Commit | Line | Data |
dbcf12a6 |
1 | package TreeOfTexts::Model::Analysis; |
2 | |
3 | use strict; |
4 | use warnings; |
5 | use Exporter 'import'; |
6 | use Text::Tradition; |
7 | use Text::Tradition::Stemma; |
8 | |
9 | use vars qw/ @EXPORT_OK /; |
10 | @EXPORT_OK = qw/ run_analysis /; |
11 | |
12 | my $tradition; # Global for convenience, ew. |
13 | # Returns |
14 | sub run_analysis { |
15 | my( $file, $stemmadot ) = @_; |
16 | # What we will return |
17 | my $svg; |
18 | my $variants = []; |
19 | |
20 | # Read in the file and stemma |
21 | my @lines; |
22 | open( INFILE, "$file" ) or die "Could not read $file"; |
23 | binmode INFILE, ':utf8'; |
24 | @lines = <INFILE>; |
25 | close INFILE; |
26 | |
27 | $tradition = Text::Tradition->new( |
28 | 'Self' => join( '', @lines ), |
29 | 'linear' => 1, |
30 | ); |
31 | my $stemma = Text::Tradition::Stemma->new( |
32 | 'collation' => $tradition->collation, |
33 | 'dot' => $stemmadot, |
34 | ); |
35 | # We will return the stemma picture |
36 | $svg = $stemma->as_svg; |
37 | |
38 | # We have the collation, so get the alignment table with witnesses in rows. |
39 | # Also return the reading objects in the table, rather than just the words. |
40 | |
41 | my $all_wits_table = $tradition->collation->make_alignment_table( 'refs' ); |
42 | |
43 | # For each column in the alignment table, we want to see if the existing |
44 | # groupings of witnesses match our stemma hypothesis. We also want, at the |
45 | # end, to produce an HTML table with all the variants. |
46 | my $html_columns = 0; |
47 | my $html_data = []; |
48 | my $total = 0; # Keep track of the total number of variant locations |
49 | |
50 | # Strip the list of sigla and save it for correlation to the readings. |
51 | my $col_wits = shift @$all_wits_table; |
52 | |
53 | # We will return a data structure, an array for each row that looks like: |
54 | # { id = X, genealogical = Y, readings = [ text = X, group = Y], empty = N } |
55 | foreach my $i ( 0 .. $#$all_wits_table ) { |
56 | # For each column in the table, group the readings by witness. |
57 | my $rdg_wits = {}; |
58 | my $col_rdgs = shift @$all_wits_table; |
59 | my $rank; |
60 | foreach my $j ( 0 .. $#{$col_rdgs} ) { |
61 | my $rdg = $col_rdgs->[$j]; |
62 | $rank = $rdg->rank if $rdg && !$rank; # Make a note of our rank |
63 | my $rdg_text = '(omitted)'; # Initialize in case of empty reading |
64 | if( $rdg ) { |
65 | $rdg_text = $rdg->is_lacuna ? undef : $rdg->text; # Don't count lacunae |
66 | } |
67 | if( defined $rdg_text ) { |
68 | # Initialize the witness array if we haven't got one yet |
69 | $rdg_wits->{$rdg_text} = [] unless $rdg_wits->{$rdg_text}; |
70 | # Add the relevant witness, subject to a.c. logic |
71 | add_variant_wit( $rdg_wits->{$rdg_text}, $col_wits->[$j] ); |
72 | } |
73 | } |
74 | |
75 | # See if this column has any potentially genealogical variants. |
76 | # If not, skip to the next. |
77 | $total++ unless scalar keys %$rdg_wits == 1; |
78 | my( $groups, $readings ) = useful_variant( $rdg_wits ); |
79 | next unless $groups && $readings; |
80 | |
81 | # Initialize the data structure for the row that we will return |
82 | my $variant_row = {'id' => $rank, 'readings' => [] }; |
83 | # Keep track of our widest row |
84 | $html_columns = scalar @$groups if scalar @$groups > $html_columns; |
85 | |
86 | # We can already look up witnesses for a reading; we also want to look |
87 | # up readings for a given witness. |
88 | my $group_readings = {}; |
89 | foreach my $x ( 0 .. $#$groups ) { |
90 | $group_readings->{wit_stringify( $groups->[$x] )} = $readings->[$x]; |
91 | } |
92 | |
93 | # For all the groups with more than one member, collect the list of all |
94 | # contiguous vertices needed to connect them. |
