Fixes for geopandas-1.0.0 (#98)

.github/workflows/tests.yml

@@ -48,6 +48,7 @@ jobs:
           pytest -v -n2 --cov --cov-append
 
       - name: Run doctest
+        if: matrix.python-version == '3.12'
         run: |
           pytest -v swn --doctest-modules
 

swn/compat.py

@@ -3,6 +3,7 @@
 import contextlib
 import warnings
 
+import geopandas
 import numpy as np
 import shapely
 from packaging.version import Version
@@ -60,3 +61,30 @@ def ignore_shapely_warnings_for_object_array():
     @contextlib.contextmanager
     def ignore_shapely_warnings_for_object_array():
         yield
+
+
+GEOPANDAS_GE_100 = Version(geopandas.__version__) >= Version("1.0.0")
+
+
+def sjoin_idx_names(left_df, right_df):
+    """Returns left and right index names from geopandas.sjoin methods.
+
+    Handles breaking change from geopandas 1.0.0.
+    """
+    left_idx_name = left_df.index.name or "index"
+    if GEOPANDAS_GE_100:
+        right_idx_name = right_df.index.name or "index"
+        # add _left/_right if needed
+        if left_df.index.name and (
+            left_idx_name == right_idx_name or left_idx_name in right_df.columns
+        ):
+            left_idx_name += "_left"
+        if (
+            right_df.index.name is None
+            or right_idx_name in left_df.columns
+            or right_idx_name == left_df.index.name
+        ):
+            right_idx_name += "_right"
+    else:
+        right_idx_name = "index_right"
+    return left_idx_name, right_idx_name

swn/core.py

@@ -14,7 +14,11 @@
 import shapely
 from shapely.geometry import LineString, Point
 
-from .compat import SHAPELY_GE_20, ignore_shapely_warnings_for_object_array
+from .compat import (
+    SHAPELY_GE_20,
+    ignore_shapely_warnings_for_object_array,
+    sjoin_idx_names,
+)
 from .spatial import bias_substring
 from .util import abbr_str
 
@@ -234,13 +238,13 @@ def from_lines(cls, lines, polygons=None):
         end_pts = obj.segments.interpolate(1.0, normalized=True)
         start_df = start_pts.to_frame("start").set_geometry("start")
         end_df = end_pts.to_frame("end").set_geometry("end")
-        segidxname = obj.segments.index.name or "index"
         # This is the main component of the algorithm
+        end_idx_name, start_idx_name = sjoin_idx_names(end_df, start_df)
         jxn = pd.DataFrame(
             geopandas.sjoin(end_df, start_df, "inner", "intersects")
             .drop(columns="end")
             .reset_index()
-            .rename(columns={segidxname: "end", "index_right": "start"})
+            .rename(columns={end_idx_name: "end", start_idx_name: "start"})
         )
         # Group end points to start points, list should only have 1 item
         to_segnum_l = jxn.groupby("end")["start"].agg(list)
@@ -281,10 +285,11 @@ def from_lines(cls, lines, polygons=None):
         # Find outlets that join to a single coodinate
         multi_outlets = set()
         out_pts = end_pts.loc[outlets].to_frame("out").set_geometry("out")
+        left_idx_name, right_idx_name = sjoin_idx_names(out_pts, out_pts)
         jout = pd.DataFrame(
             geopandas.sjoin(out_pts, out_pts, "inner")
             .reset_index()
-            .rename(columns={segidxname: "out1", "index_right": "out2"})
+            .rename(columns={left_idx_name: "out1", right_idx_name: "out2"})
         ).query("out1 != out2")
         if jout.size > 0:
             # Just evaluate 2D tuple to find segnums with same location
@@ -306,11 +311,12 @@ def from_lines(cls, lines, polygons=None):
                 obj.warnings.append(m[0] % m[1:])
                 multi_outlets |= v
         # Find outlets that join to middle of other segments
+        left_idx_name, right_idx_name = sjoin_idx_names(out_pts, obj.segments)
         joutseg = pd.DataFrame(
             geopandas.sjoin(out_pts, obj.segments[["geometry"]], "inner")
             .drop(columns="out")
             .reset_index()
-            .rename(columns={segidxname: "out", "index_right": "segnum"})
+            .rename(columns={left_idx_name: "out", right_idx_name: "segnum"})
         )
         for r in joutseg.query("out != segnum").itertuples():
             if r.out in multi_outlets:
@@ -320,10 +326,11 @@ def from_lines(cls, lines, polygons=None):
             obj.errors.append(m[0] % m[1:])
         # Find headwater that join to a single coodinate
         hw_pts = start_pts.loc[headwater].to_frame("hw").set_geometry("hw")
+        hw_idx_name, start_idx_name = sjoin_idx_names(hw_pts, start_df)
         jhw = pd.DataFrame(
             geopandas.sjoin(hw_pts, start_df, "inner")
             .reset_index()
-            .rename(columns={segidxname: "hw1", "index_right": "start"})
+            .rename(columns={hw_idx_name: "hw1", start_idx_name: "start"})
         ).query("hw1 != start")
         obj.jhw = jhw
         if jhw.size > 0:
@@ -1112,9 +1119,8 @@ def locate_geoms(
                 catchments_df = self.catchments.to_frame("geometry")
                 if catchments_df.crs is None and self.segments.crs is not None:
                     catchments_df.crs = self.segments.crs
-                match_s = geopandas.sjoin(res[sel], catchments_df, "inner")[
-                    "index_right"
-                ]
+                _, ridxn = sjoin_idx_names(res, catchments_df)
+                match_s = geopandas.sjoin(res[sel], catchments_df, "inner")[ridxn]
                 match_s.name = "segnum"
                 match_s.index.name = "gidx"
                 match = match_s.reset_index()
@@ -1200,9 +1206,10 @@ def find_downstream_in_min_stream_order(segnum):
                 )
             try:
                 # faster method, not widely available
+                _, right_idx_name = sjoin_idx_names(res, self.segments)
                 match_s = geopandas.sjoin_nearest(
                     res[sel], segments_gs.to_frame(), "inner"
-                )["index_right"]
+                )[right_idx_name]
                 has_sjoin_nearest = True
             except (AttributeError, NotImplementedError):
                 has_sjoin_nearest = False
@@ -1305,7 +1312,13 @@ def find_downstream_in_min_stream_order(segnum):
             linestring_empty = wkt.loads("LINESTRING EMPTY")
             for idx in res[~sel].index:
                 res.at[idx, "link"] = linestring_empty
-        res.set_geometry("link", drop=True, inplace=True)
+        res = (
+            res.set_geometry("link")
+            .drop(columns="geometry")
+            .rename_geometry("geometry")
+        )
+        if geom_crs:
+            res.set_crs(geom_crs, inplace=True)
         res["dist_to_seg"] = res[sel].length
         return res
 

