arkindex-base-worker 0.4.0b1__py3-none-any.whl → 0.4.0b3__py3-none-any.whl

arkindex_worker/worker/transcription.py

@@ -11,6 +11,7 @@ from peewee import IntegrityError
 from arkindex_worker import logger
 from arkindex_worker.cache import CachedElement, CachedTranscription
 from arkindex_worker.models import Element
+from arkindex_worker.utils import DEFAULT_BATCH_SIZE, batch_publication, make_batches
 
 
 class TextOrientation(Enum):
@@ -109,9 +110,11 @@ class TranscriptionMixin:
 
         return created
 
+    @batch_publication
     def create_transcriptions(
         self,
         transcriptions: list[dict[str, str | float | TextOrientation | None]],
+        batch_size: int = DEFAULT_BATCH_SIZE,
     ) -> list[dict[str, str | float]]:
         """
         Create multiple transcriptions at once on existing elements through the API,
@@ -128,6 +131,8 @@ class TranscriptionMixin:
             orientation (TextOrientation)
                 Optional. Orientation of the transcription's text.
 
+        :param batch_size: The size of each batch, which will be used to split the publication to avoid API errors.
+
         :returns: A list of dicts as returned in the ``transcriptions`` field by the ``CreateTranscriptions`` API endpoint.
         """
 
@@ -171,13 +176,19 @@ class TranscriptionMixin:
             )
             return
 
-        created_trs = self.api_client.request(
-            "CreateTranscriptions",
-            body={
-                "worker_run_id": self.worker_run_id,
-                "transcriptions": transcriptions_payload,
-            },
-        )["transcriptions"]
+        created_trs = [
+            created_tr
+            for batch in make_batches(
+                transcriptions_payload, "transcription", batch_size
+            )
+            for created_tr in self.api_client.request(
+                "CreateTranscriptions",
+                body={
+                    "worker_run_id": self.worker_run_id,
+                    "transcriptions": batch,
+                },
+            )["transcriptions"]
+        ]
 
         if self.use_cache:
             # Store transcriptions in local cache
@@ -201,11 +212,13 @@ class TranscriptionMixin:
 
         return created_trs
 
+    @batch_publication
     def create_element_transcriptions(
         self,
         element: Element | CachedElement,
         sub_element_type: str,
         transcriptions: list[dict[str, str | float]],
+        batch_size: int = DEFAULT_BATCH_SIZE,
     ) -> dict[str, str | bool]:
         """
         Create multiple elements and transcriptions at once on a single parent element through the API.
@@ -225,6 +238,8 @@ class TranscriptionMixin:
             element_confidence (float)
                 Optional. Confidence score of the element between 0 and 1.
 
+        :param batch_size: The size of each batch, which will be used to split the publication to avoid API errors.
+
         :returns: A list of dicts as returned by the ``CreateElementTranscriptions`` API endpoint.
         """
         assert element and isinstance(
@@ -291,16 +306,22 @@ class TranscriptionMixin:
             )
             return
 
-        annotations = self.api_client.request(
-            "CreateElementTranscriptions",
-            id=element.id,
-            body={
-                "element_type": sub_element_type,
-                "worker_run_id": self.worker_run_id,
-                "transcriptions": transcriptions_payload,
-                "return_elements": True,
-            },
-        )
+        annotations = [
+            annotation
+            for batch in make_batches(
+                transcriptions_payload, "transcription", batch_size
+            )
+            for annotation in self.api_client.request(
+                "CreateElementTranscriptions",
+                id=element.id,
+                body={
+                    "element_type": sub_element_type,
+                    "worker_run_id": self.worker_run_id,
+                    "transcriptions": batch,
+                    "return_elements": True,
+                },
+            )
+        ]
 
         for annotation in annotations:
             if annotation["created"]:

tests/test_elements_worker/test_classifications.py

@@ -7,6 +7,7 @@ from apistar.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
@@ -692,7 +693,8 @@ def test_create_classifications_create_ml_class(mock_elements_worker, responses)
     }
 
 
-def test_create_classifications(responses, mock_elements_worker):
+@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(
@@ -716,62 +718,98 @@ def test_create_classifications(responses, mock_elements_worker):
                 "high_confidence": False,
             },
         ],
+        batch_size=batch_size,
     )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 1
+    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/"),
-    ]
+    ] == BASE_API_CALLS + bulk_api_calls
 
-    assert json.loads(responses.calls[-1].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": [
-            {
-                "confidence": 0.75,
-                "high_confidence": False,
-                "ml_class": "0000",
-            },
-            {
-                "confidence": 0.25,
-                "high_confidence": False,
-                "ml_class": "1111",
-            },
-        ],
+        "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
+
 
-def test_create_classifications_with_cache(responses, mock_elements_worker_with_cache):
+@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")
 
-    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",
+    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,
@@ -787,32 +825,41 @@ def test_create_classifications_with_cache(responses, mock_elements_worker_with_
                 "high_confidence": False,
             },
         ],
+        batch_size=batch_size,
     )
 
-    assert len(responses.calls) == len(BASE_API_CALLS) + 1
+    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/"),
-    ]
+    ] == BASE_API_CALLS + bulk_api_calls
 
-    assert json.loads(responses.calls[-1].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": [
-            {
-                "confidence": 0.75,
-                "high_confidence": False,
-                "ml_class": "0000",
-            },
-            {
-                "confidence": 0.25,
-                "high_confidence": False,
-                "ml_class": "1111",
-            },
-        ],
+        "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(