1
|
/**
|
2
|
* log.debug(...) equal to log.trace(...) in the application-logs
|
3
|
* <p>
|
4
|
* known bug: at resumptionType 'discover' if the (resultTotal % resultSizeValue) == 0 the collecting fails -> change the resultSizeValue
|
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.net.HttpURLConnection;
|
12
|
import java.util.Iterator;
|
13
|
import java.util.Queue;
|
14
|
import java.util.concurrent.PriorityBlockingQueue;
|
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.*;
|
22
|
|
23
|
import eu.dnetlib.data.collector.plugins.oai.engine.XmlCleaner;
|
24
|
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
|
25
|
import org.apache.commons.io.IOUtils;
|
26
|
import org.apache.commons.lang3.StringUtils;
|
27
|
import org.apache.commons.logging.Log;
|
28
|
import org.apache.commons.logging.LogFactory;
|
29
|
import org.w3c.dom.Node;
|
30
|
import org.w3c.dom.NodeList;
|
31
|
import org.xml.sax.InputSource;
|
32
|
|
33
|
/**
|
34
|
* @author Jochen Schirrwagen, Aenne Loehden, Andreas Czerniak
|
35
|
* @date 2020-03-20
|
36
|
*
|
37
|
*/
|
38
|
public class RestIterator implements Iterator<String> {
|
39
|
|
40
|
// TODO: clean up the comments of replaced source code
|
41
|
private static final Log log = LogFactory.getLog(RestIterator.class); // NOPMD by marko on 11/24/08 5:02 PM
|
42
|
|
43
|
private static final String wrapName = "recordWrap";
|
44
|
private String baseUrl;
|
45
|
private String resumptionType;
|
46
|
private String resumptionParam;
|
47
|
private String resultFormatValue;
|
48
|
private String queryParams;
|
49
|
private int resultSizeValue;
|
50
|
private int resumptionInt = 0; // integer resumption token (first record to harvest)
|
51
|
private int resultTotal = -1;
|
52
|
private String resumptionStr = Integer.toString(resumptionInt); // string resumption token (first record to harvest or token scanned from results)
|
53
|
private InputStream resultStream;
|
54
|
private Transformer transformer;
|
55
|
private XPath xpath;
|
56
|
private String query;
|
57
|
private XPathExpression xprResultTotalPath;
|
58
|
private XPathExpression xprResumptionPath;
|
59
|
private XPathExpression xprEntity;
|
60
|
private String queryFormat;
|
61
|
private String querySize;
|
62
|
private String authMethod;
|
63
|
private String authToken;
|
64
|
private Queue<String> recordQueue = new PriorityBlockingQueue<String>();
|
65
|
private int discoverResultSize = 0;
|
66
|
private int pagination = 1;
|
67
|
private String resultOffsetParam;
|
68
|
|
69
|
/**
|
70
|
* RestIterator class
|
71
|
*
|
72
|
* compatible to version before 1.3.33
|
73
|
*
|
74
|
* @param baseUrl
|
75
|
* @param resumptionType
|
76
|
* @param resumptionParam
|
77
|
* @param resumptionXpath
|
78
|
* @param resultTotalXpath
|
79
|
* @param resultFormatParam
|
80
|
* @param resultFormatValue
|
81
|
* @param resultSizeParam
|
82
|
* @param resultSizeValueStr
|
83
|
* @param queryParams
|
84
|
* @param entityXpath
|
85
|
*/
|
86
|
public RestIterator(
|
87
|
final String baseUrl,
|
88
|
