1 package Text::Tradition::Parser::CollateText;
8 Text::Tradition::Parser::CollateText
12 For an overview of the package, see the documentation for the
13 Text::Tradition module.
15 This module is meant for use with a set of text files saved from Word docs,
16 which originated with the COLLATE collation program.
24 parse( $graph, $opts );
26 Takes an initialized graph and a hashref of options, which must include:
27 - 'base' - the base text referenced by the variants
28 - 'format' - the format of the variant list
29 - 'data' - the variants, in the given format.
36 my( $tradition, $opts ) = @_;
37 # String together the base text.
38 my $lineref_hash = read_stone_base( $opts->{'base'}, $tradition->collation );
40 foreach my $sigil ( @{$opts->{'sigla'}} ) {
41 $ALL_SIGLA{$sigil} = 1;
42 $tradition->add_witness( 'sigil' => $sigil );
44 # Now merge on the apparatus entries.
45 merge_stone_apparatus( $tradition->collation, $lineref_hash, $opts->{'file'} );
48 =item B<read_stone_base>
50 my $text_list = read_base( 'reference.txt', $collation );
52 Takes a text file and a (presumed empty) collation object, adds the words
53 as simple linear readings to the collation, and returns a hash of texts
54 with line keys. This collation is now the starting point for application of
55 apparatus entries in merge_base, e.g. from a CSV file or a Classical Text
58 The hash is of the form
60 { chapter_name => { line_ref => { start => node, end => node } } }
65 my( $base_file, $collation ) = @_;
67 # This array gives the first reading for each line. We put the
68 # common starting point in line zero.
69 my $last_reading = $collation->start();
70 my $lineref_hash = {};
74 open( BASE, $base_file ) or die "Could not open file $base_file: $!";
75 binmode BASE, ':utf8';
78 # Make the readings, and connect them up for the base, but
79 # also save the first reading of each line in a hash for the
82 next if /^\s+$/; # skip blank lines
83 s/^(\d)\x{589}/$1:/; # turn Armenian full stops into colons
85 # Initialize the base hash for this section.
86 $lineref_hash->{$_} = {};
87 $curr_text = $lineref_hash->{$_};
93 # The first "word" is a line reference; keep it.
94 $lineref = shift @words;
96 # Assume we are dealing with the title.
100 # Now turn the remaining words into readings.
102 foreach my $w ( @words ) {
103 my $readingref = join( ',', $lineref, ++$wordref );
104 my $reading = $collation->add_reading( $readingref );
105 $reading->text( $w );
106 unless( exists $curr_text->{$lineref}->{'start'} ) {
107 $curr_text->{$lineref}->{'start'} = $reading;
109 # Add edge paths in the graph, for easier tracking when
110 # we start applying corrections. These paths will be
111 # removed when we're done.
112 my $path = $collation->add_path( $last_reading, $reading,
113 $collation->baselabel );
114 $last_reading = $reading;
116 $curr_text->{$lineref}->{'end'} = $last_reading;
120 # Ending point for all texts
121 $collation->add_path( $last_reading, $collation->end, $collation->baselabel );
122 return( $lineref_hash );
125 =item B<merge_stone_apparatus>
127 Read an apparatus as output (presumably) by Collate. It should be reasonably
128 regular in form, I hope. Merge the apparatus variants onto the appropriate
133 sub merge_stone_apparatus {
134 my( $c, $lineref_hash, $file ) = @_;
138 open( APP, $file ) or die "Could not read apparatus file $file";
139 binmode APP, ':utf8';
144 $current_text = $lineref_hash->{$_};
148 # Otherwise, the first word of the line is the base text line reference.
151 if( s/^(\S+)\s+// ) {
154 warn "Unrecognized line $_";
156 my $baseline = $current_text->{$lineref};
157 # The start and end readings for this line are now in $baseline->{start}
158 # and $baseline->{end}.
