|
1
|
<workflow-app xmlns="uri:oozie:workflow:0.4" name="mainworkflows_common_import">
|
|
2
|
<parameters>
|
|
3
|
<!-- importing modes -->
|
|
4
|
<property>
|
|
5
|
<name>active_import_metadata</name>
|
|
6
|
<value>false</value>
|
|
7
|
<description>flag indicating HBase metadata import should be enabled, when set to false db-based project import will be performed</description>
|
|
8
|
</property>
|
|
9
|
<property>
|
|
10
|
<name>active_import_dataset</name>
|
|
11
|
<value>false</value>
|
|
12
|
<description>flag indicating dataset import should be enabled</description>
|
|
13
|
</property>
|
|
14
|
<property>
|
|
15
|
<name>active_ingest_pmc</name>
|
|
16
|
<value>false</value>
|
|
17
|
<description>flag indicating pmc metadata and citations ingestions should be performed</description>
|
|
18
|
</property>
|
|
19
|
<property>
|
|
20
|
<name>active_import_concept</name>
|
|
21
|
<value>false</value>
|
|
22
|
<description>flag indicating concept import should be executed</description>
|
|
23
|
</property>
|
|
24
|
<property>
|
|
25
|
<name>match_content_with_metadata</name>
|
|
26
|
<value>true</value>
|
|
27
|
<description>flag indicating contents should be filtered and their identifiers should be deduplicated against metadata entries retrieved from InformationSpace.
|
|
28
|
This way only contents having metadata representation will be processed.
|
|
29
|
To be disabled when processing new contents which metadata is not available in hbase or when original identifiers should be preserved (contents will not be filtered as well).</description>
|
|
30
|
</property>
|
|
31
|
<!-- import metadata related -->
|
|
32
|
<property>
|
|
33
|
<name>hbase_input_table</name>
|
|
34
|
<value>$UNDEFINED$</value>
|
|
35
|
<description>HBase input table holding InformationSpace, available on local cluster</description>
|
|
36
|
</property>
|
|
37
|
<property>
|
|
38
|
<name>hbase_approved_datasources_csv</name>
|
|
39
|
<value>$UNDEFINED$</value>
|
|
40
|
<description>CSV list of datasource ids to be approved during import. Applied on result and person entities.</description>
|
|
41
|
</property>
|
|
42
|
<property>
|
|
43
|
<name>inference_provenance_blacklist</name>
|
|
44
|
<value>iis::.*</value>
|
|
45
|
<description>list of blacklisted inference provenance which sould not be taken into account by importer, skipped when set to $UNDEFINED$</description>
|
|
46
|
</property>
|
|
47
|
<property>
|
|
48
|
<name>trust_level_threshold</name>
|
|
49
|
<value>$UNDEFINED$</value>
|
|
50
|
<description>trust level threshold represented as float value, ignored when set to $UNDEFINED$ value</description>
|
|
51
|
</property>
|
|
52
|
<property>
|
|
53
|
<name>merge_body_with_updates</name>
|
|
54
|
<value>false</value>
|
|
55
|
<description>flag indicating Oaf objects strored in body qualifier should be merged with Oaf objects stored in update qualifier</description>
|
|
56
|
</property>
|
|
57
|
<!-- import concepts related -->
|
|
58
|
<property>
|
|
59
|
<name>islookup_service_location</name>
|
|
60
|
<value>$UNDEFINED$</value>
|
|
61
|
<description>IS Lookup service location, required only when active_import_concept is set to true</description>
|
|
62
|
</property>
|
|
63
|
<property>
|
|
64
|
<name>project_concepts_context_ids_csv</name>
|
|
65
|
<value>$UNDEFINED$</value>
|
|
66
|
<description>comma separated list of concepts context identifiers to be picked by ISLookup, required only when active_import_concept is set to true</description>
|
|
67
|
</property>
|
|
68
|
<!-- import project related -->
|
|
69
|
<!-- will be used when active_import_metadata=false -->
|
|
70
|
<property>
|
|
71
|
<name>database_service_location</name>
|
|
72
|
<value>$UNDEFINED$</value>
|
|
73
|
<description>Database service (not WSDL) location URL</description>
|
|
74
|
</property>
|
|
75
|
<property>
|
|
76
|
<name>database_dbname</name>
|
|
77
|
<value>dnet_openaireplus_node6_t</value>
|
|
78
|
<description>database name</description>
|
|
79
|
</property>
|
|
80
|
<!-- import datacite related -->
|
|
81
|
<property>
|
|
82
|
<name>mdstore_service_location</name>
|
|
83
|
<value>$UNDEFINED$</value>
|
|
84
|
<description>MDStore service (not WSDL) location URL</description>
|
|
85
|
</property>
|
|
86
|
<property>
|
|
87
|
<name>dataset_mdstore_ids_csv</name>
|
|
88
|
<value>$UNDEFINED$</value>
|
|
89
|
<description>MDStore identifier</description>
|
|
90
|
</property>
|
|
91
|
<!-- import content related -->
|
|
92
|
<property>
|
|
93
|
<name>objectstore_service_location</name>
|
|
94
|
<value>$UNDEFINED$</value>
|
|
95
|
<description>object store service location required for content retrieval</description>
|
|
96
|
</property>
|
|
97
|
<property>
|
|
98
|
<name>approved_objectstores_csv</name>
|
|
99
|
<value>$UNDEFINED$</value>
|
|
100
|
<description>CSV list of object stores identifiers to be processed</description>
|
|
101
|
</property>
|
|
102
|
<property>
|
|
103
|
<name>mimetypes_pdf</name>
|
|
104
|
<description>pdf mime types</description>
|
|
105
|
</property>
|
|
106
|
<property>
|
|
107
|
<name>mimetypes_text</name>
|
|
108
|
<description>text mime types</description>
|
|
109
|
</property>
|
|
110
|
<property>
|
|
111
|
<name>mimetypes_html</name>
|
|
112
|
<description>html mime types</description>
|
|
113
|
</property>
|
|
114
|
<property>
|
|
115
|
<name>mimetypes_xml_pmc</name>
|
|
116
|
<description>xml pmc types</description>
|
|
117
|
</property>
|
|
118
|
<property>
|
|
119
|
<name>mimetypes_wos</name>
|
|
120
|
<description>wos types</description>
|
|
121
|
</property>
|
|
122
|
<!-- import timeouts related -->
|
|
123
|
<property>
|
|
