arkindex-base-worker 0.3.7rc4__py3-none-any.whl → 0.5.0a1__py3-none-any.whl

tests/test_elements_worker/{test_classifications.py → test_classification.py}

@@ -1,12 +1,14 @@
 import json
 import re
-from uuid import UUID, uuid4
+from uuid import UUID
 
 import pytest
-from apistar.exceptions import ErrorResponse
 
+from arkindex.exceptions import ErrorResponse
 from arkindex_worker.cache import CachedClassification, CachedElement
 from arkindex_worker.models import Element
+from arkindex_worker.utils import DEFAULT_BATCH_SIZE
+from tests import CORPUS_ID
 
 from . import BASE_API_CALLS
 
@@ -15,11 +17,96 @@ from . import BASE_API_CALLS
 DELETE_PARAMETER = "DELETE_PARAMETER"
 
 
+def test_load_corpus_classes_api_error(responses, mock_elements_worker):
+    responses.add(
+        responses.GET,
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        status=418,
+    )
+
+    assert not mock_elements_worker.classes
+    with pytest.raises(
+        Exception, match="Stopping pagination as data will be incomplete"
+    ):
+        mock_elements_worker.load_corpus_classes()
+
+    assert len(responses.calls) == len(BASE_API_CALLS) + 5
+    assert [
+        (call.request.method, call.request.url) for call in responses.calls
+    ] == BASE_API_CALLS + [
+        # We do 5 retries
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+    ]
+    assert not mock_elements_worker.classes
+
+
+def test_load_corpus_classes(responses, mock_elements_worker):
+    responses.add(
+        responses.GET,
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        status=200,
+        json={
+            "count": 3,
+            "next": None,
+            "results": [
+                {
+                    "id": "0000",
+                    "name": "good",
+                },
+                {
+                    "id": "1111",
+                    "name": "average",
+                },
+                {
+                    "id": "2222",
+                    "name": "bad",
+                },
+            ],
+        },
+    )
+
+    assert not mock_elements_worker.classes
+    mock_elements_worker.load_corpus_classes()
+
+    assert len(responses.calls) == len(BASE_API_CALLS) + 1
+    assert [
+        (call.request.method, call.request.url) for call in responses.calls
+    ] == BASE_API_CALLS + [
+        (
+            "GET",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
+    ]
+    assert mock_elements_worker.classes == {
+        "good": "0000",
+        "average": "1111",
+        "bad": "2222",
+    }
+
+
 def test_get_ml_class_id_load_classes(responses, mock_elements_worker):
-    corpus_id = "11111111-1111-1111-1111-111111111111"
     responses.add(
         responses.GET,
-        f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         status=200,
         json={
             "count": 1,
@@ -42,7 +129,7 @@ def test_get_ml_class_id_load_classes(responses, mock_elements_worker):
     ] == BASE_API_CALLS + [
         (
             "GET",
-            f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         ),
     ]
     assert mock_elements_worker.classes == {"good": "0000"}
@@ -51,12 +138,11 @@ def test_get_ml_class_id_load_classes(responses, mock_elements_worker):
 
 def test_get_ml_class_id_inexistant_class(mock_elements_worker, responses):
     # A missing class is now created automatically
-    corpus_id = "11111111-1111-1111-1111-111111111111"
     mock_elements_worker.classes = {"good": "0000"}
 
     responses.add(
         responses.POST,
-        f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         status=201,
         json={"id": "new-ml-class-1234"},
     )
@@ -82,12 +168,10 @@ def test_get_ml_class_id(mock_elements_worker):
 
 
 def test_get_ml_class_reload(responses, mock_elements_worker):
-    corpus_id = "11111111-1111-1111-1111-111111111111"
-
     # Add some initial classes
     responses.add(
         responses.GET,
-        f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         json={
             "count": 1,
             "next": None,
@@ -103,7 +187,7 @@ def test_get_ml_class_reload(responses, mock_elements_worker):
     # Invalid response when trying to create class2
     responses.add(
         responses.POST,
-        f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         status=400,
         json={"non_field_errors": "Already exists"},
     )
@@ -111,7 +195,7 @@ def test_get_ml_class_reload(responses, mock_elements_worker):
     # Add both classes (class2 is created by another process)
     responses.add(
         responses.GET,
-        f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         json={
             "count": 2,
             "next": None,
@@ -141,15 +225,15 @@ def test_get_ml_class_reload(responses, mock_elements_worker):
     ] == BASE_API_CALLS + [
         (
             "GET",
-            f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         ),
         (
             "POST",
-            f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         ),
         (
             "GET",
-            f"http://testserver/api/v1/corpus/{corpus_id}/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         ),
     ]
 
