|
1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="mainworkflows_preprocessing">
|
|
2
|
|
|
3
|
<parameters>
|
|
4
|
<!-- import project related -->
|
|
5
|
<property>
|
|
6
|
<name>import_database_service_location</name>
|
|
7
|
<value>$UNDEFINED$</value>
|
|
8
|
<description>Database service (not WSDL) location URL</description>
|
|
9
|
</property>
|
|
10
|
<property>
|
|
11
|
<name>import_database_dbname</name>
|
|
12
|
<value>dnet_openaireplus_node0_t</value>
|
|
13
|
<description>database name</description>
|
|
14
|
</property>
|
|
15
|
<!-- import datacite related, export datacite & wos related -->
|
|
16
|
<property>
|
|
17
|
<name>import_mdstore_service_location</name>
|
|
18
|
<value>$UNDEFINED$</value>
|
|
19
|
<description>MDStore service (not WSDL) location URL</description>
|
|
20
|
</property>
|
|
21
|
<property>
|
|
22
|
<name>import_dataset_mdstore_id</name>
|
|
23
|
<value>$UNDEFINED$</value>
|
|
24
|
<description>dataset MDStore identifier</description>
|
|
25
|
</property>
|
|
26
|
<property>
|
|
27
|
<name>import_wos_mdstore_id</name>
|
|
28
|
<value>$UNDEFINED$</value>
|
|
29
|
<description>WoS MDStore identifier</description>
|
|
30
|
</property>
|
|
31
|
<!-- import content related -->
|
|
32
|
<!-- currently disabled, input_document_content property is handled as input holding DocumentContent datastore -->
|
|
33
|
<property>
|
|
34
|
<name>import_content_object_store_location</name>
|
|
35
|
<value>$UNDEFINED$</value>
|
|
36
|
<description>object store service location required for content retrieval</description>
|
|
37
|
</property>
|
|
38
|
<property>
|
|
39
|
<name>import_content_lookup_service_location</name>
|
|
40
|
<value>$UNDEFINED$</value>
|
|
41
|
<description>lookup service location required for content retrieval, finding object store id based on repository id</description>
|
|
42
|
</property>
|
|
43
|
<property>
|
|
44
|
<name>import_content_wos_plaintext_objectstores_csv</name>
|
|
45
|
<value>$UNDEFINED$</value>
|
|
46
|
<description>CSV list of objectstore ids to be approved during WoS plaintext import.</description>
|
|
47
|
</property>
|
|
48
|
<property>
|
|
49
|
<name>import_content_datacite_objectstores_csv</name>
|
|
50
|
<value>$UNDEFINED$</value>
|
|
51
|
<description>CSV list of objectstore ids to be approved during datacite import.</description>
|
|
52
|
</property>
|
|
53
|
<!-- import content mime types -->
|
|
54
|
<property>
|
|
55
|
<name>import_content_mimetypes_pdf_csv</name>
|
|
56
|
<value>pdf,application/pdf</value>
|
|
57
|
<description>pdf mime types</description>
|
|
58
|
</property>
|
|
59
|
<property>
|
|
60
|
<name>import_content_mimetypes_text_csv</name>
|
|
61
|
<value>text,text/plain</value>
|
|
62
|
<description>text mime types</description>
|
|
63
|
</property>
|
|
64
|
<property>
|
|
65
|
<name>import_content_mimetypes_xml_pmc_csv</name>
|
|
66
|
<value>xml</value>
|
|
67
|
<description>xml pmc types</description>
|
|
68
|
</property>
|
|
69
|
<property>
|
|
70
|
<name>import_content_mimetypes_wos_text_csv</name>
|
|
71
|
<value>file::WoS</value>
|
|
72
|
<description>WoS mime types</description>
|
|
73
|
</property>
|
|
74
|
<!-- import timeouts related -->
|
|
75
|
<property>
|
|
76
|
<name>import_resultset_client_read_timeout</name>
|
|
77
|
<value>60000</value>
|
|
78
|
<description>resultset client read timeout</description>
|
|
79
|
</property>
|
|
80
|
<property>
|
|
81
|
<name>import_content_connection_timeout</name>
|
|
82
|
<value>60000</value>
|
|
83
|
<description>import content connection timeout</description>
|
|
84
|
</property>
|
|
85
|
<property>
|
|
86
|
<name>import_content_read_timeout</name>
|
|
87
|
<value>60000</value>
|
|
88
|
<description>import content read timeout</description>
|
|
89
|
</property>
|
|
90
|
<!-- metadata extraction related -->
|
|
91
|
<property>
|
|
92
|
<name>metadataextraction_excluded_ids</name>
|
|
93
|
<value>$UNDEFINED$</value>
|
|
94
|
<description>list of content identifiers excluded from metadataextraction processing</description>
|
|
95
|
</property>
|
|
96
|
<property>
|
|
97
|
<name>metadataextraction_default_cache_location</name>
|
|
98
|
<description>metadata extraction cache location, path pointing to root cache directory holding meta.json file</description>
|
|
99
|
</property>
|
|
100
|
|
|
101
|
<!-- export related -->
|
|
102
|
<property>
|
|
103
|
<name>export_action_hbase_table_name</name>
|
|
104
|
<description>action manager hbase table name</description>
|
|
105
|
</property>
|
|
106
|
<property>
|
|
107
|
<name>export_action_hbase_table_initialize</name>