95 | # TODO: deal with a.c. reading logic |
96 | my $sc = analyze_variant_location( $group_readings, $groups, $stemma->apsp ); |
97 | $variant_row->{'genealogical'} = keys %$sc ? 1 : undef; |
98 | foreach my $grp ( sort keys %$group_readings ) { |
99 | my $rdg = $group_readings->{$grp}; |
100 | push( @{$variant_row->{'readings'}}, { 'text' => $rdg, 'group' => $grp } ); |
101 | } |
102 | |
103 | # Now run the same analysis given the calculated distance tree(s). |
104 | # foreach my $tree ( 0 .. $#{$stemma->distance_trees} ) { |
105 | # my $dc = analyze_variant_location( $group_readings, $groups, |
106 | # $stemma->distance_apsps->[$tree] ); |
107 | # foreach my $rdg ( keys %$dc ) { |
108 | # my $var = $dc->{$rdg}; |
109 | # } |
110 | # } |
111 | |
112 | # Record that we used this variant in an analysis |
113 | push( @$variants, $variant_row ); |
114 | } |
115 | |
116 | # Go through our variant rows and add the number of empty columns we need. |
117 | foreach my $row ( @$variants ) { |
118 | my $empty = $html_columns - scalar @{$row->{'readings'}}; |
119 | $row->{'empty'} = $empty; |
120 | } |
121 | |
122 | return( $svg, $variants ); |
123 | } |
124 | |
125 | sub analyze_variant_location { |
126 | my( $group_readings, $groups, $apsp ) = @_; |
127 | my %contig; |
128 | my $conflict = {}; |
129 | foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { |
130 | my @members = @$g; |
131 | my $gst = wit_stringify( $g ); |
132 | map { $contig{$_} = $gst } @members; # The witnesses need themselves to be |
133 | # in their collection. |
134 | next unless @members > 1; |
135 | my $curr = pop @members; |
136 | foreach my $m ( @members ) { |
137 | foreach my $v ( $apsp->path_vertices( $curr, $m ) ) { |
138 | $contig{$v} = $gst unless exists $contig{$v}; |
139 | next if $contig{$v} eq $gst; |
140 | # print STDERR "Conflict at $v between group $gst and group " |
141 | # . $contig{$v} . "\n"; |
142 | # Record what is conflicting. |
143 | $conflict->{$group_readings->{$gst}} = $group_readings->{$contig{$v}}; |
144 | } |
145 | } |
146 | } |
147 | return $conflict; |
148 | } |
149 | |
150 | # Add the variant, subject to a.c. representation logic. |
151 | # This assumes that we will see the 'main' version before the a.c. version. |
152 | sub add_variant_wit { |
153 | my( $arr, $wit ) = @_; |
154 | my $acstr = $tradition->collation->ac_label; |
155 | my $skip; |
156 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
157 | my $real = $1; |
158 | $skip = grep { $_ =~ /^\Q$real\E$/ } @$arr; |
159 | } |
160 | push( @$arr, $wit ) unless $skip; |
161 | } |
162 | |
163 | # Return an answer if the variant is useful, i.e. if there are at least 2 variants |
164 | # with at least 2 witnesses each. |
165 | sub useful_variant { |
166 | my( $readings ) = @_; |
167 | my $total = keys %$readings; |
168 | foreach my $var ( keys %$readings ) { |
169 | $total-- if @{$readings->{$var}} == 1; |
170 | } |
171 | return( undef, undef ) if $total <= 1; |
172 | my( $groups, $text ); |
173 | foreach my $var ( keys %$readings ) { |
174 | push( @$groups, $readings->{$var} ); |
175 | push( @$text, $var ); |
176 | } |
177 | return( $groups, $text ); |
178 | } |
179 | |
180 | # Take an array of witness groupings and produce a string like |
181 | # ['A','B'] / ['C','D','E'] / ['F'] |
182 | |
183 | sub wit_stringify { |
184 | my $groups = shift; |
185 | my @gst; |
186 | # If we were passed an array of witnesses instead of an array of |
187 | # groupings, then "group" the witnesses first. |
188 | unless( ref( $groups->[0] ) ) { |
189 | my $mkgrp = [ $groups ]; |
190 | $groups = $mkgrp; |
191 | } |
192 | foreach my $g ( @$groups ) { |
193 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
194 | } |
195 | return join( ' / ', @gst ); |
196 | } |
197 | |
198 | 1; |