Merge pull request DerwenAI#219 from tomaarsen/hotfix/topicrank_no_ca…

…ndidates

Prevent exception on TopicRank when there are no noun_chunks

pytextrank/topicrank.py

@@ -181,6 +181,9 @@ def _cluster (
     returns:
 list of clusters of candidates.
         """
+        if not candidates:
+            return []
+
         bag_of_words = list(
             {
                 word.text

tests/test_base.py

@@ -90,13 +90,13 @@ def test_summary (long_doc: Doc):
     LIMIT_PHRASES = 10
     TOP_K = 5
 
-    # expected for spacy==3.2.1 and en-core-web-sm==3.2.0
+    # expected for spacy==3.3.1 and en-core-web-sm==3.3.0
     expected_trace = [
-        [0, {0, 1, 6, 8, 9}],
-        [1, {9}],
-        [2, {1}],
-        [3, {7}],
-        [6, {9, 4}],
+        [0, {8, 1, 3, 7}],
+        [2, {3}],
+        [4, {9}],
+        [6, {5}],
+        [10, {0, 2, 3}]
     ]
 
     tr = long_doc._.textrank

tests/test_topicrank.py

@@ -89,10 +89,10 @@ def test_summary (nlp: Language):
     TOP_K = 5
 
     expected_trace = [
-        [0, [2, 7]],
+        [0, [2, 6]],
         [1, [0, 1]],
-        [2, [2]],
-        [3, [8, 7]],
+        [2, [0, 2]],
+        [3, [8, 6]],
         [4, [5]]
     ]
 
@@ -226,3 +226,19 @@ def scrubber_func(text_span: Span) -> str:
     # we expect the "test" to be filtered away (due to stopwords),
     # and "scrubber" to be replaced with "modified scrubber"
     assert set(phrases) == {"modified scrubber"}
+
+def test_empty_candidates ():
+    """
+Works as a pipeline component with an empty document.
+    """
+    # given
+    nlp = spacy.load("en_core_web_sm")
+    nlp.add_pipe("topicrank", last=True)
+
+    # works as a pipeline component
+    # when
+    text = ""
+    doc = nlp(text)
+
+    # then
+    assert len(doc._.phrases) == 0