|
|
108
|
<description>flag indicating input table should be initialized</description>
|
|
109
|
</property>
|
|
110
|
<!-- action set id properties -->
|
|
111
|
<property>
|
|
112
|
<name>export_action_set_id</name>
|
|
113
|
<value>$UNDEFINED$</value>
|
|
114
|
<description>action-set identifier of exported data</description>
|
|
115
|
</property>
|
|
116
|
<property>
|
|
117
|
<name>export_action_set_id_document_referencedProjects</name>
|
|
118
|
<value>$UNDEFINED$</value>
|
|
119
|
<description>document_referencedProjects action-set identifier of exported data</description>
|
|
120
|
</property>
|
|
121
|
<property>
|
|
122
|
<name>export_action_set_id_document_referencedDatasets</name>
|
|
123
|
<value>$UNDEFINED$</value>
|
|
124
|
<description>document_referencedDatasets action-set identifier of exported data</description>
|
|
125
|
</property>
|
|
126
|
<property>
|
|
127
|
<name>export_action_set_id_entity_wos</name>
|
|
128
|
<description>action-set identifier of exported data containing wos entities</description>
|
|
129
|
</property>
|
|
130
|
<property>
|
|
131
|
<name>export_action_set_id_entity_dataset</name>
|
|
132
|
<description>action-set identifier of exported data containing dataset entities</description>
|
|
133
|
</property>
|
|
134
|
<property>
|
|
135
|
<name>export_action_hbase_remote_zookeeper_quorum</name>
|
|
136
|
<value>$UNDEFINED$</value>
|
|
137
|
<description>external hbase zookeeper quorum, set to empty value by default which means data will be exported to local hbase instance</description>
|
|
138
|
</property>
|
|
139
|
<property>
|
|
140
|
<name>export_action_hbase_remote_zookeeper_clientport</name>
|
|
141
|
<value>$UNDEFINED$</value>
|
|
142
|
<description>external hbase zookeeper client port, required only whe zookeeper quorum property is set</description>
|
|
143
|
</property>
|
|
144
|
</parameters>
|
|
145
|
|
|
146
|
<start to="import_forking" />
|
|
147
|
|
|
148
|
<fork name="import_forking">
|
|
149
|
<path start="import_project"/>
|
|
150
|
<path start="import_dataset"/>
|
|
151
|
<path start="content_importer_forking"/>
|
|
152
|
</fork>
|
|
153
|
|
|
154
|
<fork name="content_importer_forking">
|
|
155
|
<path start="wos_url_importer"/>
|
|
156
|
<path start="dataset_url_importer"/>
|
|
157
|
</fork>
|
|
158
|
|
|
159
|
<action name="wos_url_importer">
|
|
160
|
<sub-workflow>
|
|
161
|
<app-path>${wf:appPath()}/import_content_url</app-path>
|
|
162
|
<propagate-configuration/>
|
|
163
|
<configuration>
|
|
164
|
<property>
|
|
165
|
<name>workingDir</name>
|
|
166
|
<value>${workingDir}/wos_url_import/working_dir</value>
|
|
167
|
</property>
|
|
168
|
<property>
|
|
169
|
<name>objectstore_service_location</name>
|
|
170
|
<value>${import_content_object_store_location}</value>
|
|
171
|
</property>
|
|
172
|
<property>
|
|
173
|
<name>approved_objectstores_csv</name>
|
|
174
|
<value>${import_content_wos_plaintext_objectstores_csv}</value>
|
|
175
|
</property>
|
|
176
|
<property>
|
|
177
|
<name>output_dir</name>
|
|
178
|
<value>${workingDir}/wos_url_import/imported</value>
|
|
179
|
</property>
|
|
180
|
<property>
|
|
181
|
<name>mimetypes_text_csv</name>
|
|
182
|
<value>${import_content_mimetypes_wos_text_csv}</value>
|
|
183
|
</property>
|
|
184
|
<property>
|
|
185
|
<name>resultset_client_read_timeout</name>
|
|
186
|
<value>${import_resultset_client_read_timeout}</value>
|
|
187
|
</property>
|
|
188
|
</configuration>
|
|
189
|
</sub-workflow>
|
|
190
|
<ok to="wos_import_plaintext" />
|
|
191
|
<error to="fail" />
|
|
192
|
</action>
|
|
193
|
|
|
194
|
<action name="wos_import_plaintext">
|
|
195
|
<sub-workflow>
|
|
196
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
197
|
<propagate-configuration/>
|
|
198
|
<configuration>
|
|
199
|
<property>
|
|
200
|
<name>workingDir</name>
|
|
201
|
<value>${workingDir}/wos_import_plaintext/working_dir</value>
|
|
202
|
</property>
|
|
203
|
<property>
|
|
204
|
<name>input</name>
|
|
205
|
<value>${workingDir}/wos_url_import/imported/plaintext_url</value>
|
|
206
|
</property>
|
|
207
|
<property>
|
|
208
|
<name>content_connection_timeout</name>
|
|
209
|
<value>${import_content_connection_timeout}</value>
|
|
210
|
</property>
|
|
211
|
<property>
|
|
212
|
<name>content_read_timeout</name>
|
|
213
|
<value>${import_content_read_timeout}</value>
|
|
214
|
</property>
|
|
215
|
<property>
|
|
216
|
<name>output</name>
|
|
217
|
<value>${workingDir}/wos_import_plaintext/imported</value>
|
|
218
|
</property>
|
|
219
|
</configuration>
|
|
220
|
</sub-workflow>
|
|
221
|