@@ -169,7 +253,7 @@ def test_retrieve_ml_class_not_in_cache(responses, mock_elements_worker):
     """
     responses.add(
         responses.GET,
-        f"http://testserver/api/v1/corpus/{mock_elements_worker.corpus_id}/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         status=200,
         json={
             "count": 1,
@@ -189,7 +273,7 @@ def test_retrieve_ml_class_not_in_cache(responses, mock_elements_worker):
     ] == BASE_API_CALLS + [
         (
             "GET",
-            f"http://testserver/api/v1/corpus/{mock_elements_worker.corpus_id}/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         ),
     ]
 
@@ -276,7 +360,7 @@ def test_create_classification_api_error(responses, mock_elements_worker):
     responses.add(
         responses.POST,
         "http://testserver/api/v1/classifications/",
-        status=500,
+        status=418,
     )
 
     with pytest.raises(ErrorResponse):
@@ -287,17 +371,10 @@ def test_create_classification_api_error(responses, mock_elements_worker):
             high_confidence=True,
         )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 5
+    assert len(responses.calls) == len(BASE_API_CALLS) + 1
     assert [
         (call.request.method, call.request.url) for call in responses.calls
-    ] == BASE_API_CALLS + [
-        # We retry 5 times the API call
-        ("POST", "http://testserver/api/v1/classifications/"),
-        ("POST", "http://testserver/api/v1/classifications/"),
-        ("POST", "http://testserver/api/v1/classifications/"),
-        ("POST", "http://testserver/api/v1/classifications/"),
-        ("POST", "http://testserver/api/v1/classifications/"),
-    ]
+    ] == BASE_API_CALLS + [("POST", "http://testserver/api/v1/classifications/")]
 
 
 def test_create_classification_create_ml_class(mock_elements_worker, responses):
@@ -306,7 +383,7 @@ def test_create_classification_create_ml_class(mock_elements_worker, responses):
     # Automatically create a missing class!
     responses.add(
         responses.POST,
-        "http://testserver/api/v1/corpus/11111111-1111-1111-1111-111111111111/classes/",
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
         status=201,
         json={"id": "new-ml-class-1234"},
     )
@@ -325,15 +402,12 @@ def test_create_classification_create_ml_class(mock_elements_worker, responses):
     )
 
     # Check a class & classification has been created
-    for call in responses.calls:
-        print(call.request.url, call.request.body)
-
     assert [
         (call.request.url, json.loads(call.request.body))
         for call in responses.calls[-2:]
     ] == [
         (
-            "http://testserver/api/v1/corpus/11111111-1111-1111-1111-111111111111/classes/",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
             {"name": "a_class"},
         ),
         (
@@ -506,12 +580,12 @@ def test_create_classifications_wrong_data(
                 "element": Element({"id": "12341234-1234-1234-1234-123412341234"}),
                 "classifications": [
                     {
-                        "ml_class_id": "uuid1",
+                        "ml_class": "cat",
                         "confidence": 0.75,
                         "high_confidence": False,
                     },
                     {
-                        "ml_class_id": "uuid2",
+                        "ml_class": "dog",
                         "confidence": 0.25,
                         "high_confidence": False,
                     },
@@ -523,86 +597,71 @@ def test_create_classifications_wrong_data(
 
