SPlit up

pipeline/download/places.py

@@ -329,16 +329,24 @@ def _outcode_of_postcode(postcode: str) -> str:
 
 def _outcode_tree(postcodes_path: Path) -> tuple[cKDTree, list[str]]:
     """Build a nearest-neighbour index from postcode coordinates to their outcode, so each
-    street can be tagged with the outcode it sits in (used to disambiguate same-named roads)."""
+    street can be tagged with the outcode it sits in (used to disambiguate same-named roads).
+
+    The tree lives in BNG metres (like `_london_postcode_tree`): in raw degrees
+    1° of longitude is only ~0.6° of latitude at UK latitudes, which biases
+    nearest-postcode picks E-W near outcode boundaries."""
     df = (
         pl.read_parquet(
-            postcodes_path, columns=["pcds", "lat", "long", "ctry25cd", "doterm"]
+            postcodes_path,
+            columns=["pcds", "east1m", "north1m", "ctry25cd", "doterm"],
         )
         .filter((pl.col("ctry25cd") == ENGLAND_COUNTRY_CODE) & pl.col("doterm").is_null())
-        .filter(_valid_wgs84_expr())
+        .filter(_valid_bng_expr())
     )
     coords = np.column_stack(
-        [df["lat"].to_numpy().astype(np.float64), df["long"].to_numpy().astype(np.float64)]
+        [
+            df["east1m"].to_numpy().astype(np.float64),
+            df["north1m"].to_numpy().astype(np.float64),
+        ]
     )
     outcodes = [_outcode_of_postcode(pc) for pc in df["pcds"].to_list()]
     return cKDTree(coords), outcodes
@@ -354,8 +362,10 @@ def _build_street_places(
     if not streets:
         return []
 
-    coords = np.array([[street["lat"], street["lon"]] for street in streets], dtype=np.float64)
-    _, indices = tree.query(coords)
+    lons = np.array([street["lon"] for street in streets], dtype=np.float64)
+    lats = np.array([street["lat"] for street in streets], dtype=np.float64)
+    eastings, northings = WGS84_TO_BNG.transform(lons, lats)
+    _, indices = tree.query(np.column_stack([eastings, northings]))
 
     grouped: dict[tuple[str, str], dict] = {}
     for street, postcode_idx in zip(streets, indices):