src/streamlit_app.py

import streamlit as st
from streamlit_folium import st_folium
import folium
from folium.plugins import Draw
import pandas as pd
import geopandas as gpd
from shapely.geometry import Polygon, Point
import numpy as np
import re # For parsing STATEDAREA

st.set_page_config(layout="wide", page_title="Multiplex Coop Housing Filter")

st.title("🗺️ Multiplex Coop Housing Filter (Hugging Face Data)")
st.write("This app uses the `ProjectMultiplexCoop/PropertyBoundaries` dataset from Hugging Face. Draw a polygon on the map to spatially filter properties. Use the form below to apply additional filters based on property attributes. **Note: FSI, Building Coverage, Height, and Stories are synthetic for demonstration as they are not directly available in the dataset.**")

# --- 1. Load Data from Hugging Face and Process ---
@st.cache_data
def load_and_process_data():
    """
    Loads the geospatial data from Hugging Face, processes relevant columns,
    and generates synthetic data for missing attributes.
    """
    try:
        # Load the geospatial data using geopandas
        # Ensure you have 'huggingface_hub', 'geopandas', 'fiona', 'pyproj', 'shapely' installed.
        gdf = gpd.read_parquet("hf://datasets/ProjectMultiplexCoop/PropertyBoundaries/Property_Boundaries_4326.parquet")
    except Exception as e:
        st.error(f"Failed to load data from Hugging Face. Please ensure `huggingface_hub`, `geopandas`, `fiona`, and `pyproj` are installed. Error: {e}")
        st.stop()

    # Process STATEDAREA to numeric (Lot Area in Sq Metres)
    # The format is like "17366.998291 sq.m"
    def parse_stated_area(area_str):
        if pd.isna(area_str):
            return np.nan
        match = re.search(r'(\d+\.?\d*)\s*sq\.m', str(area_str))
        if match:
            return float(match.group(1))
        return np.nan

    gdf['zn_area'] = gdf['STATEDAREA'].apply(parse_stated_area)

    # Map FEATURE_TYPE to zn_type (Zoning Type)
    gdf['zn_type'] = gdf['FEATURE_TYPE']

    # Generate synthetic data for attributes not present in the Hugging Face dataset
    # but required for the filter functionality as per the original HTML.
    num_rows = len(gdf)
    gdf['fsi_total'] = np.round(np.random.uniform(0.5, 3.0, num_rows), 2)
    gdf['prcnt_cver'] = np.random.randint(20, 70, num_rows)
    gdf['height_metres'] = np.round(np.random.uniform(5, 30, num_rows), 1)
    gdf['stories'] = np.random.randint(2, 10, num_rows)

    # Add unique ID and a display name
    gdf['id'] = range(1, num_rows + 1)
    gdf['name'] = gdf['PARCELID'].apply(lambda x: f"Parcel {x}")

    # Ensure geometries are valid for centroid calculation and plotting
    # .buffer(0) is a common trick to fix minor geometry issues
    gdf['geometry'] = gdf['geometry'].buffer(0)
    # Extract centroids for point-based filtering and initial map markers
    gdf['latitude'] = gdf.geometry.centroid.y
    gdf['longitude'] = gdf.geometry.centroid.x

    # Select and reorder relevant columns for display and filtering
    df_processed = gdf[[
        'id', 'name', 'latitude', 'longitude', 'geometry',
        'zn_type', 'zn_area', 'fsi_total', 'prcnt_cver', 'height_metres', 'stories',
        'PARCELID', # Original Parcel ID for reference
        'ADDRESS_NUMBER', 'LINEAR_NAME_FULL' # For detailed address in tooltips
    ]].copy()

    return df_processed

df = load_and_process_data()

# Initialize filtered_df with the full dataframe for initial state
filtered_df = df.copy()

# --- 2. Initialize the Folium Map with Drawing Tools ---
# Center the map around the mean of the actual data's centroids
m = folium.Map(location=[df['latitude'].mean(), df['longitude'].mean()], zoom_start=12)

# Add drawing tools to the map
draw = Draw(
    export=True,
    filename="drawn_polygon.geojson",
    position="topleft",
    draw_options={
        "polyline": False, "rectangle": False, "circlemarker": False,
        "circle": False, "marker": False,
        "polygon": {
            "allowIntersection": False, # Restricts polygons to not intersect themselves
            "drawError": {"color": "#e0115f", "message": "Oups!"},
            "shapeOptions": {"color": "#ef233c", "fillOpacity": 0.5},
        },
    },
    edit_options={"edit": False, "remove": True},
)
m.add_child(draw)