 
 @pytest.mark.parametrize(
-    ("arg_name", "data", "error_message", "error_type"),
+    ("arg_name", "data", "error_message"),
     [
-        # Wrong classifications > ml_class_id
+        # Wrong classifications > ml_class
         (
-            "ml_class_id",
+            "ml_class",
             DELETE_PARAMETER,
-            "ml_class_id shouldn't be null and should be of type str",
-            AssertionError,
-        ),  # Updated
+            "ml_class shouldn't be null and should be of type str",
+        ),
         (
-            "ml_class_id",
+            "ml_class",
             None,
-            "ml_class_id shouldn't be null and should be of type str",
-            AssertionError,
+            "ml_class shouldn't be null and should be of type str",
         ),
         (
-            "ml_class_id",
+            "ml_class",
             1234,
-            "ml_class_id shouldn't be null and should be of type str",
-            AssertionError,
-        ),
-        (
-            "ml_class_id",
-            "not_an_uuid",
-            "ml_class_id is not a valid uuid.",
-            ValueError,
+            "ml_class shouldn't be null and should be of type str",
         ),
         # Wrong classifications > confidence
         (
             "confidence",
             DELETE_PARAMETER,
             "confidence shouldn't be null and should be a float in [0..1] range",
-            AssertionError,
         ),
         (
             "confidence",
             None,
             "confidence shouldn't be null and should be a float in [0..1] range",
-            AssertionError,
         ),
         (
             "confidence",
             "wrong confidence",
             "confidence shouldn't be null and should be a float in [0..1] range",
-            AssertionError,
         ),
         (
             "confidence",
             0,
             "confidence shouldn't be null and should be a float in [0..1] range",
-            AssertionError,
         ),
         (
             "confidence",
             2.00,
             "confidence shouldn't be null and should be a float in [0..1] range",
-            AssertionError,
         ),
         # Wrong classifications > high_confidence
         (
             "high_confidence",
             "wrong high_confidence",
             "high_confidence should be of type bool",
-            AssertionError,
         ),
     ],
 )
 def test_create_classifications_wrong_classifications_data(
-    arg_name, data, error_message, error_type, mock_elements_worker
+    arg_name, data, error_message, mock_elements_worker
 ):
     all_data = {
         "element": Element({"id": "12341234-1234-1234-1234-123412341234"}),
         "classifications": [
             {
-                "ml_class_id": str(uuid4()),
+                "ml_class": "cat",
                 "confidence": 0.75,
                 "high_confidence": False,
             },
             {
-                "ml_class_id": str(uuid4()),
+                "ml_class": "dog",
                 "confidence": 0.25,
                 "high_confidence": False,
                 # Overwrite with wrong data
@@ -614,7 +673,7 @@ def test_create_classifications_wrong_classifications_data(
         del all_data["classifications"][1][arg_name]
 
     with pytest.raises(
-        error_type,
+        AssertionError,
         match=re.escape(
             f"Classification at index 1 in classifications: {error_message}"
         ),
@@ -623,20 +682,21 @@ def test_create_classifications_wrong_classifications_data(
 
 
 def test_create_classifications_api_error(responses, mock_elements_worker):
+    mock_elements_worker.classes = {"cat": "0000", "dog": "1111"}
     responses.add(
         responses.POST,
         "http://testserver/api/v1/classification/bulk/",
-        status=500,
+        status=418,
     )
     elt = Element({"id": "12341234-1234-1234-1234-123412341234"})
     classes = [
         {
-            "ml_class_id": str(uuid4()),
+            "ml_class": "cat",
             "confidence": 0.75,
             "high_confidence": False,
         },
         {
-            "ml_class_id": str(uuid4()),
+            "ml_class": "dog",
             "confidence": 0.25,
             "high_confidence": False,
         },
@@ -647,192 +707,245 @@ def test_create_classifications_api_error(responses, mock_elements_worker):
             element=elt, classifications=classes
         )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 5
+    assert len(responses.calls) == len(BASE_API_CALLS) + 1
     assert [
         (call.request.method, call.request.url) for call in responses.calls
-    ] == BASE_API_CALLS + [
-        # We retry 5 times the API call
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-    ]
-
+    ] == BASE_API_CALLS + [("POST", "http://testserver/api/v1/classification/bulk/")]
 
