1
|
/**
|
2
|
*
|
3
|
*
|
4
|
* log.debug(...) equal to log.trace(...) in the application-logs
|
5
|
*/
|
6
|
package eu.dnetlib.data.collector.plugins.rest;
|
7
|
|
8
|
import java.io.InputStream;
|
9
|
import java.io.StringWriter;
|
10
|
import java.net.URL;
|
11
|
import java.util.Iterator;
|
12
|
import java.util.Queue;
|
13
|
import java.util.concurrent.PriorityBlockingQueue;
|
14
|
|
15
|
import javax.xml.transform.OutputKeys;
|
16
|
import javax.xml.transform.Transformer;
|
17
|
import javax.xml.transform.TransformerConfigurationException;
|
18
|
import javax.xml.transform.TransformerFactory;
|
19
|
import javax.xml.transform.dom.DOMSource;
|
20
|
import javax.xml.transform.stream.StreamResult;
|
21
|
import javax.xml.xpath.XPath;
|
22
|
import javax.xml.xpath.XPathConstants;
|
23
|
import javax.xml.xpath.XPathExpression;
|
24
|
import javax.xml.xpath.XPathExpressionException;
|
25
|
import javax.xml.xpath.XPathFactory;
|
26
|
|
27
|
import org.apache.commons.io.IOUtils;
|
28
|
import org.apache.commons.lang3.StringUtils;
|
29
|
import org.apache.commons.logging.Log;
|
30
|
import org.apache.commons.logging.LogFactory;
|
31
|
import org.w3c.dom.Node;
|
32
|
import org.w3c.dom.NodeList;
|
33
|
import org.xml.sax.InputSource;
|
34
|
|
35
|
import eu.dnetlib.data.collector.plugins.oai.OaiIterator;
|
36
|
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
|
37
|
|
38
|
/**
|
39
|
* @author Jochen Schirrwagen, Aenne Loehden, Andreas Czerniak
|
40
|
* @date 2018-08-06
|
41
|
*
|
42
|
*/
|
43
|
public class RestIterator implements Iterator<String> {
|
44
|
|
45
|
// TODO: clean up the comments of replaced source code
|
46
|
private static final Log log = LogFactory.getLog(OaiIterator.class); // NOPMD by marko on 11/24/08 5:02 PM
|
47
|
|
48
|
private static final String wrapName = "recordWrap";
|
49
|
private String baseUrl;
|
50
|
private String resumptionType;
|
51
|
private String resumptionParam;
|
52
|
private String resultFormatValue;
|
53
|
private String queryParams;
|
54
|
private int resultSizeValue;
|
55
|
private int resumptionInt = 0; // integer resumption token (first record to harvest)
|
56
|
private int resultTotal = -1;
|
57
|
private String resumptionStr = Integer.toString(resumptionInt); // string resumption token (first record to harvest or token scanned from results)
|
58
|
private InputStream resultStream;
|
59
|
private Transformer transformer;
|
60
|
private XPath xpath;
|
61
|
private String query;
|
62
|
private XPathExpression xprResultTotalPath;
|
63
|
private XPathExpression xprResumptionPath;
|
64
|
private XPathExpression xprEntity;
|
65
|
private String queryFormat;
|
66
|
private String querySize;
|
67
|
private Queue<String> recordQueue = new PriorityBlockingQueue<String>();
|
68
|
private int discoverResultSize = 0;
|
69
|
|
70
|
public RestIterator(
|
71
|
final String baseUrl,
|
72
|
final String resumptionType,
|
73
|
final String resumptionParam,
|
74
|
final String resumptionXpath,
|
75
|
final String resultTotalXpath,
|
76
|
final String resultFormatParam,
|
77
|
final String resultFormatValue,
|
78
|
final String resultSizeParam,
|
79
|
final String resultSizeValueStr,
|
80
|
final String queryParams,
|
81
|
final String entityXpath
|
82
|
) {
|
83
|
this.baseUrl = baseUrl;
|
84
|
this.resumptionType = resumptionType;
|
85
|
this.resumptionParam = resumptionParam;
|
86
|
this.resultFormatValue = resultFormatValue;
|
87
|
this.queryParams = queryParams;
|
88
|
this.resultSizeValue = Integer.valueOf(resultSizeValueStr);
|
89
|
|
90
|
queryFormat = StringUtils.isNotBlank(resultFormatParam) ? "&" + resultFormatParam + "=" + resultFormatValue : "";
|
91
|
querySize = StringUtils.isNotBlank(resultSizeParam) ? "&" + resultSizeParam + "=" + resultSizeValueStr : "";
|
92
|
|
93
|
try {
|
94
|
initXmlTransformation(resultTotalXpath, resumptionXpath, entityXpath);
|
95
|
} catch(Exception e) {
|
96
|
throw new IllegalStateException("xml transformation init failed: " + e.getMessage());
