 | 'input' => 'Self', |
29 | 'file' => $file, |
d71100ed |
30 | 'linear' => 1, |
31 | ); |
32 | my $stemma = Text::Tradition::Stemma->new( |
33 | 'collation' => $tradition->collation, |
34 | 'dot' => $stemmadot, |
35 | ); |
36 | # We will return the stemma picture |
37 | $svg = $stemma->as_svg; |
38 | ### DIRTY HACK |
39 | $svg =~ s/transform=\"scale\(1 1\)/transform=\"scale\(0.7 0.7\)/; |
40 | |
41 | # We have the collation, so get the alignment table with witnesses in rows. |
42 | # Also return the reading objects in the table, rather than just the words. |
43 | |
44 | my $all_wits_table = $tradition->collation->make_alignment_table( 'refs' ); |
45 | |
46 | # For each column in the alignment table, we want to see if the existing |
47 | # groupings of witnesses match our stemma hypothesis. We also want, at the |
48 | # end, to produce an HTML table with all the variants. |
49 | my $html_columns = 0; |
50 | my $html_data = []; |
51 | my $total = 0; # Keep track of the total number of variant locations |
52 | |
53 | # Strip the list of sigla and save it for correlation to the readings. |
54 | my $col_wits = shift @$all_wits_table; |
55 | |
56 | # We will return a data structure, an array for each row that looks like: |
57 | # { id = X, genealogical = Y, readings = [ text = X, group = Y], empty = N } |
58 | foreach my $i ( 0 .. $#$all_wits_table ) { |
59 | # For each column in the table, group the readings by witness. |
60 | my $rdg_wits = {}; |
61 | my $col_rdgs = shift @$all_wits_table; |
62 | my $rank; |
63 | foreach my $j ( 0 .. $#{$col_rdgs} ) { |
64 | my $rdg = $col_rdgs->[$j]; |
d71100ed |
65 | my $rdg_text = '(omitted)'; # Initialize in case of empty reading |
66 | if( $rdg ) { |
67 | $rdg_text = $rdg->is_lacuna ? undef : $rdg->text; # Don't count lacunae |
c7d0f253 |
68 | # Get the rank from any real reading; they should be identical. |
69 | $rank = $rdg->rank unless $rank || $rdg->is_lacuna; |
d71100ed |
70 | } |
71 | if( defined $rdg_text ) { |
72 | # Initialize the witness array if we haven't got one yet |
73 | $rdg_wits->{$rdg_text} = [] unless $rdg_wits->{$rdg_text}; |
74 | # Add the relevant witness, subject to a.c. logic |
75 | add_variant_wit( $rdg_wits->{$rdg_text}, $col_wits->[$j], |
76 | $tradition->collation->ac_label ); |
77 | } |
78 | } |
79 | |
80 | # See if this column has any potentially genealogical variants. |
81 | # If not, skip to the next. |
82 | $total++ unless scalar keys %$rdg_wits == 1; |
83 | my( $groups, $readings ) = useful_variant( $rdg_wits ); |
84 | next unless $groups && $readings; |
85 | |
86 | # Initialize the data structure for the row that we will return |
87 | my $variant_row = {'id' => $rank, 'readings' => [] }; |
88 | # Keep track of our widest row |
89 | $html_columns = scalar @$groups if scalar @$groups > $html_columns; |
90 | |
91 | # We can already look up witnesses for a reading; we also want to look |
92 | # up readings for a given witness. |
93 | my $group_readings = {}; |
94 | foreach my $x ( 0 .. $#$groups ) { |
95 | $group_readings->{wit_stringify( $groups->[$x] )} = $readings->[$x]; |
96 | } |
97 | |
98 | # For all the groups with more than one member, collect the list of all |
99 | # contiguous vertices needed to connect them. |
100 | # TODO: deal with a.c. reading logic |
101 | my $sc = analyze_variant_location( $group_readings, $groups, $stemma->apsp ); |
102 | $variant_row->{'genealogical'} = keys %$sc ? 1 : undef; |
103 | foreach my $grp ( sort keys %$group_readings ) { |