-def test_create_classifications(responses, mock_elements_worker_with_cache):
-    # Set MLClass in cache
-    portrait_uuid = str(uuid4())
-    landscape_uuid = str(uuid4())
-    mock_elements_worker_with_cache.classes = {
-        "portrait": portrait_uuid,
-        "landscape": landscape_uuid,
-    }
 
-    elt = CachedElement.create(id="12341234-1234-1234-1234-123412341234", type="thing")
-    classes = [
-        {
-            "ml_class_id": portrait_uuid,
-            "confidence": 0.75,
-            "high_confidence": False,
-        },
-        {
-            "ml_class_id": landscape_uuid,
-            "confidence": 0.25,
-            "high_confidence": False,
-        },
-    ]
+def test_create_classifications_create_ml_class(mock_elements_worker, responses):
+    elt = Element({"id": "12341234-1234-1234-1234-123412341234"})
 
+    # Automatically create a missing class!
+    responses.add(
+        responses.POST,
+        f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        status=201,
+        json={"id": "new-ml-class-1234"},
+    )
     responses.add(
         responses.POST,
         "http://testserver/api/v1/classification/bulk/",
-        status=200,
+        status=201,
         json={
             "parent": str(elt.id),
             "worker_run_id": "56785678-5678-5678-5678-567856785678",
             "classifications": [
                 {
                     "id": "00000000-0000-0000-0000-000000000000",
-                    "ml_class": portrait_uuid,
+                    "ml_class": "new-ml-class-1234",
                     "confidence": 0.75,
                     "high_confidence": False,
                     "state": "pending",
                 },
-                {
-                    "id": "11111111-1111-1111-1111-111111111111",
-                    "ml_class": landscape_uuid,
-                    "confidence": 0.25,
-                    "high_confidence": False,
-                    "state": "pending",
-                },
             ],
         },
     )
-
-    mock_elements_worker_with_cache.create_classifications(
-        element=elt, classifications=classes
+    mock_elements_worker.classes = {"another_class": "0000"}
+    mock_elements_worker.create_classifications(
+        element=elt,
+        classifications=[
+            {
+                "ml_class": "a_class",
+                "confidence": 0.75,
+                "high_confidence": False,
+            }
+        ],
     )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 1
+    # Check a class & classification has been created
+    assert len(responses.calls) == len(BASE_API_CALLS) + 2
     assert [
         (call.request.method, call.request.url) for call in responses.calls
     ] == BASE_API_CALLS + [
+        (
+            "POST",
+            f"http://testserver/api/v1/corpus/{CORPUS_ID}/classes/",
+        ),
         ("POST", "http://testserver/api/v1/classification/bulk/"),
     ]
 
+    assert json.loads(responses.calls[-2].request.body) == {"name": "a_class"}
     assert json.loads(responses.calls[-1].request.body) == {
-        "parent": str(elt.id),
+        "parent": "12341234-1234-1234-1234-123412341234",
         "worker_run_id": "56785678-5678-5678-5678-567856785678",
-        "classifications": classes,
+        "classifications": [
+            {
+                "ml_class": "new-ml-class-1234",
+                "confidence": 0.75,
+                "high_confidence": False,
+            }
+        ],
     }
 
-    # Check that created classifications were properly stored in SQLite cache
-    assert list(CachedClassification.select()) == [
-        CachedClassification(
-            id=UUID("00000000-0000-0000-0000-000000000000"),
-            element_id=UUID(elt.id),
-            class_name="portrait",
-            confidence=0.75,
-            state="pending",
-            worker_run_id=UUID("56785678-5678-5678-5678-567856785678"),
-        ),
-        CachedClassification(
-            id=UUID("11111111-1111-1111-1111-111111111111"),
-            element_id=UUID(elt.id),
-            class_name="landscape",
-            confidence=0.25,
-            state="pending",
-            worker_run_id=UUID("56785678-5678-5678-5678-567856785678"),
-        ),
-    ]
-
-
-def test_create_classifications_not_in_cache(
-    responses, mock_elements_worker_with_cache
-):
-    """
-    CreateClassifications using ID that are not in `.classes` attribute.
-    Will load corpus MLClass to insert the corresponding name in Cache.
-    """
-    portrait_uuid = str(uuid4())
-    landscape_uuid = str(uuid4())
-    elt = CachedElement.create(id="12341234-1234-1234-1234-123412341234", type="thing")
-    classes = [
-        {
-            "ml_class_id": portrait_uuid,
-            "confidence": 0.75,
-            "high_confidence": False,
-        },
-        {
-            "ml_class_id": landscape_uuid,
-            "confidence": 0.25,
-            "high_confidence": False,
-        },
-    ]
 
