Merge pull request #4223 from nooraangelva/test-change

bug fix: unicode error with organisations in arxiv
inspirehep · Sep 29, 2022 · cfa855b · cfa855b
2 parents f90e85d + ee96740
commit cfa855b
Show file tree

Hide file tree

Showing 3 changed files with 37 additions and 1 deletion.
diff --git a/inspirehep/modules/workflows/tasks/arxiv.py b/inspirehep/modules/workflows/tasks/arxiv.py
@@ -307,7 +307,8 @@ def extract_authors_from_xml(xml_content):
 
         # Gets all the names for affiliated organizations using the organization ids from author
         for affiliation in author.xpath("./authorAffiliations/authorAffiliation/@organizationid").getall():
-            orgName = str(content.xpath(u'//organizations/Organization[@id="{}"]/orgName[@source="spiresICN" or @source="INSPIRE" and text()!="" ]/text()'.format(affiliation)).get())
+            orgName = content.xpath(u'string(//organizations/Organization[@id="{}"]/orgName[@source="spiresICN" or @source="INSPIRE" and text()!="" ]/text())'.format(affiliation)).get()
+
             cleaned_org_name = re.sub(remove_new_line_regex, '', orgName)
             if orgName and not re.match(undefined_or_none_value_regex, cleaned_org_name):
                 affiliations.append(cleaned_org_name)

diff --git a/tests/unit/workflows/fixtures/2202.12988.tar.gz b/tests/unit/workflows/fixtures/2202.12988.tar.gz
diff --git a/tests/unit/workflows/test_workflows_tasks_arxiv.py b/tests/unit/workflows/test_workflows_tasks_arxiv.py
@@ -1421,3 +1421,38 @@ def test_arxiv_ignores_random_xml_files():
 
     arxiv_author_list(obj, eng)
     assert obj.data.get('authors', None) is None
+
+
+def test_arxiv_handles_non_ascii_organization_names():
+    schema = load_schema('hep')
+    eprints_subschema = schema['properties']['arxiv_eprints']
+    filename = pkg_resources.resource_filename(
+        __name__, os.path.join('fixtures', '2202.12988.tar.gz'))
+
+    data = {
+        '$schema': 'http://localhost:5000/hep.json',
+        'arxiv_eprints': [
+            {
+                'categories': [
+                    'hep-ex',
+                ],
+                'value': '2202.12988',
+            },
+        ],
+    }
+    validate(data['arxiv_eprints'], eprints_subschema)
+
+    extra_data = {}
+    files = MockFiles({
+        '2202.12988.tar.gz': AttrDict({
+            'file': AttrDict({
+                'uri': filename,
+            })
+        })
+    })
+
+    obj = MockObj(data, extra_data, files=files)
+    eng = MockEng()
+
+    arxiv_author_list(obj, eng)
+    assert obj.data.get('authors', None) is not None