# Add a sample of points to the initial map for responsiveness
# Plotting all 500k+ polygons/points at once can cause performance issues.
sample_df_for_initial_map = df.sample(min(1000, len(df)), random_state=42) # Sample up to 1000 points

for idx, row in sample_df_for_initial_map.iterrows():
    folium.CircleMarker(
        location=[row['latitude'], row['longitude']],
        radius=3, # Smaller radius for denser data points
        color='blue',
        fill=True,
        fill_color='blue',
        fill_opacity=0.5,
        tooltip=(
            f"Parcel ID: {row['PARCELID']}<br>Name: {row['name']}<br>Zoning: {row['zn_type']}<br>"
            f"Area: {row['zn_area'] if pd.notna(row['zn_area']) else 'N/A'} m²<br>"
            f"FSI: {row['fsi_total']}<br>Coverage: {row['prcnt_cver']}%<br>"
            f"Height: {row['height_metres']}m<br>Stories: {row['stories']}<br>"
            f"Address: {row['ADDRESS_NUMBER'] if pd.notna(row['ADDRESS_NUMBER']) else ''} {row['LINEAR_NAME_FULL'] if pd.notna(row['LINEAR_NAME_FULL']) else ''}"
        )
    ).add_to(m)

st.subheader("Draw a Polygon on the Map")
st.info(f"Displaying a sample of {len(sample_df_for_initial_map)} points on the map for responsiveness. All {len(df)} properties will be used for filtering.")
output = st_folium(m, width=1000, height=600, returned_objects=["all_draw_features"])

polygon_drawn = False
shapely_polygon = None
polygon_coords = None

if output and output["all_draw_features"]:
    polygons = [
        feature["geometry"]["coordinates"]
        for feature in output["all_draw_features"]
        if feature["geometry"]["type"] == "Polygon"
    ]

    if polygons:
        polygon_coords = polygons[-1][0] # Get the coordinates of the last drawn polygon
        # Shapely Polygon expects (lon, lat) tuples, Folium provides (lat, lon)
        shapely_polygon = Polygon([(lon, lat) for lat, lon in polygon_coords])
        polygon_drawn = True

        # Apply spatial filter to the full dataframe based on centroid containment
        filtered_df = df[
            df.apply(
                lambda row: shapely_polygon.contains(Point(row['longitude'], row['latitude'])),
                axis=1
            )
        ].copy() # Use .copy() to avoid SettingWithCopyWarning
        st.success(f"Initially filtered {len(filtered_df)} properties within the drawn polygon.")
    else:
        st.info("Draw a polygon on the map to spatially filter properties.")
else:
    st.info("Draw a polygon on the map to spatially filter properties.")

# --- 3. Attribute Filtering Form ---
st.subheader("Filter Property Attributes")

with st.form("attribute_filters"):
    col1, col2 = st.columns(2)

    with col1:
        # Zoning Type filter
        # Get unique zoning types from the loaded data, including a default 'All' option
        all_zoning_types = ['All Resdidential Zoning (0, 101, 6)'] + sorted(df['zn_type'].unique().tolist())
        selected_zn_type = st.selectbox("Zoning Type", all_zoning_types, key="zn_type_select")

        # Lot Area in Sq Metres filter
        # Use actual min/max from data for number input range
        min_zn_area = st.number_input(
            "Minimum Lot Area in Sq Metres",
            min_value=float(df['zn_area'].min() if pd.notna(df['zn_area'].min()) else 0),
            value=float(df['zn_area'].min() if pd.notna(df['zn_area'].min()) else 0),
            step=100.0,
            key="zn_area_input"
        )

        # Floor Space Index (FSI) filter - Synthetic data
        min_fsi_total = st.number_input("Minimum Floor Space Index (FSI)", min_value=0.0, value=0.0, step=0.1, format="%.2f", key="fsi_total_input")

    with col2:
        # Building Percent Coverage filter - Synthetic data
        max_prcnt_cver = st.number_input("Maximum Building Percent Coverage (%)", min_value=0, value=100, step=1, key="prcnt_cver_input")

        # Height or Stories selection - Synthetic data
        height_stories_option = st.radio(
            "Filter by",
            ("Height", "Stories"),
            index=0, # Default to Height
            key="height_stories_radio"
        )

        # Single input field for height/stories, label changes dynamically
        if height_stories_option == "Height":
            min_height_value = st.number_input("Minimum Height in Metres", min_value=0.0, value=0.0, step=0.1, format="%.1f", key="height_input")
        else: # Stories
            min_stories_value = st.number_input("Minimum Stories", min_value=0, value=0, step=1, key="stories_input")

    submitted = st.form_submit_button("Apply Attribute Filters")

    if submitted:
        # Apply attribute filters to the already spatially filtered_df
        if selected_zn_type != 'All Resdidential Zoning (0, 101, 6)':
            filtered_df = filtered_df[filtered_df['zn_type'] == selected_zn_type]