124
|
<name>resultset_client_read_timeout</name>
|
|
125
|
<value>60000</value>
|
|
126
|
<description>resultset client read timeout</description>
|
|
127
|
</property>
|
|
128
|
<property>
|
|
129
|
<name>content_connection_timeout</name>
|
|
130
|
<value>60000</value>
|
|
131
|
<description>import content connection timeout</description>
|
|
132
|
</property>
|
|
133
|
<property>
|
|
134
|
<name>content_read_timeout</name>
|
|
135
|
<value>60000</value>
|
|
136
|
<description>import content read timeout</description>
|
|
137
|
</property>
|
|
138
|
<!-- metadata extraction related -->
|
|
139
|
<property>
|
|
140
|
<name>metadataextraction_excluded_checksums</name>
|
|
141
|
<value>$UNDEFINED$</value>
|
|
142
|
<description>list of content checksums excluded from metadataextraction processing</description>
|
|
143
|
</property>
|
|
144
|
<property>
|
|
145
|
<name>metadataextraction_max_file_size_mb</name>
|
|
146
|
<value>$UNDEFINED$</value>
|
|
147
|
<description>maximum allowed file size in Megabytes</description>
|
|
148
|
</property>
|
|
149
|
<property>
|
|
150
|
<name>metadataextraction_default_cache_location</name>
|
|
151
|
<value>/cache/metadataextraction</value>
|
|
152
|
<description>metadata extraction HDFS cache location</description>
|
|
153
|
</property>
|
|
154
|
<!-- metadata import output subdirectory names -->
|
|
155
|
<property>
|
|
156
|
<name>metadataimport_output_name_document_meta</name>
|
|
157
|
<value>docmeta</value>
|
|
158
|
<description>metadata import docmeta output subdirectory name</description>
|
|
159
|
</property>
|
|
160
|
<property>
|
|
161
|
<name>metadataimport_output_name_document_project</name>
|
|
162
|
<value>docproject</value>
|
|
163
|
<description>metadata import document to project relation subdirectory name</description>
|
|
164
|
</property>
|
|
165
|
<property>
|
|
166
|
<name>metadataimport_output_name_project</name>
|
|
167
|
<value>project</value>
|
|
168
|
<description>metadata import project output subdirectory name</description>
|
|
169
|
</property>
|
|
170
|
<property>
|
|
171
|
<name>metadataimport_output_name_person</name>
|
|
172
|
<value>person</value>
|
|
173
|
<description>metadata import person output subdirectory name</description>
|
|
174
|
</property>
|
|
175
|
<property>
|
|
176
|
<name>metadataimport_output_name_dedup_mapping</name>
|
|
177
|
<value>dedupmapping</value>
|
|
178
|
<description>metadata import deduplication mapping output subdirectory name</description>
|
|
179
|
</property>
|
|
180
|
<!-- output parameters -->
|
|
181
|
<property>
|
|
182
|
<name>output_extracted_document_metadata</name>
|
|
183
|
<description>extracted document metadata output directory</description>
|
|
184
|
</property>
|
|
185
|
<property>
|
|
186
|
<name>output_metadataimport_root</name>
|
|
187
|
<value>$UNDEFINED$</value>
|
|
188
|
<description>metadata importer output root directory, required when ${active_import_metadata}=true</description>
|
|
189
|
</property>
|
|
190
|
<property>
|
|
191
|
<name>output_citation_pmc</name>
|
|
192
|
<description>PMC citation output directory, required when ${active_import_metadata}=true</description>
|
|
193
|
</property>
|
|
194
|
<property>
|
|
195
|
<name>output_dataset</name>
|
|
196
|
<description>dataset importer output directory holding dataset metadata, required when ${active_import_dataset}=true</description>
|
|
197
|
</property>
|
|
198
|
<property>
|
|
199
|
<name>output_dataset_to_mdstore</name>
|
|
200
|
<description>dataset importer output directory holding dataset to mdstore mappings, required when ${active_import_dataset}=true</description>
|
|
201
|
</property>
|
|
202
|
<property>
|
|
203
|
<name>output_document_text</name>
|
|
204
|
<description>text import output directory. merged from three different sources</description>
|
|
205
|
</property>
|
|
206
|
<property>
|
|
207
|
<name>output_wos</name>
|
|
208
|
<description>wos import output directory</description>
|
|
209
|
</property>
|
|
210
|
<property>
|
|
211
|
<name>output_project_concept</name>
|
|
212
|
<description>project concepts output directory</description>
|
|
213
|
</property>
|
|
214
|
<property>
|
|
215
|
<name>output_faults</name>
|
|
216
|
<description>processing faults output directory</description>
|
|
217
|
</property>
|
|
218
|
<property>
|
|
219
|
<name>remove_sideproducts</name>
|
|
220
|
<value>true</value>
|
|
221
|
<description>flag indicating whole workingDir will be erased.
|
|
222
|
Notice: do not provide any output directory location pointing to workingDir subdirectory!</description>
|
|
223
|
</property>
|
|
224
|
</parameters>
|
|
225
|
|
|
226
|
<global>
|
|
227
|
<job-tracker>${jobTracker}</job-tracker>
|
|
228
|
<name-node>${nameNode}</name-node>
|
|
229
|
<configuration>
|
|
230
|
<property>
|
|
231
|
<name>mapred.job.queue.name</name>
|
|
232
|
<value>${queueName}</value>
|
|
233
|
</property>
|
|
234
|
</configuration>
|
|
235
|
</global>
|
|
236
|
|
|
237
|
<start to="import_forking" />
|
|
238
|
|
|
239
|
<fork name="import_forking">
|
|
240
|
<path start="decision-import_concept"/>
|
|
241
|
<path start="decision-metadata_importer"/>
|
|
242
|
<path start="decision-import_dataset"/>
|
|
243
|
</fork>
|
|
244
|
|
|
245
|
<decision name="decision-import_concept">
|
|
246
|
<switch>
|
|
247
|
<case to="import_concept">${active_import_concept eq "true"}</case>
|
|
248
|
<default to="skip-import_concept"/>
|
|
249
|
</switch>
|
|
250
|
</decision>
|
|
251
|
|
|
252
|
<action name="import_concept">
|
|
253
|
<sub-workflow>
|
|
254
|
<app-path>${wf:appPath()}/import_concept</app-path>
|
|
255
|
<propagate-configuration/>
