Retain unique attribute values when consolidating nodes

CHANGELOG.md

@@ -16,6 +16,9 @@ Read the v2 [migration guide](https://github.com/gboeing/osmnx/issues/1123)
 - remove save_graph_xml function's node_tags, node_attrs, edge_tags, edge_attrs, merge_edges, oneway, api_version, and precision parameters (#1135)
 - make save_graph_xml function accept only an unsimplified MultiDiGraph as its input data (#1135)
 - replace save_graph_xml function's edge_tag_aggs tuple parameter with way_tag_aggs dict parameter (#1135)
+- make consolidate_intersections function retain unique attribute values when consolidating nodes (#1144)
+- add OSM junction and railway tags to the default settings.useful_tags_node (#1144)
+- fix graph projection creating useless lat and lon node attributes (#1144)
 - make optional function parameters keyword-only throughout package (#1134)
 - make dist function parameters required rather than optional throughout package (#1134)
 - make which_result function parameter consistently able to accept a list throughout package (#1113)

osmnx/_osm_xml.py

@@ -212,17 +212,14 @@ def _save_graph_xml(
             else:
                 gdf[col] = gdf[col].fillna(value)
 
-    # warn user if graph is projected then remove lat/lon gdf_nodes columns if
-    # they exist, as x/y cols will be saved as lat/lon node attributes instead
+    # warn user if graph is projected
     if projection.is_projected(G.graph["crs"]):
         msg = (
-            "Graph should be unprojected: the existing lat-lon node attributes will "
-            "be discarded and the projected x-y coordinates will be saved as lat-lon "
-            "node attributes instead. Project your graph back to lat-lon to avoid this."
+            "Graph should be unprojected: the existing projected x-y coordinates "
+            "will be saved as lat-lon node attributes. Project your graph back to "
+            "lat-lon to avoid this."
         )
         warn(msg, category=UserWarning, stacklevel=2)
-    for col in set(gdf_nodes.columns) & {"lat", "lon"}:
-        gdf_nodes = gdf_nodes.drop(columns=[col])
 
     # transform nodes gdf to meet OSM XML spec
     # 1) reset index (osmid) then rename osmid, x, and y columns
@@ -279,7 +276,12 @@ def _add_nodes_xml(
         node_element = SubElement(parent, "node", attrib=attrs)
 
         # add each node tag dict as its own SubElement of the node SubElement
-        tags = ({"k": k, "v": str(node[k])} for k in node_tags & node.keys() if pd.notna(node[k]))
+        # for vals that are non-null (or list if node consolidation was done)
+        tags = (
+            {"k": k, "v": str(node[k])}
+            for k in node_tags & node.keys()
+            if isinstance(node[k], list) or pd.notna(node[k])
+        )
         for tag in tags:
             _ = SubElement(node_element, "tag", attrib=tag)
 

osmnx/io.py

@@ -190,8 +190,6 @@ def load_graphml(
     default_node_dtypes = {
         "elevation": float,
         "elevation_res": float,
-        "lat": float,
-        "lon": float,
         "osmid": int,
         "street_count": int,
         "x": float,

osmnx/projection.py

@@ -163,12 +163,6 @@ def project_graph(
     # STEP 1: PROJECT THE NODES
     gdf_nodes = utils_graph.graph_to_gdfs(G, edges=False)
 
-    # create new lat/lon columns to preserve lat/lon for later reference if
-    # cols do not already exist (ie, don't overwrite in later re-projections)
-    if "lon" not in gdf_nodes.columns or "lat" not in gdf_nodes.columns:
-        gdf_nodes["lon"] = gdf_nodes["x"]
-        gdf_nodes["lat"] = gdf_nodes["y"]
-
     # project the nodes GeoDataFrame and extract the projected x/y values
     gdf_nodes_proj = project_gdf(gdf_nodes, to_crs=to_crs)
     gdf_nodes_proj["x"] = gdf_nodes_proj["geometry"].x

