| 1 |
26600
|
sandro.lab
|
package eu.dnetlib.data.collector.plugins.oai;
|
| 2 |
|
|
|
| 3 |
28959
|
michele.ar
|
import java.io.StringReader;
|
| 4 |
26600
|
sandro.lab
|
import java.util.Iterator;
|
| 5 |
|
|
import java.util.Queue;
|
| 6 |
|
|
import java.util.concurrent.PriorityBlockingQueue;
|
| 7 |
|
|
|
| 8 |
|
|
import org.apache.commons.logging.Log;
|
| 9 |
|
|
import org.apache.commons.logging.LogFactory;
|
| 10 |
|
|
import org.dom4j.Document;
|
| 11 |
28959
|
michele.ar
|
import org.dom4j.DocumentException;
|
| 12 |
26600
|
sandro.lab
|
import org.dom4j.Node;
|
| 13 |
|
|
import org.dom4j.io.SAXReader;
|
| 14 |
|
|
|
| 15 |
28959
|
michele.ar
|
import eu.dnetlib.data.collector.plugins.oai.engine.HttpConnector;
|
| 16 |
|
|
import eu.dnetlib.data.collector.plugins.oai.engine.XmlCleaner;
|
| 17 |
29018
|
michele.ar
|
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
|
| 18 |
26600
|
sandro.lab
|
|
| 19 |
|
|
public class OaiIterator implements Iterator<String> {
|
| 20 |
|
|
|
| 21 |
|
|
private static final Log log = LogFactory.getLog(OaiIterator.class); // NOPMD by marko on 11/24/08 5:02 PM
|
| 22 |
|
|
|
| 23 |
|
|
private Queue<String> queue = new PriorityBlockingQueue<String>();
|
| 24 |
|
|
private SAXReader reader = new SAXReader();
|
| 25 |
28959
|
michele.ar
|
|
| 26 |
26600
|
sandro.lab
|
private String baseUrl;
|
| 27 |
27189
|
michele.ar
|
private String set;
|
| 28 |
|
|
private String mdFormat;
|
| 29 |
27303
|
michele.ar
|
private String fromDate;
|
| 30 |
|
|
private String untilDate;
|
| 31 |
26600
|
sandro.lab
|
private String token;
|
| 32 |
27189
|
michele.ar
|
private boolean started;
|
| 33 |
28959
|
michele.ar
|
private HttpConnector httpConnector;
|
| 34 |
26600
|
sandro.lab
|
|
| 35 |
28959
|
michele.ar
|
public OaiIterator(final String baseUrl, final String mdFormat, final String set, final String fromDate, final String untilDate, final HttpConnector httpConnector) {
|
| 36 |
26600
|
sandro.lab
|
this.baseUrl = baseUrl;
|
| 37 |
27189
|
michele.ar
|
this.mdFormat = mdFormat;
|
| 38 |
|
|
this.set = set;
|
| 39 |
27303
|
michele.ar
|
this.fromDate = fromDate;
|
| 40 |
|
|
this.untilDate = untilDate;
|
| 41 |
27189
|
michele.ar
|
this.started = false;
|
| 42 |
28959
|
michele.ar
|
this.httpConnector = httpConnector;
|
| 43 |
27189
|
michele.ar
|
}
|
| 44 |
|
|
|
| 45 |
|
|
private void verifyStarted() {
|
| 46 |
|
|
if (!this.started) {
|
| 47 |
|
|
this.started = true;
|
| 48 |
28959
|
michele.ar
|
try {
|
| 49 |
|
|
this.token = firstPage();
|
| 50 |
29018
|
michele.ar
|
} catch (CollectorServiceException e) {
|
| 51 |
28959
|
michele.ar
|
throw new RuntimeException(e);
|
| 52 |
|
|
}
|
| 53 |
26600
|
sandro.lab
|
}
|
| 54 |
|
|
}
|
| 55 |
|
|
|
| 56 |
|
|
@Override
|
| 57 |
|
|
public boolean hasNext() {
|
| 58 |
|
|
synchronized (queue) {
|
| 59 |
27189
|
michele.ar
|
verifyStarted();
|
| 60 |
26600
|
sandro.lab
|
return !queue.isEmpty();
|
| 61 |
|
|
}
|
| 62 |
|
|
}
|
| 63 |
|
|
|
| 64 |
|
|
@Override
|
| 65 |
|
|