|
|
256
|
<configuration>
|
|
257
|
<property>
|
|
258
|
<name>workingDir</name>
|
|
259
|
<value>${workingDir}/import_concept/working_dir</value>
|
|
260
|
</property>
|
|
261
|
<property>
|
|
262
|
<name>islookup_service_location</name>
|
|
263
|
<value>${islookup_service_location}</value>
|
|
264
|
</property>
|
|
265
|
<property>
|
|
266
|
<name>context_ids_csv</name>
|
|
267
|
<value>${project_concepts_context_ids_csv}</value>
|
|
268
|
</property>
|
|
269
|
<property>
|
|
270
|
<name>output</name>
|
|
271
|
<value>${output_project_concept}</value>
|
|
272
|
</property>
|
|
273
|
</configuration>
|
|
274
|
</sub-workflow>
|
|
275
|
<ok to="import_joining" />
|
|
276
|
<error to="fail" />
|
|
277
|
</action>
|
|
278
|
|
|
279
|
<action name="skip-import_concept">
|
|
280
|
<java>
|
|
281
|
<prepare>
|
|
282
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
283
|
<delete path="${nameNode}${workingDir}/import_concept" />
|
|
284
|
<delete path="${nameNode}${output_project_concept}" />
|
|
285
|
<mkdir path="${nameNode}${workingDir}/import_concept" />
|
|
286
|
</prepare>
|
|
287
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
288
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
289
|
<arg>-C{concept,
|
|
290
|
eu.dnetlib.iis.importer.schemas.Concept,
|
|
291
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
292
|
<arg>-Oconcept=${output_project_concept}</arg>
|
|
293
|
</java>
|
|
294
|
<ok to="import_joining"/>
|
|
295
|
<error to="fail"/>
|
|
296
|
</action>
|
|
297
|
|
|
298
|
<decision name="decision-metadata_importer">
|
|
299
|
<switch>
|
|
300
|
<case to="metadata_importer">${active_import_metadata eq "true"}</case>
|
|
301
|
<default to="import_project"/>
|
|
302
|
</switch>
|
|
303
|
</decision>
|
|
304
|
|
|
305
|
<action name="metadata_importer">
|
|
306
|
<sub-workflow>
|
|
307
|
<app-path>${wf:appPath()}/import_mapred</app-path>
|
|
308
|
<propagate-configuration/>
|
|
309
|
<configuration>
|
|
310
|
<property>
|
|
311
|
<name>workingDir</name>
|
|
312
|
<value>${workingDir}/import/working_dir</value>
|
|
313
|
</property>
|
|
314
|
<property>
|
|
315
|
<name>approved_datasources_csv</name>
|
|
316
|
<value>${hbase_approved_datasources_csv}</value>
|
|
317
|
</property>
|
|
318
|
<property>
|
|
319
|
<name>output</name>
|
|
320
|
<value>${output_metadataimport_root}</value>
|
|
321
|
</property>
|
|
322
|
<!-- subdirectory names -->
|
|
323
|
<property>
|
|
324
|
<name>output_name_document_meta</name>
|
|
325
|
<value>${metadataimport_output_name_document_meta}</value>
|
|
326
|
</property>
|
|
327
|
<property>
|
|
328
|
<name>output_name_document_project</name>
|
|
329
|
<value>${metadataimport_output_name_document_project}</value>
|
|
330
|
</property>
|
|
331
|
<property>
|
|
332
|
<name>output_name_project</name>
|
|
333
|
<value>${metadataimport_output_name_project}</value>
|
|
334
|
</property>
|
|
335
|
<property>
|
|
336
|
<name>output_name_person</name>
|
|
337
|
<value>${metadataimport_output_name_person}</value>
|
|
338
|
</property>
|
|
339
|
<property>
|
|
340
|
<name>output_name_dedup_mapping</name>
|
|
341
|
<value>${metadataimport_output_name_dedup_mapping}</value>
|
|
342
|
</property>
|
|
343
|
<!-- all the other properties are autmatically propagated-->
|
|
344
|
</configuration>
|
|
345
|
</sub-workflow>
|
|
346
|
<ok to="transformers-idextractor" />
|
|
347
|
<error to="fail" />
|
|
348
|
</action>
|
|
349
|
|
|
350
|
<action name="transformers-idextractor">
|
|
351
|
<sub-workflow>
|
|
352
|
<app-path>${wf:appPath()}/transformers_idextractor</app-path>
|
|
353
|
<propagate-configuration/>
|
|
354
|
<configuration>
|
|
355
|
<property>
|
|
356
|
<name>workingDir</name>
|
|
357
|
<value>${workingDir}/transformers_idextractor/working_dir</value>
|
|
358
|
</property>
|
|
359
|
<property>
|
|
360
|
<name>input_document_metadata</name>
|
|
361
|
<value>${output_metadataimport_root}/${metadataimport_output_name_document_meta}</value>
|
|
362
|
</property>
|
|
363
|
<property>
|
|
364
|
<name>output_identifier</name>
|
|
365
|
<value>${workingDir}/transformers_idextractor/output</value>
|
|
366
|
</property>
|
|
367
|
</configuration>
|
|
368
|
</sub-workflow>
|
|
369
|
<ok to="decision-import_content_url"/>
|
|
370
|
<error to="fail"/>
|
|
371
|
</action>
|
|
372
|
|
|
373
|
<action name="import_project">
|
|
374
|
<sub-workflow>
|
|
375
|
<app-path>${wf:appPath()}/import_project</app-path>
|
|
376
|
<propagate-configuration/>
|
|
377
|
<configuration>
|
|
378
|
<property>
|
|
379
|
<name>workingDir</name>
|
|
380
|
<value>${workingDir}/import_project/working_dir</value>
|
|
381
|
</property>
|
|
382
|
<property>
|
|
383
|
<name>output</name>
|
|
384
|
<value>${output_metadataimport_root}/${metadataimport_output_name_project}</value>
|
|
385
|
</property>
|
|
386
|
<!-- all the other properties are autmatically propagated-->
|
|
387
|
</configuration>
|
|
388
|
</sub-workflow>
|
|
389
|
<ok to="decision-import_content_url" />
|
|
390
|
<error to="fail" />
|
|
391
|
</action>
|
|
392
|
|
|
393
|
<decision name="decision-import_dataset">
|
|
394
|
<switch>
|
|
395
|
<case to="import_dataset">${active_import_dataset eq "true"}</case>
|
|
396
|
<default to="skip-import_dataset"/>
|
|
397
|
</switch>
|
|
398
|
</decision>
|
|
399
|
|
|
400
|
<action name="import_dataset">
|
|
401
|
<sub-workflow>
|
|
402
|
<app-path>${wf:appPath()}/import_dataset</app-path>
|
|
403
|
<propagate-configuration/>
|
|
404
|
<configuration>
|
|
405
|
<property>
|
|
406
|
<name>workingDir</name>
|
|
407
|
<value>${workingDir}/import_dataset/working_dir</value>
|
|
408
|
</property>
|
|
409
|
<property>
|
