use strict;
use warnings;
+use Encode qw/ decode /;
use Text::Tradition::Parser::Util qw/ collate_variants /;
use XML::LibXML;
use XML::LibXML::XPathContext;
my %sigil_for; # Save the XML IDs for witnesses.
my %apps; # Save the apparatus XML for a given ID.
my %has_ac; # Keep track of witnesses that have corrections.
-my %group_sigla = ( ## HACK HACK HACK
- '#M38' => [ qw( #M23 #M24 #M25 #M27 #M30 #M26 #M31 #M32 #M33 ) ], # l -> L*
- );
sub parse {
my( $tradition, $opts ) = @_;
my $c = $tradition->collation; # Some shorthand
# First, parse the XML.
- my $parser = XML::LibXML->new();
- my $doc;
- if( exists $opts->{'string'} ) {
- $doc = $parser->parse_string( $opts->{'string'} );
- } elsif ( exists $opts->{'file'} ) {
- $doc = $parser->parse_file( $opts->{'file'} );
- } else {
- warn "Could not find string or file option to parse";
- return;
- }
- my $tei = $doc->documentElement();
- my $xpc = XML::LibXML::XPathContext->new( $tei );
+ my( $tei, $xpc ) = _remove_formatting( $opts );
+ return unless $tei; # we have already warned.
# CTE uses a DTD rather than any xmlns-based parsing. Thus we
# need no namespace handling.
-
# Get the witnesses and create the witness objects.
foreach my $wit_el ( $xpc->findnodes( '//sourceDesc/listWit/witness' ) ) {
# The witness xml:id is used internally, and is *not* the sigil name.
my $id= $wit_el->getAttribute( 'xml:id' );
- my @sig_parts = $xpc->findnodes( './abbr/descendant::text()', $wit_el );
+ my @sig_parts = $xpc->findnodes( 'descendant::text()', $wit_el );
my $sig = _stringify_sigil( @sig_parts );
- unless( exists $group_sigla{'#'.$id} ) { ## More HACKY
- $tradition->add_witness( sigil => $sig, source => $wit_el->toString() );
- $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax
- }
+ print STDERR "Adding witness $sig\n";
+ $tradition->add_witness( sigil => $sig, sourcetype => 'collation' );
+ $sigil_for{'#'.$id} = $sig; # Make life easy by keying on the ID ref syntax
}
-
+
# Now go through the text and find the base tokens, apparatus tags, and
# anchors. Make a giant array of all of these things in sequence.
# TODO consider combining this with creation of graph below
push( @base_text, _get_base( $xn ) );
}
}
-
# We now have to work through this array applying the alternate
# apparatus readings to the base text. Essentially we will put
# everything on the graph, from which we will delete the apps and
$r = $c->add_reading( { id => 'n'.$counter++,
text => $item->{'content'} } );
} elsif ( $item->{'type'} eq 'anchor' ) {
- $r = $c->add_reading( { id => '#ANCHOR_' . $item->{'content'} . '#',
+ $r = $c->add_reading( { id => '__ANCHOR_' . $item->{'content'} . '__',
is_ph => 1 } );
} elsif ( $item->{'type'} eq 'app' ) {
- my $tag = '#APP_' . $counter++ . '#';
+ my $tag = '__APP_' . $counter++ . '__';
$r = $c->add_reading( { id => $tag, is_ph => 1 } );
$apps{$tag} = $item->{'content'};
}
# Finally, add explicit witness paths, remove the base paths, and remove
# the app/anchor tags.
expand_all_paths( $c );
+
+ # Save the text for each witness so that we can ensure consistency
+ # later on
+ $tradition->collation->text_from_paths();
+ $tradition->collation->calculate_ranks();
+ $tradition->collation->flatten_ranks();
}
sub _stringify_sigil {
return join( '', @parts );
}
-## Recursive little helper function to help us navigate through nested
-## XML, picking out the words, the apparatus, and the anchors.
+# Get rid of all the formatting elements that get in the way of tokenization.
