Improved DRuKoLa meta data handling
Change-Id: I50a79baa595429abeb8b56b11b1942f6e2ed8374
diff --git a/lib/KorAP/XML/Krill.pm b/lib/KorAP/XML/Krill.pm
index 29b1291..398c045 100644
--- a/lib/KorAP/XML/Krill.pm
+++ b/lib/KorAP/XML/Krill.pm
@@ -69,12 +69,14 @@
else {
# Load file
$file = b(Mojo::File->new($data_xml)->slurp);
+
try {
local $SIG{__WARN__} = sub {
$error = 1;
};
$rt = xml2hash($file, text => '#text', attr => '-')->{raw_text};
+
} catch {
$self->log->warn($unable);
$error = 1;
diff --git a/lib/KorAP/XML/Meta/I5.pm b/lib/KorAP/XML/Meta/I5.pm
index 0dd5330..34d99e5 100644
--- a/lib/KorAP/XML/Meta/I5.pm
+++ b/lib/KorAP/XML/Meta/I5.pm
@@ -63,7 +63,8 @@
my $translator;
if ($editor && $editor->attr('role') && $editor->attr('role') eq 'translator') {
# Translator is only supported on the text level currently
- $self->{translator} = _squish $editor->all_text;
+ $translator = _squish $editor->all_text;
+ $self->{translator} = $translator if $translator;
$editor = undef;
}
else {
@@ -112,7 +113,7 @@
# Corpus title not yet given
unless ($self->{corpus_title}) {
if ($title = $dom->at('fileDesc > titleStmt > c\.title')) {
- $title = $title->all_text;
+ $title = _squish($title->all_text);
if ($title) {
$self->{corpus_title} = _remove_prefix($title, $self->corpus_sigle);
@@ -125,7 +126,7 @@
elsif ($type eq 'doc') {
unless ($self->{doc_title}) {
if ($title = $dom->at('fileDesc > titleStmt > d\.title')) {
- $title = $title->all_text;
+ $title = _squish($title->all_text);
if ($title) {
$self->{doc_title} = _remove_prefix($title, $self->doc_sigle);
@@ -138,7 +139,7 @@
elsif ($type eq 'text') {
unless ($self->{title}) {
if ($title = $dom->at('fileDesc > titleStmt > t\.title')) {
- $title = $title->all_text;
+ $title = _squish($title->all_text);
if ($title) {
$self->{title} = _remove_prefix($title, $self->text_sigle);
};
@@ -242,14 +243,14 @@
$temp->find("catRef")->each(
sub {
- my ($ign, @ttopic) = split('\.', $_->attr('target'));
+ my ($ign, @ttopic) = grep { $_ } map { _squish($_) } split('\.', $_->attr('target'));
push(@topic, @ttopic);
}
);
$self->{text_class} = [@topic] if @topic > 0;
my $kws = $self->{keywords};
- my @keywords = $temp->find("h\.keywords > keyTerm")->each;
+ my @keywords = $temp->find("h\.keywords > keyTerm")->map(sub {_squish($_) })->grep(sub { $_ })->each;
push(@$kws, @keywords) if @keywords > 0;
};
diff --git a/t/corpus/BBU2/Blog/83701_a_82376/data.xml b/t/corpus/BBU2/Blog/83701_a_82376/data.xml
new file mode 100644
index 0000000..630b2dc
--- /dev/null
+++ b/t/corpus/BBU2/Blog/83701_a_82376/data.xml
@@ -0,0 +1,7 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<?xml-model href="text.rng" type="application/xml" schematypens="http://relaxng.org/ns/structure/1.0"?>
+
+<raw_text docid="BBU_BLOG.83701_a_82376" xmlns="http://ids-mannheim.de/ns/KorAP">
+ <metadata file="metadata.xml" />
+ <text>.</text>
+</raw_text>
diff --git a/t/corpus/BBU2/Blog/83701_a_82376/header.xml b/t/corpus/BBU2/Blog/83701_a_82376/header.xml
new file mode 100644
index 0000000..94f5be2
--- /dev/null
+++ b/t/corpus/BBU2/Blog/83701_a_82376/header.xml
@@ -0,0 +1,38 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<?xml-model href="header.rng" type="application/xml" schematypens="http://relaxng.org/ns/structure/1.0"?>
+<!DOCTYPE idsCorpus PUBLIC "-//IDS//DTD IDS-XCES 1.0//EN" "http://corpora.ids-mannheim.de/idsxces1/DTD/ids.xcesdoc.dtd">
+<idsHeader type="text" pattern="text" status="new" version="1.1" TEIform="teiHeader">
+ <fileDesc>