|
410
|
<name>mdstore_ids_csv</name>
|
|
411
|
<value>${dataset_mdstore_ids_csv}</value>
|
|
412
|
</property>
|
|
413
|
<!-- all the other properties are autmatically propagated-->
|
|
414
|
</configuration>
|
|
415
|
</sub-workflow>
|
|
416
|
<ok to="import_joining" />
|
|
417
|
<error to="fail" />
|
|
418
|
</action>
|
|
419
|
|
|
420
|
<action name="skip-import_dataset">
|
|
421
|
<java>
|
|
422
|
<prepare>
|
|
423
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
424
|
<delete path="${nameNode}${workingDir}/import_dataset" />
|
|
425
|
<delete path="${nameNode}${output_dataset}" />
|
|
426
|
<delete path="${nameNode}${output_dataset_to_mdstore}" />
|
|
427
|
<mkdir path="${nameNode}${workingDir}/import_dataset" />
|
|
428
|
</prepare>
|
|
429
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
430
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
431
|
<arg>-C{dataset,
|
|
432
|
eu.dnetlib.iis.importer.schemas.DataSetReference,
|
|
433
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
434
|
<arg>-C{dataset_to_mdstore,
|
|
435
|
eu.dnetlib.iis.importer.schemas.DocumentToMDStore,
|
|
436
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
437
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
438
|
<arg>-Odataset=${output_dataset}</arg>
|
|
439
|
<arg>-Odataset_to_mdstore=${output_dataset_to_mdstore}</arg>
|
|
440
|
</java>
|
|
441
|
<ok to="import_joining"/>
|
|
442
|
<error to="fail"/>
|
|
443
|
</action>
|
|
444
|
|
|
445
|
<decision name="decision-import_content_url">
|
|
446
|
<switch>
|
|
447
|
<case to="skip-import_content_url">${objectstore_service_location eq "$UNDEFINED$"}</case>
|
|
448
|
<default to="input_id_mapping-path-setter"/>
|
|
449
|
</switch>
|
|
450
|
</decision>
|
|
451
|
|
|
452
|
<action name='input_id_mapping-path-setter'>
|
|
453
|
<java>
|
|
454
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
455
|
<arg>eu.dnetlib.iis.common.oozie.property.ConditionalPropertySetter</arg>
|
|
456
|
<arg>-Pcondition=${active_import_metadata eq "true" and match_content_with_metadata eq "true"}</arg>
|
|
457
|
<arg>-PinCaseOfTrue=${output_metadataimport_root}/${metadataimport_output_name_dedup_mapping}</arg>
|
|
458
|
<arg>-PelseCase=$UNDEFINED$</arg>
|
|
459
|
<capture-output />
|
|
460
|
</java>
|
|
461
|
<ok to="input_id-path-setter" />
|
|
462
|
<error to="fail" />
|
|
463
|
</action>
|
|
464
|
|
|
465
|
<action name='input_id-path-setter'>
|
|
466
|
<java>
|
|
467
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
468
|
<arg>eu.dnetlib.iis.common.oozie.property.ConditionalPropertySetter</arg>
|
|
469
|
<arg>-Pcondition=${active_import_metadata eq "true" and match_content_with_metadata eq "true"}</arg>
|
|
470
|
<arg>-PinCaseOfTrue=${workingDir}/transformers_idextractor/output</arg>
|
|
471
|
<arg>-PelseCase=$UNDEFINED$</arg>
|
|
472
|
<capture-output />
|
|
473
|
</java>
|
|
474
|
<ok to="import_content_url" />
|
|
475
|
<error to="fail" />
|
|
476
|
</action>
|
|
477
|
|
|
478
|
<action name="import_content_url">
|
|
479
|
<sub-workflow>
|
|
480
|
<app-path>${wf:appPath()}/import_content_url</app-path>
|
|
481
|
<propagate-configuration/>
|
|
482
|
<configuration>
|
|
483
|
<property>
|
|
484
|
<name>workingDir</name>
|
|
485
|
<value>${workingDir}/import_content_url/working_dir</value>
|
|
486
|
</property>
|
|
487
|
<property>
|
|
488
|
<name>input_id</name>
|
|
489
|
<value>${wf:actionData('input_id-path-setter')['result']}</value>
|
|
490
|
</property>
|
|
491
|
<property>
|
|
492
|
<name>input_id_mapping</name>
|
|
493
|
<value>${wf:actionData('input_id_mapping-path-setter')['result']}</value>
|
|
494
|
</property>
|
|
495
|
<property>
|
|
496
|
<name>output_root</name>
|
|
497
|
<value>${workingDir}/import_content_url/imported</value>
|
|
498
|
</property>
|
|
499
|
<property>
|
|
500
|
<name>output_name_pdf</name>
|
|
501
|
<value>pdf</value>
|
|
502
|
</property>
|
|
503
|
<property>
|
|
504
|
<name>output_name_text</name>
|
|
505
|
<value>text</value>
|
|
506
|
</property>
|
|
507
|
<property>
|
|
508
|
<name>output_name_html</name>
|
|
509
|
<value>html</value>
|
|
510
|
</property>
|
|
511
|
<property>
|
|
512
|
<name>output_name_xml_pmc</name>
|
|
513
|
<value>xmlpmc</value>
|
|
514
|
</property>
|
|
515
|
<property>
|
|
516
|
<name>output_name_wos</name>
|
|
517
|
<value>wos</value>
|
|
518
|
</property>
|
|
519
|
<!-- all the other properties are autmatically propagated-->
|
|
520
|
</configuration>
|
|
521
|
</sub-workflow>
|
|
522
|
<ok to="import_urlbased_forking" />
|
|
523
|
<error to="fail" />
|
|
524
|
</action>
|
|
525
|
|
|
526
|
<action name="skip-import_content_url">
|
|
527
|
<java>
|
|
528
|
<prepare>
|
|
529
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
530
|
<delete path="${nameNode}${workingDir}/import_content_url" />
|
|
531
|
<delete path="${nameNode}${output_document_text}" />
|
|
532
|
<delete path="${nameNode}${output_extracted_document_metadata}" />
|
|
533
|
<mkdir path="${nameNode}${workingDir}/import_content_url" />
|
|
534
|
</prepare>
|
|
535
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
536
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
537
|
<arg>-C{document_text,
|
|
538
|
eu.dnetlib.iis.metadataextraction.schemas.DocumentText,
|
|
539
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
540
|
<arg>-C{extracted_document_metadata,
|
|
541
|
eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadata,
|
|
542
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
543
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
544
|
<arg>-Odocument_text=${output_document_text}</arg>
|
|
545
|