<ok to="content_importer_joining" />
|
|
222
|
<error to="fail" />
|
|
223
|
</action>
|
|
224
|
|
|
225
|
<action name="dataset_url_importer">
|
|
226
|
<sub-workflow>
|
|
227
|
<app-path>${wf:appPath()}/import_content_url</app-path>
|
|
228
|
<propagate-configuration/>
|
|
229
|
<configuration>
|
|
230
|
<property>
|
|
231
|
<name>workingDir</name>
|
|
232
|
<value>${workingDir}/dataset_url_import/working_dir</value>
|
|
233
|
</property>
|
|
234
|
<property>
|
|
235
|
<name>objectstore_service_location</name>
|
|
236
|
<value>${import_content_object_store_location}</value>
|
|
237
|
</property>
|
|
238
|
<property>
|
|
239
|
<name>approved_objectstores_csv</name>
|
|
240
|
<value>${import_content_datacite_objectstores_csv}</value>
|
|
241
|
</property>
|
|
242
|
<property>
|
|
243
|
<name>mimetypes_pdf_csv</name>
|
|
244
|
<value>${import_content_mimetypes_pdf_csv}</value>
|
|
245
|
</property>
|
|
246
|
<property>
|
|
247
|
<name>mimetypes_text_csv</name>
|
|
248
|
<value>${import_content_mimetypes_text_csv}</value>
|
|
249
|
</property>
|
|
250
|
<property>
|
|
251
|
<name>mimetypes_xml_pmc_csv</name>
|
|
252
|
<value>${import_content_mimetypes_xml_pmc_csv}</value>
|
|
253
|
</property>
|
|
254
|
<property>
|
|
255
|
<name>resultset_client_read_timeout</name>
|
|
256
|
<value>${import_resultset_client_read_timeout}</value>
|
|
257
|
</property>
|
|
258
|
<property>
|
|
259
|
<name>output_dir</name>
|
|
260
|
<value>${workingDir}/dataset_url_import/imported</value>
|
|
261
|
</property>
|
|
262
|
</configuration>
|
|
263
|
</sub-workflow>
|
|
264
|
<ok to="dataset_import_urlbased_forking" />
|
|
265
|
<error to="fail" />
|
|
266
|
</action>
|
|
267
|
|
|
268
|
<fork name="dataset_import_urlbased_forking">
|
|
269
|
<path start="dataset_import_plaintext"/>
|
|
270
|
<path start="dataset_import_plaintext_pmc"/>
|
|
271
|
<path start="metadataextraction_cached"/>
|
|
272
|
</fork>
|
|
273
|
|
|
274
|
<action name="dataset_import_plaintext">
|
|
275
|
<sub-workflow>
|
|
276
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
277
|
<propagate-configuration/>
|
|
278
|
<configuration>
|
|
279
|
<property>
|
|
280
|
<name>workingDir</name>
|
|
281
|
<value>${workingDir}/dataset_import_plaintext/working_dir</value>
|
|
282
|
</property>
|
|
283
|
<property>
|
|
284
|
<name>input</name>
|
|
285
|
<value>${workingDir}/dataset_url_import/imported/plaintext_url</value>
|
|
286
|
</property>
|
|
287
|
<property>
|
|
288
|
<name>content_connection_timeout</name>
|
|
289
|
<value>${import_content_connection_timeout}</value>
|
|
290
|
</property>
|
|
291
|
<property>
|
|
292
|
<name>content_read_timeout</name>
|
|
293
|
<value>${import_content_read_timeout}</value>
|
|
294
|
</property>
|
|
295
|
<property>
|
|
296
|
<name>output</name>
|
|
297
|
<value>${workingDir}/dataset_import_plaintext/imported</value>
|
|
298
|
</property>
|
|
299
|
</configuration>
|
|
300
|
</sub-workflow>
|
|
301
|
<ok to="dataset_import_urlbased_joining" />
|
|
302
|
<error to="fail" />
|
|
303
|
</action>
|
|
304
|
|
|
305
|
<action name="dataset_import_plaintext_pmc">
|
|
306
|
<sub-workflow>
|
|
307
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
308
|
<propagate-configuration/>
|
|
309
|
<configuration>
|
|
310
|
<property>
|
|
311
|
<name>workingDir</name>
|
|
312
|
<value>${workingDir}/import_plaintext_pmc/working_dir</value>
|
|
313
|
</property>
|
|
314
|
<property>
|
|
315
|
<name>input</name>
|
|
316
|
<value>${workingDir}/dataset_url_import/imported/xml_pmc_url</value>
|
|
317
|
</property>
|
|
318
|
<property>
|
|
319
|
<name>content_connection_timeout</name>
|
|
320
|
<value>${import_content_connection_timeout}</value>
|
|
321
|
</property>
|
|
322
|
<property>
|
|
323
|
<name>content_read_timeout</name>
|
|
324
|
<value>${import_content_read_timeout}</value>
|
|
325
|
</property>
|
|
326
|
<property>
|
|
327
|
<name>output</name>
|
|
328
|
<value>${workingDir}/import_plaintext_pmc/imported</value>
|
|
329
|
</property>
|
|
330
|
</configuration>
|
|
331
|
</sub-workflow>
|
|
332
|
<ok to="import_pmc" />
|
|
333
|
<error to="fail" />
|
|
334
|
</action>
|
|
335
|
|
|
336
|
<action name="import_pmc">
|
|
337
|
<sub-workflow>
|
|
338
|
<app-path>${wf:appPath()}/import_pmc</app-path>
|
|
339
|
<propagate-configuration/>
|
|
340
|
<configuration>
|
|
341
|
<property>
|
|
342
|
<name>workingDir</name>
|
|
343
|
<value>${workingDir}/import_pmc/working_dir</value>
|
|
344
|
</property>
|
|
345
|
<property>
|
|
346
|
<name>input_document_nlm</name>
|
|
347
|
<value>${workingDir}/import_plaintext_pmc/imported</value>
|
