Batch search selection data queries

Co-Authored-By: Codex <267193182+codex@users.noreply.github.com>

src/documents/permissions.py

@@ -576,12 +576,9 @@ def annotate_document_count_by_ids(
                        or CustomFieldInstance)
         related_object_field: field on the relation pointing back to queryset pk
         document_ids: the document ids to count against -- a concrete list/set,
-                       or a simple (already resolved) queryset of ids. Callers
-                       that need this filtered by a complex condition (e.g. a
-                       permission check) should resolve it to a concrete list
-                       first if the same ids will be reused across multiple
-                       calls, rather than passing the complex queryset itself
-                       into each -- see ``_get_selection_data_for_queryset``.
+                       or a simple (already resolved) queryset of ids. Keep
+                       the list small: it is embedded in the statement as
+                       literal parameters.
         target_field: field on the relation pointing to Document id
     """
 

src/documents/tests/test_api_bulk_edit.py

@@ -1141,8 +1141,8 @@ class TestBulkEditAPI(DirectoriesMixin, APITestCase):
         self.assertCountEqual(
             response.data["selected_document_types"],
             [
-                {"id": self.c1.id, "document_count": 1},
-                {"id": self.c2.id, "document_count": 0},
+                {"id": self.dt1.id, "document_count": 1},
+                {"id": self.dt2.id, "document_count": 0},
             ],
         )
 

src/documents/tests/test_api_documents.py

@@ -49,6 +49,7 @@ from documents.models import WorkflowTrigger
 from documents.search import get_backend
 from documents.signals.handlers import run_workflows
 from documents.tests.utils import ConsumeTaskMixin
+from documents.views import DocumentViewSet
 from paperless_testing.dirs import DirectoriesMixin
 from paperless_testing.factories import DocumentFactory
 from paperless_testing.factories import TagFactory
@@ -1755,6 +1756,106 @@ class TestDocumentApi(DirectoriesMixin, ConsumeTaskMixin, APITestCase):
         }
         self.assertEqual(counts_by_field[custom_field.id], 1)
 
+    def test_selection_data_counts_with_shared_by_filter(self) -> None:
+        """
+        GIVEN:
+            - Several documents owned by one user with the same correspondent
+              and document type, only one of which is shared
+        WHEN:
+            - Requesting the document list filtered by shared_by__id with
+              include_selection_data=true
+        THEN:
+            - Counts reflect only the shared document, not every document the
+              owner has with that correspondent or document type, regardless of
+              the batch size
+        """
+        owner = User.objects.create_user(username="owner")
+        other = User.objects.create_user(username="other")
+        correspondent = Correspondent.objects.create(name="c1")
+        doc_type = DocumentType.objects.create(name="dt1")
+        tag = Tag.objects.create(name="tag")
+        docs = [
+            Document.objects.create(
+                checksum=str(i),
+                owner=owner,
+                correspondent=correspondent,
+                document_type=doc_type,
+            )
+            for i in range(3)
+        ]
+        for doc in docs:
+            doc.tags.add(tag)
+        grant_object(other, docs[0], "view_document")
+
+        for batch_size in (10_000, 1):
+            with (
+                self.subTest(batch_size=batch_size),
+                mock.patch.object(
+                    DocumentViewSet,
+                    "selection_data_batch_size",
+                    batch_size,
+                ),
+            ):
+                response = self.client.get(
+                    f"/api/documents/?shared_by__id={owner.id}"
+                    "&include_selection_data=true",
+                )
+
+                self.assertEqual(response.status_code, status.HTTP_200_OK)
+                self.assertEqual(response.data["count"], 1)
+                selection_data = response.data["selection_data"]
+                for key, obj in (
+                    ("selected_correspondents", correspondent),
+                    ("selected_document_types", doc_type),
+                    ("selected_tags", tag),
+                ):
+                    counts = {
+                        item["id"]: item["document_count"]
+                        for item in selection_data[key]
+                    }
+                    self.assertEqual(counts[obj.id], 1, key)
+
+    def test_selection_data_counts_documents_once_when_filter_joins(self) -> None:
+        """
+        GIVEN:
+            - A document matching a multi-valued filter through two tags
+        WHEN:
+            - Requesting the document list with include_selection_data=true
+        THEN:
+            - The document is counted once per correspondent, not once per
+              matching tag, regardless of the batch size
+        """
+        correspondent = Correspondent.objects.create(name="c1")
+        tag_a = Tag.objects.create(name="a")
+        tag_b = Tag.objects.create(name="b")
+        both = Document.objects.create(checksum="1", correspondent=correspondent)
+        both.tags.add(tag_a, tag_b)
+        only_a = Document.objects.create(checksum="2", correspondent=correspondent)
+        only_a.tags.add(tag_a)
+
+        for batch_size in (10_000, 1):
+            with (
+                self.subTest(batch_size=batch_size),
+                mock.patch.object(
+                    DocumentViewSet,
+                    "selection_data_batch_size",
+                    batch_size,
+                ),
+            ):
+                response = self.client.get(
+                    f"/api/documents/?tags__id__in={tag_a.id},{tag_b.id}"
+                    "&include_selection_data=true",
+                )
+
+                self.assertEqual(response.status_code, status.HTTP_200_OK)
+                counts = {
+                    item["id"]: item["document_count"]
+                    for item in response.data["selection_data"][
+                        "selected_correspondents"
+                    ]
+                }
+                self.assertEqual(counts[correspondent.id], 2)
+
     def test_statistics(self) -> None:
         doc1 = Document.objects.create(
             title="none1",