<arg>-Oextracted_document_metadata=${output_extracted_document_metadata}</arg>
|
|
546
|
</java>
|
|
547
|
<ok to="import_joining"/>
|
|
548
|
<error to="fail"/>
|
|
549
|
</action>
|
|
550
|
|
|
551
|
<fork name="import_urlbased_forking">
|
|
552
|
<path start="import_plaintext"/>
|
|
553
|
<path start="import_wos"/>
|
|
554
|
<path start="import_plaintext_pmc"/>
|
|
555
|
<path start="import_html"/>
|
|
556
|
<path start="decision-metadata_extractor_use_cache"/>
|
|
557
|
</fork>
|
|
558
|
|
|
559
|
<action name="import_plaintext">
|
|
560
|
<sub-workflow>
|
|
561
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
562
|
<propagate-configuration/>
|
|
563
|
<configuration>
|
|
564
|
<property>
|
|
565
|
<name>workingDir</name>
|
|
566
|
<value>${workingDir}/import_plaintext/working_dir</value>
|
|
567
|
</property>
|
|
568
|
<property>
|
|
569
|
<name>input</name>
|
|
570
|
<value>${workingDir}/import_content_url/imported/text</value>
|
|
571
|
</property>
|
|
572
|
<property>
|
|
573
|
<name>output</name>
|
|
574
|
<value>${workingDir}/import_plaintext/imported</value>
|
|
575
|
</property>
|
|
576
|
<!-- all the other properties are autmatically propagated-->
|
|
577
|
</configuration>
|
|
578
|
</sub-workflow>
|
|
579
|
<ok to="import_urlbased_joining" />
|
|
580
|
<error to="fail" />
|
|
581
|
</action>
|
|
582
|
|
|
583
|
<action name="import_wos">
|
|
584
|
<sub-workflow>
|
|
585
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
586
|
<propagate-configuration/>
|
|
587
|
<configuration>
|
|
588
|
<property>
|
|
589
|
<name>workingDir</name>
|
|
590
|
<value>${workingDir}/import_wos/working_dir</value>
|
|
591
|
</property>
|
|
592
|
<property>
|
|
593
|
<name>input</name>
|
|
594
|
<value>${workingDir}/import_content_url/imported/wos</value>
|
|
595
|
</property>
|
|
596
|
<property>
|
|
597
|
<name>output</name>
|
|
598
|
<value>${output_wos}</value>
|
|
599
|
</property>
|
|
600
|
<!-- all the other properties are autmatically propagated-->
|
|
601
|
</configuration>
|
|
602
|
</sub-workflow>
|
|
603
|
<ok to="import_urlbased_joining" />
|
|
604
|
<error to="fail" />
|
|
605
|
</action>
|
|
606
|
|
|
607
|
<action name="import_plaintext_pmc">
|
|
608
|
<sub-workflow>
|
|
609
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
610
|
<propagate-configuration/>
|
|
611
|
<configuration>
|
|
612
|
<property>
|
|
613
|
<name>workingDir</name>
|
|
614
|
<value>${workingDir}/import_plaintext_pmc/working_dir</value>
|
|
615
|
</property>
|
|
616
|
<property>
|
|
617
|
<name>input</name>
|
|
618
|
<value>${workingDir}/import_content_url/imported/xmlpmc</value>
|
|
619
|
</property>
|
|
620
|
<property>
|
|
621
|
<name>output</name>
|
|
622
|
<value>${workingDir}/import_plaintext_pmc/imported</value>
|
|
623
|
</property>
|
|
624
|
<!-- all the other properties are autmatically propagated-->
|
|
625
|
</configuration>
|
|
626
|
</sub-workflow>
|
|
627
|
<ok to="ingest_pmc_forking" />
|
|
628
|
<error to="fail" />
|
|
629
|
</action>
|
|
630
|
|
|
631
|
<fork name="ingest_pmc_forking">
|
|
632
|
<path start="ingest_pmc_plaintext"/>
|
|
633
|
<path start="decision-ingest_pmc_metadata"/>
|
|
634
|
</fork>
|
|
635
|
|
|
636
|
<action name="ingest_pmc_plaintext">
|
|
637
|
<sub-workflow>
|
|
638
|
<app-path>${wf:appPath()}/ingest_pmc_plaintext</app-path>
|
|
639
|
<propagate-configuration/>
|
|
640
|
<configuration>
|
|
641
|
<property>
|
|
642
|
<name>workingDir</name>
|
|
643
|
<value> /working_dir</value>
|
|
644
|
</property>
|
|
645
|
<property>
|
|
646
|
<name>input_document_nlm</name>
|
|
647
|
<value>${workingDir}/import_plaintext_pmc/imported</value>
|
|
648
|
</property>
|
|
649
|
<property>
|
|
650
|
<name>output_document_plaintext</name>
|
|
651
|
<value>${workingDir}/ingest_pmc_plaintext/imported</value>
|
|
652
|
</property>
|
|
653
|
</configuration>
|
|
654
|
</sub-workflow>
|
|
655
|
<ok to="ingest_pmc_joining" />
|
|
656
|
<error to="fail" />
|
|
657
|
</action>
|
|
658
|
|
|
659
|
<decision name="decision-ingest_pmc_metadata">
|
|
660
|
<switch>
|
|
661
|
<!-- define ingest_pmc_metadata_joining here when introducing pmc metadata ingestion -->
|
|
662
|
<case to="transformers-doitooaid">${active_import_metadata eq "true" and active_ingest_pmc eq "true"}</case>
|
|
663
|
<default to="skip-ingest_pmc_citations"/>
|
|
664
|
</switch>
|
|
665
|
</decision>
|
|
666
|
|
|
667
|
<action name="transformers-doitooaid">
|
|
668
|
<sub-workflow>
|
|
669
|
<app-path>${wf:appPath()}/transformers_externalidtooaid</app-path>
|
|
670
|
<propagate-configuration/>
|
|
671
|
<configuration>
|
|
672
|
<property>
|
|
673
|
<name>workingDir</name>
|
|
674
|
<value>${workingDir}/transformers_doitooaid/working_dir</value>
|
|
675
|
</property>
|
|
676
|
<property>
|
|
677
|
<name>input_document_metadata</name>
|
|
678
|
<value>${output_metadataimport_root}/${metadataimport_output_name_document_meta}</value>
|
|
679
|
</property>
|
|
680
|
<property>
|
|
681
|
<name>external_id_type</name>
|
|
682
|
<value>doi</value>
|
|
683
|
</property>
|
|
684
|
<property>
|
|
685
|
<name>output</name>
|
|
686
|
<value>${workingDir}/transformers_doitooaid/out</value>
|
|
687
|
</property>
|
|
688
|
</configuration>
|
|
689
|
</sub-workflow>
|
|
690
|
<ok to="ingest_pmc_metadata"/>
|
|
691
|
<error to="fail"/>
|
|
692
|
</action>
|
|
693
|
|
|
694
|
<action name="ingest_pmc_metadata">
|
|
695
|
<sub-workflow>
|
|
696
|
<app-path>${wf:appPath()}/ingest_pmc_metadata</app-path>
|
|
697
|
<propagate-configuration/>
|
|
698
|
<configuration>
|
|
699
|
<property>
|
|
700
|
<name>workingDir</name>
|
|
701
|
<value>${workingDir}/ingest_pmc_metadata/working_dir</value>
|
|
702
|
</property>
