Faster tests

pynsee/geodata/translate_and_zoom.py

@@ -1,5 +1,6 @@
 import logging
 import math
+import re
 from typing import Optional
 
 import pandas as pd
@@ -10,12 +11,55 @@
 from ._make_offshore_points import _make_offshore_points
 from ._rescale_geom import _rescale_geom
 from ._get_center import _get_center
-from ._add_insee_dep import _add_insee_dep
-
+from ._get_geodata_with_backup import _get_geodata_with_backup
+from pynsee.utils.save_df import save_df
 
 logger = logging.getLogger(__name__)
 
 
+@save_df(cls=GeoDataFrame, day_lapse_max=30)
+def _deps_with_valid_coverage() -> GeoDataFrame:
+    """
+    Inner function used to create a geodataframe of french departments, safe to
+    use for a spatial join.
+
+    First (if overlaps detected) a valid coverage is enforced (meaning polygons
+    do not overlap and are sharing edges) with a simplification of 1 meter.
+    After that, a negative 10 meters buffer is applied to prevent duplicates
+    when running a spatial join with intersects predicate.
+    This function uses a 30 day cache storage.
+
+    Returns
+    -------
+    GeoDataFrame
+        Departments' geodataset
+        Only two columns : code_insee_du_departement, insee_dep_geometry
+
+    """
+    dataset_id = "ADMINEXPRESS-COG-CARTO.LATEST:departement"
+    dep = _get_geodata_with_backup(dataset_id).to_crs("EPSG:3857")
+    dep = dep[["code_insee", "geometry"]].rename(
+        columns={
+            "code_insee": "code_insee_du_departement",
+            "geometry": "insee_dep_geometry",
+        }
+    )
+    dep = dep.set_geometry("insee_dep_geometry")
+
+    x = dep.sjoin(dep, how="left", predicate="overlaps").query(
+        "~code_insee_du_departement_right.isnull()"
+    )
+    if not x.empty:
+        # Geometries overlap:
+        # Force coverage validity (non-overlapping, edge-matched polygons)
+        # (This shouldn't be necessary with modern ADMINEXPRESS datasets and
+        # is here as a backup safe)
+        dep["insee_dep_geometry"] = dep.simplify_coverage(1)
+    dep["insee_dep_geometry"] = dep["insee_dep_geometry"].buffer(-1)
+
+    return dep
+
+
 def transform_overseas(
     gdf: GeoDataFrame,
     departement: tuple[str, ...] = ("971", "972", "974", "973", "976", "NR"),
@@ -74,7 +118,46 @@ def transform_overseas(
         gdf["code_insee_du_departement"] = gdf["code_insee"]
         gdf["insee_dep_geometry"] = gdf["geometry"]
     else:
-        gdf = _add_insee_dep(gdf.copy()).reset_index(drop=True)
+
+        # retrieve safe deps geometries
+        dep_cov = _deps_with_valid_coverage()
+
+        # detect INSEE department's codes
+        # (available in some of IGN's geodatasets, but with different patterns)
+        pattern = "codes?.*?insee.*?departement"
+        dep = gdf.columns[gdf.columns.str.match(pattern, flags=re.IGNORECASE)]
+        if not dep.empty:
+            dep = dep[0]
+            deps = gdf[dep].str.extractall("([0-9AB]{2,3})")
+            deps.index = deps.index.droplevel(-1)
+            deps = deps.loc[~deps.index.duplicated(keep="first")]
+            gdf["code_insee_du_departement"] = deps[0]
+
+        else:
+
+            # retrieve department's codes using a spatial join. The negative
+            # buffer on deps should be safe to be used without any duplication
+            # of gdf, except when there is a valid overlapping (ie regions
+            # covering multiple deps, interdep epcis...).
+            # In case of duplication (for instance, starting from REGION) keep
+            # only first dep geometry
+            gdf = gdf.sjoin(dep_cov, how="left")
+            gdf = gdf.loc[~gdf.index.duplicated(keep="first")]
+            gdf = gdf.drop("index_right", axis=1)
+
+        gdf["code_insee_du_departement"] = gdf[
+            "code_insee_du_departement"
+        ].fillna("NR")
+
+        # Retrieve simplified geometries for deps. Note that it used a negative
+        # buffer (10 meters) which should not alter geographic transformations
+        # given it's range
+        gdf = gdf.merge(dep_cov, on="code_insee_du_departement", how="left")
+
+        # where dep geom is still missing, use initial geometry
+        gdf["insee_dep_geometry"] = gdf["insee_dep_geometry"].combine_first(
+            gdf.geometry
+        )
 
     offshore_points = _make_offshore_points(
         center=Point(center),

pynsee/localdata/get_population.py

@@ -1,6 +1,7 @@
 # -*- coding: utf-8 -*-
 
 from functools import lru_cache
+import warnings
 
 from pynsee.geodata.get_geodata import get_geodata
 
@@ -14,6 +15,13 @@ def get_population():
         >>> pop = get_population()
     """
 
+    warnings.warn(
+        "get_population is deprecated. Please use "
+        "`from pynsee import get_geodata;get_geodata('ADMINEXPRESS-COG-CARTO.LATEST:commune')` "
+        "instead",
+        category=FutureWarning,
+    )
+
     df = get_geodata("ADMINEXPRESS-COG-CARTO.LATEST:commune")
 
     return df

pyproject.toml

@@ -20,7 +20,7 @@ keywords = [
     "INSEE", "IGN", "statistics", "demography", "geospatial", "France", "GIS",
     "statistique", "démographie", "géospatial", "SIG"
 ]
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 dynamic = ["dependencies", "optional-dependencies", "version"]
 classifiers = [
     "Development Status :: 5 - Production/Stable",

requirements.txt

@@ -1,11 +1,11 @@
 pandas >= 2.2.0, < 4.0
-geopandas >= 1.0, < 2.0
+geopandas >= 1.1.0, < 2.0
+shapely >= 2.1
 pyarrow >= 17.0
 tqdm >= 4.56
 platformdirs >= 4
 unidecode >= 1.3
 urllib3 >= 2.2
-shapely >= 2
 requests >= 2.32
 requests-ratelimiter >= 0.7
 openpyxl

tests/localdata/test_pynsee_localdata.py

@@ -14,7 +14,6 @@
 from pynsee.localdata.get_local_data import get_local_data
 from pynsee.localdata.get_nivgeo_list import get_nivgeo_list
 from pynsee.localdata.get_local_metadata import get_local_metadata
-from pynsee.localdata.get_population import get_population
 from pynsee.localdata.get_old_city import get_old_city
 from pynsee.localdata.get_new_city import get_new_city
 from pynsee.localdata.get_area_projection import get_area_projection
@@ -29,11 +28,6 @@
 
 class TestFunction(TestCase):
 
-    def test_get_population(self):
-        df = get_population()
-        test = isinstance(df, pd.DataFrame)
-        self.assertTrue(test)
-
     def test_get_insee_one_area_1(self):
         def get_insee_one_area_test(area_type="derf", codearea="c"):
             _get_insee_one_area(area_type=area_type, codearea=codearea)