src/streamlit_app.py

"""
Streamlit UI for the Bernalillo County Urban Opportunity Index,
designed for legal aid staff and community members in Bernalillo County.

Run with:
    streamlit run ui.py
"""

import os

import geopandas as gpd
import pandas as pd
import streamlit as st
import folium
from streamlit_folium import st_folium

from bern_uoi_pipeline import build_bern_uoi, ACS_YEAR, OUTPUT_DIR

DATA_GPKG = os.path.join(OUTPUT_DIR, "bern_uoi_tracts.gpkg")
DATA_LAYER = "tracts"


# =========================
# DATA LOADING
# =========================

@st.cache_data(show_spinner=True)
def load_data() -> gpd.GeoDataFrame:
    """
    Load Bernalillo UOI data from GeoPackage if present.
    If key fields (like eviction_resilience_score) are missing, rebuild.
    """
    if os.path.exists(DATA_GPKG):
        gdf = gpd.read_file(DATA_GPKG, layer=DATA_LAYER)
        print(f"Loaded existing {DATA_GPKG}")

        # If this is an older file without eviction columns, rebuild
        if "eviction_resilience_score" not in gdf.columns:
            print("Old data detected → rebuilding Bernalillo UOI dataset...")
            gdf = build_bern_uoi(refresh=True)
    else:
        gdf = build_bern_uoi(refresh=True)

    return gdf


# =========================
# MAP BUILDER
# =========================

def make_choropleth(gdf: gpd.GeoDataFrame, color_col: str, legend_name: str) -> folium.Map:
    """
    Build a simple choropleth map for the chosen indicator.
    Values are clipped to the 5th–95th percentile range so that
    a few extreme tracts do not distort the color scale.
    """
    # Map center = mean of tract centroids
    centroids = gdf.geometry.centroid
    center_lat = centroids.y.mean()
    center_lon = centroids.x.mean()

    # Copy and create a clipped plotting column
    gdf_plot = gdf.copy()

    # Use to_numeric to handle strings like "<NA>"
    series = pd.to_numeric(gdf_plot[color_col], errors="coerce")

    if series.notna().sum() > 0:
        q_low = series.quantile(0.05)
        q_high = series.quantile(0.95)
        gdf_plot["_plot_val"] = series.clip(lower=q_low, upper=q_high)
    else:
        gdf_plot["_plot_val"] = series

    m = folium.Map(
        location=[center_lat, center_lon],
        zoom_start=11,
        tiles="CartoDB positron",
    )

    folium.Choropleth(
        geo_data=gdf_plot.to_json(),
        data=gdf_plot,
        columns=["GEOID", "_plot_val"],
        key_on="feature.properties.GEOID",
        fill_color="YlGnBu",
        fill_opacity=0.8,
        line_opacity=0.3,
        nan_fill_opacity=0.15,
        legend_name=legend_name,
    ).add_to(m)

    # Hover tooltip uses the original (unclipped) values
    folium.GeoJson(
        gdf,
        style_function=lambda x: {"fillOpacity": 0, "color": "#444", "weight": 0.4},
        tooltip=folium.features.GeoJsonTooltip(
            fields=["NAME", color_col],
            aliases=["Tract:", legend_name + ":"],
            localize=True,
        ),
    ).add_to(m)

    folium.LayerControl().add_to(m)
    return m


# =========================
# APP LAYOUT
# =========================

def main():
    st.set_page_config(
        page_title="Bernalillo County Urban Opportunity Index",
        layout="wide",
    )

    st.title("Bernalillo County Opportunity Map")

    st.markdown(
        """
        This map shows how different parts of Bernalillo County compare on everyday basics like:

        - **Internet at home**
        - **Housing and rent**
        - **Health coverage**
        - **Poverty and income**
        - **Disability and education**
        - **Eviction risk**

        It’s meant to help **community members and legal aid staff** see where needs are greatest and
        where people may need extra support.
        """
    )

    gdf = load_data()

    # ----- SIDEBAR CONTROLS -----
    st.sidebar.header("Pick what to see")

    group = st.sidebar.radio(
        "What do you want to look at?",
        (
            "Overall opportunity",
            "Internet access",
            "Housing costs",
            "Health coverage",
            "Money & poverty",
            "Income",
            "Disability",
            "Education",
            "Eviction risk",
        ),
    )

    value_type = st.sidebar.radio(
        "How should the numbers be shown?",
        (
            "Opportunity score (0–1, higher = better)",
            "Original value (as collected)",
        ),
    )