|
348
|
</property>
|
|
349
|
<property>
|
|
350
|
<name>output_document_plaintext</name>
|
|
351
|
<value>${workingDir}/import_pmc/imported</value>
|
|
352
|
</property>
|
|
353
|
</configuration>
|
|
354
|
</sub-workflow>
|
|
355
|
<ok to="dataset_import_urlbased_joining" />
|
|
356
|
<error to="fail" />
|
|
357
|
</action>
|
|
358
|
|
|
359
|
<action name="metadataextraction_cached">
|
|
360
|
<sub-workflow>
|
|
361
|
<app-path>${wf:appPath()}/metadataextraction_cached</app-path>
|
|
362
|
<propagate-configuration/>
|
|
363
|
<configuration>
|
|
364
|
<property>
|
|
365
|
<name>workingDir</name>
|
|
366
|
<value>${workingDir}/metadataextraction_cached/working_dir</value>
|
|
367
|
</property>
|
|
368
|
<property>
|
|
369
|
<name>input</name>
|
|
370
|
<value>${workingDir}/dataset_url_import/imported/content_url</value>
|
|
371
|
</property>
|
|
372
|
<property>
|
|
373
|
<name>excluded_ids</name>
|
|
374
|
<value>${metadataextraction_excluded_ids}</value>
|
|
375
|
</property>
|
|
376
|
<property>
|
|
377
|
<name>default_cache_location</name>
|
|
378
|
<value>${metadataextraction_default_cache_location}</value>
|
|
379
|
</property>
|
|
380
|
<property>
|
|
381
|
<name>mapred_max_split_size</name>
|
|
382
|
<value>10000</value>
|
|
383
|
</property>
|
|
384
|
<property>
|
|
385
|
<name>content_connection_timeout</name>
|
|
386
|
<value>${import_content_connection_timeout}</value>
|
|
387
|
</property>
|
|
388
|
<property>
|
|
389
|
<name>content_read_timeout</name>
|
|
390
|
<value>${import_content_read_timeout}</value>
|
|
391
|
</property>
|
|
392
|
<property>
|
|
393
|
<name>output_root</name>
|
|
394
|
<value>${workingDir}/metadataextraction_cached/out</value>
|
|
395
|
</property>
|
|
396
|
</configuration>
|
|
397
|
</sub-workflow>
|
|
398
|
<ok to="dataset_import_urlbased_joining" />
|
|
399
|
<error to="fail" />
|
|
400
|
</action>
|
|
401
|
|
|
402
|
<join name="dataset_import_urlbased_joining" to="transformers_common_union_plaintext_merge_outputs"/>
|
|
403
|
|
|
404
|
<!-- merging document text datastores:
|
|
405
|
1) retrieved directly from objectstore
|
|
406
|
2) generated by metadataextraction
|
|
407
|
3) imported from PMC XMLs -->
|
|
408
|
<action name="transformers_common_union_plaintext_merge_outputs">
|
|
409
|
<sub-workflow>
|
|
410
|
<app-path>${wf:appPath()}/transformers_common_union3</app-path>
|
|
411
|
<propagate-configuration/>
|
|
412
|
<configuration>
|
|
413
|
<property>
|
|
414
|
<name>input_a</name>
|
|
415
|
<value>${workingDir}/dataset_import_plaintext/imported</value>
|
|
416
|
</property>
|
|
417
|
<property>
|
|
418
|
<name>input_b</name>
|
|
419
|
<value>${workingDir}/metadataextraction_cached/out/plaintext</value>
|
|
420
|
</property>
|
|
421
|
<property>
|
|
422
|
<name>input_c</name>
|
|
423
|
<value>${workingDir}/import_pmc/imported</value>
|
|
424
|
</property>
|
|
425
|
<property>
|
|
426
|
<name>output</name>
|
|
427
|
<value>${workingDir}/dataset_plaintext</value>
|
|
428
|
</property>
|
|
429
|
<property>
|
|
430
|
<name>schema</name>
|
|
431
|
<value>eu.dnetlib.iis.metadataextraction.schemas.DocumentText</value>
|
|
432
|
</property>
|
|
433
|
</configuration>
|
|
434
|
</sub-workflow>
|
|
435
|
<ok to="content_importer_joining"/>
|
|
436
|
<error to="fail" />
|
|
437
|
</action>
|
|
438
|
|
|
439
|
<join name="content_importer_joining" to="import_joining"/>
|
|
440
|
|
|
441
|
<action name="import_project">
|
|
442
|
<sub-workflow>
|
|
443
|
<app-path>${wf:appPath()}/import_project</app-path>
|
|
444
|
<propagate-configuration/>
|
|
445
|
<configuration>
|
|
446
|
<property>
|
|
447
|
<name>workingDir</name>
|
|
448
|
<value>${workingDir}/import_project/working_dir</value>
|
|
449
|
</property>
|
|
450
|
<property>
|
|
451
|
<name>input_database_service_location</name>
|
|
452
|
<value>${import_database_service_location}</value>
|
|
453
|
</property>
|
|
454
|
<property>
|
|
455
|
<name>input_database_name</name>
|
|
456
|
<value>${import_database_dbname}</value>
|
|
457
|
</property>
|
|
458
|
<property>
|
|
459
|
<name>resultset_client_read_timeout</name>
|
|
460
|
<value>${import_resultset_client_read_timeout}</value>
|
|
461
|
</property>
|
|
462
|
<property>
|
|
463
|
<name>output_dir</name>
|
|
464
|
<value>${workingDir}/import_project/imported</value>
|
|
465
|
</property>
|
|
466
|
</configuration>
|
|
467
|
</sub-workflow>
|
|
468
|
<ok to="import_joining" />
|
|
469
|
<error to="fail" />
|
|
470
|
</action>
|
|
471
|
|
|
472
|
<action name="import_dataset">
|
|
473
|
<sub-workflow>
|
|
474
|
<app-path>${wf:appPath()}/import_dataset</app-path>
