diff --git a/libcorpus2/io/cclreader.cpp b/libcorpus2/io/cclreader.cpp
index 37d6cd2e999c306e6e9f1a9613804db5b02cb87c..117526b1194e97b95e836738bbfecef4f134b2a3 100644
--- a/libcorpus2/io/cclreader.cpp
+++ b/libcorpus2/io/cclreader.cpp
@@ -36,7 +36,8 @@ class CclReaderImpl : public XmlReader
public:
CclReaderImpl(const TokenReader& base_reader,
std::deque< boost::shared_ptr<Chunk> >& obuf,
- bool disamb_only, bool disamb_sh, bool autogen_sent_id);
+ bool disamb_only, bool disamb_sh, bool autogen_sent_id,
+ bool autogen_chunk_id);
~CclReaderImpl();
@@ -48,6 +49,14 @@ public:
return autogen_sent_id_;
}
+ void set_autogen_chunk_id(bool autogen_chunk_id) {
+ autogen_chunk_id_ = autogen_chunk_id;
+ }
+
+ bool get_autogen_chunk_id() const {
+ return autogen_chunk_id_;
+ }
+
protected:
bool process_start_element(const Glib::ustring & name,
const AttributeList& attributes);
@@ -86,21 +95,30 @@ private:
bool autogen_sent_id_;
unsigned int sent_number_; /// Sentence number, automatically generated
static const std::string SENT_ID_PREFFIX;
+ // and same for chunk...
+ bool autogen_chunk_id_;
+ unsigned int chunk_number_;
+ static const std::string CHUNK_ID_PREFFIX;
};
const std::string CclReaderImpl::SENT_ID_PREFFIX = "sentence";
+const std::string CclReaderImpl::CHUNK_ID_PREFFIX = "chunk";
CclReader::CclReader(const Tagset& tagset, std::istream& is,
- bool disamb_only, bool disamb_sh, bool autogen_sent_id)
+ bool disamb_only, bool disamb_sh, bool autogen_sent_id,
+ bool autogen_chunk_id)
: BufferedChunkReader(tagset),
- impl_(new CclReaderImpl(*this, chunk_buf_, disamb_only, disamb_sh, autogen_sent_id))
+ impl_(new CclReaderImpl(*this, chunk_buf_, disamb_only, disamb_sh,
+ autogen_sent_id, autogen_chunk_id))
{
this->is_ = &is;
}
CclReader::CclReader(const Tagset& tagset, const std::string& filename,
- bool disamb_only, bool disamb_sh, bool autogen_sent_id)
+ bool disamb_only, bool disamb_sh, bool autogen_sent_id,
+ bool autogen_chunk_id)
: BufferedChunkReader(tagset),
- impl_(new CclReaderImpl(*this, chunk_buf_, disamb_only, disamb_sh, autogen_sent_id))
+ impl_(new CclReaderImpl(*this, chunk_buf_, disamb_only, disamb_sh,
+ autogen_sent_id, autogen_chunk_id))
{
this->is_owned_.reset(new std::ifstream(filename.c_str(), std::ifstream::in));
@@ -131,14 +149,17 @@ void CclReader::ensure_more()
CclReaderImpl::CclReaderImpl(const TokenReader& base_reader,
std::deque< boost::shared_ptr<Chunk> >& obuf,
- bool disamb_only, bool disamb_sh, bool autogen_sent_id)
+ bool disamb_only, bool disamb_sh, bool autogen_sent_id,
+ bool autogen_chunk_id)
: XmlReader(base_reader, obuf)
{
XmlReader::set_disamb_only(disamb_only);
XmlReader::set_disamb_sh(disamb_sh);
sentence_tag_name_ = "sentence";
sent_number_ = 0;
+ chunk_number_ = 0;
autogen_sent_id_ = autogen_sent_id;
+ autogen_chunk_id_ = autogen_chunk_id;
}
CclReaderImpl::~CclReaderImpl()
@@ -147,8 +168,16 @@ CclReaderImpl::~CclReaderImpl()
void CclReaderImpl::start_chunk(const AttributeList& attributes)
{
- // TODO: Autogenerating chunk identifiers
+ std::string id = get_id_from_attributes(attributes);
+ if (id.empty() && autogen_chunk_id_) {
+ std::ostringstream ss;
+ ss << ++chunk_number_;
+ id = CclReaderImpl::CHUNK_ID_PREFFIX + ss.str();
+ }
+
chunk_ = boost::make_shared<Chunk>();
+ chunk_->set_attribute("id", id);
+
std::string type = get_type_from_attributes(attributes);
if (type == "s") {
throw XcesError("Trying to parse XCES as CCL (<chunk type=\"s\">)");
@@ -278,6 +307,8 @@ void CclReader::set_option(const std::string& option)
impl_->set_disamb_only(true);
} else if (option == "autogen_sent_id") {
impl_->set_autogen_sent_id(true);
+ } else if (option == "autogen_chunk_id") {
+ impl_->set_autogen_chunk_id(true);
}
else {
BufferedChunkReader::set_option(option);
@@ -292,6 +323,8 @@ std::string CclReader::get_option(const std::string& option) const
return impl_->get_warn_on_inconsistent() ? option : "";
