Merge origin/master into uk-local-area-breakdown

Brings in #4026 (raise the economy cache schema version from 3 to 4)
and the 5.7.0 "Update PolicyEngine API" release commit.

Conflicts:

- tests/unit/services/test_reform_impacts_service.py: both sides added
  a new test at the end of the file. This branch added
  test_a_briefly_kept_error_expires_and_leaves_other_records; master
  added test_impact_cached_under_previous_economy_schema_version_is_not_matched.
  Kept both, this branch's test first. The imports master added for its
  test (constants, get_economy_impact_cache_version) merged without
  conflict.

Merged without conflict, checked by hand:

- policyengine_api/constants.py: RUNTIME_CACHE_SCHEMA_VERSIONS
  ["economy_impact"] is 4 (master), and UK_COUNTRY_REGIONS,
  UK_REGION_REGISTRY_CODES and LOCAL_AREA_BREAKDOWN_REGIONS are kept
  (this branch).
- pyproject.toml: package version 5.7.0 (master) and the
  policyengine[models]==6.10.0 pin (this branch).
- uv.lock: policyengine-api 5.7.0 (master) with this branch's .py
  6.10.0 dependency set (policyengine 6.10.0, policyengine-core 3.33.3,
  policyengine-uk 2.132.6). `uv lock --check` passes.
- tests/unit/test_constants.py: both sides' additions kept.
- CHANGELOG.md and changelog.d/: master's 5.7.0 entry and its removal
  of the released remove-economy-comparison fragment, plus master's
  economy-cache-decile-fractions fragment; this branch's fragments are
  unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Author: anth-volk

CHANGELOG.md

@@ -1,3 +1,10 @@
+## [5.7.0] - 2026-10-22
+
+### Removed
+
+- Removed the unused economy comparison module (`policyengine_api/services/economy_comparison.py`) with its tests and snapshot data. Nothing has called it since the Public API stopped falling back to its own society-wide simulations (#2569); the route-boundary rule now covers the calculation route alone.
+
+
 ## [5.6.4] - 2026-10-21
 
 ### Changed

changelog.d/economy-cache-decile-fractions.fixed.md

@@ -0,0 +1 @@
+Raised the economy result cache schema version from 3 to 4, so economy results cached while the simulation API returned decile and UK wealth decile relative changes as percentages (x100) instead of fractions are no longer served, and are recomputed on the next request. Budget-window results share this cache version, so they are recomputed too.

changelog.d/remove-economy-comparison.removed.md

@@ -194,7 +194,13 @@ def get_bundle_default_dataset_option(country_id: str) -> dict:
 ) or _resolve_distribution_version(_dist_versions, "policyengine-core", "policyengine")
 
 RUNTIME_CACHE_SCHEMA_VERSIONS = {
-    "economy_impact": 3,
+    # 4: economy results cached under 3 can hold decile.relative and the UK
+    # wealth_decile.relative as percentages (x100) instead of fractions, from
+    # the simulation API (PolicyEngine/policyengine-sim-api#827). Economy cache
+    # keys do not change when only the simulation API's code changes, so this
+    # version was raised to stop serving those results. The budget-window
+    # cache key uses the same version.
+    "economy_impact": 4,
     "report_output": 3,
 }
 # Runtime cache versions are stored in legacy VARCHAR(10) columns, and keep at

policyengine_api/constants.py

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "policyengine-api"
-version = "5.6.4"
+version = "5.7.0"
 description = "PolicyEngine API"
 readme = "README.md"
 license = "AGPL-3.0-only"