+@pytest.mark.parametrize("batch_size", [DEFAULT_BATCH_SIZE, 1])
+def test_create_classifications(batch_size, responses, mock_elements_worker):
+    mock_elements_worker.classes = {"portrait": "0000", "landscape": "1111"}
+    elt = Element({"id": "12341234-1234-1234-1234-123412341234"})
     responses.add(
         responses.POST,
         "http://testserver/api/v1/classification/bulk/",
         status=200,
-        json={
-            "parent": str(elt.id),
-            "worker_run_id": "56785678-5678-5678-5678-567856785678",
-            "classifications": [
-                {
-                    "id": "00000000-0000-0000-0000-000000000000",
-                    "ml_class": portrait_uuid,
-                    "confidence": 0.75,
-                    "high_confidence": False,
-                    "state": "pending",
-                },
-                {
-                    "id": "11111111-1111-1111-1111-111111111111",
-                    "ml_class": landscape_uuid,
-                    "confidence": 0.25,
-                    "high_confidence": False,
-                    "state": "pending",
-                },
-            ],
-        },
+        json={"classifications": []},
     )
-    responses.add(
-        responses.GET,
-        f"http://testserver/api/v1/corpus/{mock_elements_worker_with_cache.corpus_id}/classes/",
-        status=200,
-        json={
-            "count": 2,
-            "next": None,
-            "results": [
-                {
-                    "id": portrait_uuid,
-                    "name": "portrait",
-                },
-                {"id": landscape_uuid, "name": "landscape"},
-            ],
-        },
+
+    mock_elements_worker.create_classifications(
+        element=elt,
+        classifications=[
+            {
+                "ml_class": "portrait",
+                "confidence": 0.75,
+                "high_confidence": False,
+            },
+            {
+                "ml_class": "landscape",
+                "confidence": 0.25,
+                "high_confidence": False,
+            },
+        ],
+        batch_size=batch_size,
     )
 