swn/modflow/_base.py

@@ -930,10 +930,8 @@ def do_linemerge(ij, df, drop_reach_ids):
             if is_spatial:
                 try:
                     match_s = geopandas.sjoin_nearest(
-                        diversions_in_model, obj.grid_cells, "inner"
-                    )[["index_right0", "index_right1"]].rename(
-                        columns={"index_right0": "i", "index_right1": "j"}
-                    )
+                        diversions_in_model, obj.grid_cells.reset_index(), "inner"
+                    )[["i", "j"]]
                     match_s.index.name = "divid"
                     match = match_s.reset_index()
                     has_sjoin_nearest = True

swn/spatial.py

@@ -511,10 +511,10 @@ def location_pair_geoms(pairs, loc_df, n):
     >>> pair_gdf.sort_values("length", ascending=False, inplace=True)
     >>> pair_gdf
                                                     geometry      length
-    11 14  LINESTRING (378.491 404.717, 420.000 330.000, ...  282.359779
-    12 13  LINESTRING (728.462 227.692, 710.000 160.000, ...  184.465938
-    15 13  LINESTRING (692.027 172.838, 710.000 160.000, ...  136.388347
-    14 15      LINESTRING (595.730 241.622, 692.027 172.838)  118.340096
+    11 14  LINESTRING (378.491 404.717, 420 330, 584 250,...  282.359779
+    12 13  LINESTRING (728.462 227.692, 710 160, 770 100,...  184.465938
+    15 13  LINESTRING (692.027 172.838, 710 160, 770 100,...  136.388347
+    14 15       LINESTRING (595.73 241.622, 692.027 172.838)  118.340096
 
