shapely error for invalid polygon

[srose89]

Received this error: TopologyException: unable to assign free hole to a shell at 5718.16259765625 3264.425048828125

Appears this could be due to self intersecting polygons.

Updated fix_invalid_geometry to the following to handle this and now sample completes successfully.

def fix_self_intersection(poly):
    """
    Attempts to fix self-intersecting polygons using buffer(0).
    Returns the fixed Polygon, or None if fixing failed or result is not a Polygon.
    """
    if poly.is_valid:
        return poly

    try:
        # buffer(0) is a common trick to fix self-intersections
        fixed_poly = poly.buffer(0)

        # buffer(0) might return MultiPolygon - take largest component
        if fixed_poly.geom_type == 'MultiPolygon':
            fixed_poly = max(fixed_poly.geoms, key=lambda p: p.area)
        
        # Ensure the result is actually a Polygon (not Point/LineString/Empty)
        if fixed_poly.geom_type == 'Polygon':
            return fixed_poly
        
        return None
    except Exception:
        return None

def fix_invalid_geometry(gdf: gpd.GeoDataFrame):
    """
    Fix invalid geometries by first attempting to resort coordinates,
    and then attempting to fix self-intersections via buffer(0) for those
    that remain invalid.
    """
    # 1. Identify initial invalid geometries
    mask = ~gdf.geometry.is_valid
    if not mask.any():
        return gdf
    
    # 2. First attempt: Resort coordinates
    fixed_step1 = gdf.loc[mask].geometry.apply(resort_coordinates)
    gdf.loc[mask, gdf.geometry.name] = fixed_step1

    # 3. Check if any are STILL invalid after resort
    sub_gdf = gdf.loc[mask]
    mask_still_invalid = ~sub_gdf.geometry.is_valid
    
    # If everything is fixed, we are done
    if not mask_still_invalid.any():
        return gdf

    # Get the global indices of rows that are still invalid
    problem_indices = sub_gdf[mask_still_invalid].index

    # 4. Second attempt: Fix self-intersections (buffer(0))
    fixed_step2 = gdf.loc[problem_indices].geometry.apply(fix_self_intersection)
    
    # Update the GeoDataFrame
    # Note: this will assign None if the fix failed.
    gdf.loc[problem_indices, gdf.geometry.name] = fixed_step2

    return gdf