+ <titleStmt>
+ <textSigle>Corola-Bucurenci/Blog.83701_a_82376</textSigle>
+ <t.title>Confesiunile unui misogin</t.title>
+ </titleStmt>
+ <publicationStmt>
+ <distributor/>
+ <pubAddress/>
+ <availability region="world">QAO-NC</availability>
+ <pubDate/>
+ </publicationStmt>
+ <sourceDesc>
+ <biblStruct>
+ <monogr>
+ <h.author>Dragoș Bucurenci</h.author>
+ <editor role="translator">-</editor>
+ <imprint>
+ <pubDate type="year">2013</pubDate>
+ <pubDate type="month">11</pubDate>
+ <pubDate type="day">18</pubDate>
+ <pubPlace>URL:http://www.bucurenci.ro</pubPlace>
+ </imprint>
+ </monogr>
+ </biblStruct>
+ </sourceDesc>
+ </fileDesc>
+ <profileDesc>
+ <textDesc>
+ <textType>-.-</textType>
+ <textClass>
+ <catRef target="-.-" scheme="topic"/>
+ </textClass>
+ </profileDesc>
+ </idsHeader>
\ No newline at end of file
diff --git a/t/corpus/BBU2/Blog/header.xml b/t/corpus/BBU2/Blog/header.xml
new file mode 100644
index 0000000..a6afb53
--- /dev/null
+++ b/t/corpus/BBU2/Blog/header.xml
@@ -0,0 +1,30 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<?xml-model href="header.rng" type="application/xml" schematypens="http://relaxng.org/ns/structure/1.0"?>
+<!DOCTYPE idsCorpus PUBLIC "-//IDS//DTD IDS-XCES 1.0//EN" "http://corpora.ids-mannheim.de/idsxces1/DTD/ids.xcesdoc.dtd">
+<idsHeader type="document" pattern="text" status="new" version="1.1" TEIform="teiHeader">
+ <fileDesc>
+ <titleStmt>
+ <dokumentSigle>Corola-Bucurenci/Blog</dokumentSigle>
+ <d.title>-</d.title>
+ </titleStmt>
+ <publicationStmt>
+ <distributor/>
+ <pubAddress/>
+ <availability region="world">[...]</availability>
+ <pubDate/>
+ </publicationStmt>
+ <sourceDesc>
+ <biblStruct>
+ <monogr>
+ <h.title type="main"/>
+ <edition>
+ <further/>
+ <kind/>
+ <appearance/>
+ </edition>
+ <imprint/>
+ </monogr>
+ </biblStruct>
+ </sourceDesc>
+ </fileDesc>
+ </idsHeader>
\ No newline at end of file
diff --git a/t/corpus/BBU2/header.xml b/t/corpus/BBU2/header.xml
new file mode 100644
index 0000000..7901a2e
--- /dev/null
+++ b/t/corpus/BBU2/header.xml
@@ -0,0 +1,22 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<?xml-model href="header.rng" type="application/xml" schematypens="http://relaxng.org/ns/structure/1.0"?>
+<!DOCTYPE idsCorpus PUBLIC "-//IDS//DTD IDS-XCES 1.0//EN" "http://corpora.ids-mannheim.de/idsxces1/DTD/ids.xcesdoc.dtd">
+<idsHeader type="corpus" status="new" version="1.1" TEIform="teiHeader">
+ <fileDesc>
+ <titleStmt>
+ <korpusSigle>Corola-Bucurenci</korpusSigle>
+ <c.title>Corola-Bucurenci</c.title>
+ </titleStmt>
+ </fileDesc>
+ <profileDesc>
+ <langUsage>
+ <language id="ro" usage="100">Romanian</language>
+ </langUsage>
+ <!--
+ <textDesc>
+ This element is suppressed in p5!
+ <channel mode="w">written</channel>
+ </textDesc>
+ -->
+ </profileDesc>
+ </idsHeader>
\ No newline at end of file
diff --git a/t/real/drukola.t b/t/real/drukola.t
index 4656f18..1eb4587 100644
--- a/t/real/drukola.t
+++ b/t/real/drukola.t
@@ -106,5 +106,23 @@
like($token, qr!drukola/m:gender:feminine!, 'data');
like($token, qr!drukola/p:NOUN!, 'data');
+
+# New
+# BBU2/BLOG/83709_a_82384
+$path = catdir(dirname(__FILE__), '../corpus/BBU2/Blog/83701_a_82376');
+
+
+
+ok($doc = KorAP::XML::Krill->new( path => $path . '/' ), 'Load Korap::Document');
+ok($doc->parse, 'Parse document');
+
+$meta = $doc->meta;
+
+ok(!exists $meta->{doc_title}, 'No doc title');
+ok(!exists $meta->{translator}, 'No translator');
+
+ok(!exists $meta->{text_class}, 'No translator');
+
+
done_testing;
__END__