    # ----- COLUMN SELECTION -----
    if group == "Overall opportunity":
        col_norm = "uoi_score"
        col_raw = "uoi_score"  # same numbers, just labeled differently
        label = "Overall opportunity score (0–1)"

    elif group == "Internet access":
        col_norm = "norm_broadband"
        col_raw = "pct_broadband"
        label = "Households with home internet (%)"

    elif group == "Housing costs":
        col_norm = "norm_rent_burdened"
        col_raw = "pct_rent_burdened"
        label = "Households with high housing costs (%)"

    elif group == "Health coverage":
        col_norm = "norm_uninsured"
        col_raw = "pct_uninsured"
        label = "People without health insurance (%)"

    elif group == "Money & poverty":
        col_norm = "norm_poverty"
        col_raw = "pct_poverty"
        label = "People living below the poverty line (%)"

    elif group == "Income":
        col_norm = "norm_income"
        col_raw = "median_hh_income"
        label = "Median household income (dollars)"

    elif group == "Disability":
        col_norm = "norm_disability"
        col_raw = "pct_disability"
        label = "People living with a disability (%)"

    elif group == "Education":
        col_norm = "norm_hs_or_higher"
        col_raw = "pct_hs_or_higher"
        label = "Adults with high school or higher (%)"

    else:  # Eviction risk
        # “Opportunity score” view: resilience (higher = safer)
        # “Original value” view: risk (higher = more pressure)
        col_norm = "eviction_resilience_score"
        col_raw = "eviction_risk_score"
        label = "Eviction risk score (0–1, higher = more risk)"

    if value_type.startswith("Opportunity"):
        color_col = col_norm
        legend_name = f"{group} (score 0–1)"
    else:
        color_col = col_raw
        legend_name = label

    # ----- SAFETY CHECKS -----
    if color_col not in gdf.columns:
        st.warning(
            f"The column **{color_col}** is not available in the current data. "
            "This may happen if the file is from an older version of the pipeline. "
            "Try restarting after deleting old files in the `outputs/` folder, or pick a different topic."
        )
        return

    plot_gdf = gdf[~gdf[color_col].isna()].copy()

    if plot_gdf.empty:
        st.warning(
            "No data are available yet for this topic. "
            "Try picking a different item in the sidebar."
        )
        return

    # ----- MAIN LAYOUT: MAP + TABLE -----
    col_map, col_table = st.columns([2, 1], gap="large")

    with col_map:
        st.subheader("Map")
        m = make_choropleth(plot_gdf, color_col=color_col, legend_name=legend_name)
        st_folium(m, width="100%", height=600)

    with col_table:
        st.subheader("Neighborhood list")

        df_table = plot_gdf[["GEOID", "NAME", color_col]].copy()
        df_table = df_table.sort_values(color_col, ascending=False).reset_index(drop=True)
        df_table.index = df_table.index + 1
        df_table.rename(
            columns={
                "NAME": "Tract name",
                color_col: legend_name,
            },
            inplace=True,
        )

        st.dataframe(
            df_table,
            use_container_width=True,
            hide_index=False,
        )

        st.caption(
            "If you pick **Opportunity score**, higher numbers mean better access and lower hardship. "
            "If you pick **Original value**, you’ll see the raw numbers (percentages, dollar amounts, or distances) "
            "for each neighborhood."
        )

    # ----- FOOTER / METHODS BLURB -----
    with st.expander("How this map was made", expanded=False):
        st.markdown(
            """
            **Where the data come from**

            - Public data from the U.S. Census Bureau’s American Community Survey ({})  
            - We look at each small area (“census tract”) in Bernalillo County.

            **What we measure**

            - **Internet access:** share of households that have any kind of home internet subscription  
            - **Housing costs:** share of households paying 30% or more of their income on housing  
            - **Health coverage:** share of people who do *not* have health insurance  
            - **Poverty:** share of people living below the federal poverty line  
            - **Income:** typical (median) household income in dollars  
            - **Disability:** share of people living with a disability  
            - **Education:** share of adults (25+) with a high school diploma, GED, or higher  
            - **Eviction risk (proxy):** a 0–1 score built from rent burden and poverty  

            **How the score works**

            - For each measure, we put neighborhoods on the same 0–1 scale so they can be compared.  
            - Areas with **better internet**, **lower housing burden**, **more people insured**, **less poverty**, 
              **higher incomes**, **fewer disability-related barriers**, and **higher education levels** get **higher scores**.  
            - The overall opportunity score is the average of all of these pieces.

            In short: **higher scores = more opportunity and easier access to basics.**
            """.format(ACS_YEAR)
        )


if __name__ == "__main__":
    main()