src/documents/tests/test_api_search.py

@@ -25,6 +25,7 @@ from documents.models import Tag
 from documents.models import Workflow
 from documents.search import get_backend
 from documents.search import reset_backend
+from documents.views import DocumentViewSet
 from paperless_mail.models import MailAccount
 from paperless_mail.models import MailRule
 from paperless_testing.dirs import DirectoriesMixin
@@ -345,38 +346,39 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
         storage_path = StoragePath.objects.create(name="sp1")
         tag = Tag.objects.create(name="tag")
 
-        matching_doc = Document.objects.create(
-            title="bank statement",
-            content="bank content",
-            checksum="A",
-            correspondent=correspondent,
-            document_type=doc_type,
-            storage_path=storage_path,
-        )
-        matching_doc.tags.add(tag)
+        for checksum, title in (
+            ("A", "bank statement"),
+            ("B", "bank account"),
+            ("C", "credit card"),
+        ):
+            doc = Document.objects.create(
+                title=title,
+                checksum=checksum,
+                correspondent=correspondent,
+                document_type=doc_type,
+                storage_path=storage_path,
+            )
+            doc.tags.add(tag)
+            get_backend().add_or_update(doc)
 
-        get_backend().add_or_update(matching_doc)
-
-        response = self.client.get(
-            "/api/documents/?query=bank&include_selection_data=true",
-        )
+        with mock.patch.object(DocumentViewSet, "selection_data_batch_size", 1):
+            response = self.client.get(
+                "/api/documents/?query=bank&page_size=1&include_selection_data=true",
+            )
 
         self.assertEqual(response.status_code, status.HTTP_200_OK)
-        self.assertIn("selection_data", response.data)
-
-        selected_correspondent = next(
-            item
-            for item in response.data["selection_data"]["selected_correspondents"]
-            if item["id"] == correspondent.id
-        )
-        selected_tag = next(
-            item
-            for item in response.data["selection_data"]["selected_tags"]
-            if item["id"] == tag.id
-        )
-
-        self.assertEqual(selected_correspondent["document_count"], 1)
-        self.assertEqual(selected_tag["document_count"], 1)
+        self.assertEqual(response.data["count"], 2)
+        self.assertEqual(len(response.data["results"]), 1)
+        for key, pk in (
+            ("selected_correspondents", correspondent.pk),
+            ("selected_document_types", doc_type.pk),
+            ("selected_storage_paths", storage_path.pk),
+            ("selected_tags", tag.pk),
+        ):
+            self.assertEqual(
+                response.data["selection_data"][key],
+                [{"id": pk, "document_count": 2}],
+            )
 
     def test_search_custom_field_ordering(self) -> None:
         custom_field = CustomField.objects.create(

src/documents/views.py

@@ -5,11 +5,13 @@ import platform
 import re
 import tempfile
 import zipfile
+from collections import Counter
 from collections import deque
 from collections.abc import Callable
 from datetime import datetime
 from datetime import timedelta
 from http import HTTPStatus
+from itertools import islice
 from pathlib import Path
 from time import mktime
 from time import sleep
@@ -168,7 +170,6 @@ from documents.permissions import PaperlessNotePermissions
 from documents.permissions import PaperlessObjectPermissions
 from documents.permissions import TrashPermissions
 from documents.permissions import ViewDocumentsPermissions
-from documents.permissions import annotate_document_count_by_ids
 from documents.permissions import annotate_document_count_for_related_queryset
 from documents.permissions import get_document_count_filter_for_user
 from documents.permissions import get_permission_grantees
@@ -1022,71 +1023,80 @@ class DocumentViewSet(
         "custom_field_",
     )
 
-    def _get_selection_data_for_queryset(self, queryset):
-        # Resolve once instead of once per model below. `queryset` can carry an
-        # arbitrarily expensive WHERE clause (user filters plus the permission
-        # filter); re-embedding it as a subquery inside 5 separate Count(...)
-        # calls forces the database to re-evaluate that whole thing 5 times, and
-        # -- for FK relations especially -- can defeat semi-join planning
-        # entirely at scale. A concrete id list is cheap to reuse.
-        # order_by() drops the default/user ordering -- irrelevant for a plain
-        # id list, but left in place it forces a sort over the full filtered
-        # set before the ids can even be collected.
-        document_ids = list(queryset.order_by().values_list("pk", flat=True))
+    # Bound the IN clauses used for search results and through-table lookups.
+    selection_data_batch_size = 10_000
 