160 # Now look at the apparatus entries for this line. They are
162 my @apps = split( /\s+\|\s+/ );
164 foreach my $app ( @apps ) {
165 my( $lemma, $rest ) = split( /\s+\]\s+/, $app );
166 next unless $rest; # Skip lines e.g. 'Chapter 2'
167 # Find the lemma reading.
168 my( $lemma_start, $lemma_end ) =
169 _find_reading_on_line( $c, $lemma, $baseline );
170 unless( $lemma_start && $lemma_end ) {
171 print STDERR "Lemma $lemma not found; skipping readings $rest\n";
174 my( $rdg_start, $rdg_end, @lemma_chain );
175 if( $lemma_start eq '__PRIOR__' ) {
176 # Deal with 'inc' readings: lemma chain should be empty, rdg_start
177 # is a placeholder, rdg_end is $lemma_end.
178 $rdg_start = _add_reading_placeholders( $c, $lemma_end );
179 $rdg_end = $lemma_end;
181 @lemma_chain = $c->reading_sequence( $lemma_start, $lemma_end );
182 # Splice in "start" and "end" placeholders on either
184 ( $rdg_start, $rdg_end ) =
185 _add_reading_placeholders( $c, $lemma_start, $lemma_end );
187 # For each reading, attach it to the lemma.
188 my @indiv = split( / /, $rest );
191 map { $seen_sigla{$_} = 0 } keys %ALL_SIGLA;
192 foreach my $rdg ( @indiv ) {
194 my( $words, $sigla, $recurse ) = parse_app_entry( $rdg );
196 # Do something really very dodgy indeed.
197 if( exists $sigla->{'__REL__'} && !$has_rel ) {
198 # Handling this has to be deferred until the end, so push it
199 # back onto @indiv and note that we've done so.
201 push( @indiv, $rdg );
206 foreach my $rdg_word ( @$words ) {
207 next if $rdg_word =~ /^__/;
208 my $reading_id = ref( $lemma_start )
209 ? $lemma_start->name : $lemma_start;
210 $reading_id .= '_' . $lemma_end->name . '/' . $rdg_ctr++;
211 my $reading = $c->add_reading( $reading_id );
212 $reading->text( $rdg_word );
213 push( @readings, $reading );
216 # Deal with any specials.
218 if( @$words && $words->[0] eq '__LEMMA__'
219 && $lemma_end ne $rdg_end ) {
220 # It's an addition (unless lemma_end eq rdg_end, in which case
221 # it's an 'inc'.) Start from lemma rather than from placeholder.
222 $lemma_sequence = [ $lemma_end, $rdg_end ];
223 } elsif ( @$words && $words->[0] eq '__TRANSPOSE__' ) {
224 # Hope it is only two or three words in the lemma.
225 # TODO figure out how we really want to handle this
226 @readings = reverse @lemma_chain;
228 $lemma_sequence = [ $rdg_start, @lemma_chain, $rdg_end ]
229 unless $lemma_sequence;
231 # Note which sigla we are actually dealing with.
232 if( $sigla->{'__REL__'} ) {
233 delete $sigla->{'__REL__'};
234 map { $sigla->{$_} = 1 }
235 grep { $seen_sigla{$_} == 0 } keys %seen_sigla;
237 map { $seen_sigla{$_} = 1 } keys %$sigla;
240 # Now hook up the paths.
241 unshift( @readings, $lemma_sequence->[0] );
242 push( @readings, $lemma_sequence->[-1] );
243 foreach my $i ( 1 .. $#readings ) {
244 if( $recurse->{$i} ) {
245 my( $rwords, $rsig ) = parse_app_entry( $recurse->{$i} );
246 # Get the local "lemma" sequence
247 my $llseq = [ $readings[$i], $readings[$i+1] ];
248 if( $rwords->[0] ne '__LEMMA__' ) {
249 unshift( @$llseq, $readings[$i-1] );
250 } # Otherwise treat it as an addition to the last word
251 # Create the reading nodes in $rwords
252 # TODO Hope we don't meet ~ in a recursion
254 foreach my $i ( 0 .. $#$rwords ) {
255 next if $i == 0 && $rwords->[$i] =~ /^__/;
256 my $reading_id = $llseq->[0]->name . '_' .