|
|
475
|
<propagate-configuration/>
|
|
476
|
<configuration>
|
|
477
|
<property>
|
|
478
|
<name>workingDir</name>
|
|
479
|
<value>${workingDir}/import_dataset/working_dir</value>
|
|
480
|
</property>
|
|
481
|
<property>
|
|
482
|
<name>input_mdstore_service_location</name>
|
|
483
|
<value>${import_mdstore_service_location}</value>
|
|
484
|
</property>
|
|
485
|
<property>
|
|
486
|
<name>input_mdstore_id</name>
|
|
487
|
<value>${import_dataset_mdstore_id}</value>
|
|
488
|
</property>
|
|
489
|
<property>
|
|
490
|
<name>resultset_client_read_timeout</name>
|
|
491
|
<value>${import_resultset_client_read_timeout}</value>
|
|
492
|
</property>
|
|
493
|
<property>
|
|
494
|
<name>output_dir</name>
|
|
495
|
<value>${workingDir}/import_dataset/imported</value>
|
|
496
|
</property>
|
|
497
|
</configuration>
|
|
498
|
</sub-workflow>
|
|
499
|
<ok to="import_joining" />
|
|
500
|
<error to="fail" />
|
|
501
|
</action>
|
|
502
|
|
|
503
|
<join name="import_joining" to="referenceextraction_forking"/>
|
|
504
|
|
|
505
|
<fork name="referenceextraction_forking">
|
|
506
|
<path start="referenceextraction_dataset"/>
|
|
507
|
<path start="referenceextraction_project"/>
|
|
508
|
</fork>
|
|
509
|
|
|
510
|
<action name="referenceextraction_dataset">
|
|
511
|
<sub-workflow>
|
|
512
|
<app-path>${wf:appPath()}/referenceextraction_dataset</app-path>
|
|
513
|
<propagate-configuration/>
|
|
514
|
<configuration>
|
|
515
|
<property>
|
|
516
|
<name>workingDir</name>
|
|
517
|
<value>${workingDir}/referenceextraction_dataset/working_dir</value>
|
|
518
|
</property>
|
|
519
|
<property>
|
|
520
|
<name>input_document_text</name>
|
|
521
|
<value>${workingDir}/dataset_plaintext</value>
|
|
522
|
</property>
|
|
523
|
<property>
|
|
524
|
<name>input_dataset</name>
|
|
525
|
<value>${workingDir}/import_dataset/imported</value>
|
|
526
|
</property>
|
|
527
|
<property>
|
|
528
|
<name>output_document_to_dataset</name>
|
|
529
|
<value>${workingDir}/referenceextraction_dataset/document_datasets</value>
|
|
530
|
</property>
|
|
531
|
</configuration>
|
|
532
|
</sub-workflow>
|
|
533
|
<ok to="referenceextraction_joining"/>
|
|
534
|
<error to="fail" />
|
|
535
|
</action>
|
|
536
|
|
|
537
|
<action name="referenceextraction_project">
|
|
538
|
<sub-workflow>
|
|
539
|
<app-path>${wf:appPath()}/referenceextraction_project</app-path>
|
|
540
|
<propagate-configuration/>
|
|
541
|
<configuration>
|
|
542
|
<property>
|
|
543
|
<name>workingDir</name>
|
|
544
|
<value>${workingDir}/referenceextraction_project/working_dir</value>
|
|
545
|
</property>
|
|
546
|
<property>
|
|
547
|
<name>input_document_text</name>
|
|
548
|
<value>${workingDir}/wos_import_plaintext/imported</value>
|
|
549
|
</property>
|
|
550
|
<property>
|
|
551
|
<name>input_project</name>
|
|
552
|
<value>${workingDir}/import_project/imported</value>
|
|
553
|
</property>
|
|
554
|
<property>
|
|
555
|
<name>output_document_to_project</name>
|
|
556
|
<value>${workingDir}/referenceextraction_project/document_projects</value>
|
|
557
|
</property>
|
|
558
|
</configuration>
|
|
559
|
</sub-workflow>
|
|
560
|
<ok to="referenceextraction_joining"/>
|
|
561
|
<error to="fail" />
|
|
562
|
</action>
|
|
563
|
|
|
564
|
<join name="referenceextraction_joining" to="transformers_export_document_producer"/>
|
|
565
|
|
|
566
|
<!-- this node is required due to the PIG limitation
|
|
567
|
disallowing empty directories as input avro storages -->
|
|
568
|
<action name="transformers_export_document_producer">
|
|
569
|
<java>
|
|
570
|
<job-tracker>${jobTracker}</job-tracker>
|
|
571
|
<name-node>${nameNode}</name-node>
|
|
572
|
<!-- The data generated by this node is deleted in this section -->
|
|
573
|
<prepare>
|
|
574
|
<delete path="${nameNode}${workingDir}/producer" />
|
|
575
|
<mkdir path="${nameNode}${workingDir}/producer" />
|
|
576
|
</prepare>
|
|
577
|
<configuration>
|
|
578
|
<property>
|
|
579
|
<name>mapred.job.queue.name</name>
|
|
580
|
<value>${queueName}</value>
|
|
581
|
</property>
|
|
582
|
</configuration>
|
|
583
|
<!-- This is simple wrapper for the Java code -->
|
|
584
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
585
|
<!-- The business Java code that gets to be executed -->
|
|
586
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
587
|
<!-- Specification of the output ports -->
|
|
588
|
<arg>-C{extracted_document_metadata,
|
|
589
|
eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadata,
|
|
590
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
591
|
<arg>-C{citation,
|