final String resumptionType,
|
89
|
final String resumptionParam,
|
90
|
final String resumptionXpath,
|
91
|
final String resultTotalXpath,
|
92
|
final String resultFormatParam,
|
93
|
final String resultFormatValue,
|
94
|
final String resultSizeParam,
|
95
|
final String resultSizeValueStr,
|
96
|
final String queryParams,
|
97
|
final String entityXpath
|
98
|
) {
|
99
|
this(baseUrl,resumptionType,resumptionParam,resumptionXpath,resultTotalXpath,resultFormatParam,resultFormatValue,resultSizeParam,resultSizeValueStr,queryParams,entityXpath,"", "","");
|
100
|
}
|
101
|
|
102
|
/** RestIterator class
|
103
|
* compatible to version 1.3.33
|
104
|
* @param baseUrl
|
105
|
* @param resumptionType
|
106
|
* @param resumptionParam
|
107
|
* @param resumptionXpath
|
108
|
* @param resultTotalXpath
|
109
|
* @param resultFormatParam
|
110
|
* @param resultFormatValue
|
111
|
* @param resultSizeParam
|
112
|
* @param resultSizeValueStr
|
113
|
* @param queryParams
|
114
|
* @param entityXpath
|
115
|
* @param authMethod
|
116
|
* @param authToken
|
117
|
*/
|
118
|
public RestIterator(
|
119
|
final String baseUrl,
|
120
|
final String resumptionType,
|
121
|
final String resumptionParam,
|
122
|
final String resumptionXpath,
|
123
|
final String resultTotalXpath,
|
124
|
final String resultFormatParam,
|
125
|
final String resultFormatValue,
|
126
|
final String resultSizeParam,
|
127
|
final String resultSizeValueStr,
|
128
|
final String queryParams,
|
129
|
final String entityXpath,
|
130
|
final String authMethod,
|
131
|
final String authToken
|
132
|
) {
|
133
|
this(baseUrl,resumptionType,resumptionParam,resumptionXpath,resultTotalXpath,resultFormatParam,resultFormatValue,resultSizeParam,resultSizeValueStr,queryParams,entityXpath,"", "","");
|
134
|
}
|
135
|
|
136
|
public RestIterator(
|
137
|
final String baseUrl,
|
138
|
final String resumptionType,
|
139
|
final String resumptionParam,
|
140
|
final String resumptionXpath,
|
141
|
final String resultTotalXpath,
|
142
|
final String resultFormatParam,
|
143
|
final String resultFormatValue,
|
144
|
final String resultSizeParam,
|
145
|
final String resultSizeValueStr,
|
146
|
final String queryParams,
|
147
|
final String entityXpath,
|
148
|
final String authMethod,
|
149
|
final String authToken,
|
150
|
final String resultOffsetParam
|
151
|
) {
|
152
|
this.baseUrl = baseUrl;
|
153
|
this.resumptionType = resumptionType;
|
154
|
this.resumptionParam = resumptionParam;
|
155
|
this.resultFormatValue = resultFormatValue;
|
156
|
this.queryParams = queryParams;
|
157
|
this.resultSizeValue = Integer.valueOf(resultSizeValueStr);
|
158
|
this.authMethod = authMethod;
|
159
|
this.authToken = authToken;
|
160
|
this.resultOffsetParam = resultOffsetParam;
|
161
|
|
162
|
queryFormat = StringUtils.isNotBlank(resultFormatParam) ? "&" + resultFormatParam + "=" + resultFormatValue : "";
|
163
|
querySize = StringUtils.isNotBlank(resultSizeParam) ? "&" + resultSizeParam + "=" + resultSizeValueStr : "";
|
164
|
|
165
|
try {
|
166
|