pyproject.toml

@@ -3,6 +3,8 @@
 
 import pytest
 
+from policyengine_api import constants
+from policyengine_api.constants import get_economy_impact_cache_version
 from policyengine_api.runtime_cache.core import CacheNamespace
 from policyengine_api.runtime_cache.fake import InMemoryCacheBackend
 from policyengine_api.runtime_cache.reform_impacts import (
@@ -226,3 +228,58 @@ def test_a_briefly_kept_error_expires_and_leaves_other_records() -> None:
     assert [(impact.execution_id, impact.status) for impact in after_expiry] == [
         ("next-job", "computing")
     ]
+
+
+def test_impact_cached_under_previous_economy_schema_version_is_not_matched(
+    service,
+):
+    # Economy results cached under schema version 3 can hold decile relative
+    # changes as percentages (x100) from the simulation API
+    # (PolicyEngine/policyengine-sim-api#827). A lookup built from the
+    # current cache version must not find them.
+    current_cache_version = get_economy_impact_cache_version(SCOPE.country_id)
+    with pytest.MonkeyPatch.context() as schema_version_patch:
+        schema_version_patch.setitem(
+            constants.RUNTIME_CACHE_SCHEMA_VERSIONS, "economy_impact", 3
+        )
+        percentage_decile_cache_version = get_economy_impact_cache_version(
+            SCOPE.country_id
+        )
+
+    options_hash = "decile-hash"
+    stale_identity = ReformImpactIdentity(
+        scope=SCOPE,
+        api_version=percentage_decile_cache_version,
+        options_hash=options_hash,
+    )
+    service.set_reform_impact(
+        stale_identity,
+        ReformImpactExecution(
+            execution_id="percentage-decile-job",
+            start_time=datetime(2026, 7, 1),
+        ),
+        options={"hash": options_hash},
+        status="computing",
+        reform_impact_json={},
+    )
+    service.set_complete_reform_impact(
+        "percentage-decile-job",
+        reform_impact_json={"decile": {"relative": {"1": 1.5}}},
+    )
+
+    current_identity = ReformImpactIdentity(
+        scope=SCOPE,
+        api_version=current_cache_version,
+        options_hash=options_hash,
+    )
+
+    assert [
+        impact.execution_id for impact in service.get_all_reform_impacts(stale_identity)
+    ] == ["percentage-decile-job"]
+    assert service.get_all_reform_impacts(current_identity) == []
+    assert (
+        service.get_all_reform_impacts_by_options_hash_prefix(
+            current_identity, f"{options_hash}%"
+        )
+        == []
+    )

tests/unit/services/test_reform_impacts_service.py

@@ -9,12 +9,15 @@
     POLICYENGINE_CORE_VERSION,
     POLICYENGINE_VERSION,
     REGION_PREFIXES,
+    RUNTIME_CACHE_SCHEMA_VERSIONS,
+    RUNTIME_CACHE_VERSION_LENGTH,
     UK_REGION_TYPES,
     US_REGION_TYPES,
     _bundle_dataset_display_label,
     _load_policyengine_bundle,
     _normalize_distribution_name,
     _resolve_distribution_version,
+    get_economy_impact_cache_version,
     get_py_manifest,
 )
 
@@ -113,6 +116,24 @@ def test__has_exactly_three_prefixes(self):
             assert len(REGION_PREFIXES["us"]) == 3
 
 
+class TestEconomyImpactCacheVersion:
+    # Since its move to .py 4.x, the simulation API returned decile.relative
+    # and the UK wealth_decile.relative as percentages (x100) instead of
+    # fractions (PolicyEngine/policyengine-sim-api#827). Economy results
+    # cached under schema version 3 can hold those values, and the cache key
+    # does not change when the simulation API's code does, so version 4 makes
+    # every economy cache lookup miss them and recompute.
+    def test__economy_impact_schema_version_is_4(self):
+        assert RUNTIME_CACHE_SCHEMA_VERSIONS["economy_impact"] == 4
+
+    @pytest.mark.parametrize("country_id", ["us", "uk"])
+    def test__economy_impact_cache_version_carries_schema_version_4(self, country_id):
+        cache_version = get_economy_impact_cache_version(country_id)
+
+        assert cache_version.startswith("e4")
+        assert len(cache_version) == RUNTIME_CACHE_VERSION_LENGTH
+
+
 class TestDistributionVersionHelpers:
     def test__normalize_distribution_name(self):
         assert _normalize_distribution_name("policyengine_core") == (