134 lines
3.9 KiB
Java
134 lines
3.9 KiB
Java
|
package eu.dnetlib.data.collector.plugins.oaisets;
|
||
|
|
||
|
import java.io.StringReader;
|
||
|
import java.util.Iterator;
|
||
|
import java.util.Queue;
|
||
|
import java.util.Set;
|
||
|
import java.util.concurrent.PriorityBlockingQueue;
|
||
|
|
||
|
import org.apache.commons.logging.Log;
|
||
|
import org.apache.commons.logging.LogFactory;
|
||
|
import org.dom4j.Document;
|
||
|
import org.dom4j.DocumentException;
|
||
|
import org.dom4j.Element;
|
||
|
import org.dom4j.Node;
|
||
|
import org.dom4j.io.SAXReader;
|
||
|
|
||
|
import com.google.common.collect.Sets;
|
||
|
|
||
|
import eu.dnetlib.data.collector.plugins.HttpConnector;
|
||
|
import eu.dnetlib.data.collector.plugins.oai.engine.XmlCleaner;
|
||
|
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
|
||
|
|
||
|
public class OaiSetsIterator implements Iterator<String> {
|
||
|
|
||
|
private static final Log log = LogFactory.getLog(OaiSetsIterator.class);
|
||
|
|
||
|
private Queue<String> queue = new PriorityBlockingQueue<String>();
|
||
|
private SAXReader reader = new SAXReader();
|
||
|
|
||
|
private String baseUrl;
|
||
|
|
||
|
private String token;
|
||
|
private boolean started;
|
||
|
private HttpConnector httpConnector;
|
||
|
|
||
|
private Set<String> setsAlreadySeen = Sets.newHashSet();
|
||
|
|
||
|
public OaiSetsIterator(final String baseUrl, final HttpConnector httpConnector) {
|
||
|
this.baseUrl = baseUrl;
|
||
|
this.started = false;
|
||
|
this.httpConnector = httpConnector;
|
||
|
}
|
||
|
|
||
|
private void verifyStarted() {
|
||
|
if (!this.started) {
|
||
|
this.started = true;
|
||
|
try {
|
||
|
this.token = firstPage();
|
||
|
} catch (CollectorServiceException e) {
|
||
|
throw new RuntimeException(e);
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
@Override
|
||
|
public boolean hasNext() {
|
||
|
synchronized (queue) {
|
||
|
verifyStarted();
|
||
|
return !queue.isEmpty();
|
||
|
}
|
||
|
}
|
||
|
|
||
|
@Override
|
||
|
public String next() {
|
||
|
synchronized (queue) {
|
||
|
verifyStarted();
|
||
|
final String res = queue.poll();
|
||
|
while (queue.isEmpty() && (token != null) && !token.isEmpty()) {
|
||
|
try {
|
||
|
token = otherPages(token);
|
||
|
} catch (CollectorServiceException e) {
|
||
|
throw new RuntimeException(e);
|
||
|
}
|
||
|
}
|
||
|
return res;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
@Override
|
||
|
public void remove() {}
|
||
|
|
||
|
private String firstPage() throws CollectorServiceException {
|
||
|
final String url = baseUrl + "?verb=ListSets";
|
||
|
log.info("Start harvesting using url: " + url);
|
||
|
return downloadPage(url);
|
||
|
}
|
||
|
|
||
|
private String otherPages(final String resumptionToken) throws CollectorServiceException {
|
||
|
return downloadPage(baseUrl + "?verb=ListSets&resumptionToken=" + resumptionToken);
|
||
|
}
|
||
|
|
||
|
private String downloadPage(final String url) throws CollectorServiceException {
|
||
|
|
||
|
final String xml = httpConnector.getInputSource(url);
|
||
|
|
||
|
Document doc;
|
||
|
try {
|
||
|
doc = reader.read(new StringReader(xml));
|
||
|
} catch (DocumentException e) {
|
||
|
log.warn("Error parsing xml, I try to clean it: " + xml, e);
|
||
|
final String cleaned = XmlCleaner.cleanAllEntities(xml);
|
||
|
try {
|
||
|
doc = reader.read(new StringReader(cleaned));
|
||
|
} catch (DocumentException e1) {
|
||
|
throw new CollectorServiceException("Error parsing cleaned document:" + cleaned, e1);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
final Node errorNode = doc.selectSingleNode("/*[local-name()='OAI-PMH']/*[local-name()='error']");
|
||
|
if (errorNode != null) {
|
||
|
final String code = errorNode.valueOf("@code");
|
||
|
if ("noRecordsMatch".equalsIgnoreCase(code.trim())) {
|
||
|
log.warn("noRecordsMatch for oai call: " + url);
|
||
|
return null;
|
||
|
} else throw new CollectorServiceException(code + " - " + errorNode.getText());
|
||
|
}
|
||
|
|
||
|
boolean sawAllSets = true;
|
||
|
for (Object o : doc.selectNodes("//*[local-name()='ListSets']/*[local-name()='set']")) {
|
||
|
String set = ((Element) o).valueOf("./*[local-name()='setSpec']");
|
||
|
if (!setsAlreadySeen.contains(set)) {
|
||
|
sawAllSets = false;
|
||
|
setsAlreadySeen.add(set);
|
||
|
queue.add(((Node) o).asXML());
|
||
|
}
|
||
|
}
|
||
|
if (sawAllSets) {
|
||
|
log.warn("URL " + baseUrl + " keeps returning the same OAI sets. Please contact the repo admin.");
|
||
|
System.out.println("URL " + baseUrl + " keeps returning the same OAI sets. Please contact the repo admin.");
|
||
|
return null;
|
||
|
} else return doc.valueOf("//*[local-name()='resumptionToken']");
|
||
|
}
|
||
|
}
|