257 $llseq->[-1]->name . '/' . $i;
258 $reading_id =~ s/ATTACH//g;
259 my $reading = $c->add_reading( $reading_id );
260 $reading->text( $rwords->[$i] );
261 push( @$local_rdg, $reading );
263 unshift( @$local_rdg, $llseq->[0] );
264 push( @$local_rdg, $llseq->[-1] );
265 # Add the path(s) necessary
266 _add_sigil_path( $c, $rsig, $llseq, $local_rdg );
269 _add_sigil_path( $c, $sigla, $lemma_sequence, \@readings );
270 } # end processing of $app
271 } # end foreach my $app in line
275 # Now reconcile all the paths in the collation, and delete our
276 # temporary anchor nodes.
277 expand_all_paths( $c );
279 # Finally, calculate the ranks we've got.
280 # $c->calculate_ranks;
283 sub _find_reading_on_line {
284 my( $c, $lemma, $baseline, $prior ) = @_;
286 if( $lemma eq 'totum' ) {
287 # We want the whole line.
288 return( $baseline->{'start'}, $baseline->{'end'} );
289 } elsif( $lemma eq 'inc' ) {
290 # We want to shove things in before the line begins.
291 return( '__PRIOR__', $baseline->{'start'} );
294 my $lemma_start = $baseline->{'start'};
296 my $too_far = $c->next_reading( $baseline->{'end'} );
297 my @lemma_words = split( /\s+/, $lemma );
300 my $scrutinize = ''; # DEBUG variable
301 my ( $lw, $seq ) = _get_seq( $lemma_words[0] );
302 while( $lemma_start ne $too_far ) {
304 if( $seen{ $lemma_start->name() } ) {
305 warn "Detected loop at " . $lemma_start->name . " for lemma $lemma";
308 $seen{ $lemma_start->name() } = 1;
310 # Try to match the lemma.
311 # TODO move next/prior reading methods into the reading classes,
312 # to make this more self-contained and not need to pass $c.
314 print STDERR "Matching ".$lemma_start->text." against $lw...\n"
316 if( _norm( $lemma_start->text ) eq _norm( $lw ) ) {
317 # Skip it if we need a match that is not the first.
319 # Now we have to compare the rest of the words here.
320 if( scalar( @lemma_words ) > 1 ) {
321 my $next_reading = next_real_reading( $c, $lemma_start );
323 foreach my $w ( @lemma_words[1..$#lemma_words] ) {
329 # This should be the word after a --- now, and the
331 my( $wst, $wend ) = _find_reading_on_line( $c, $w,
332 $baseline, $lemma_start );
333 warn "Something unexpected" unless $wst eq $wend;
338 # If we got this far, there is no wildcard. We must
339 # match each word in sequence.
340 my( $nlw, $nseq ) = _get_seq( $w );
341 printf STDERR "Now matching %s against %s\n",
342 $next_reading->text, $nlw
344 if( _norm( $nlw ) eq _norm( $next_reading->text ) ) {
345 $lemma_end = $next_reading;
346 $next_reading = $c->next_reading( $lemma_end );
352 } else { # single-word match, easy.
353 $lemma_end = $lemma_start;
355 } else { # we need the Nth match and aren't there yet
358 $unmatch = 1 if $prior && !$seen{$prior->name};
360 last unless ( $unmatch || !defined( $lemma_end ) );
362 $lemma_start = $c->next_reading( $lemma_start );
365 unless( $lemma_end ) {
366 warn "No match found for @lemma_words";
369 return( $lemma_start, $lemma_end );
372 sub _add_reading_placeholders {
373 my( $collation, $lemma_start, $lemma_end ) = @_;
374 # We will splice in a 'begin' and 'end' marker on either side of the
375 # lemma, as sort of a double-endpoint attachment in the graph.
376 # Note that all of this assumes we have a linear base graph at this
377 # point, and no diverging readings on the lemmas.