     """
     from shapely.ops import linemerge, substring, unary_union

tests/test_basic.py

@@ -9,7 +9,7 @@
 from shapely.geometry import LineString, Point
 
 import swn
-from swn.compat import ignore_shapely_warnings_for_object_array
+from swn.compat import GEOPANDAS_GE_100, ignore_shapely_warnings_for_object_array
 from swn.spatial import force_2d, round_coords
 
 from .conftest import matplotlib, plt
@@ -1237,7 +1237,11 @@ def test_locate_geoms_in_basic_swn(caplog):
     a = r2.geometry.apply(lambda x: Point(*x.coords[0]))
     assert (a.distance(gs.drop(index=16)) == 0.0).all()
     b = r2.geometry.apply(lambda x: Point(*x.coords[-1]))
-    seg_mls = n.segments.geometry[r2.segnum].unary_union
+    seg_geoms = n.segments.geometry[r2.segnum]
+    if GEOPANDAS_GE_100:
+        seg_mls = seg_geoms.union_all()
+    else:
+        seg_mls = seg_geoms.unary_union
     assert (b.distance(seg_mls) < 1e-10).all()
     # now check the empty geometry
     for k in e.keys():
@@ -1342,7 +1346,11 @@ def test_locate_geoms_only_lines(coastal_geom, coastal_swn):
     assert (r.geometry.apply(lambda g: len(g.coords)) == 2).all()
     a = r.geometry.interpolate(0.0)
     b = r.geometry.interpolate(1.0, normalized=True)
-    seg_mls = coastal_swn.segments.geometry[r.segnum].unary_union
+    seg_geoms = coastal_swn.segments.geometry[r.segnum]
+    if GEOPANDAS_GE_100:
+        seg_mls = seg_geoms.union_all()
+    else:
+        seg_mls = seg_geoms.unary_union
     assert (a.distance(coastal_geom) < 1e-10).all()
     assert (a.distance(seg_mls) > 0.0).all()
     assert (b.distance(coastal_geom) > 0.0).all()
@@ -1407,7 +1415,11 @@ def test_locate_geoms_with_catchments(coastal_geom, coastal_swn_w_poly):
     assert (r.geometry.apply(lambda g: len(g.coords)) == 2).all()
     a = r.geometry.interpolate(0.0)
     b = r.geometry.interpolate(1.0, normalized=True)
-    seg_mls = coastal_swn_w_poly.segments.geometry[r.segnum].unary_union
+    seg_geoms = coastal_swn_w_poly.segments.geometry[r.segnum]
+    if GEOPANDAS_GE_100:
+        seg_mls = seg_geoms.union_all()
+    else:
+        seg_mls = seg_geoms.unary_union
     assert (a.distance(coastal_geom) < 1e-10).all()
     assert (a.distance(seg_mls) > 0.0).all()
     assert (b.distance(coastal_geom) > 0.0).all()

tests/test_compat.py

@@ -0,0 +1,50 @@
+"""Test compat module."""
+
+import geopandas
+from shapely.geometry import Point
+
+from swn import compat
+
+
+def test_sjoin_idx_names():
+    noname_idx_df = geopandas.GeoDataFrame(geometry=[Point(0, 1)])
+    left_idx_name, right_idx_name = compat.sjoin_idx_names(noname_idx_df, noname_idx_df)
+    assert left_idx_name == "index"
+    assert right_idx_name == "index_right"
+    sj = geopandas.sjoin(noname_idx_df, noname_idx_df)
+    assert sj.index.name is None
+    assert left_idx_name not in sj.columns
+    assert right_idx_name in sj.columns
+
+    name_idx_df = noname_idx_df.copy()
+    name_idx_df.index.name = "idx"
+    left_idx_name, right_idx_name = compat.sjoin_idx_names(name_idx_df, name_idx_df)
+    if compat.GEOPANDAS_GE_100:
+        assert left_idx_name == "idx_left"
+        assert right_idx_name == "idx_right"
+    else:
+        assert left_idx_name == "idx"
+        assert right_idx_name == "index_right"
+    sj = geopandas.sjoin(name_idx_df, name_idx_df)
+    assert sj.index.name == left_idx_name
+    assert left_idx_name not in sj.columns
+    assert right_idx_name in sj.columns
+
+    left_idx_name, right_idx_name = compat.sjoin_idx_names(noname_idx_df, name_idx_df)
+    assert left_idx_name == "index"
+    if compat.GEOPANDAS_GE_100:
+        assert right_idx_name == "idx"
+    else:
+        assert right_idx_name == "index_right"
+    sj = geopandas.sjoin(noname_idx_df, name_idx_df)
+    assert sj.index.name is None
+    assert left_idx_name not in sj.columns
+    assert right_idx_name in sj.columns
+
+    left_idx_name, right_idx_name = compat.sjoin_idx_names(name_idx_df, noname_idx_df)
+    assert left_idx_name == "idx"
+    assert right_idx_name == "index_right"
+    sj = geopandas.sjoin(name_idx_df, noname_idx_df)
+    assert sj.index.name == left_idx_name
+    assert left_idx_name not in sj.columns
+    assert right_idx_name in sj.columns