Mock sentence-transformers and nltk in tests (elastic#3059) (elastic#3062)

* Mock sentence-transformers and nltk in tests

* Update test_elasticsearch/test_dsl/conftest.py



* switch to a local mock that only affects the one test

---------


(cherry picked from commit e05d7f1)

Co-authored-by: Miguel Grinberg <[email protected]>
Co-authored-by: Quentin Pradet <[email protected]>

Author: 3 people

pyproject.toml

@@ -78,8 +78,6 @@ dev = [
     "pandas",
     "mapbox-vector-tile",
     "jinja2",
-    "nltk",
-    "sentence_transformers",
     "tqdm",
     "mypy",
     "pyright",

test_elasticsearch/test_dsl/test_integration/test_examples/_async/test_vectors.py

@@ -15,27 +15,27 @@
 #  specific language governing permissions and limitations
 #  under the License.
 
+import sys
 from hashlib import md5
 from typing import Any, List, Tuple
 from unittest import SkipTest
+from unittest.mock import Mock, patch
 
 import pytest
 
 from elasticsearch import AsyncElasticsearch
 
-from ..async_examples import vectors
-
 
 @pytest.mark.asyncio
 async def test_vector_search(
-    async_write_client: AsyncElasticsearch, es_version: Tuple[int, ...], mocker: Any
+    async_write_client: AsyncElasticsearch, es_version: Tuple[int, ...]
 ) -> None:
     # this test only runs on Elasticsearch >= 8.11 because the example uses
     # a dense vector without specifying an explicit size
     if es_version < (8, 11):
         raise SkipTest("This test requires Elasticsearch 8.11 or newer")
 
-    class MockModel:
+    class MockSentenceTransformer:
         def __init__(self, model: Any):
             pass
 
@@ -44,9 +44,22 @@ def encode(self, text: str) -> List[float]:
             total = sum(vector)
             return [float(v) / total for v in vector]
 
-    mocker.patch.object(vectors, "SentenceTransformer", new=MockModel)
+    def mock_nltk_tokenize(content: str):
+        return content.split("\n")
+
+    # mock sentence_transformers and nltk, because they are quite big and
+    # irrelevant for testing the example logic
+    with patch.dict(
+        sys.modules,
+        {
+            "sentence_transformers": Mock(SentenceTransformer=MockSentenceTransformer),
+            "nltk": Mock(sent_tokenize=mock_nltk_tokenize),
+        },
+    ):
+        # import the example after the dependencies are mocked
+        from ..async_examples import vectors
 
-    await vectors.create()
-    await vectors.WorkplaceDoc._index.refresh()
-    results = await (await vectors.search("Welcome to our team!")).execute()
-    assert results[0].name == "New Employee Onboarding Guide"
+        await vectors.create()
+        await vectors.WorkplaceDoc._index.refresh()
+        results = await (await vectors.search("Welcome to our team!")).execute()
+        assert results[0].name == "Intellectual Property Policy"

test_elasticsearch/test_dsl/test_integration/test_examples/_sync/test_vectors.py

@@ -15,27 +15,27 @@
 #  specific language governing permissions and limitations
 #  under the License.
 
+import sys
 from hashlib import md5
 from typing import Any, List, Tuple
 from unittest import SkipTest
+from unittest.mock import Mock, patch
 
 import pytest
 
 from elasticsearch import Elasticsearch
 
-from ..examples import vectors
-
 
 @pytest.mark.sync
 def test_vector_search(
-    write_client: Elasticsearch, es_version: Tuple[int, ...], mocker: Any
+    write_client: Elasticsearch, es_version: Tuple[int, ...]
 ) -> None:
     # this test only runs on Elasticsearch >= 8.11 because the example uses
     # a dense vector without specifying an explicit size
     if es_version < (8, 11):
         raise SkipTest("This test requires Elasticsearch 8.11 or newer")
 
-    class MockModel:
+    class MockSentenceTransformer:
         def __init__(self, model: Any):
             pass
 
@@ -44,9 +44,22 @@ def encode(self, text: str) -> List[float]:
             total = sum(vector)
             return [float(v) / total for v in vector]
 
-    mocker.patch.object(vectors, "SentenceTransformer", new=MockModel)
+    def mock_nltk_tokenize(content: str):
+        return content.split("\n")
+
+    # mock sentence_transformers and nltk, because they are quite big and
+    # irrelevant for testing the example logic
+    with patch.dict(
+        sys.modules,
+        {
+            "sentence_transformers": Mock(SentenceTransformer=MockSentenceTransformer),
+            "nltk": Mock(sent_tokenize=mock_nltk_tokenize),
+        },
+    ):
+        # import the example after the dependencies are mocked
+        from ..examples import vectors
 
-    vectors.create()
-    vectors.WorkplaceDoc._index.refresh()
-    results = (vectors.search("Welcome to our team!")).execute()
-    assert results[0].name == "New Employee Onboarding Guide"
+        vectors.create()
+        vectors.WorkplaceDoc._index.refresh()
+        results = (vectors.search("Welcome to our team!")).execute()
+        assert results[0].name == "Intellectual Property Policy"