initXmlTransformation(resultTotalXpath, resumptionXpath, entityXpath);
|
167
|
} catch (Exception e) {
|
168
|
throw new IllegalStateException("xml transformation init failed: " + e.getMessage());
|
169
|
}
|
170
|
initQueue();
|
171
|
}
|
172
|
|
173
|
|
174
|
private void initXmlTransformation(String resultTotalXpath, String resumptionXpath, String entityXpath)
|
175
|
throws TransformerConfigurationException, XPathExpressionException {
|
176
|
transformer = TransformerFactory.newInstance().newTransformer();
|
177
|
transformer.setOutputProperty(OutputKeys.INDENT, "yes");
|
178
|
transformer.setOutputProperty("{http://xml.apache.org/xslt}indent-amount", "3");
|
179
|
xpath = XPathFactory.newInstance().newXPath();
|
180
|
xprResultTotalPath = xpath.compile(resultTotalXpath);
|
181
|
xprResumptionPath = xpath.compile(StringUtils.isBlank(resumptionXpath) ? "/" : resumptionXpath);
|
182
|
xprEntity = xpath.compile(entityXpath);
|
183
|
}
|
184
|
|
185
|
private void initQueue() {
|
186
|
query = baseUrl + "?" + queryParams + querySize + queryFormat;
|
187
|
}
|
188
|
|
189
|
private void disconnect() {
|
190
|
// TODO close inputstream
|
191
|
}
|
192
|
|
193
|
/* (non-Javadoc)
|
194
|
* @see java.util.Iterator#hasNext()
|
195
|
*/
|
196
|
@Override
|
197
|
public boolean hasNext() {
|
198
|
if (recordQueue.isEmpty() && query.isEmpty()) {
|
199
|
disconnect();
|
200
|
return false;
|
201
|
} else {
|
202
|
return true;
|
203
|
}
|
204
|
}
|
205
|
|
206
|
/* (non-Javadoc)
|
207
|
* @see java.util.Iterator#next()
|
208
|
*/
|
209
|
@Override
|
210
|
public String next() {
|
211
|
synchronized (recordQueue) {
|
212
|
while (recordQueue.isEmpty() && !query.isEmpty()) {
|
213
|
try {
|
214
|
log.info("get Query: " + query);
|
215
|
query = downloadPage(query);
|
216
|
log.debug("next queryURL from downloadPage(): " + query);
|
217
|
} catch (CollectorServiceException e) {
|
218
|
log.debug("CollectorPlugin.next()-Exception: " + e);
|
219
|
throw new RuntimeException(e);
|
220
|
}
|
221
|
}
|
222
|
return recordQueue.poll();
|
223
|
}
|
224
|
}
|
225
|
|
226
|
/*
|
227
|
* download page and return nextQuery
|
228
|
*/
|
229
|
private String downloadPage(String query) throws CollectorServiceException {
|
230
|
String resultJson;
|
231
|
String resultXml = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>";
|
232
|
String nextQuery = "";
|
233
|
String emptyXml = resultXml + "<" + wrapName + "></" + wrapName + ">";
|
234
|
Node resultNode = null;
|
235
|
NodeList nodeList = null;
|
236
|
String qUrlArgument = "";
|
237
|
int urlOldResumptionSize = 0;
|
238
|
InputStream theHttpInputStream;
|
239
|
|
240
|
try {
|
241
|
URL qUrl = new URL(query);
|
242
|
|
243
|
if (this.authMethod == "bearer") {
|
244
|
log.trace("authMethode before inputStream: " + resultXml);
|
245
|
|
246
|
HttpURLConnection conn = (HttpURLConnection) qUrl.openConnection();
|
247
|
conn.setRequestProperty("Authorization","Bearer "+authToken);
|
248
|
conn.setRequestProperty("Content-Type","application/json");
|
249
|
