assign_geo_identifiers raises RuntimeError: driverName computed but not passed to load_spatial_data

[jinskeep-morpc]

Summary

morpc.assign_geo_identifiers(points, ["county"]) (and ["tract"], ["zcta"], ["place"]) raises a bare RuntimeError on every call. It computes a driverName for the polygon source but never passes it to load_spatial_data(), which then cannot infer a driver for the .zip URL and raises.

Version: 0.6.2 (396e47c).

Reproduce

import morpc, geopandas as gpd
pts = gpd.GeoDataFrame({"id": [1]},
                       geometry=gpd.points_from_xy([-83.0007], [39.9612]), crs="EPSG:4326")
morpc.assign_geo_identifiers(pts, ["county"])

File ".../morpc/morpc.py", line 1778, in assign_geo_identifiers
    polys = load_spatial_data(sourcePath = filePath, layerName = layerName, verbose=False)
File ".../morpc/morpc.py", line 1455, in load_spatial_data
    raise RuntimeError
RuntimeError:

(preceded by: File extension is unsupported: .zip. It is possible to load zipped spatial data, but you must specify the driver name.)

Cause

In assign_geo_identifiers each supported geography sets a local driverName, e.g.

# morpc/morpc.py ~1730-1755
if(geography == "county"):
    filePath = "https://www2.census.gov/geo/tiger/TIGER2020PL/LAYER/COUNTY/2020/tl_2020_39_county20.zip"
    layerName = None
    driverName = "Census Shapefile"   # Custom driver name for load_spatial_data
    polyIdField = "GEOID20"

but the call at line 1778 drops it:

polys = load_spatial_data(sourcePath = filePath, layerName = layerName, verbose=False)
#                                                                       ^ no driverName=

so load_spatial_data falls into its driverName is None branch, sees .zip, and raise RuntimeError (line 1454-1455).

Two things need fixing:

1. Line 1778 should pass driverName = driverName.

2. load_spatial_data has no "Census Shapefile" branch (only GPKG / ESRI Shapefile / OpenFileGDB / SQLite, else generic gpd.read_file). So even once passed, "Census Shapefile" would land in the generic else. That path does happen to work:

   >>> gpd.read_file("https://www2.census.gov/geo/tiger/TIGER2020PL/LAYER/COUNTY/2020/tl_2020_39_county20.zip")
   # 88 rows, EPSG:4269, ~2s

   so either add an explicit "Census Shapefile" case (→ gpd.read_file(sourcePath, engine="pyogrio"), no layer=), or have the geographies pass a driver load_spatial_data already understands.

Also

assign_geo_identifiers renames with "id_{}".format(geography, polyIdField) — the second arg is unused (leftover {1}?), and the sjoin leaves an index_right column (only fid_* is dropped).

Workaround

Point-in-polygon against the TIGER shapefile directly:

counties = gpd.read_file(COUNTY_ZIP_URL)[["GEOID20", "geometry"]].to_crs(pts.crs)
pts.sjoin(counties, how="left", predicate="within")