import streamlit as st
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

# Apply the default theme and activate color codes
sns.set_theme()
sns.set(color_codes=True)

################### import dataset
tips = sns.load_dataset("tips")
tips["tip_percentage"] = tips["tip"] / tips["total_bill"] * 100

########### set the title and subtitle
st.title("How does the amount of tip / percentage of tips differ across different days of the week?")
st.subheader("This app shows which days of the week bring in higher tip percentages and tip amounts, helping restaurants and staff adapt to customer tipping behavior and optimize their business.")

############## create filters for our interactive plot
with st.sidebar:
    st.subheader("Filters")

    # Day selection
    all_days = sorted(tips["day"].unique())
    selected_days = st.multiselect(
        "Days to show",
        options=all_days,
        default=all_days,
    )

    # Feature options
    feature_options = {
        "Tip": "tip",
        "Tip Percentage": "tip_percentage"
    }
    feature_label = st.selectbox("Feature (x-axis)", list(feature_options.keys()))
    x_col = feature_options[feature_label]

    # KDE options
    fill = st.checkbox("Shade area", value=True)

if not selected_days:
    st.info("Select at least one day to display the plot.")
else:
    # Filter the data
    data = tips[tips["day"].isin(selected_days)].dropna(subset=[x_col])

    # KPI (numeric summary)
    avg_value = data[x_col].mean()
    unit = "$" if x_col == "tip" else "%"
    st.metric(
        label=f"Average {feature_label} for selected days",
        value=f"{avg_value:.2f} {unit}"
    )

    # KDE plot
    g = sns.displot(
        data=data,
        x=x_col,
        hue="day",
        kind="kde",
        fill=fill
    )

    fig = g.fig
    st.pyplot(fig)
    plt.close(fig)

    # Dynamic insight text
    max_day = (
        data.groupby("day")[x_col].mean()
        .sort_values(ascending=False)
        .index[0]
    )
    st.success(
        f"💡 On average, **{max_day}** has the highest {feature_label.lower()} among the selected days."
    )