|
|
703
|
<property>
|
|
704
|
<name>input</name>
|
|
705
|
<value>${workingDir}/import_plaintext_pmc/imported</value>
|
|
706
|
</property>
|
|
707
|
<property>
|
|
708
|
<name>output</name>
|
|
709
|
<value>${workingDir}/ingest_pmc_metadata/out</value>
|
|
710
|
</property>
|
|
711
|
</configuration>
|
|
712
|
</sub-workflow>
|
|
713
|
<ok to="collapse_pmc_metadata"/>
|
|
714
|
<error to="fail"/>
|
|
715
|
</action>
|
|
716
|
|
|
717
|
<action name="collapse_pmc_metadata">
|
|
718
|
<sub-workflow>
|
|
719
|
<app-path>${wf:appPath()}/basic_collapser</app-path>
|
|
720
|
<propagate-configuration/>
|
|
721
|
<configuration>
|
|
722
|
<property>
|
|
723
|
<name>workingDir</name>
|
|
724
|
<value>${workingDir}/collapse_pmc_metadata/working_dir</value>
|
|
725
|
</property>
|
|
726
|
<property>
|
|
727
|
<name>input</name>
|
|
728
|
<value>${workingDir}/ingest_pmc_metadata/out</value>
|
|
729
|
</property>
|
|
730
|
<property>
|
|
731
|
<name>output</name>
|
|
732
|
<value>${workingDir}/collapse_pmc_metadata/out</value>
|
|
733
|
</property>
|
|
734
|
<property>
|
|
735
|
<name>schema</name>
|
|
736
|
<value>eu.dnetlib.iis.ingest.pmc.metadata.schemas.ExtractedDocumentMetadata</value>
|
|
737
|
</property>
|
|
738
|
<property>
|
|
739
|
<name>blocking_field</name>
|
|
740
|
<value>id</value>
|
|
741
|
</property>
|
|
742
|
<property>
|
|
743
|
<name>significant_fields</name>
|
|
744
|
<value>journal,references,pages</value>
|
|
745
|
</property>
|
|
746
|
</configuration>
|
|
747
|
</sub-workflow>
|
|
748
|
<ok to="transformers_ingest_pmc_metadata"/>
|
|
749
|
<error to="fail"/>
|
|
750
|
</action>
|
|
751
|
|
|
752
|
<action name="transformers_ingest_pmc_metadata">
|
|
753
|
<sub-workflow>
|
|
754
|
<app-path>${wf:appPath()}/transformers_ingest_pmc_metadata</app-path>
|
|
755
|
<propagate-configuration/>
|
|
756
|
<configuration>
|
|
757
|
<property>
|
|
758
|
<name>workingDir</name>
|
|
759
|
<value>${workingDir}/transformers_ingest_pmc_metadata/working_dir</value>
|
|
760
|
</property>
|
|
761
|
<property>
|
|
762
|
<name>input</name>
|
|
763
|
<value>${workingDir}/collapse_pmc_metadata/out</value>
|
|
764
|
</property>
|
|
765
|
<property>
|
|
766
|
<name>output</name>
|
|
767
|
<value>${workingDir}/transformers_ingest_pmc_metadata/out</value>
|
|
768
|
</property>
|
|
769
|
</configuration>
|
|
770
|
</sub-workflow>
|
|
771
|
<ok to="ingest_pmc_idmapping_pmidtooaid"/>
|
|
772
|
<error to="fail"/>
|
|
773
|
</action>
|
|
774
|
|
|
775
|
<action name="ingest_pmc_idmapping_pmidtooaid">
|
|
776
|
<sub-workflow>
|
|
777
|
<app-path>${wf:appPath()}/ingest_pmc_idmapping_pmidtooaid</app-path>
|
|
778
|
<propagate-configuration/>
|
|
779
|
<configuration>
|
|
780
|
<property>
|
|
781
|
<name>workingDir</name>
|
|
782
|
<value>${workingDir}/ingest_pmc_idmapping_pmidtooaid/working_dir</value>
|
|
783
|
</property>
|
|
784
|
<property>
|
|
785
|
<name>input</name>
|
|
786
|
<value>${workingDir}/collapse_pmc_metadata/out</value>
|
|
787
|
</property>
|
|
788
|
<property>
|
|
789
|
<name>output</name>
|
|
790
|
<value>${workingDir}/ingest_pmc_idmapping_pmidtooaid/out</value>
|
|
791
|
</property>
|
|
792
|
</configuration>
|
|
793
|
</sub-workflow>
|
|
794
|
<ok to="ingest_pmc_citations"/>
|
|
795
|
<error to="fail"/>
|
|
796
|
</action>
|
|
797
|
|
|
798
|
<action name="ingest_pmc_citations">
|
|
799
|
<sub-workflow>
|
|
800
|
<app-path>${wf:appPath()}/ingest_pmc_citations</app-path>
|
|
801
|
<propagate-configuration/>
|
|
802
|
<configuration>
|
|
803
|
<property>
|
|
804
|
<name>workingDir</name>
|
|
805
|
<value>${workingDir}/ingest_pmc_citations/working_dir</value>
|
|
806
|
</property>
|
|
807
|
<property>
|
|
808
|
<name>input_extracted_document_metadata</name>
|
|
809
|
<value>${workingDir}/collapse_pmc_metadata/out</value>
|
|
810
|
</property>
|
|
811
|
<property>
|
|
812
|
<name>input_dedup_map</name>
|
|
813
|
<value>${output_metadataimport_root}/${metadataimport_output_name_dedup_mapping}</value>
|
|
814
|
</property>
|
|
815
|
<property>
|
|
816
|
<name>input_doi_to_oaid</name>
|
|
817
|
<value>${workingDir}/transformers_doitooaid/out</value>
|
|
818
|
</property>
|
|
819
|
<property>
|
|
820
|
<name>input_pmid_to_oaid</name>
|
|
821
|
<value>${workingDir}/ingest_pmc_idmapping_pmidtooaid/out</value>
|
|
822
|
</property>
|
|
823
|
<property>
|
|
824
|
<name>output_citation</name>
|
|
825
|
<value>${output_citation_pmc}</value>
|
|
826
|
</property>
|
|
827
|
</configuration>
|
|
828
|
</sub-workflow>
|
|
829
|
<ok to="ingest_pmc_joining" />
|
|
830
|
<error to="fail" />
|
|
831
|
</action>
|
|
832
|
|
|
833
|
<action name="skip-ingest_pmc_citations">
|
|
834
|
<java>
|
|
835
|
<prepare>
|
|
836
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
837
|
<delete path="${nameNode}${workingDir}/ingest_pmc_citations" />
|
|
838
|
<delete path="${nameNode}${workingDir}/transformers_ingest_pmc_metadata"/>
|
|
839
|
<delete path="${nameNode}${output_citation_pmc}"/>
|
|
840
|
<mkdir path="${nameNode}${workingDir}/ingest_pmc_citations" />
|
|
841
|
<mkdir path="${nameNode}${workingDir}/transformers_ingest_pmc_metadata" />
|
|
842
|
<mkdir path="${nameNode}${output_citation_pmc}"/>
|
|
843
|
</prepare>
|
|
844
|
<main-class>eu.dnetlib.iis.core.java.ProcessWrapper</main-class>
|
|
845
|
<arg>eu.dnetlib.iis.core.java.jsonworkflownodes.Producer</arg>
|
|
846
|
<arg>-C{citation_pmc,
|
|
847
|
eu.dnetlib.iis.ingest.pmc.citations.schemas.Citation,
|
|
848
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
849
|
<arg>-C{metadata_pmc,
|
|
850
|
eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadata,
|
|
851
|