} else if (option == "autogen_sent_id") {
return impl_->get_autogen_sent_id() ? "autogen_sent_id" : "";
+ } else if (option == "autogen_chunk_id") {
+ return impl_->get_autogen_chunk_id() ? "autogen_chunk_id" : "";
}
return BufferedChunkReader::get_option(option);
}
diff --git a/libcorpus2/io/cclreader.h b/libcorpus2/io/cclreader.h
index daac508b14b64d50aef9e8de5da07d85ead09ff3..59c8e8b1b85ad4cd42bdb071f2da4570a2551a0e 100644
--- a/libcorpus2/io/cclreader.h
+++ b/libcorpus2/io/cclreader.h
@@ -32,11 +32,11 @@ class CclReader : public BufferedChunkReader
public:
CclReader(const Tagset& tagset, std::istream& is,
bool disamb_only = false, bool disamb_sh = false,
- bool autogen_sent_id = false);
+ bool autogen_sent_id = false, bool autogen_chunk_id_ = false);
CclReader(const Tagset& tagset, const std::string& filename,
bool disamb_only = false, bool disamb_sh = false,
- bool autogen_sent_id = false);
+ bool autogen_sent_id = false, bool autogen_chunk_id_ = false);
~CclReader();
diff --git a/libcorpus2/io/docreader.cpp b/libcorpus2/io/docreader.cpp
index cccb62583ecb230d48d5c0cf1cc2e30ba747b2d5..ca038e7170574a0cd8f3cfe64a1b31dcdb3eb5aa 100644
--- a/libcorpus2/io/docreader.cpp
+++ b/libcorpus2/io/docreader.cpp
@@ -61,7 +61,19 @@ namespace Corpus2 {
{
if (option == "autogen_sent_id") {
ccl_reader_->set_option("autogen_sent_id");
+ } else if (option == "autogen_chunk_id") {
+ ccl_reader_->set_option("autogen_chunk_id");
}
}
+ std::string DocumentReader::get_option(const std::string& option) const {
+ if (option == "autogen_sent_id") {
+ return ccl_reader_->get_option("autogen_sent_id");
+ }
+ else if (option == "autogen_chunk_id") {
+ return ccl_reader_->get_option("autogen_chunk_id");
+ }
+ return "";
+ }
+
} /* end ns Corpus2 */
diff --git a/libcorpus2/io/docreader.h b/libcorpus2/io/docreader.h
index fe683c9cfd44b981b0d0a8b0c5a47a92114af9c6..65422e2534c0e2725b9433579f9325b61ece4a91 100644
--- a/libcorpus2/io/docreader.h
+++ b/libcorpus2/io/docreader.h
@@ -59,6 +59,11 @@ public:
*/
void set_option(const std::string& option);
+ /**
+ * @return option
+ */
+ std::string get_option(const std::string& option) const;
+
private:
/**
* Makes CclReader and RelationReader for given paths to files.
diff --git a/libcorpus2/io/xmlreader.cpp b/libcorpus2/io/xmlreader.cpp
index 5b34b5cdf52f9c20b10609d1605fbfbed1d44a0c..773eb857cb85273f6298cc0d683d4ef0bd6beb72 100644
--- a/libcorpus2/io/xmlreader.cpp
+++ b/libcorpus2/io/xmlreader.cpp
@@ -52,6 +52,16 @@ std::string XmlReader::get_type_from_attributes(const AttributeList& attributes)
return type;
}
+std::string XmlReader::get_id_from_attributes(const AttributeList& attributes) const
+{
+ foreach (const Attribute& a, attributes) {
+ if (a.name == "id") {
+ return a.value;
+ }
+ }
+ return "";
+}
+
void XmlReader::on_start_element(const Glib::ustring &name,
const AttributeList& attributes)
diff --git a/libcorpus2/io/xmlreader.h b/libcorpus2/io/xmlreader.h
index d9234d4c38fd8e437f15bd8bebf7ab17dea8ad16..4517e511c6881969d8d7d21305c67e158190e317 100644
--- a/libcorpus2/io/xmlreader.h
+++ b/libcorpus2/io/xmlreader.h
@@ -49,6 +49,7 @@ public:
protected:
std::string get_type_from_attributes(const AttributeList& attributes) const;
+ std::string get_id_from_attributes(const AttributeList& attributes) const;
void on_start_element(const Glib::ustring & name,
const AttributeList& attributes);
diff --git a/swig/documentreader.i b/swig/documentreader.i
index 69229e8265c99d60496f8997b66fa2a16d9b62ba..6a538f840ece454a738a28c033d5aba46e4388ee 100644
--- a/swig/documentreader.i
+++ b/swig/documentreader.i
@@ -34,6 +34,7 @@ namespace Corpus2 {
boost::shared_ptr<Document> read();
void set_option(const std::string& option);
+ std::string get_option(const std::string& option) const;
/* --------------------------------------------------------------------- */
~DocumentReader();