From b3e0cf0e216fd3ca1a77a6c9752e912514ebeea8 Mon Sep 17 00:00:00 2001
From: stumpylog <797416+stumpylog@users.noreply.github.com>
Date: Tue, 28 Jul 2026 16:36:44 -0700
Subject: [PATCH] chore(profiling): confirm Stage 2 loop-resolution reduces
 query count and wall time at scale

---
 profiling/results/history.jsonl         |  2 +
 profiling/test_stage2_document_loops.py | 64 +++++++++++++++++++++++++
 2 files changed, 66 insertions(+)
 create mode 100644 profiling/test_stage2_document_loops.py

diff --git a/profiling/results/history.jsonl b/profiling/results/history.jsonl
index f35df5290..b78923969 100644
--- a/profiling/results/history.jsonl
+++ b/profiling/results/history.jsonl
@@ -4,3 +4,5 @@
 {"timestamp": "2026-07-28T01:38:32.706519+00:00", "code_ref": "e80ede4cb", "stage": "stage1", "scenario": "get_objects_for_user_owner_aware_document_baseline", "scale": "medium", "best_seconds": 31.46481539506931, "query_count": 1}
 {"timestamp": "2026-07-28T03:53:46.770669+00:00", "code_ref": "d4fa85237", "stage": "stage1", "scenario": "get_objects_for_user_owner_aware_document_baseline", "scale": "medium", "best_seconds": 31.671881378046237, "query_count": 1}
 {"timestamp": "2026-07-28T04:02:41.942168+00:00", "code_ref": "d4fa85237", "stage": "stage1", "scenario": "permitted_document_ids_after_stage1", "scale": "medium", "best_seconds": 0.07956207206007093, "query_count": 1}
+{"timestamp": "2026-07-28T23:30:28.602501+00:00", "code_ref": "86846255f", "stage": "stage2", "scenario": "per_row_checker_vs_resolved_set_old", "scale": "medium", "best_seconds": 3.4628250940004364, "query_count": 692}
+{"timestamp": "2026-07-28T23:30:28.606987+00:00", "code_ref": "86846255f", "stage": "stage2", "scenario": "per_row_checker_vs_resolved_set_new", "scale": "medium", "best_seconds": 0.10436739504802972, "query_count": 1}
diff --git a/profiling/test_stage2_document_loops.py b/profiling/test_stage2_document_loops.py
new file mode 100644
index 000000000..9e38d5c18
--- /dev/null
+++ b/profiling/test_stage2_document_loops.py
@@ -0,0 +1,64 @@
+# profiling/test_stage2_document_loops.py
+from __future__ import annotations
+
+import pytest
+
+from documents.permissions import has_perms_owner_aware
+from documents.permissions import permitted_document_ids
+from profiling.harness import append_profiling_history
+from profiling.harness import require_postgres
+from profiling.harness import run_profile
+from profiling.seed import seed_permission_dataset
+
+
+@pytest.mark.profiling
+@pytest.mark.django_db
+def test_profile_per_row_checker_vs_resolved_set_for_bulk_operation() -> None:
+    require_postgres()
+    data = seed_permission_dataset(scale="medium")
+    user = data.users[0]
+    # simulate a bulk operation over 500 documents the user can actually see
+    # (bulk-edit/share/trash all operate on a user's own selection) -- built
+    # from confirmed-visible documents so all() can't short-circuit unfairly
+    # in either old_style or new_style, see amendment note above
+    permitted_ids = set(permitted_document_ids(user))
+    batch = [doc for doc in data.documents if doc.pk in permitted_ids][:500]
+    assert len(batch) == 500, (
+        f"expected at least 500 visible documents to build a fair batch, "
+        f"got {len(batch)} -- scale profile or seed ratios may need adjusting"
+    )
+
+    def old_style():
+        return all(has_perms_owner_aware(user, "view_document", doc) for doc in batch)
+
+    def new_style():
+        permitted = set(permitted_document_ids(user))
+        return all(doc.pk in permitted for doc in batch)
+
+    old_profile = run_profile(old_style, repeat=3)
+    new_profile = run_profile(new_style, repeat=3)
+
+    append_profiling_history(
+        stage="stage2",
+        scenario="per_row_checker_vs_resolved_set_old",
+        best_seconds=old_profile.best_seconds,
+        query_count=old_profile.query_count,
+        scale="medium",
+    )
+    append_profiling_history(
+        stage="stage2",
+        scenario="per_row_checker_vs_resolved_set_new",
+        best_seconds=new_profile.best_seconds,
+        query_count=new_profile.query_count,
+        scale="medium",
+    )
+
+    print(  # noqa: T201
+        f"\nOLD (per-row checker): best={old_profile.best_seconds:.4f}s queries={old_profile.query_count}",
+    )
+    print(  # noqa: T201
+        f"NEW (resolved set):    best={new_profile.best_seconds:.4f}s queries={new_profile.query_count}",
+    )
+
+    assert new_profile.query_count < old_profile.query_count
+    assert new_profile.best_seconds < old_profile.best_seconds