+sub _remove_formatting {
+ my( $opts ) = @_;
+
+ # First, parse the original XML
+ my $parser = XML::LibXML->new();
+ my $doc;
+ if( exists $opts->{'string'} ) {
+ $doc = $parser->parse_string( $opts->{'string'} );
+ } elsif ( exists $opts->{'file'} ) {
+ $doc = $parser->parse_file( $opts->{'file'} );
+ } else {
+ warn "Could not find string or file option to parse";
+ return;
+ }
+
+ # Second, remove the formatting
+ my $xpc = XML::LibXML::XPathContext->new( $doc->documentElement );
+ my @useless = $xpc->findnodes( '//hi' );
+ foreach my $n ( @useless ) {
+ my $parent = $n->parentNode();
+ my @children = $n->childNodes();
+ my $first = shift @children;
+ $parent->replaceChild( $first, $n );
+ foreach my $c ( @children ) {
+ $parent->insertAfter( $c, $first );
+ $first = $c;
+ }
+ }
+
+ # Third, write out and reparse to merge the text nodes.
+ my $result = decode( $doc->encoding, $doc->toString() );
+ my $tei = $parser->parse_string( $result )->documentElement;
+ $xpc = XML::LibXML::XPathContext->new( $tei );
+ return( $tei, $xpc );
+}
+
+## Helper function to help us navigate through nested XML, picking out
+## the words, the apparatus, and the anchors.
sub _get_base {
my( $xn ) = @_;
if( $xn->nodeType == XML_TEXT_NODE ) {
# Base text, just split the words on whitespace and add them
# to our sequence.
- # TODO consider that XML markup might appear mid-token.
my $str = $xn->data;
$str =~ s/^\s+//;
- foreach my $w ( split( /\s+/, $str ) ) {
- # HACK to cope with mismatched doublequotes
- $w =~ s/\"//g;
- push( @readings, { 'type' => 'token', 'content' => $w } );
- }
- } elsif( $xn->nodeName eq 'hi' ) {
- # Recurse as if the hi weren't there.
- foreach( $xn->childNodes ) {
- push( @readings, _get_base( $_ ) );
- }
+ my @tokens = split( /\s+/, $str );
+ push( @readings, map { { 'type' => 'token', 'content' => $_ } } @tokens );
} elsif( $xn->nodeName eq 'app' ) {
# Apparatus, just save the entire XML node.
push( @readings, { 'type' => 'app', 'content' => $xn } );
return @readings;
}
+sub _append_tokens {
+ my( $list, @tokens ) = @_;
+ if( @$list && $list->[-1]->{'content'} =~ /\#JOIN\#$/ ) {
+ # The list evidently ended mid-word; join the next token onto it.
+ my $t = shift @tokens;
+ if( ref $t && $t->{'type'} eq 'token' ) {
+ # Join the word
+ $t = $t->{'content'};
+ } elsif( ref $t ) {
+ # An app or anchor intervened; end the word.
+ unshift( @tokens, $t );
+ $t = '';
+ }
+ $list->[-1]->{'content'} =~ s/\#JOIN\#$/$t/;
+ }
+ foreach my $t ( @tokens ) {
+ unless( ref( $t ) ) {
+ $t = { 'type' => 'token', 'content' => $t };
+ }
+ push( @$list, $t );
+ }
+}
+
sub _add_readings {
my( $c, $app_id ) = @_;
my $xn = $apps{$app_id};
# Get the lemma, which is all the readings between app and anchor,
# excluding other apps or anchors.
my @lemma = _return_lemma( $c, $app_id, $anchor );
- my $lemma_str = join( ' ', grep { $_ !~ /^\#/ } map { $_->text } @lemma );
+ my $lemma_str = join( ' ', grep { $_ !~ /^__/ } map { $_->text } @lemma );
# For each reading, send its text to 'interpret' along with the lemma,
# and then save the list of witnesses that these tokens belong to.
my %wit_rdgs; # Maps from witnesses to the variant text
my $ctr = 0;
my $tag = $app_id;
- $tag =~ s/^\#APP_(.*)\#$/$1/;
+ $tag =~ s/^\__APP_(.*)\__$/$1/;
+
foreach my $rdg ( $xn->getChildrenByTagName( 'rdg' ) ) {
my @text;
foreach ( $rdg->childNodes ) {
( $interpreted, $flag ) = interpret(
join( ' ', map { $_->{'content'} } @text ), $lemma_str );
}
+ next if( $interpreted eq $lemma_str ) && !$flag; # Reading is lemma.
