Better

pipeline/transform/poi_proximity.py

@@ -56,6 +56,22 @@ DYNAMIC_FILTER_ALL_GROUPS = {"Public Transport", "Leisure", "Health"}
 DYNAMIC_FILTER_COUNT_THRESHOLD_GROUPS = {"Groceries"}
 DYNAMIC_FILTER_EXCLUDED_CATEGORIES = {"Park"}
 
+# Combined "nearest of any rail mode" distance: the distance to the closest of a
+# Rail station, Tube station, DLR station, or Tram & Metro stop. Materialized as
+# its own real column so the server loads it like any other per-category
+# distance. min_distance_per_postcode already supports multi-category groups, so
+# this is a genuine nearest-of-the-union query (identical to the per-mode
+# minimum). Distance only: a combined "within Xkm" count is not meaningful, so
+# it is intentionally omitted from the count groups.
+COMBINED_STATION_GROUP_KEY = "poi_any_station"
+COMBINED_STATION_DISPLAY_NAME = "Any station"
+COMBINED_STATION_SOURCE_CATEGORIES = [
+    "Rail station",
+    "Tube station",
+    "DLR station",
+    "Tram & Metro stop",
+]
+
 
 def _poi_category_slug(category: str) -> str:
     ascii_text = (
@@ -227,10 +243,21 @@ def main():
     dynamic_counts_5km = count_pois_per_postcode(
         postcodes, pois, groups=poi_category_groups, radius_km=5
     )
+    # Distances additionally include the combined "Any station" group (nearest of
+    # any rail mode). It is distance-only, so it is added here but not to the
+    # 2km/5km count groups above.
+    distance_groups = {
+        **poi_category_groups,
+        COMBINED_STATION_GROUP_KEY: COMBINED_STATION_SOURCE_CATEGORIES,
+    }
+    distance_display_names = {
+        **poi_display_names,
+        COMBINED_STATION_GROUP_KEY: COMBINED_STATION_DISPLAY_NAME,
+    }
     dynamic_distances = min_distance_per_postcode(
-        postcodes, pois, groups=poi_category_groups
+        postcodes, pois, groups=distance_groups
     )
-    dynamic_renames = _dynamic_poi_metric_renames(poi_display_names)
+    dynamic_renames = _dynamic_poi_metric_renames(distance_display_names)
     dynamic_counts_2km = dynamic_counts_2km.rename(
         {k: v for k, v in dynamic_renames.items() if k in dynamic_counts_2km.columns}
     )

pipeline/transform/test_poi_proximity.py

@@ -1,6 +1,9 @@
 import polars as pl
 
 from pipeline.transform.poi_proximity import (
+    COMBINED_STATION_DISPLAY_NAME,
+    COMBINED_STATION_GROUP_KEY,
+    COMBINED_STATION_SOURCE_CATEGORIES,
     GREENSPACE_PARK_FUNCTIONS,
     POI_GROUPS_2KM,
     _build_poi_category_groups,
@@ -8,7 +11,7 @@ from pipeline.transform.poi_proximity import (
     _greenspace_count_frame,
     _groceries_categories,
 )
-from pipeline.utils.poi_counts import count_pois_per_postcode
+from pipeline.utils.poi_counts import count_pois_per_postcode, min_distance_per_postcode
 
 
 def test_groceries_2km_counts_geolytix_brand_categories() -> None:
@@ -84,6 +87,42 @@ def test_dynamic_poi_groups_include_requested_categories_only() -> None:
     assert "poi_school" not in groups
 
 
+def test_combined_station_distance_is_nearest_of_any_rail_mode() -> None:
+    """The combined "Any station" distance is the nearest of Rail/Tube/DLR/Tram,
+    materialized as its own column so the server loads it directly (no runtime
+    synthesis). A far rail station and a near DLR stop: the combined distance
+    must follow the DLR stop."""
+    postcodes = pl.DataFrame(
+        {"postcode": ["E14 5AB"], "lat": [51.5054], "lon": [-0.0235]}
+    )
+    pois = pl.DataFrame(
+        {
+            "category": ["Rail station", "DLR station"],
+            "group": ["Public Transport", "Public Transport"],
+            "lat": [51.6000, 51.5055],
+            "lng": [-0.2000, -0.0236],
+        }
+    )
+
+    distance_groups = {
+        "poi_rail_station": ["Rail station"],
+        COMBINED_STATION_GROUP_KEY: COMBINED_STATION_SOURCE_CATEGORIES,
+    }
+    result = min_distance_per_postcode(postcodes, pois, groups=distance_groups)
+    renames = _dynamic_poi_metric_renames(
+        {COMBINED_STATION_GROUP_KEY: COMBINED_STATION_DISPLAY_NAME}
+    )
+    result = result.rename({k: v for k, v in renames.items() if k in result.columns})
+
+    combined = result["Distance to nearest amenity (Any station) (km)"][0]
+    rail_only = result["poi_rail_station_nearest_km"][0]
+    # The near DLR stop is within a few hundred metres; the only rail station is
+    # kilometres away. The combined distance must follow the DLR stop.
+    assert combined < 0.2
+    assert rail_only > 5.0
+    assert combined < rail_only
+
+
 def test_dynamic_poi_metric_renames_support_park_count_options() -> None:
     assert _dynamic_poi_metric_renames({"parks": "Park"}) == {
         "parks_nearest_km": "Distance to nearest amenity (Park) (km)",