+    bulk_api_calls = [("POST", "http://testserver/api/v1/classification/bulk/")]
+    if batch_size != DEFAULT_BATCH_SIZE:
+        bulk_api_calls.append(("POST", "http://testserver/api/v1/classification/bulk/"))
+
+    assert len(responses.calls) == len(BASE_API_CALLS) + len(bulk_api_calls)
+    assert [
+        (call.request.method, call.request.url) for call in responses.calls
+    ] == BASE_API_CALLS + bulk_api_calls
+
+    first_cl = {"confidence": 0.75, "high_confidence": False, "ml_class": "0000"}
+    second_cl = {"confidence": 0.25, "high_confidence": False, "ml_class": "1111"}
+    empty_payload = {
+        "parent": str(elt.id),
+        "worker_run_id": "56785678-5678-5678-5678-567856785678",
+        "classifications": [],
+    }
+
+    bodies = []
+    first_call_idx = None
+    if batch_size > 1:
+        first_call_idx = -1
+        bodies.append({**empty_payload, "classifications": [first_cl, second_cl]})
+    else:
+        first_call_idx = -2
+        bodies.append({**empty_payload, "classifications": [first_cl]})
+        bodies.append({**empty_payload, "classifications": [second_cl]})
+
+    assert [
+        json.loads(bulk_call.request.body)
+        for bulk_call in responses.calls[first_call_idx:]
+    ] == bodies
+
+
+@pytest.mark.parametrize("batch_size", [DEFAULT_BATCH_SIZE, 1])
+def test_create_classifications_with_cache(
+    batch_size, responses, mock_elements_worker_with_cache
+):
+    mock_elements_worker_with_cache.classes = {"portrait": "0000", "landscape": "1111"}
+    elt = CachedElement.create(id="12341234-1234-1234-1234-123412341234", type="thing")
+
+    if batch_size > 1:
+        responses.add(
+            responses.POST,
+            "http://testserver/api/v1/classification/bulk/",
+            status=200,
+            json={
+                "parent": str(elt.id),
+                "worker_run_id": "56785678-5678-5678-5678-567856785678",
+                "classifications": [
+                    {
+                        "id": "00000000-0000-0000-0000-000000000000",
+                        "ml_class": "0000",
+                        "confidence": 0.75,
+                        "high_confidence": False,
+                        "state": "pending",
+                    },
+                    {
+                        "id": "11111111-1111-1111-1111-111111111111",
+                        "ml_class": "1111",
+                        "confidence": 0.25,
+                        "high_confidence": False,
+                        "state": "pending",
+                    },
+                ],
+            },
+        )
+    else:
+        for cl_id, cl_class, cl_conf in [
+            ("00000000-0000-0000-0000-000000000000", "0000", 0.75),
+            ("11111111-1111-1111-1111-111111111111", "1111", 0.25),
+        ]:
+            responses.add(
+                responses.POST,
+                "http://testserver/api/v1/classification/bulk/",
+                status=200,
+                json={
+                    "parent": str(elt.id),
+                    "worker_run_id": "56785678-5678-5678-5678-567856785678",
+                    "classifications": [
+                        {
+                            "id": cl_id,
+                            "ml_class": cl_class,
+                            "confidence": cl_conf,
+                            "high_confidence": False,
+                            "state": "pending",
+                        },
+                    ],
+                },
+            )
+
     mock_elements_worker_with_cache.create_classifications(
-        element=elt, classifications=classes
+        element=elt,
+        classifications=[
+            {
+                "ml_class": "portrait",
+                "confidence": 0.75,
+                "high_confidence": False,
+            },
+            {
+                "ml_class": "landscape",
+                "confidence": 0.25,
+                "high_confidence": False,
+            },
+        ],
+        batch_size=batch_size,
     )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 2
+    bulk_api_calls = [("POST", "http://testserver/api/v1/classification/bulk/")]
+    if batch_size != DEFAULT_BATCH_SIZE:
+        bulk_api_calls.append(("POST", "http://testserver/api/v1/classification/bulk/"))
+
+    assert len(responses.calls) == len(BASE_API_CALLS) + len(bulk_api_calls)
     assert [
         (call.request.method, call.request.url) for call in responses.calls
-    ] == BASE_API_CALLS + [
-        ("POST", "http://testserver/api/v1/classification/bulk/"),
-        (
-            "GET",
-            f"http://testserver/api/v1/corpus/{mock_elements_worker_with_cache.corpus_id}/classes/",
-        ),
-    ]
+    ] == BASE_API_CALLS + bulk_api_calls
 
-    assert json.loads(responses.calls[-2].request.body) == {
+    first_cl = {"confidence": 0.75, "high_confidence": False, "ml_class": "0000"}
+    second_cl = {"confidence": 0.25, "high_confidence": False, "ml_class": "1111"}
+    empty_payload = {
         "parent": str(elt.id),
         "worker_run_id": "56785678-5678-5678-5678-567856785678",
-        "classifications": classes,
+        "classifications": [],
     }
 
+    bodies = []
+    first_call_idx = None
+    if batch_size > 1:
+        first_call_idx = -1
+        bodies.append({**empty_payload, "classifications": [first_cl, second_cl]})
+    else:
+        first_call_idx = -2
+        bodies.append({**empty_payload, "classifications": [first_cl]})
+        bodies.append({**empty_payload, "classifications": [second_cl]})
+
+    assert [
+        json.loads(bulk_call.request.body)
+        for bulk_call in responses.calls[first_call_idx:]
+    ] == bodies
+
     # Check that created classifications were properly stored in SQLite cache
     assert list(CachedClassification.select()) == [
         CachedClassification(