+
my @rdg_nodes;
- foreach my $w ( split( /\s+/, $interpreted ) ) {
- my $r = $c->add_reading( { id => $tag . "/" . $ctr++,
- text => $w } );
- push( @rdg_nodes, $r );
+ if( $interpreted eq '#LACUNA#' ) {
+ push( @rdg_nodes, $c->add_reading( { id => 'r'.$tag.".".$ctr++,
+ is_lacuna => 1 } ) );
+ } else {
+ foreach my $w ( split( /\s+/, $interpreted ) ) {
+ my $r = $c->add_reading( { id => 'r'.$tag.".".$ctr++,
+ text => $w } );
+ push( @rdg_nodes, $r );
+ }
}
-
# For each listed wit, save the reading.
foreach my $wit ( split( /\s+/, $rdg->getAttribute( 'wit' ) ) ) {
$wit .= $flag if $flag;
# Now collate the variant readings, since it is not done for us.
collate_variants( $c, \@lemma, values %wit_rdgs );
-
- # HACKY HACKY Expand "group" sigla.
- # Does not work for nested groups; also does not work with a modifier
- # on the group sigil.
- foreach my $wit_id ( keys %group_sigla ) {
- if ( exists $wit_rdgs{$wit_id} ) {
- my $rdg = $wit_rdgs{$wit_id};
- foreach my $w ( @{$group_sigla{$wit_id}} ) {
- if( exists $wit_rdgs{$w} ) {
- $DB::single = 1;
- warn "Had reading for individual member $w of group $wit_id at $xn";
- }
- $wit_rdgs{$w} = $rdg;
- }
- delete $wit_rdgs{$wit_id};
- }
- }
-
+
# Now add the witness paths for each reading.
foreach my $wit_id ( keys %wit_rdgs ) {
my $witstr = get_sigil( $wit_id, $c );
sub _anchor_name {
my $xmlid = shift;
$xmlid =~ s/^\#//;
- return sprintf( "#ANCHOR_%s#", $xmlid );
+ return sprintf( "__ANCHOR_%s__", $xmlid );
}
sub _return_lemma {
my( $c, $app, $anchor ) = @_;
- my @nodes = grep { $_->id !~ /^\#A(PP|NCHOR)/ }
+ my @nodes = grep { $_->id !~ /^__A(PP|NCHOR)/ }
$c->reading_sequence( $c->reading( $app ), $c->reading( $anchor ),
$c->baselabel );
return @nodes;
return $reading if $reading eq $lemma;
my $oldreading = $reading;
# $lemma =~ s/\s+[[:punct:]]+$//;
- # $reading =~ s/\s*\(?sic([\s\w.]+)?\)?$//;
my $flag; # In case of p.c. indications
my @words = split( /\s+/, $lemma );
if( $reading =~ /^(.*) praem.$/ ) {
$reading = "$1 $lemma";
} elsif( $reading =~ /^(.*) add.$/ ) {
$reading = "$lemma $1";
- } elsif( $reading =~ /add. alia manu/ ) {
+ } elsif( $reading =~ /add. alia manu/
+ || $reading =~ /inscriptionem compegi e/ # TODO huh?
+ || $reading eq 'inc.' # TODO huh?
+ ) {
# Ignore it.
$reading = $lemma;
- } elsif( $reading eq 'om.'
- || $reading =~ /locus [uv]acuus/
- || $reading =~ /inscriptionem compegi e/ # TODO huh?
- || $reading eq 'def.' # TODO huh?
+ } elsif( $reading =~ /locus [uv]acuus/
+ || $reading eq 'def.'
+ || $reading eq 'illeg.'
+ || $reading eq 'onleesbar'
) {
+ $reading = '#LACUNA#';
+ } elsif( $reading eq 'om.' ) {
$reading = '';
- } elsif( $reading =~ /^in[uv]\.$/ ) {
+ } elsif( $reading =~ /^in[uv]\.$/
+ || $reading eq 'transp.' ) {
# Hope it is two words.
print STDERR "WARNING: want to invert a lemma that is not two words\n"
unless scalar( @words ) == 2;
# There was nothing before a correction.
$reading = '';
$flag = '_ac';
+ } elsif( $reading =~ /^(.*?)\s*\(?sic([\s\w.]+)?\)?$/ ) {
+ # Discard any 'sic' notation; indeed, indeed.
+ $reading = $1;
} elsif( $reading =~ /^(.*) \.\.\. (.*)$/ ) {
# The first and last N words captured should replace the first and
# last N words of the lemma.
$wit->path( \@path );
if( $has_ac{$sig} ) {
my @ac_path = grep { !$_->is_ph }
- $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label, $sig );
+ $c->reading_sequence( $c->start, $c->end, $sig.$c->ac_label );
$wit->uncorrected_path( \@ac_path );
}
}
projects / scpubgit/stemmatology.git / blobdiff