-        correspondents = Correspondent.objects.annotate(
-            document_count=Count(
-                "documents",
-                filter=Q(documents__id__in=document_ids),
-                distinct=True,
-            ),
-        )
-        document_types = DocumentType.objects.annotate(
-            document_count=Count(
-                "documents",
-                filter=Q(documents__id__in=document_ids),
-                distinct=True,
-            ),
-        )
-        storage_paths = StoragePath.objects.annotate(
-            document_count=Count(
-                "documents",
-                filter=Q(documents__id__in=document_ids),
-                distinct=True,
-            ),
-        )
-        # Tag and CustomField reach Document through an M2M/through-model table;
-        # a plain Count(filter=...) there is a much more expensive plan than the
-        # FK relations above once the bridge table is large -- see
-        # annotate_document_count_by_ids() for why.
-        tags = annotate_document_count_by_ids(
-            Tag.objects.all(),
-            through_model=Document.tags.through,
-            related_object_field="tag_id",
-            document_ids=document_ids,
-        )
-        custom_fields = annotate_document_count_by_ids(
-            CustomField.objects.all(),
-            through_model=CustomFieldInstance,
-            related_object_field="field_id",
-            document_ids=document_ids,
+    def _get_selection_data_for_queryset(
+        self,
+        queryset: QuerySet[Document],
+        *,
+        search_ids: list[int] | None = None,
+    ) -> dict[str, list[dict[str, int]]]:
+        # Keep permission and filter predicates at document granularity.
+        # Reusing the filtered queryset as a subquery for each count is slow.
+        rows = (
+            queryset.order_by()
+            .prefetch_related(None)
+            .values_list(
+                "pk",
+                "correspondent_id",
+                "document_type_id",
+                "storage_path_id",
+            )
         )
+        batch_size = self.selection_data_batch_size
+
+        def document_rows():
+            if search_ids is None:
+                yield from rows.iterator()
+            else:
+                for start in range(0, len(search_ids), batch_size):
+                    yield from rows.filter(
+                        pk__in=search_ids[start : start + batch_size],
+                    ).iterator()
+
+        correspondents: Counter[int | None] = Counter()
+        document_types: Counter[int | None] = Counter()
+        storage_paths: Counter[int | None] = Counter()
+        document_ids: set[int] = set()
+        for pk, correspondent, document_type, storage_path in document_rows():
+            # Filters that join multi-valued relations can repeat a document.
+            if pk in document_ids:
+                continue
+            document_ids.add(pk)
+            correspondents[correspondent] += 1
+            document_types[document_type] += 1
+            storage_paths[storage_path] += 1
+
+        tags: Counter[int | None] = Counter()
+        custom_fields: Counter[int | None] = Counter()
+        ids = iter(document_ids)
+        while batch := list(islice(ids, batch_size)):
+            for model, field, counts in (
+                (Document.tags.through, "tag_id", tags),
+                (CustomFieldInstance, "field_id", custom_fields),
+            ):
+                counts.update(
+                    dict(
+                        model.objects.filter(document_id__in=batch)
+                        .order_by()
+                        .values_list(field)
+                        .annotate(document_count=Count("document_id")),
+                    ),
+                )
+
+        def selection(model: type[Model], counts: Counter[int | None]):
+            return [
+                {"id": pk, "document_count": counts[pk]}
+                for pk in model._default_manager.values_list("pk", flat=True)
+            ]
+
         return {
-            "selected_correspondents": [
-                {"id": t.id, "document_count": t.document_count} for t in correspondents
-            ],
-            "selected_tags": [
-                {"id": t.id, "document_count": t.document_count} for t in tags
-            ],
-            "selected_document_types": [
-                {"id": t.id, "document_count": t.document_count} for t in document_types
-            ],
-            "selected_storage_paths": [
-                {"id": t.id, "document_count": t.document_count} for t in storage_paths
-            ],
-            "selected_custom_fields": [
-                {"id": t.id, "document_count": t.document_count} for t in custom_fields
-            ],
+            "selected_correspondents": selection(Correspondent, correspondents),
+            "selected_tags": selection(Tag, tags),
+            "selected_document_types": selection(DocumentType, document_types),
+            "selected_storage_paths": selection(StoragePath, storage_paths),
+            "selected_custom_fields": selection(CustomField, custom_fields),
         }
 
     @classmethod
@@ -2707,13 +2717,10 @@ class UnifiedSearchViewSet(DocumentViewSet):
                 if get_boolean(
                     str(request.query_params.get("include_selection_data", "false")),
                 ):
-                    # NOTE: pk__in=ordered_ids generates a large SQL IN clause
-                    # for big result sets.  Acceptable today but may need a temp
-                    # table or chunked approach if selection_data becomes slow
-                    # at scale (tens of thousands of matching documents).
                     response.data["selection_data"] = (
                         self._get_selection_data_for_queryset(
-                            filtered_qs.filter(pk__in=result.ordered_ids),
+                            filtered_qs,
+                            search_ids=result.ordered_ids,
                         )
                     )
                 return response