Fixture of corpus of docs with one subject only for SVC backend training

tests/conftest.py

@@ -96,6 +96,17 @@ def document_corpus(subject_index):
     return doc_corpus
 
 
+@pytest.fixture(scope='module')
+def document_corpus_single_subject(document_corpus):
+    docs_single_subj = []
+    for doc in document_corpus.documents:
+        uri = list(doc.uris)[0] if len(doc.uris) > 0 else None
+        label = list(doc.labels)[0] if len(doc.labels) > 0 else None
+        docs_single_subj.append(
+            annif.corpus.Document(doc.text, {uri}, {label}))
+    return annif.corpus.DocumentList(docs_single_subj)
+
+
 @pytest.fixture(scope='module')
 def pretrained_vectors():
     return py.path.local(os.path.join(

tests/test_backend_svc.py

@@ -34,26 +34,26 @@ def test_svc_suggest_no_vectorizer(project):
         svc.suggest("example text")
 
 
-def test_svc_train(datadir, document_corpus, project):
+def test_svc_train(datadir, document_corpus_single_subject, project):
     svc_type = annif.backend.get_backend('svc')
     svc = svc_type(
         backend_id='svc',
         config_params={},
         project=project)
 
-    svc.train(document_corpus)
+    svc.train(document_corpus_single_subject)
     assert svc._model is not None
     assert datadir.join('svc-model.gz').exists()
 
 
-def test_svc_train_ngram(datadir, document_corpus, project):
+def test_svc_train_ngram(datadir, document_corpus_single_subject, project):
     svc_type = annif.backend.get_backend('svc')
     svc = svc_type(
         backend_id='svc',
         config_params={'ngram': 2},
         project=project)
 
-    svc.train(document_corpus)
+    svc.train(document_corpus_single_subject)
     assert svc._model is not None
     assert datadir.join('svc-model.gz').exists()