104 | my $rdg = $group_readings->{$grp}; |
105 | push( @{$variant_row->{'readings'}}, { 'text' => $rdg, 'group' => $grp } ); |
106 | } |
107 | |
108 | # Now run the same analysis given the calculated distance tree(s). |
109 | # foreach my $tree ( 0 .. $#{$stemma->distance_trees} ) { |
110 | # my $dc = analyze_variant_location( $group_readings, $groups, |
111 | # $stemma->distance_apsps->[$tree] ); |
112 | # foreach my $rdg ( keys %$dc ) { |
113 | # my $var = $dc->{$rdg}; |
114 | # } |
115 | # } |
116 | |
117 | # Record that we used this variant in an analysis |
118 | push( @$variants, $variant_row ); |
119 | } |
120 | |
121 | # Go through our variant rows and add the number of empty columns we need. |
122 | foreach my $row ( @$variants ) { |
123 | my $empty = $html_columns - scalar @{$row->{'readings'}}; |
124 | $row->{'empty'} = $empty; |
125 | } |
126 | |
127 | return( $svg, $variants ); |
128 | } |
129 | |
130 | sub analyze_variant_location { |
131 | my( $group_readings, $groups, $apsp ) = @_; |
132 | my %contig; |
133 | my $conflict = {}; |
134 | foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { |
135 | my @members = @$g; |
136 | my $gst = wit_stringify( $g ); |
137 | map { $contig{$_} = $gst } @members; # The witnesses need themselves to be |
138 | # in their collection. |
139 | next unless @members > 1; |
140 | my $curr = pop @members; |
141 | foreach my $m ( @members ) { |
142 | foreach my $v ( $apsp->path_vertices( $curr, $m ) ) { |
143 | $contig{$v} = $gst unless exists $contig{$v}; |
144 | next if $contig{$v} eq $gst; |
145 | # print STDERR "Conflict at $v between group $gst and group " |
146 | # . $contig{$v} . "\n"; |
147 | # Record what is conflicting. |
148 | $conflict->{$group_readings->{$gst}} = $group_readings->{$contig{$v}}; |
149 | } |
150 | } |
151 | } |
152 | return $conflict; |
153 | } |
154 | |
155 | # Add the variant, subject to a.c. representation logic. |
156 | # This assumes that we will see the 'main' version before the a.c. version. |
157 | sub add_variant_wit { |
158 | my( $arr, $wit, $acstr ) = @_; |
159 | my $skip; |
160 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
161 | my $real = $1; |
162 | $skip = grep { $_ =~ /^\Q$real\E$/ } @$arr; |
163 | } |
164 | push( @$arr, $wit ) unless $skip; |
165 | } |
166 | |
167 | # Return an answer if the variant is useful, i.e. if there are at least 2 variants |
168 | # with at least 2 witnesses each. |
169 | sub useful_variant { |
170 | my( $readings ) = @_; |
171 | my $total = keys %$readings; |
172 | foreach my $var ( keys %$readings ) { |
173 | $total-- if @{$readings->{$var}} == 1; |
174 | } |
175 | return( undef, undef ) if $total <= 1; |
176 | my( $groups, $text ); |
177 | foreach my $var ( keys %$readings ) { |
178 | push( @$groups, $readings->{$var} ); |
179 | push( @$text, $var ); |
180 | } |
181 | return( $groups, $text ); |
182 | } |
183 | |
184 | # Take an array of witness groupings and produce a string like |
185 | # ['A','B'] / ['C','D','E'] / ['F'] |
186 | |
187 | sub wit_stringify { |
188 | my $groups = shift; |
189 | my @gst; |
190 | # If we were passed an array of witnesses instead of an array of |
191 | # groupings, then "group" the witnesses first. |
192 | unless( ref( $groups->[0] ) ) { |
193 | my $mkgrp = [ $groups ]; |
194 | $groups = $mkgrp; |
195 | } |
196 | foreach my $g ( @$groups ) { |
197 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
198 | } |
199 | return join( ' / ', @gst ); |
200 | } |
201 | |
202 | 1; |