|
97
|
}
|
98
|
initQueue();
|
99
|
}
|
100
|
|
101
|
private void initXmlTransformation(String resultTotalXpath, String resumptionXpath, String entityXpath) throws TransformerConfigurationException, XPathExpressionException{
|
102
|
transformer = TransformerFactory.newInstance().newTransformer();
|
103
|
transformer.setOutputProperty(OutputKeys.INDENT,"yes");
|
104
|
transformer.setOutputProperty("{http://xml.apache.org/xslt}indent-amount","3");
|
105
|
xpath = XPathFactory.newInstance().newXPath();
|
106
|
xprResultTotalPath = xpath.compile(resultTotalXpath);
|
107
|
xprResumptionPath = xpath.compile(StringUtils.isBlank(resumptionXpath) ? "/" : resumptionXpath);
|
108
|
xprEntity = xpath.compile(entityXpath);
|
109
|
}
|
110
|
|
111
|
private void initQueue() {
|
112
|
query = baseUrl + "?" + queryParams + querySize + "&" + resumptionParam + "=" + resumptionStr + queryFormat;
|
113
|
}
|
114
|
|
115
|
private void disconnect() {
|
116
|
// TODO close inputstream
|
117
|
}
|
118
|
|
119
|
/* (non-Javadoc)
|
120
|
* @see java.util.Iterator#hasNext()
|
121
|
*/
|
122
|
@Override
|
123
|
public boolean hasNext() {
|
124
|
if (recordQueue.isEmpty() && query.isEmpty()) {
|
125
|
disconnect();
|
126
|
return false;
|
127
|
} else {
|
128
|
return true;
|
129
|
}
|
130
|
}
|
131
|
|
132
|
/* (non-Javadoc)
|
133
|
* @see java.util.Iterator#next()
|
134
|
*/
|
135
|
@Override
|
136
|
public String next() {
|
137
|
synchronized (recordQueue) {
|
138
|
while (recordQueue.isEmpty() && !query.isEmpty() ) {
|
139
|
try {
|
140
|
log.info("get Query: " + query);
|
141
|
query = downloadPage(query);
|
142
|
log.debug("next query from downloadPage method: " + query);
|
143
|
} catch(CollectorServiceException e) {
|
144
|
log.debug("CollectorPlugin.next()-Exception: " + e);
|
145
|
throw new RuntimeException(e);
|
146
|
}
|
147
|
}
|
148
|
return recordQueue.poll();
|
149
|
}
|
150
|
}
|
151
|
|
152
|
|
153
|
/*
|
154
|
* download page and return nextQuery
|
155
|
*/
|
156
|
private String downloadPage(String query) throws CollectorServiceException{
|
157
|
String resultJson;
|
158
|
String resultXml = "";
|
159
|
String nextQuery = "";
|
160
|
String emptyXml = "<"+wrapName+"></"+wrapName+">";
|
161
|
Node resultNode = null;
|
162
|
NodeList nodeList = null;
|
163
|
String qUrlArgument = "";
|
164
|
int urlOldResumptionSize = 0;
|
165
|
|
166
|
try {
|
167
|
URL qUrl = new URL(query);
|
168
|
|
169
|
resultStream = qUrl.openStream();
|
170
|
if("json".equals(resultFormatValue.toLowerCase())){
|
171
|
resultJson = IOUtils.toString(resultStream,"UTF-8");
|
172
|
|
173
|
//TODO move regex definitions as constant fields
|
174
|
// pre-clean json - rid spaces of element names (misinterpreted as elements with attributes in xml)
|
175
|
resultJson = syntaxConvertJsonKeyNamens(resultJson);
|
176
|
// while(resultJson.matches(".*\"([^\"]*)\\s+([^\"]*)\":.*")){
|
177
|
// resultJson = resultJson.replaceAll("\"([^\"]*)\\s+([^\"]*)\":", "\"$1_$2\":");
|
178
|
// }
|
179
|
org.json.JSONObject jsonObject = new org.json.JSONObject(resultJson);
|
180
|
resultXml = org.json.XML.toString(jsonObject,wrapName); // wrap xml in single root element
|
181
|
log.trace(resultXml);
|
182
|
resultStream = IOUtils.toInputStream(resultXml,"UTF-8");
|
183
|
}
|
184
|
|
185
|
if (!(emptyXml.toLowerCase()).equals(resultXml.toLowerCase())) {
|
186
|
resultNode = (Node) xpath.evaluate("/", new InputSource(resultStream), XPathConstants.NODE);
|
187
|
nodeList = (NodeList) xprEntity.evaluate(resultNode, XPathConstants.NODESET);
|
188
|
log.debug("nodeList.length: " + nodeList.getLength());
|
189
|
for (int i = 0; i < nodeList.getLength(); i++) {
|
190
|
StringWriter sw = new StringWriter();
|
191
|
transformer.transform(new DOMSource(nodeList.item(i)), new StreamResult(sw));
|
192
|
recordQueue.add(sw.toString());
|
193
|
}
|
194
|
} else { log.info("resultXml is equal with emptyXml"); }
|
195
|
|
196
|
resumptionInt += resultSizeValue;