|
592
|
eu.dnetlib.iis.citationmatching.schemas.Citation,
|
|
593
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
594
|
<arg>-C{document_to_project,
|
|
595
|
eu.dnetlib.iis.referenceextraction.project.schemas.DocumentToProject,
|
|
596
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
597
|
<arg>-C{document_to_dataset,
|
|
598
|
eu.dnetlib.iis.referenceextraction.dataset.schemas.DocumentToDataSet,
|
|
599
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
600
|
<arg>-C{document_to_research_initiative,
|
|
601
|
eu.dnetlib.iis.referenceextraction.researchinitiative.schemas.DocumentToConceptId,
|
|
602
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
603
|
<arg>-C{document_to_document_clusters,
|
|
604
|
eu.dnetlib.iis.documentsclustering.schemas.DocumentToDocumentClusters,
|
|
605
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
606
|
<arg>-C{document_to_document_classes,
|
|
607
|
eu.dnetlib.iis.documentsclassification.schemas.DocumentToDocumentClasses,
|
|
608
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
609
|
<arg>-C{document_to_document_statistics,
|
|
610
|
eu.dnetlib.iis.statistics.schemas.DocumentToDocumentStatistics,
|
|
611
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
612
|
<arg>-C{document_with_website_usage_similarities,
|
|
613
|
eu.dnetlib.iis.websiteusage.schemas.DocumentsWithWebsiteUsageSimilarities,
|
|
614
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
615
|
<!-- this one is required by dataset id generator, it is impossible to check
|
|
616
|
existing dataset records therefore generating empty datastore -->
|
|
617
|
<arg>-C{dataset_existing_id,
|
|
618
|
eu.dnetlib.iis.importer.schemas.DocumentId,
|
|
619
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
620
|
<!-- this one is required by WoS document exporter, it is impossible to check
|
|
621
|
existing WoS publications therefore generating empty datastore -->
|
|
622
|
<arg>-C{document_existing_meta,
|
|
623
|
eu.dnetlib.iis.importer.schemas.DocumentMetadata,
|
|
624
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
625
|
<!-- All input and output ports have to be bound to paths in HDFS, working
|
|
626
|
directory has to be specified as well -->
|
|
627
|
<arg>-SworkingDir=${workingDir}/producer/working_dir</arg>
|
|
628
|
<arg>-Oextracted_document_metadata=${workingDir}/producer/extracted_document_metadata</arg>
|
|
629
|
<arg>-Ocitation=${workingDir}/producer/citation</arg>
|
|
630
|
<arg>-Odocument_to_project=${workingDir}/producer/document_to_project</arg>
|
|
631
|
<arg>-Odocument_to_dataset=${workingDir}/producer/document_to_dataset</arg>
|
|
632
|
<arg>-Odocument_to_research_initiative=${workingDir}/producer/document_to_research_initiative</arg>
|
|
633
|
<arg>-Odocument_to_document_clusters=${workingDir}/producer/document_to_document_clusters</arg>
|
|
634
|
<arg>-Odocument_to_document_classes=${workingDir}/producer/document_to_document_classes</arg>
|
|
635
|
<arg>-Odocument_to_document_statistics=${workingDir}/producer/document_to_document_statistics</arg>
|
|
636
|
<arg>-Odocument_with_website_usage_similarities=${workingDir}/producer/document_with_website_usage_similarities</arg>
|
|
637
|
<arg>-Odataset_existing_id=${workingDir}/producer/dataset_existing_id</arg>
|
|
638
|
<arg>-Odocument_existing_meta=${workingDir}/producer/document_existing_meta</arg>
|
|
639
|
</java>
|
|
640
|
<ok to="transformers_export_document"/>
|
|
641
|
<error to="fail"/>
|
|
642
|
</action>
|
|
643
|
|
|
644
|
<action name="transformers_export_document">
|
|
645
|
<sub-workflow>
|
|
646
|
<app-path>${wf:appPath()}/transformers_export_document</app-path>
|
|
647
|
<propagate-configuration/>
|
|
648
|
<configuration>
|
|
649
|
<property>
|
|
650
|
<name>workingDir</name>
|
|
651
|
<value>${workingDir}/transformers_export_document/working_dir</value>
|
|
652
|
</property>
|
|
653
|
<!-- real input datastores -->
|
|
654
|
<property>
|
|
655
|
<name>input_document_to_project</name>
|
|
656
|
<value>${workingDir}/referenceextraction_project/document_projects</value>
|
|
657
|
</property>
|
|
658
|
<property>
|
|
659
|
<name>input_document_to_dataset</name>
|
|
660
|
<value>${workingDir}/referenceextraction_dataset/document_datasets</value>
|
|
661
|
</property>
|
|
662
|
<!-- fake input datastores -->
|
|
663
|
<property>
|
|
664
|
<name>input_document_to_research_initiative</name>
|
|
665
|
<value>${workingDir}/producer/document_to_research_initiative</value>
|
|
666
|
</property>
|
|
667
|
<property>
|
|
668
|
<name>input_extracted_document_metadata</name>
|
|
669
|