public String next() {
|
| 66 |
|
|
synchronized (queue) {
|
| 67 |
27189
|
michele.ar
|
verifyStarted();
|
| 68 |
26600
|
sandro.lab
|
final String res = queue.poll();
|
| 69 |
|
|
while (queue.isEmpty() && (token != null) && !token.isEmpty()) {
|
| 70 |
28959
|
michele.ar
|
try {
|
| 71 |
|
|
token = otherPages(token);
|
| 72 |
29018
|
michele.ar
|
} catch (CollectorServiceException e) {
|
| 73 |
28959
|
michele.ar
|
throw new RuntimeException(e);
|
| 74 |
|
|
}
|
| 75 |
26600
|
sandro.lab
|
}
|
| 76 |
|
|
return res;
|
| 77 |
|
|
}
|
| 78 |
|
|
}
|
| 79 |
|
|
|
| 80 |
|
|
@Override
|
| 81 |
|
|
public void remove() {}
|
| 82 |
|
|
|
| 83 |
29018
|
michele.ar
|
private String firstPage() throws CollectorServiceException {
|
| 84 |
27189
|
michele.ar
|
|
| 85 |
26600
|
sandro.lab
|
String url = baseUrl + "?verb=ListRecords&metadataPrefix=" + mdFormat;
|
| 86 |
|
|
if ((set != null) && !set.isEmpty()) {
|
| 87 |
|
|
url += "&set=" + set;
|
| 88 |
|
|
}
|
| 89 |
27189
|
michele.ar
|
|
| 90 |
27303
|
michele.ar
|
if ((fromDate != null) && fromDate.matches("\\d{4}-\\d{2}-\\d{2}")) {
|
| 91 |
|
|
url += "&from=" + fromDate;
|
| 92 |
|
|
}
|
| 93 |
|
|
|
| 94 |
|
|
if ((untilDate != null) && untilDate.matches("\\d{4}-\\d{2}-\\d{2}")) {
|
| 95 |
|
|
url += "&until=" + untilDate;
|
| 96 |
|
|
}
|
| 97 |
|
|
|
| 98 |
28959
|
michele.ar
|
log.info("Start harvesting using url: " + url);
|
| 99 |
27189
|
michele.ar
|
|
| 100 |
26600
|
sandro.lab
|
return downloadPage(url);
|
| 101 |
|
|
}
|
| 102 |
|
|
|
| 103 |
29018
|
michele.ar
|
private String otherPages(final String resumptionToken) throws CollectorServiceException {
|
| 104 |
26600
|
sandro.lab
|
return downloadPage(baseUrl + "?verb=ListRecords&resumptionToken=" + resumptionToken);
|
| 105 |
|
|
}
|
| 106 |
|
|
|
| 107 |
29018
|
michele.ar
|
private String downloadPage(final String url) throws CollectorServiceException {
|
| 108 |
28959
|
michele.ar
|
|
| 109 |
|
|
final String xml = httpConnector.getInputSource(url);
|
| 110 |
27189
|
michele.ar
|
|
| 111 |
28959
|
michele.ar
|
Document doc;
|
| 112 |
26600
|
sandro.lab
|
try {
|
| 113 |
28959
|
michele.ar
|
doc = reader.read(new StringReader(xml));
|
| 114 |
|
|
} catch (DocumentException e) {
|
| 115 |
|
|
log.warn("Error parsing xml, I try to clean it: " + xml, e);
|
| 116 |
|
|
final String cleaned = XmlCleaner.cleanAllEntities(xml);
|
| 117 |
|
|
try {
|
| 118 |
|
|
doc = reader.read(new StringReader(cleaned));
|
| 119 |
|
|
} catch (DocumentException e1) {
|
| 120 |
29018
|
michele.ar
|
throw new CollectorServiceException("Error parsing cleaned document:" + cleaned, e1);
|
| 121 |
26600
|
sandro.lab
|
}
|
| 122 |
|
|
}
|
| 123 |
28959
|
michele.ar
|
|
| 124 |
|
|
for (Object o : doc.selectNodes("//*[local-name()='ListRecords']/*[local-name()='record']")) {
|
| 125 |
|
|
queue.add(((Node) o).asXML());
|
| 126 |
|
|
}
|
| 127 |
|
|
|
| 128 |
|
|
return doc.valueOf("//*[local-name()='resumptionToken']");
|
| 129 |
|
|
|
| 130 |
26600
|
sandro.lab
|
}
|
| 131 |
|
|
|
| 132 |
|
|
}
|