dc-l1-bangkok-attribute-coercion

Rerun

import geopandas as gpd

def investigate_geojson(file_path):
    try:
        gdf = gpd.read_file(file_path)
        print("Columns in the GeoDataFrame:")
        print(gdf.columns)
        print("\nData Types:")
        print(gdf.dtypes)
        print("\nFirst few rows:")
        print(gdf.head())
        print("\nSummary statistics for numeric columns:")
        print(gdf.describe())
        
        # Check for non-numeric values in supposedly numeric columns
        # We'll try to identify which columns are intended to be numeric but are currently objects
        for col in gdf.columns:
            if col != 'geometry':
                # Check if the column is an object (string) but contains numbers
                if gdf[col].dtype == 'object':
                    sample_values = gdf[col].dropna().unique()[:10]
                    print(f"\nColumn '{col}' is type object. Sample values: {sample_values}")
                    
        return gdf
    except Exception as e:
        print(f"Error: {e}")
        return None

gdf = investigate_geojson('bangkok_aq_stations.geojson')

import geopandas as gpd
import pandas as pd

def investigate_geojson(file_path):
    try:
        gdf = gpd.read_file(file_path)
        print("Columns in the GeoDataFrame:")
        print(gdf.columns)
        print("\nData Types:")
        print(gdf.dtypes)
        print("\nFirst few rows:")
        print(gdf.head())
        print("\nSummary statistics for numeric columns:")
        print(gdf.describe())
        
        # Check for non-numeric values in supposedly numeric columns
        # We'll try to identify which columns are intended to be numeric but are currently objects
        for col in gdf.columns:
            if col != 'geometry':
                # Check if the column is an object (string) but contains numbers
                if gdf[col].dtype == 'object':
                    sample_values = gdf[col].dropna().unique()[:10]
                    print(f"\nColumn '{col}' is type object. Sample values: {sample_values}")
                    
        return gdf
    except Exception as e:
        print(f"Error: {e}")
        return None

gdf = investigate_geojson('bangkok_aq_stations.geojson')

import geopandas as gpd
import pandas as pd

def fix_data(file_path, output_path):
    print(f"Reading {file_path}...")
    gdf = gpd.read_file(file_path)
    
    # Based on investigation, sensor_value, pm25_ug_m3, and elevation_m are strings but should be numeric.
    # station_id might be numeric too, but let's see if we should keep it as string or convert.
    # Usually IDs can be kept as strings if they have leading zeros, but here they look like integers.
    # However, the prompt says "Leave the string columns ... alone". 
    # It's ambiguous if 'station_id' is a string column or should be numeric. 
    # In the output of investigate.py, station_id was str.
    # Let's focus on the columns that definitely represent measurements.
    
    numeric_cols = ['sensor_value', 'pm25_ug_m3', 'elevation_m']
    
    for col in numeric_cols:
        if col in gdf.columns:
            print(f"Converting {col} to numeric...")
            # to_numeric will convert errors to NaN
            gdf[col] = pd.to_numeric(gdf[col], errors='coerce')
    
    # Check if station_id should be numeric. 
    # If we convert it to numeric, it might be better for some tools, 
    # but we'll stick to the clearly measurement-oriented ones first.
    # Actually, looking at the output:
    # station_id           str
    # name_th              str
    # name_en              str
    # sensor_value         str
    # pm25_ug_m3           str
    # elevation_m          str
    
    # Let's also try converting station_id to numeric if possible, 
    # but typically IDs are kept as they are. 
    # The prompt says "Leave the string columns and the geometry alone".
    # This could mean "don't change the content of the strings" OR "don't change the types of columns that are meant to be strings".
    # The core issue is "averages ... are coming back wrong ... means show up as NaN".
    # This happens because the numeric columns are currently typed as objects (strings).
    
    print("Saving to", output_path)
    gdf.to_file(output_path, driver='GeoJSON')
    print("Done.")

if __name__ == "__main__":
    fix_data('bangkok_aq_stations.geojson', 'bangkok_aq_typed.geojson')

import geopandas as gpd

def verify_fix(file_path):
    print(f"Verifying {file_path}...")
    gdf = gpd.read_file(file_path)
    print("Data Types:")
    print(gdf.dtypes)
    print("\nSummary statistics for numeric columns:")
    print(gdf.describe())
    
    # Check if any of the measurement columns are still objects
    numeric_cols = ['sensor_value', 'pm25_ug_m3', 'elevation_m']
    for col in numeric_cols:
        if col in gdf.columns:
            if not pd.api.types.is_numeric_dtype(gdf[col]):
                print(f"WARNING: {col} is still {gdf[col].dtype}")
            else:
                print(f"SUCCESS: {col} is {gdf[col].dtype}")

import pandas as pd
verify_fix('bangkok_aq_typed.geojson')