conn.setRequestMethod("GET");
|
250
|
theHttpInputStream = conn.getInputStream();
|
251
|
} else {
|
252
|
theHttpInputStream = qUrl.openStream();
|
253
|
}
|
254
|
|
255
|
resultStream = theHttpInputStream;
|
256
|
if ("json".equals(resultFormatValue.toLowerCase())) {
|
257
|
resultJson = IOUtils.toString(resultStream, "UTF-8");
|
258
|
resultJson = syntaxConvertJsonKeyNamens(resultJson);
|
259
|
org.json.JSONObject jsonObject = new org.json.JSONObject(resultJson);
|
260
|
resultXml += org.json.XML.toString(jsonObject, wrapName); // wrap xml in single root element
|
261
|
log.trace("before inputStream: " + resultXml);
|
262
|
resultXml = XmlCleaner.cleanAllEntities(resultXml);
|
263
|
log.trace("after cleaning: " + resultXml);
|
264
|
resultStream = IOUtils.toInputStream(resultXml, "UTF-8");
|
265
|
}
|
266
|
|
267
|
if (!(emptyXml.toLowerCase()).equals(resultXml.toLowerCase())) {
|
268
|
resultNode = (Node) xpath.evaluate("/", new InputSource(resultStream), XPathConstants.NODE);
|
269
|
nodeList = (NodeList) xprEntity.evaluate(resultNode, XPathConstants.NODESET);
|
270
|
log.debug("nodeList.length: " + nodeList.getLength());
|
271
|
for (int i = 0; i < nodeList.getLength(); i++) {
|
272
|
StringWriter sw = new StringWriter();
|
273
|
transformer.transform(new DOMSource(nodeList.item(i)), new StreamResult(sw));
|
274
|
recordQueue.add(sw.toString());
|
275
|
}
|
276
|
} else { log.info("resultXml is equal with emptyXml"); }
|
277
|
|
278
|
resumptionInt += resultSizeValue;
|
279
|
|
280
|
switch (resumptionType.toLowerCase()) {
|
281
|
case "scan": // read of resumptionToken , evaluate next results, e.g. OAI, iterate over items
|
282
|
resumptionStr = xprResumptionPath.evaluate(resultNode);
|
283
|
break;
|
284
|
|
285
|
case "count": // begin at one step for all records, iterate over items
|
286
|
resumptionStr = Integer.toString(resumptionInt);
|
287
|
break;
|
288
|
|
289
|
case "discover": // size of result items unknown, iterate over items (for openDOAR - 201808)
|
290
|
if (resultSizeValue < 2) {throw new CollectorServiceException("Mode: discover, Param 'resultSizeValue' is less than 2");}
|
291
|
qUrlArgument = qUrl.getQuery();
|
292
|
String[] arrayQUrlArgument = qUrlArgument.split("&");
|
293
|
for (String arrayUrlArgStr : arrayQUrlArgument) {
|
294
|
if (arrayUrlArgStr.startsWith(resumptionParam)) {
|
295
|
String[] resumptionKeyValue = arrayUrlArgStr.split("=");
|
296
|
if(isInteger(resumptionKeyValue[1])) {
|
297
|
urlOldResumptionSize = Integer.parseInt(resumptionKeyValue[1]);
|
298
|
log.debug("discover OldResumptionSize from Url (int): " + urlOldResumptionSize);
|
299
|
} else {
|
300
|
log.debug("discover OldResumptionSize from Url (str): " + resumptionKeyValue[1]);
|
301
|
}
|
302
|
}
|
303
|
}
|
304
|
|
305
|
if (((emptyXml.toLowerCase()).equals(resultXml.toLowerCase()))
|
306
|
|| ((nodeList != null) && (nodeList.getLength() < resultSizeValue))
|
307
|
) {
|
308
|
// resumptionStr = "";
|
309
|
if (nodeList != null) { discoverResultSize += nodeList.getLength(); }
|
310
|
