workflows: timeouts refextract

Worksaround inspirehep/refextract#26 by interrupting the running away
refextract process.

Signed-off-by: Samuele Kaplun <[email protected]>

inspirehep/modules/workflows/tasks/actions.py

@@ -28,6 +28,7 @@
 
 from flask import current_app
 from werkzeug import secure_filename
+from timeout_decorator import TimeoutError
 
 from inspirehep.modules.workflows.utils import (
     get_pdf_in_workflow,
@@ -233,11 +234,14 @@ def _prepare_update_payload(obj, eng):
 def refextract(obj, eng):
     uri = get_pdf_in_workflow(obj)
     if uri:
-        mapped_references = extract_references(uri)
-        if mapped_references:
-            obj.data['references'] = mapped_references
-            obj.log.info('Extracted %d references', len(mapped_references))
-        else:
-            obj.log.info('No references extracted')
+        try:
+            mapped_references = extract_references(uri)
+            if mapped_references:
+                obj.data['references'] = mapped_references
+                obj.log.info('Extracted %d references', len(mapped_references))
+            else:
+                obj.log.info('No references extracted')
+        except TimeoutError as err:
+            obj.log.error('Timeout when extracting references from the PDF')
     else:
         obj.log.error('Not able to download and process the PDF')

inspirehep/modules/workflows/tasks/refextract.py

@@ -25,6 +25,7 @@
 from __future__ import absolute_import, division, print_function
 
 import json
+from timeout_decorator import decorator
 
 from refextract import extract_journal_reference, extract_references_from_file
 
@@ -84,6 +85,7 @@ def extract_journal_info(obj, eng):
     obj.data["publication_info"] = new_publication_info
 
 
+@timeout(5*60)
 def extract_references(filepath):
     """Extract references from PDF and return in INSPIRE format."""
     references = extract_references_from_file(

setup.py

@@ -103,6 +103,7 @@
     'python-redis-lock~=3.2',
     'backoff~=1.4.2',
     'requests~=2.15.1',
+    'timeout-decorator>=0.3.3,~=0.3',
 ]
 
 tests_require = [