<value>${workingDir}/producer/extracted_document_metadata</value>
|
|
670
|
</property>
|
|
671
|
<property>
|
|
672
|
<name>input_citation</name>
|
|
673
|
<value>${workingDir}/producer/citation</value>
|
|
674
|
</property>
|
|
675
|
<property>
|
|
676
|
<name>input_document_to_document_statistics</name>
|
|
677
|
<value>${workingDir}/producer/document_to_document_statistics</value>
|
|
678
|
</property>
|
|
679
|
<property>
|
|
680
|
<name>input_document_to_document_clusters</name>
|
|
681
|
<value>${workingDir}/producer/document_to_document_clusters</value>
|
|
682
|
</property>
|
|
683
|
<property>
|
|
684
|
<name>input_document_to_document_classes</name>
|
|
685
|
<value>${workingDir}/producer/document_to_document_classes</value>
|
|
686
|
</property>
|
|
687
|
<property>
|
|
688
|
<name>input_document_with_website_usage_similarities</name>
|
|
689
|
<value>${workingDir}/producer/document_with_website_usage_similarities</value>
|
|
690
|
</property>
|
|
691
|
<!-- output datastore -->
|
|
692
|
<property>
|
|
693
|
<name>output_document_with_inferenced_data</name>
|
|
694
|
<value>${workingDir}/exported/document_with_inferenced_data</value>
|
|
695
|
</property>
|
|
696
|
</configuration>
|
|
697
|
</sub-workflow>
|
|
698
|
<ok to="export_actionmanager"/>
|
|
699
|
<error to="fail" />
|
|
700
|
</action>
|
|
701
|
|
|
702
|
<action name="export_actionmanager">
|
|
703
|
<sub-workflow>
|
|
704
|
<app-path>${wf:appPath()}/export_actionmanager</app-path>
|
|
705
|
<propagate-configuration/>
|
|
706
|
<configuration>
|
|
707
|
<property>
|
|
708
|
<name>workingDir</name>
|
|
709
|
<value>${workingDir}/export_actionmanager/working_dir</value>
|
|
710
|
</property>
|
|
711
|
<property>
|
|
712
|
<name>input_document</name>
|
|
713
|
<value>${workingDir}/exported/document_with_inferenced_data</value>
|
|
714
|
</property>
|
|
715
|
<property>
|
|
716
|
<name>action_hbase_table_name</name>
|
|
717
|
<value>${export_action_hbase_table_name}</value>
|
|
718
|
</property>
|
|
719
|
<property>
|
|
720
|
<name>action_hbase_table_initialize</name>
|
|
721
|
<value>${export_action_hbase_table_initialize}</value>
|
|
722
|
</property>
|
|
723
|
<property>
|
|
724
|
<name>action_set_id</name>
|
|
725
|
<value>${export_action_set_id}</value>
|
|
726
|
</property>
|
|
727
|
<property>
|
|
728
|
<name>action_set_id_document_referencedProjects</name>
|
|
729
|
<value>${export_action_set_id_document_referencedProjects}</value>
|
|
730
|
</property>
|
|
731
|
<property>
|
|
732
|
<name>action_set_id_document_referencedDatasets</name>
|
|
733
|
<value>${export_action_set_id_document_referencedDatasets}</value>
|
|
734
|
</property>
|
|
735
|
<property>
|
|
736
|
<name>action_hbase_remote_zookeeper_quorum</name>
|
|
737
|
<value>${export_action_hbase_remote_zookeeper_quorum}</value>
|
|
738
|
</property>
|
|
739
|
<property>
|
|
740
|
<name>action_hbase_remote_zookeeper_clientport</name>
|
|
741
|
<value>${export_action_hbase_remote_zookeeper_clientport}</value>
|
|
742
|
</property>
|
|
743
|
</configuration>
|
|
744
|
</sub-workflow>
|
|
745
|
<ok to="export_entities_forking"/>
|
|
746
|
<error to="fail" />
|
|
747
|
</action>
|
|
748
|
|
|
749
|
<fork name="export_entities_forking">
|
|
750
|
<path start="transformers_export_identifier_referenceddatasets"/>
|
|
751
|
<path start="transformers_export_identifier_document_to_project"/>
|
|
752
|
</fork>
|
|
753
|
|
|
754
|
<!-- dataset entities export section -->
|
|
755
|
<action name="transformers_export_identifier_referenceddatasets">
|
|
756
|
<sub-workflow>
|
|
757
|
<app-path>${wf:appPath()}/transformers_export_identifier_referenceddatasets</app-path>
|
|
758
|
<propagate-configuration/>
|
|
759
|
<configuration>
|
|
760
|
<property>
|
|
761
|
<name>workingDir</name>
|
|
762
|
<value>${workingDir}/export_identifier_referenceddatasets/working_dir</value>
|
|
763
|
</property>
|
|
764
|
<property>
|
|
765
|
<name>input_document_id</name>
|
|
766
|
<!-- this is generated empty datastore -->
|
|
767
|
<value>${workingDir}/producer/dataset_existing_id</value>
|
|
768
|
</property>
|
|
769
|
<property>
|
|
770
|
<name>input_document_with_inferenced_data</name>
|
|
771
|
<value>${workingDir}/exported/document_with_inferenced_data</value>
|
|
772
|
</property>
|
|
773
|
<property>
|
|
774
|
<name>output_identifier</name>
|
|
775
|
<value>${workingDir}/identifier/datasets</value>
|
|
776
|
</property>
|
|
777
|
</configuration>
|
|
778
|
</sub-workflow>
|
|
779
|
<ok to="exporter-dataset-entities"/>
|
|
780
|
<error to="fail" />
|
|
781
|
</action>
|
|
782
|
|
|
783
|