resultTotal = discoverResultSize;
|
311
|
} else {
|
312
|
resumptionStr = Integer.toString(resumptionInt);
|
313
|
resultTotal = resumptionInt + 1;
|
314
|
if (nodeList != null) { discoverResultSize += nodeList.getLength(); }
|
315
|
}
|
316
|
log.info("discoverResultSize: " + discoverResultSize);
|
317
|
break;
|
318
|
|
319
|
case "pagination":
|
320
|
case "page": // pagination, iterate over page numbers
|
321
|
pagination += 1;
|
322
|
if (nodeList != null) {
|
323
|
discoverResultSize += nodeList.getLength();
|
324
|
} else {
|
325
|
resultTotal = discoverResultSize;
|
326
|
pagination = discoverResultSize;
|
327
|
}
|
328
|
resumptionInt = pagination;
|
329
|
resumptionStr = Integer.toString(resumptionInt);
|
330
|
break;
|
331
|
|
332
|
case "deep-cursor": // size of result items unknown, iterate over items (for supporting deep cursor in solr)
|
333
|
// isn't relevant -- if (resultSizeValue < 2) {throw new CollectorServiceException("Mode: deep-cursor, Param 'resultSizeValue' is less than 2");}
|
334
|
// if ( (this.resultOffsetParam.length()) < 2) {throw new CollectorServiceException("Mode: deep-cursor, Param 'resultOffsetParam' is less than 2");}
|
335
|
|
336
|
resumptionStr = xprResumptionPath.evaluate(resultNode);
|
337
|
queryParams.replace("cursor=*", "");
|
338
|
queryParams.replace("&&", "&");
|
339
|
|
340
|
resumptionStr += "&" + resultOffsetParam + "=" + Integer.toString(resumptionInt);
|
341
|
|
342
|
log.debug("downloadPage().deep-cursor: resumptionStr=" + resumptionStr + " ; queryParams=" + queryParams);
|
343
|
|
344
|
break;
|
345
|
|
346
|
default: // otherwise: abort
|
347
|
// resultTotal = resumptionInt;
|
348
|
break;
|
349
|
}
|
350
|
|
351
|
} catch (Exception e) {
|
352
|
log.error(e);
|
353
|
throw new IllegalStateException("collection failed: " + e.getMessage());
|
354
|
}
|
355
|
|
356
|
try {
|
357
|
if (resultTotal == -1) {
|
358
|
resultTotal = Integer.parseInt(xprResultTotalPath.evaluate(resultNode));
|
359
|
if (resumptionType.toLowerCase().equals("page")) { resultTotal += 1; } // to correct the upper bound
|
360
|
log.info("resultTotal was -1 is now: " + resultTotal);
|
361
|
}
|
362
|
} catch(Exception e) {
|
363
|
log.error(e);
|
364
|
throw new IllegalStateException("downloadPage resultTotal couldn't parse: " + e.getMessage());
|
365
|
}
|
366
|
log.info("resultTotal: " + resultTotal);
|
367
|
log.info("resInt: " + resumptionInt);
|
368
|
if (resumptionInt <= resultTotal) {
|
369
|
nextQuery = baseUrl + "?" + queryParams + querySize + "&" + resumptionParam + "=" + resumptionStr + queryFormat;
|
370
|
} else
|
371
|
nextQuery = "";
|
372
|
|
373
|
log.debug("nextQueryUrl: " + nextQuery);
|
374
|
return nextQuery;
|
375
|
|
376
|
|
377
|
}
|
378
|
|
379
|
/**
|
380
|
* convert in JSON-KeyName 'whitespace(s)' to '_' and '/' to '_', '(' and ')' to ''
|
381
|
* check W3C XML syntax: https://www.w3.org/TR/2006/REC-xml11-20060816/#sec-starttags for valid tag names
|
382
|
* and work-around for the JSON to XML converting of org.json.XML-package.