osmnx/settings.py

@@ -114,7 +114,7 @@
     API repeatedly for the same request. Default is `True`.
 useful_tags_node : list[str]
     OSM "node" tags to add as graph node attributes, when present in the data
-    retrieved from OSM. Default is `["highway", "ref"]`.
+    retrieved from OSM. Default is `["highway", "junction", "railway", "ref"]`.
 useful_tags_way : list[str]
     OSM "way" tags to add as graph edge attributes, when present in the data
     retrieved from OSM. Default is `["access", "area", "bridge", "est_width",
@@ -162,7 +162,7 @@
 requests_kwargs: dict[str, Any] = {}
 requests_timeout: float = 180
 use_cache: bool = True
-useful_tags_node: list[str] = ["highway", "ref"]
+useful_tags_node: list[str] = ["highway", "junction", "railway", "ref"]
 useful_tags_way: list[str] = [
     "access",
     "area",

osmnx/simplification.py

@@ -13,6 +13,7 @@
 from shapely.geometry import Point
 from shapely.geometry import Polygon
 
+from . import settings
 from . import stats
 from . import utils
 from . import utils_graph
@@ -511,7 +512,7 @@ def _merge_nodes_geometric(G: nx.MultiDiGraph, tolerance: float) -> gpd.GeoSerie
     return gpd.GeoSeries(merged.geoms, crs=G.graph["crs"])
 
 
-def _consolidate_intersections_rebuild_graph(  # noqa: PLR0912,PLR0915
+def _consolidate_intersections_rebuild_graph(  # noqa: C901,PLR0912,PLR0915
     G: nx.MultiDiGraph,
     tolerance: float,
     reconnect_edges: bool,  # noqa: FBT001
@@ -563,7 +564,9 @@ def _consolidate_intersections_rebuild_graph(  # noqa: PLR0912,PLR0915
     # attach each node to its cluster of merged nodes. first get the original
     # graph's node points then spatial join to give each node the label of
     # cluster it's within. make cluster labels type string.
-    node_points = utils_graph.graph_to_gdfs(G, edges=False)[["geometry"]]
+    node_points = utils_graph.graph_to_gdfs(G, edges=False)
+    cols = set(node_points.columns).intersection(["geometry", *settings.useful_tags_node])
+    node_points = node_points[list(cols)]
     gdf = gpd.sjoin(node_points, node_clusters, how="left", predicate="within")
     gdf = gdf.drop(columns="geometry").rename(columns={"index_right": "cluster"})
     gdf["cluster"] = gdf["cluster"].astype(str)
@@ -608,14 +611,23 @@ def _consolidate_intersections_rebuild_graph(  # noqa: PLR0912,PLR0915
             osmid = osmids[0]
             H.add_node(cluster_label, osmid_original=osmid, **G.nodes[osmid])
         else:
-            # if cluster is multiple merged nodes, create one new node to
-            # represent them
-            H.add_node(
-                cluster_label,
-                osmid_original=str(osmids),
-                x=nodes_subset["x"].iloc[0],
-                y=nodes_subset["y"].iloc[0],
-            )
+            # if cluster is multiple merged nodes, create one new node with
+            # attributes to represent them
+            node_attrs = {
+                "osmid_original": osmids,
+                "x": nodes_subset["x"].iloc[0],
+                "y": nodes_subset["y"].iloc[0],
+            }
+            for col in set(nodes_subset.columns).intersection(settings.useful_tags_node):
+                # get the unique non-null values (we won't add null attrs)
+                unique_vals = list(set(nodes_subset[col].dropna()))
+                if len(unique_vals) == 1:
+                    # if there's 1 unique value for this attribute, keep it
+                    node_attrs[col] = unique_vals[0]
+                elif len(unique_vals) > 1:
+                    # if there are multiple unique values, keep all uniques
+                    node_attrs[col] = unique_vals
+            H.add_node(cluster_label, **node_attrs)
 
     # calculate street_count attribute for all nodes lacking it
     null_nodes = [n for n, sc in H.nodes(data="street_count") if sc is None]