379 my $start_node = $collation->prior_reading( $lemma_start );
380 unless( $start_node->name =~ /ATTACH/ ) {
381 my $sn_id = '#ATTACH_' . $lemma_start->name . '_START#';
382 my $prior = $start_node;
383 $start_node = $collation->add_reading( $sn_id );
384 $start_node->is_meta( 1 );
385 $collation->graph->del_edge( $collation->graph->edge( $prior, $lemma_start ) );
386 $collation->add_path( $prior, $start_node, $collation->baselabel );
387 $collation->add_path( $start_node, $lemma_start, $collation->baselabel );
389 return $start_node unless $lemma_end;
391 # Now the converse for the end.
392 my $end_node = $collation->next_reading( $lemma_end );
393 unless( $end_node->name =~ /ATTACH/ ) {
394 my $en_id = '#ATTACH_' . $lemma_end->name . '_END#';
395 my $next = $end_node;
396 $end_node = $collation->add_reading( $en_id );
397 $end_node->is_meta( 1 );
398 $collation->graph->del_edge( $collation->graph->edge( $lemma_end, $next ) );
399 $collation->add_path( $lemma_end, $end_node, $collation->baselabel );
400 $collation->add_path( $end_node, $next, $collation->baselabel );
403 return( $start_node, $end_node );
406 # Function to parse an apparatus reading string, with reference to no other
407 # data. Need to do this separately as readings can include readings (ugh).
408 # Try to give whatever information we might need, including recursive app
409 # entries that might need to be parsed.
411 sub parse_app_entry {
415 next unless $rdg; # just in case
416 my @words = split( /\s+/, $rdg );
417 # Zero or more sigils e.g. +, followed by Armenian,
418 # followed by (possibly modified) sigla, followed by
419 # optional : with note.
422 my $is_transposition;
426 my $reading_sigla = {};
428 my $sig_regex = join( '|', sort { length $b <=> length $a } keys %ALL_SIGLA );
430 my $bit = shift @words;
433 } elsif( $bit eq 'om' ) {
435 } elsif( $bit eq '~' ) {
436 $is_transposition = 1;
437 } elsif( $bit =~ /\p{Armenian}/ ) {
438 warn "Found text in omission?!" if $is_omission;
439 push( @reading, $bit );
440 } elsif( $bit eq ':' ) {
443 } elsif( $bit =~ /^\(/ ) {
444 # It's a recursive reading within a reading. Lemmatize what we
445 # have so far and grab the extra.
447 until( $new[-1] =~ /\)$/ ) {
448 push( @new, shift @words );
450 my $recursed_reading = join( ' ', @new );
451 $recursed_reading =~ s/^\((.*)\)/$1/;
452 # This recursive entry refers to the last reading word(s) we
453 # saw. Push its index+1. We will have to come back to parse
454 # it when we are dealing with the main reading.
455 # TODO handle () as first element
456 # TODO handle - as suffix to add, i.e. make new word
457 $recursed->{@reading} = $recursed_reading;
458 } elsif( $bit =~ /^($sig_regex)(.*)$/ ) {
459 # It must be a sigil.
460 my( $sigil, $mod ) = ( $1, $2 );
461 if( $mod eq "\x{80}" ) {
462 $reading_sigla->{$sigil} = '_PC_';
463 $ALL_SIGLA{$sigil} = 2; # a pre- and post-corr version exists
464 } elsif( $mod eq '*' ) {
465 $reading_sigla->{$sigil} = '_AC_';
466 $ALL_SIGLA{$sigil} = 2; # a pre- and post-corr version exists
468 $reading_sigla->{$sigil} = 1 unless $mod; # skip secondhand corrections
470 } elsif( $bit eq 'rel' ) {
471 # The anti-reading. All sigla except those cited.
472 $reading_sigla->{'__REL__'} = 1;
473 } elsif( $bit eq 'ed' ) {
474 # An emendation. TODO make sure all other sigla appear in readings?
477 } elsif( $bit =~ /transpos/ ) {
478 # There are some transpositions not coded rigorously; skip them.