|
383
|
*
|
384
|
* known bugs: doesn't prevent "key name":" ["sexy name",": penari","erotic dance"],
|
385
|
*
|
386
|
* @param jsonInput
|
387
|
* @return convertedJsonKeynameOutput
|
388
|
*/
|
389
|
private String syntaxConvertJsonKeyNamens(String jsonInput) {
|
390
|
|
391
|
log.trace("before convertJsonKeyNames: " + jsonInput);
|
392
|
// pre-clean json - rid spaces of element names (misinterpreted as elements with attributes in xml)
|
393
|
// replace ' 's in JSON Namens with '_'
|
394
|
while (jsonInput.matches(".*\"([^\"]*)\\s+([^\"]*)\":.*")) {
|
395
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)\\s+([^\"]*)\":", "\"$1_$2\":");
|
396
|
}
|
397
|
|
398
|
// replace forward-slash (sign '/' ) in JSON Names with '_'
|
399
|
while (jsonInput.matches(".*\"([^\"]*)/([^\"]*)\":.*")) {
|
400
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)/([^\"]*)\":", "\"$1_$2\":");
|
401
|
}
|
402
|
|
403
|
// replace '(' in JSON Names with ''
|
404
|
while (jsonInput.matches(".*\"([^\"]*)[(]([^\"]*)\":.*")) {
|
405
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)[(]([^\"]*)\":", "\"$1$2\":");
|
406
|
}
|
407
|
|
408
|
// replace ')' in JSON Names with ''
|
409
|
while (jsonInput.matches(".*\"([^\"]*)[)]([^\"]*)\":.*")) {
|
410
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)[)]([^\"]*)\":", "\"$1$2\":");
|
411
|
}
|
412
|
|
413
|
// add prefix of startNumbers in JSON Keynames with 'n_'
|
414
|
while (jsonInput.matches(".*\"([^\"][0-9])([^\"]*)\":.*")) {
|
415
|
jsonInput = jsonInput.replaceAll("\"([^\"][0-9])([^\"]*)\":", "\"n_$1$2\":");
|
416
|
}
|
417
|
// add prefix of only numbers in JSON Keynames with 'm_'
|
418
|
while (jsonInput.matches(".*\"([0-9]+)\":.*")) {
|
419
|
jsonInput = jsonInput.replaceAll("\"([0-9]+)\":", "\"m_$1\":");
|
420
|
}
|
421
|
|
422
|
// replace ':' between number like '2018-08-28T11:05:00Z' in JSON keynames with ''
|
423
|
while (jsonInput.matches(".*\"([^\"]*[0-9]):([0-9][^\"]*)\":.*")) {
|
424
|
jsonInput = jsonInput.replaceAll("\"([^\"]*[0-9]):([0-9][^\"]*)\":", "\"$1$2\":");
|
425
|
}
|
426
|
|
427
|
// replace ',' in JSON Keynames with '.' to prevent , in xml tagnames.
|
428
|
// while (jsonInput.matches(".*\"([^\"]*),([^\"]*)\":.*")) {
|
429
|
// jsonInput = jsonInput.replaceAll("\"([^\"]*),([^\"]*)\":", "\"$1.$2\":");
|
430
|
// }
|
431
|
|
432
|
// replace '=' in JSON Keynames with '-'
|
433
|
while (jsonInput.matches(".*\"([^\"]*)=([^\"]*)\":.*")) {
|
434
|
jsonInput = jsonInput.replaceAll("\"([^\"]*)=([^\"]*)\":", "\"$1-$2\":");
|
435
|
}
|
436
|
|
437
|
log.trace("after syntaxConvertJsonKeyNames: " + jsonInput);
|
438
|
return jsonInput;
|
439
|
}
|
440
|
|
441
|
/**
|
442
|
*
|
443
|
* https://www.w3.org/TR/REC-xml/#charencoding shows character enoding in entities
|
444
|
* *
|
445
|
* @param bufferStr - XML string
|
446
|
* @return
|
447
|
*/
|
448
|
private static String cleanUnwantedJsonCharsInXmlTagnames(String bufferStr) {
|
449
|
|
450
|
while (bufferStr.matches(".*<([^<>].*),(.)>.*")) {
|
451
|
bufferStr = bufferStr.replaceAll("<([^<>.*),(.*)>", "<$1$2>");
|
452
|
}
|
453
|
|
454
|
// replace [#x10-#x1f] with ''
|
455
|
// while (bufferStr.matches(".*[0-9a-f].*")) {
|
456
|
// bufferStr = bufferStr.replaceAll("([0-9a-fA-F])", "");
|
457
|
// }
|
458
|
|
459
|
return bufferStr;
|
460
|
}
|
461
|
|
462
|
private boolean isInteger(String s) {
|
463
|
boolean isValidInteger = false;
|
464
|
try {
|
465
|
Integer.parseInt(s);
|
466
|
|
467
|
// s is a valid integer
|
468
|
|
469
|
isValidInteger = true;
|
470
|
} catch (NumberFormatException ex) {
|
471
|
// s is not an integer
|
472
|
}
|
473
|
|
474
|
return isValidInteger;
|
475
|
}
|
476
|
|
477
|
}
|