|
197
|
|
198
|
/* replaced by switch statement as follow
|
199
|
if("scan".equals(resumptionType.toLowerCase())) { resumptionStr = xprResumptionPath.evaluate(resultNode);}
|
200
|
if("count".equals(resumptionType.toLowerCase())){ resumptionStr = Integer.toString(resumptionInt); }
|
201
|
*/
|
202
|
switch(resumptionType.toLowerCase()) {
|
203
|
case "scan": // read of resumptionToken , evaluate next results, e.g. OAI
|
204
|
resumptionStr = xprResumptionPath.evaluate(resultNode);
|
205
|
break;
|
206
|
|
207
|
case "count": // begin at one step for all records
|
208
|
resumptionStr = Integer.toString(resumptionInt);
|
209
|
break;
|
210
|
|
211
|
case "discover": // length of results unknown
|
212
|
qUrlArgument = qUrl.getQuery();
|
213
|
String[] arrayQUrlArgument = qUrlArgument.split("&");
|
214
|
for(String arrayUrlArgStr : arrayQUrlArgument ) {
|
215
|
if(arrayUrlArgStr.startsWith(resumptionParam)) {
|
216
|
String[] resumptionKeyValue = arrayUrlArgStr.split("=");
|
217
|
urlOldResumptionSize = Integer.parseInt(resumptionKeyValue[1]);
|
218
|
log.debug("discover OldResumptionSize from Url: " + urlOldResumptionSize);
|
219
|
}
|
220
|
}
|
221
|
|
222
|
//
|
223
|
if( ( (emptyXml.toLowerCase()).equals(resultXml.toLowerCase()) )
|
224
|
|| ( (nodeList != null) && (nodeList.getLength() < resultSizeValue) )
|
225
|
) {
|
226
|
// resumptionStr = "";
|
227
|
if(nodeList != null) { discoverResultSize += nodeList.getLength(); }
|
228
|
resultTotal = discoverResultSize;
|
229
|
} else {
|
230
|
resumptionStr = Integer.toString(resumptionInt);
|
231
|
resultTotal = resumptionInt+1;
|
232
|
if(nodeList != null) { discoverResultSize += nodeList.getLength(); }
|
233
|
}
|
234
|
log.info("discoverResultSize: " + discoverResultSize);
|
235
|
break;
|
236
|
|
237
|
default: // otherwise: abort
|
238
|
// resultTotal = resumptionInt;
|
239
|
break;
|
240
|
}
|
241
|
|
242
|
if (resultTotal == -1) {
|
243
|
resultTotal = Integer.parseInt(xprResultTotalPath.evaluate(resultNode));
|
244
|
log.info("resultTotal was -1 is now: " + resultTotal);
|
245
|
}
|
246
|
log.info("resultTotal: " + resultTotal);
|
247
|
log.info("resInt: " + resumptionInt);
|
248
|
if (resumptionInt < resultTotal) {
|
249
|
nextQuery = baseUrl + "?" + queryParams + querySize + "&" + resumptionParam + "=" + resumptionStr + queryFormat;
|
250
|
} else
|
251
|
nextQuery = "";
|
252
|
|
253
|
log.debug("nextQueryUrl: " + nextQuery);
|
254
|
return nextQuery;
|
255
|
|
256
|
} catch(Exception e) {
|
257
|
log.error(e);
|
258
|
throw new IllegalStateException("collection failed: " + e.getMessage());
|
259
|
}
|
260
|
}
|
261
|
|
262
|
/**
|
263
|
* convert in Json-KeyName 'whitespace(s)' to '_' and '/' to '_', '(' and ')' to ''
|
264
|
*
|
265
|
* @param jsonInput
|
266
|
* @return
|
267
|
*/
|
268
|
private String syntaxConvertJsonKeyNamens(String jsonInput) {
|
269
|
|
270
|
log.trace("before convertJsonKeyNames: " + jsonInput);
|
271
|
// replace ' 's in JSON Namens with '_'
|
272
|
while (jsonInput.matches(".*\"([^\"]*)\\s+([^\"]*)\":.*")) {
|
273
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)\\s+([^\"]*)\":", "\"$1_$2\":");
|
274
|
}
|
275
|
|
276
|
// replace forward-slash (sign '/' ) in JSON Names with '_'
|
277
|
while (jsonInput.matches(".*\"([^\"]*)/([^\"]*)\":.*")) {
|
278
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)/([^\"]*)\":", "\"$1_$2\":");
|
279
|
}
|
280
|
|
281
|
// replace '(' in JSON Names with ''
|
282
|
while (jsonInput.matches(".*\"([^\"]*)[(]([^\"]*)\":.*")) {
|
283
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)[(]([^\"]*)\":", "\"$1$2\":");
|
284
|
}
|
285
|
// replace ')' in JSON Names with ''
|
286
|
while (jsonInput.matches(".*\"([^\"]*)[)]([^\"]*)\":.*")) {
|
287
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)[)]([^\"]*)\":", "\"$1$2\":");
|
288
|
}
|
289
|
|
290
|
log.trace("after syntaxConvertJsonKeyNames: " + jsonInput);
|
291
|
return jsonInput;
|
292
|
}
|
293
|
|
294
|
}
|