479 warn "Found hard transposition in $rdg; fix manually";
482 warn "Not sure what to do with bit $bit in $rdg";
488 return( [], {}, {} ) if $skip;
489 # Transmogrify the reading if necessary.
490 unshift( @reading, '__LEMMA__' ) if $is_add;
491 unshift( @reading, '__TRANSPOSE__' ) if $is_transposition;
492 @reading = () if $is_omission;
493 unless( @reading || $is_omission ) {
494 # It was just sigla on a line, meaning the base changed. Thus
495 # the reading is the lemma.
496 unshift( @reading, '__LEMMA__' );
499 return( \@reading, $reading_sigla, $recursed );
502 # Add a path for the specified sigla to connect the reading sequence.
503 # Add an a.c. path to the base sequence if we have an explicitly p.c.
505 # Also handle the paths for sigla we have already added in recursive
506 # apparatus readings (i.e. don't add a path if one already exists.)
508 sub _add_sigil_path {
509 my( $c, $sigla, $base_sequence, $reading_sequence ) = @_;
511 foreach my $sig ( keys %$sigla ) {
512 my $use_sig = $sigla->{$sig} eq '_AC_' ? $sig.$c->ac_label : $sig;
513 foreach my $i ( 0 .. $#{$reading_sequence}-1 ) {
514 if( $skip{$use_sig} ) {
515 next if !_has_prior_reading( $reading_sequence->[$i], $use_sig );
518 if( _has_next_reading( $reading_sequence->[$i], $use_sig ) ) {
522 $c->add_path( $reading_sequence->[$i], $reading_sequence->[$i+1], $use_sig );
524 if( $sigla->{$sig} eq '_PC_') {
525 $use_sig = $sig.$c->ac_label;
526 foreach my $i ( 0 .. $#{$base_sequence}-1 ) {
527 if( $skip{$use_sig} ) {
528 next if !_has_prior_reading( $reading_sequence->[$i], $use_sig );
531 if( _has_next_reading( $reading_sequence->[$i], $use_sig ) ) {
535 $c->add_path( $base_sequence->[$i], $base_sequence->[$i+1], $use_sig );
541 # Walk the collation for all witness paths, delete the ATTACH anchor nodes,
542 # and then nuke and re-draw all edges (thus getting rid of the base).
544 sub expand_all_paths {
547 # Walk the collation and fish out the paths for each witness
548 foreach my $sig ( keys %ALL_SIGLA ) {
549 my $wit = $c->tradition->witness( $sig );
550 my @path = grep { $_->name !~ /ATTACH/ }
551 $c->reading_sequence( $c->start, $c->end, $sig );
552 $wit->path( \@path );
553 if( $ALL_SIGLA{$sig} > 1 ) {
554 my @ac_path = grep { $_->name !~ /ATTACH/ }
555 $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label, $sig );
556 $wit->uncorrected_path( \@ac_path );
561 foreach my $anchor ( grep { $_->name =~ /ATTACH/ } $c->readings ) {
562 $c->del_reading( $anchor );
565 map { $c->del_path( $_ ) } $c->paths;
567 # Make the path edges
568 $c->make_witness_paths();
575 if( $str =~ /^(.*)(\d)\x{b0}$/ ) {
576 ( $lw, $seq) = ( $1, $2 );
581 # Normalize to lowercase, no punct
584 $str =~ s/[^[:alnum:]]//g;
588 sub _has_next_reading {
589 my( $rdg, $sigil ) = @_;
590 return grep { $_->label eq $sigil } $rdg->outgoing();
592 sub _has_prior_reading {
593 my( $rdg, $sigil ) = @_;
594 return grep { $_->label eq $sigil } $rdg->incoming();
596 sub next_real_reading {
598 while( my $r = $c->next_reading( $rdg ) ) {
599 return $r unless $r->is_meta;
600 return $r if $r eq $c->end;
607 if( ref( $_[0] ) eq 'ARRAY' ) {
610 my $str = join( ' ', map { $_->text } @l );