eu/dnetlib/iis/mainworkflows/data/empty.json}</arg>
|
|
852
|
<!-- notice: directory have to aligned with skipped action output -->
|
|
853
|
<arg>-Ocitation_pmc=${output_citation_pmc}</arg>
|
|
854
|
<arg>-Ometadata_pmc=${workingDir}/transformers_ingest_pmc_metadata/out</arg>
|
|
855
|
</java>
|
|
856
|
<ok to="ingest_pmc_joining"/>
|
|
857
|
<error to="fail"/>
|
|
858
|
</action>
|
|
859
|
|
|
860
|
<join name="ingest_pmc_joining" to="import_urlbased_joining"/>
|
|
861
|
|
|
862
|
<!-- html import and plaintext ingestion section -->
|
|
863
|
<action name="import_html">
|
|
864
|
<sub-workflow>
|
|
865
|
<app-path>${wf:appPath()}/import_plaintext</app-path>
|
|
866
|
<propagate-configuration/>
|
|
867
|
<configuration>
|
|
868
|
<property>
|
|
869
|
<name>workingDir</name>
|
|
870
|
<value>${workingDir}/import_html/working_dir</value>
|
|
871
|
</property>
|
|
872
|
<property>
|
|
873
|
<name>input</name>
|
|
874
|
<value>${workingDir}/import_content_url/imported/html</value>
|
|
875
|
</property>
|
|
876
|
<property>
|
|
877
|
<name>output</name>
|
|
878
|
<value>${workingDir}/import_html/imported</value>
|
|
879
|
</property>
|
|
880
|
<!-- all the other properties are autmatically propagated-->
|
|
881
|
</configuration>
|
|
882
|
</sub-workflow>
|
|
883
|
<ok to="ingest_html_plaintext" />
|
|
884
|
<error to="fail" />
|
|
885
|
</action>
|
|
886
|
|
|
887
|
|
|
888
|
<action name="ingest_html_plaintext">
|
|
889
|
<sub-workflow>
|
|
890
|
<app-path>${wf:appPath()}/ingest_html_plaintext</app-path>
|
|
891
|
<propagate-configuration/>
|
|
892
|
<configuration>
|
|
893
|
<property>
|
|
894
|
<name>workingDir</name>
|
|
895
|
<value>${workingDir}/ingest_html_plaintext/working_dir</value>
|
|
896
|
</property>
|
|
897
|
<property>
|
|
898
|
<name>input</name>
|
|
899
|
<value>${workingDir}/import_html/imported</value>
|
|
900
|
</property>
|
|
901
|
<property>
|
|
902
|
<name>output</name>
|
|
903
|
<value>${workingDir}/ingest_html_plaintext/imported</value>
|
|
904
|
</property>
|
|
905
|
</configuration>
|
|
906
|
</sub-workflow>
|
|
907
|
<ok to="import_urlbased_joining" />
|
|
908
|
<error to="fail" />
|
|
909
|
</action>
|
|
910
|
|
|
911
|
<!-- metadata extraction section -->
|
|
912
|
<decision name="decision-metadata_extractor_use_cache">
|
|
913
|
<switch>
|
|
914
|
<case to="metadata_extractor">${metadataextraction_default_cache_location eq "$UNDEFINED$"}</case>
|
|
915
|
<default to="metadata_extractor_cached"/>
|
|
916
|
</switch>
|
|
917
|
</decision>
|
|
918
|
|
|
919
|
<action name="metadata_extractor_cached">
|
|
920
|
<sub-workflow>
|
|
921
|
<app-path>${wf:appPath()}/metadataextraction_cached</app-path>
|
|
922
|
<propagate-configuration/>
|
|
923
|
<configuration>
|
|
924
|
<property>
|
|
925
|
<name>workingDir</name>
|
|
926
|
<value>${workingDir}/metadata_extractor/working_dir</value>
|
|
927
|
</property>
|
|
928
|
<property>
|
|
929
|
<name>input</name>
|
|
930
|
<value>${workingDir}/import_content_url/imported/pdf</value>
|
|
931
|
</property>
|
|
932
|
<property>
|
|
933
|
<name>excluded_ids</name>
|
|
934
|
<value>${metadataextraction_excluded_checksums}</value>
|
|
935
|
</property>
|
|
936
|
<property>
|
|
937
|
<name>max_file_size_mb</name>
|
|
938
|
<value>${metadataextraction_max_file_size_mb}</value>
|
|
939
|
</property>
|
|
940
|
<property>
|
|
941
|
<name>default_cache_location</name>
|
|
942
|
<value>${metadataextraction_default_cache_location}</value>
|
|
943
|
</property>
|
|
944
|
<property>
|
|
945
|
<name>output_name_meta</name>
|
|
946
|
<value>meta</value>
|
|
947
|
</property>
|
|
948
|
<property>
|
|
949
|
<name>output_name_plaintext</name>
|
|
950
|
<value>plaintext</value>
|
|
951
|
</property>
|
|
952
|
<property>
|
|
953
|
<name>output_name_fault</name>
|
|
954
|
<value>fault</value>
|
|
955
|
</property>
|
|
956
|
<property>
|
|
957
|
<name>output_root</name>
|
|
958
|
<value>${workingDir}/metadata_extractor/out</value>
|
|
959
|
</property>
|
|
960
|
<!-- all the other properties are autmatically propagated-->
|
|
961
|
</configuration>
|
|
962
|
</sub-workflow>
|
|
963
|
<ok to="import_urlbased_joining"/>
|
|
964
|
<error to="fail" />
|
|
965
|
</action>
|
|
966
|
|
|
967
|
<action name="metadata_extractor">
|
|
968
|
<sub-workflow>
|
|
969
|
<app-path>${wf:appPath()}/metadataextraction</app-path>
|
|
970
|
<propagate-configuration/>
|
|
971
|
<configuration>
|
|
972
|
<property>
|
|
973
|
<name>workingDir</name>
|
|
974
|
<value>${workingDir}/metadata_extractor/working_dir</value>
|
|
975
|
</property>
|
|
976
|
<!-- enabling streaming mode -->
|
|
977
|
<property>
|
|
978
|
<name>processing_mode</name>
|
|
979
|
<value>StreamingMetadataExtractorMapper</value>
|
|
980
|
</property>
|
|
981
|
<property>
|
|
982
|
<name>inputport_classname</name>
|
|
983
|
<value>eu.dnetlib.iis.importer.auxiliary.schemas.DocumentContentUrl</value>
|
|
984
|
</property>
|
|
985
|
<property>
|
|
986
|
<name>input</name>
|
|
987
|
<value>${workingDir}/import_content_url/imported/pdf</value>
|
|
988
|
</property>
|
|
989
|
<property>
|
|
990
|
<name>excluded_ids</name>
|
|
991
|
<value>${metadataextraction_excluded_checksums}</value>
|
|
992
|
</property>
|
|
993
|
<property>
|
|
994
|
<name>max_file_size_mb</name>
|
|
995
|
<value>${metadataextraction_max_file_size_mb}</value>
|
|
996
|
</property>
|
|
997
|
<property>
|
|
998
|
<name>output_name_meta</name>
|
|
999
|
<value>meta</value>
|
|
1000
|
</property>
|
|
1001
|
<property>
|
|
1002
|
<name>output_name_plaintext</name>
|
|
1003
|
<value>plaintext</value>
|
|
1004
|
</property>
|
|
1005
|
<property>
|
|
1006
|
<name>output_name_fault</name>
|
|
1007
|
