dc-l1-tokyo-ring-orientation

Rerun

import geopandas as gpd

# Load the data
gdf = gpd.read_file('tokyo_buildings_legacy.geojson')

# Print initial info
print(f"Loaded {len(gdf)} features.")
print(gdf.head())

# Check for invalid geometries
invalid_count = (~gdf.is_valid).sum()
print(f"Invalid geometries: {invalid_count}")

# The problem description says "building interiors are shaded incorrectly: holes appear filled and exteriors render as empty space."
# This usually means the winding order of the rings is incorrect.
# For GeoJSON (RFC 7946), the exterior ring should be counter-clockwise and interior rings should be clockwise.
# However, many GIS tools use different conventions.
# Let's try to fix the orientation.

# Shapely's orient function can fix winding order.
from shapely.geometry import Polygon, MultiPolygon
from shapely.ops import orient

def fix_geometry(geom):
    if geom.is_empty:
        return geom
    if isinstance(geom, Polygon):
        # orient(polygon, sign=1.0) makes exterior ring CCW and interior rings CW
        return orient(geom, sign=1.0)
    elif isinstance(geom, MultiPolygon):
        return MultiPolygon([orient(p, sign=1.0) for p in geom.geoms])
    else:
        return geom

# Apply the orientation fix
gdf['geometry'] = gdf['geometry'].apply(fix_geometry)

# Also ensure they are valid (buffer(0) is a common trick, but orient should be enough for winding)
gdf['geometry'] = gdf['geometry'].buffer(0)

# Check invalid geometries again
invalid_count = (~gdf.is_valid).sum()
print(f"Invalid geometries after fix: {invalid_count}")

# Save the fixed file
gdf.to_file('tokyo_buildings_fixed.geojson', driver='GeoJSON')
print("Successfully saved tokyo_buildings_fixed.geojson")

import geopandas as gpd
import json

# Load original
gdf_orig = gpd.read_file('tokyo_buildings_legacy.geojson')
# Load fixed
gdf_fixed = gpd.read_file('tokyo_buildings_fixed.geojson')

# 1. Check feature count
print(f"Original count: {len(gdf_orig)}")
print(f"Fixed count: {len(gdf_fixed)}")

# 2. Check feature_id preservation (and alignment)
# We'll assume feature_id is the join key as per instructions.
gdf_orig = gdf_orig.sort_values('feature_id')
gdf_fixed = gdf_fixed.sort_values('feature_id')

if not gdf_orig['feature_id'].equals(gdf_fixed['feature_id']):
    print("ERROR: feature_id mismatch!")
else:
    print("SUCCESS: feature_id preserved.")

# 3. Check attributes
# Get column names (excluding geometry)
orig_cols = [c for c in gdf_orig.columns if c != 'geometry']
fixed_cols = [c for c in gdf_fixed.columns if c != 'geometry']

if set(orig_cols) != set(fixed_cols):
    print(f"ERROR: Attribute mismatch. Original: {orig_cols}, Fixed: {fixed_cols}")
else:
    print("SUCCESS: Attributes preserved.")

# Compare a sample of attributes
for i in range(min(5, len(gdf_orig))):
    for col in orig_cols:
        if gdf_orig.iloc[i][col] != gdf_fixed.iloc[i][col]:
             print(f"ERROR: Attribute mismatch at index {i}, col {col}")
             break
    else:
        continue
    break
else:
    print("SUCCESS: Sample attributes match.")

# 4. Check geometry structure (not values, but types)
# We want to ensure they didn't flatten holes.
def check_rings(geom):
    if geom.is_empty: return 0
    if geom.geom_type == 'Polygon':
        return 1 + len(geom.interiors)
    if geom.geom_type == 'MultiPolygon':
        return sum(1 + len(p.interiors) for p in geom.geoms)
    return 0

orig_rings = gdf_orig['geometry'].apply(check_rings).sum()
fixed_rings = gdf_fixed['geometry'].apply(check_rings).sum()

print(f"Total rings (exterior + interior) original: {orig_rings}")
print(f"Total rings (exterior + interior) fixed: {fixed_rings}")

if orig_rings != fixed_rings:
    print("ERROR: Geometry structure changed (rings lost)!")
else:
    print("SUCCESS: Geometry structure preserved.")

import geopandas as gpd
import json

# Load original
gdf_orig = gpd.read_file('tokyo_buildings_legacy.geojson')
# Load fixed
gdf_fixed = gpd.read_file('tokyo_buildings_fixed.geojson')

# Print all attributes of the first feature to inspect
print("Original feature 0 attributes:")
print(gdf_orig.iloc[0].to_dict())
print("\nFixed feature 0 attributes:")
print(gdf_fixed.iloc[0].to_dict())

# Let's re-examine the mismatch
idx = 0
col = 'height'
val_orig = gdf_orig.iloc[idx][col]
val_fixed = gdf_fixed.iloc[idx][col]
print(f"\nMismatch in '{col}' at index {idx}: {val_orig} vs {val_fixed}")

# Check types
print(f"Type orig: {type(val_orig)}, Type fixed: {type(val_fixed)}")

import geopandas as gpd
import numpy as np

# Load original
gdf_orig = gpd.read_file('tokyo_buildings_legacy.geojson')
# Load fixed
gdf_fixed = gpd.read_file('tokyo_buildings_fixed.geojson')

# Check if nan == nan is causing the "mismatch" in my validation script
def compare_dfs(df1, df2):
    # Comparing only non-geometry columns
    cols = [c for c in df1.columns if c != 'geometry']
    
    # Use pandas testing to see if they are identical
    import pandas as pd
    try:
        pd.testing.assert_frame_equal(df1[cols], df2[cols], check_dtype=False)
        print("SUCCESS: DataFrames are identical (using pd.testing.assert_frame_equal)")
    except AssertionError as e:
        print("ERROR: DataFrames are not identical:")
        print(e)

compare_dfs(gdf_orig, gdf_fixed)