        # Handle NaN values for zn_area before comparison by treating NaN as 0 for min comparison
        filtered_df = filtered_df[filtered_df['zn_area'].fillna(0) >= min_zn_area]

        if min_fsi_total > 0:
            filtered_df = filtered_df[filtered_df['fsi_total'] >= min_fsi_total]

        if max_prcnt_cver < 100: # Assuming 100% means no upper limit applied
            filtered_df = filtered_df[filtered_df['prcnt_cver'] <= max_prcnt_cver]

        if height_stories_option == "Height" and min_height_value > 0:
            filtered_df = filtered_df[filtered_df['height_metres'] >= min_height_value]
        elif height_stories_option == "Stories" and min_stories_value > 0:
            filtered_df = filtered_df[filtered_df['stories'] >= min_stories_value]

        st.success(f"Applied attribute filters. Total properties after all filters: {len(filtered_df)}")
    else:
        st.info("Adjust filters and click 'Apply Attribute Filters'.")


# --- 4. Display Filtered Data on a New Map and as a Table ---
st.subheader("Filtered Properties")

if not filtered_df.empty:
    # Create a new map to show only the filtered properties
    if len(filtered_df) > 0:
        # Calculate bounds for filtered data to set appropriate zoom
        min_lat, max_lat = filtered_df['latitude'].min(), filtered_df['latitude'].max()
        min_lon, max_lon = filtered_df['longitude'].min(), filtered_df['longitude'].max()

        # Adjust map center and zoom dynamically based on filtered data extent
        if min_lat == max_lat and min_lon == max_lon:
            filtered_map_center = [min_lat, min_lon]
            filtered_map_zoom = 18 # Very close zoom for single point
        else:
            filtered_map_center = [filtered_df['latitude'].mean(), filtered_df['longitude'].mean()]
            # Simple heuristic for zoom level based on spatial extent
            lat_diff = max_lat - min_lat
            lon_diff = max_lon - min_lon
            if max(lat_diff, lon_diff) < 0.001: filtered_map_zoom = 18
            elif max(lat_diff, lon_diff) < 0.01: filtered_map_zoom = 16
            elif max(lat_diff, lon_diff) < 0.1: filtered_map_zoom = 14
            else: filtered_map_zoom = 12
    else:
        # Fallback to original map center if no data is filtered
        filtered_map_center = [df['latitude'].mean(), df['longitude'].mean()]
        filtered_map_zoom = 12

    filtered_m = folium.Map(location=filtered_map_center, zoom_start=filtered_map_zoom)

    # Add the drawn polygon to the new map if it exists
    if polygon_drawn and polygon_coords:
        folium.Polygon(
            locations=polygon_coords, # Use original (lat,lon) for folium
            color="#ef233c",
            fill=True,
            fill_color="#ef233c",
            fill_opacity=0.5
        ).add_to(filtered_m)

    # Convert filtered_df back to GeoDataFrame for direct plotting of geometries
    filtered_gdf = gpd.GeoDataFrame(filtered_df, geometry='geometry')

    # Add filtered polygons to the map as GeoJSON layer
    folium.GeoJson(
        filtered_gdf.to_json(),
        style_function=lambda x: {
            'fillColor': 'green',
            'color': 'darkgreen',
            'weight': 1,
            'fillOpacity': 0.7
        },
        tooltip=folium.GeoJsonTooltip(
            fields=['PARCELID', 'zn_type', 'zn_area', 'fsi_total', 'prcnt_cver', 'height_metres', 'stories', 'ADDRESS_NUMBER', 'LINEAR_NAME_FULL'],
            aliases=['Parcel ID:', 'Zoning Type:', 'Lot Area (m²):', 'FSI:', 'Coverage (%):', 'Height (m):', 'Stories:', 'Address Num:', 'Street:'],
            localize=True
        )
    ).add_to(filtered_m)

    st_folium(filtered_m, width=1000, height=500)

    st.subheader("Filtered Properties Table")
    # Display relevant columns in the table
    display_cols = ['PARCELID', 'zn_type', 'zn_area', 'fsi_total', 'prcnt_cver', 'height_metres', 'stories', 'ADDRESS_NUMBER', 'LINEAR_NAME_FULL']
    st.dataframe(filtered_df[display_cols])

    # --- 5. Export Data Button ---
    csv = filtered_df.to_csv(index=False).encode('utf-8')
    st.download_button(
        label="Export Filtered Data to CSV",
        data=csv,
        file_name="multiplex_coop_filtered_properties.csv",
        mime="text/csv",
    )

else:
    st.warning("No properties match the current filters. Try adjusting your criteria or drawing a different polygon.")

st.markdown("---")
st.markdown("This app demonstrates spatial and attribute filtering on the ProjectMultiplexCoop/PropertyBoundaries dataset from Hugging Face. FSI, Building Coverage, Height, and Stories are synthetic for demonstration.")