<action name="exporter-dataset-entities">
|
|
784
|
<java>
|
|
785
|
<job-tracker>${jobTracker}</job-tracker>
|
|
786
|
<name-node>${nameNode}</name-node>
|
|
787
|
<prepare>
|
|
788
|
<delete path="${nameNode}${workingDir}/export-dataset-entities" />
|
|
789
|
<mkdir path="${nameNode}${workingDir}/export-dataset-entities" />
|
|
790
|
</prepare>
|
|
791
|
<configuration>
|
|
792
|
<property>
|
|
793
|
<name>mapred.job.queue.name</name>
|
|
794
|
<value>${queueName}</value>
|
|
795
|
</property>
|
|
796
|
</configuration>
|
|
797
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
798
|
<arg>eu.dnetlib.iis.export.actionmanager.entity.dataset.DatasetExporterProcess</arg>
|
|
799
|
<arg>-SworkingDir=${workingDir}</arg>
|
|
800
|
<arg>-Iinput=${workingDir}/identifier/datasets</arg>
|
|
801
|
|
|
802
|
<arg>-Pimport.datacite.mdstore.service.location=${import_mdstore_service_location}</arg>
|
|
803
|
<arg>-Pimport.datacite.mdstore.id=${import_dataset_mdstore_id}</arg>
|
|
804
|
<arg>-Pexport.action.setid=${export_action_set_id_entity_dataset}</arg>
|
|
805
|
<arg>-Pexport.action.hbase.table.name=${export_action_hbase_table_name}</arg>
|
|
806
|
<arg>-Pexport.action.hbase.remote.zookeeper.quorum=${export_action_hbase_remote_zookeeper_quorum}</arg>
|
|
807
|
<arg>-Pexport.action.hbase.remote.zookeeper.clientport=${export_action_hbase_remote_zookeeper_clientport}</arg>
|
|
808
|
<arg>-Pexport.action.hbase.table.initialize=${export_action_hbase_table_initialize}</arg>
|
|
809
|
</java>
|
|
810
|
<ok to="export_entities_joining" />
|
|
811
|
<error to="fail" />
|
|
812
|
</action>
|
|
813
|
<!-- end of dataset entities export section -->
|
|
814
|
|
|
815
|
<!-- WoS entities export section -->
|
|
816
|
<action name="transformers_export_identifier_document_to_project">
|
|
817
|
<sub-workflow>
|
|
818
|
<app-path>${wf:appPath()}/transformers_export_identifier_document_to_project</app-path>
|
|
819
|
<propagate-configuration/>
|
|
820
|
<configuration>
|
|
821
|
<property>
|
|
822
|
<name>workingDir</name>
|
|
823
|
<value>${workingDir}/export_identifier_documents/working_dir</value>
|
|
824
|
</property>
|
|
825
|
<property>
|
|
826
|
<name>input_document_to_project</name>
|
|
827
|
<value>${workingDir}/referenceextraction_project/document_projects</value>
|
|
828
|
</property>
|
|
829
|
<property>
|
|
830
|
<name>output_identifier</name>
|
|
831
|
<value>${workingDir}/identifier/documents</value>
|
|
832
|
</property>
|
|
833
|
</configuration>
|
|
834
|
</sub-workflow>
|
|
835
|
<ok to="exporter-document-entities"/>
|
|
836
|
<error to="fail" />
|
|
837
|
</action>
|
|
838
|
|
|
839
|
<action name="exporter-document-entities">
|
|
840
|
<java>
|
|
841
|
<job-tracker>${jobTracker}</job-tracker>
|
|
842
|
<name-node>${nameNode}</name-node>
|
|
843
|
<prepare>
|
|
844
|
<delete path="${nameNode}${workingDir}/export-document-entities" />
|
|
845
|
<mkdir path="${nameNode}${workingDir}/export-document-entities" />
|
|
846
|
</prepare>
|
|
847
|
<configuration>
|
|
848
|
<property>
|
|
849
|
<name>mapred.job.queue.name</name>
|
|
850
|
<value>${queueName}</value>
|
|
851
|
</property>
|
|
852
|
</configuration>
|
|
853
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
854
|
<arg>eu.dnetlib.iis.export.actionmanager.entity.document.DocumentExporterProcess</arg>
|
|
855
|
<arg>-SworkingDir=${workingDir}</arg>
|
|
856
|
<arg>-Iinput=${workingDir}/identifier/documents</arg>
|
|
857
|
|
|
858
|
<arg>-Pimport.document.mdstore.service.location=${import_mdstore_service_location}</arg>
|
|
859
|
<arg>-Pimport.document.mdstore.id=${import_wos_mdstore_id}</arg>
|
|
860
|
<arg>-Pexport.action.setid=${export_action_set_id_entity_wos}</arg>
|
|
861
|
<arg>-Pexport.action.hbase.table.name=${export_action_hbase_table_name}</arg>
|
|
862
|
<arg>-Pexport.action.hbase.remote.zookeeper.quorum=${export_action_hbase_remote_zookeeper_quorum}</arg>
|
|
863
|
<arg>-Pexport.action.hbase.remote.zookeeper.clientport=${export_action_hbase_remote_zookeeper_clientport}</arg>
|
|
864
|
<arg>-Pexport.action.hbase.table.initialize=${export_action_hbase_table_initialize}</arg>
|
|
865
|
</java>
|
|
866
|
<ok to="export_entities_joining" />
|
|
867
|
<error to="fail" />
|
|
868
|
</action>
|
|
869
|
<!-- end of WoS entities export section -->
|
|
870
|
|
|
871
|
<join name="export_entities_joining" to="end"/>
|
|
872
|
|
|
873
|
<kill name="fail">
|
|
874
|
<message>Unfortunately, the process failed -- error message:
|
|
875
|
[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
876
|
</kill>
|
|
877
|
<end name="end" />
|
|
878
|
</workflow-app>
|