<value>fault</value>
|
|
1008
|
</property>
|
|
1009
|
<property>
|
|
1010
|
<name>output_root</name>
|
|
1011
|
<value>${workingDir}/metadata_extractor/out</value>
|
|
1012
|
</property>
|
|
1013
|
<!-- all the other properties are autmatically propagated-->
|
|
1014
|
</configuration>
|
|
1015
|
</sub-workflow>
|
|
1016
|
<ok to="import_urlbased_joining"/>
|
|
1017
|
<error to="fail" />
|
|
1018
|
</action>
|
|
1019
|
<!-- end of metadata extraction section -->
|
|
1020
|
|
|
1021
|
<join name="import_urlbased_joining" to="transformers_common_union_document_text"/>
|
|
1022
|
|
|
1023
|
<!-- merging document text datastores:
|
|
1024
|
1) retrieved directly from objectstore
|
|
1025
|
2) generated by metadataextraction
|
|
1026
|
3) ingested from PMC XMLs
|
|
1027
|
3) ingested from HTML
|
|
1028
|
-->
|
|
1029
|
<action name="transformers_common_union_document_text">
|
|
1030
|
<sub-workflow>
|
|
1031
|
<app-path>${wf:appPath()}/transformers_common_union4</app-path>
|
|
1032
|
<propagate-configuration/>
|
|
1033
|
<configuration>
|
|
1034
|
<property>
|
|
1035
|
<name>workingDir</name>
|
|
1036
|
<value>${workingDir}/transformers_common_union_document_text/working_dir</value>
|
|
1037
|
</property>
|
|
1038
|
<property>
|
|
1039
|
<name>input_a</name>
|
|
1040
|
<value>${workingDir}/import_plaintext/imported</value>
|
|
1041
|
</property>
|
|
1042
|
<property>
|
|
1043
|
<name>input_b</name>
|
|
1044
|
<value>${workingDir}/metadata_extractor/out/plaintext</value>
|
|
1045
|
</property>
|
|
1046
|
<property>
|
|
1047
|
<name>input_c</name>
|
|
1048
|
<value>${workingDir}/ingest_pmc_plaintext/imported</value>
|
|
1049
|
</property>
|
|
1050
|
<property>
|
|
1051
|
<name>input_d</name>
|
|
1052
|
<value>${workingDir}/ingest_html_plaintext/imported</value>
|
|
1053
|
</property>
|
|
1054
|
<property>
|
|
1055
|
<name>output</name>
|
|
1056
|
<value>${output_document_text}</value>
|
|
1057
|
</property>
|
|
1058
|
<property>
|
|
1059
|
<name>schema</name>
|
|
1060
|
<value>eu.dnetlib.iis.metadataextraction.schemas.DocumentText</value>
|
|
1061
|
</property>
|
|
1062
|
</configuration>
|
|
1063
|
</sub-workflow>
|
|
1064
|
<ok to="extracted_document_metadata_collapser"/>
|
|
1065
|
<error to="fail" />
|
|
1066
|
</action>
|
|
1067
|
|
|
1068
|
<!-- merging extracted document metadata datastores:
|
|
1069
|
1) extracted from PDF documents
|
|
1070
|
2) ingested from PMC documents
|
|
1071
|
-->
|
|
1072
|
<action name="extracted_document_metadata_collapser">
|
|
1073
|
<sub-workflow>
|
|
1074
|
<app-path>${wf:appPath()}/multiple_input_collapser</app-path>
|
|
1075
|
<propagate-configuration/>
|
|
1076
|
<configuration>
|
|
1077
|
<property>
|
|
1078
|
<name>workingDir</name>
|
|
1079
|
<value>${workingDir}/extracted_document_metadata_collapser/working_dir</value>
|
|
1080
|
</property>
|
|
1081
|
<property>
|
|
1082
|
<name>origin_1</name>
|
|
1083
|
<value>pmc_ingestion</value>
|
|
1084
|
</property>
|
|
1085
|
<property>
|
|
1086
|
<name>input_1</name>
|
|
1087
|
<value>${workingDir}/transformers_ingest_pmc_metadata/out</value>
|
|
1088
|
</property>
|
|
1089
|
<property>
|
|
1090
|
<name>origin_2</name>
|
|
1091
|
<value>cermine</value>
|
|
1092
|
</property>
|
|
1093
|
<property>
|
|
1094
|
<name>input_2</name>
|
|
1095
|
<value>${workingDir}/metadata_extractor/out/meta</value>
|
|
1096
|
</property>
|
|
1097
|
<property>
|
|
1098
|
<name>output</name>
|
|
1099
|
<value>${output_extracted_document_metadata}</value>
|
|
1100
|
</property>
|
|
1101
|
<property>
|
|
1102
|
<name>blocking_field</name>
|
|
1103
|
<value>id</value>
|
|
1104
|
</property>
|
|
1105
|
<property>
|
|
1106
|
<name>schema_input</name>
|
|
1107
|
<value>eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadata</value>
|
|
1108
|
</property>
|
|
1109
|
<property>
|
|
1110
|
<name>schema_input_envelope</name>
|
|
1111
|
<value>eu.dnetlib.iis.metadataextraction.schemas.ExtractedDocumentMetadataEnvelope</value>
|
|
1112
|
</property>
|
|
1113
|
</configuration>
|
|
1114
|
</sub-workflow>
|
|
1115
|
<ok to="import_joining"/>
|
|
1116
|
<error to="fail"/>
|
|
1117
|
</action>
|
|
1118
|
|
|
1119
|
<join name="import_joining" to="init-faults-dir"/>
|
|
1120
|
|
|
1121
|
<action name="init-faults-dir">
|
|
1122
|
<fs>
|
|
1123
|
<delete path="${nameNode}${output_faults}" />
|
|
1124
|
<mkdir path="${nameNode}${output_faults}" />
|
|
1125
|
</fs>
|
|
1126
|
<ok to="preserve-faults"/>
|
|
1127
|
<error to="fail"/>
|
|
1128
|
</action>
|
|
1129
|
|
|
1130
|
<action name="preserve-faults">
|
|
1131
|
<distcp xmlns="uri:oozie:distcp-action:0.1">
|
|
1132
|
<job-tracker>${jobTracker}</job-tracker>
|
|
1133
|
<name-node>${nameNode}</name-node>
|
|
1134
|
<arg>${nameNode}${workingDir}/metadata_extractor/out/fault</arg>
|
|
1135
|
<arg>${nameNode}${output_faults}/metadataextraction</arg>
|
|
1136
|
</distcp>
|
|
1137
|
<ok to="finalize"/>
|
|
1138
|
<error to="fail"/>
|
|
1139
|
</action>
|
|
1140
|
|
|
1141
|
<decision name="finalize">
|
|
1142
|
<switch>
|
|
1143
|
<case to="remove_sideproducts">${remove_sideproducts eq "true"}</case>
|
|
1144
|
<default to="end" />
|
|
1145
|
</switch>
|
|
1146
|
</decision>
|
|
1147
|
|
|
1148
|
<action name="remove_sideproducts">
|
|
1149
|
<fs>
|
|
1150
|
<delete path="${nameNode}${workingDir}" />
|
|
1151
|
</fs>
|
|
1152
|
<ok to="end" />
|
|
1153
|
<error to="fail" />
|
|
1154
|
</action>
|
|
1155
|
|
|
1156
|
<kill name="fail">
|
|
1157
|
<message>Unfortunately, the process failed -- error message:
|
|
1158
|
[${wf:errorMessage(wf:lastErrorNode())}]</message>
|
|
1159
|
</kill>
|
|
1160
|
<end name="end" />
|
|
1161
|
</workflow-app>
|