"""
Point Plot Examples
===================

This example demonstrates point plot functionality in PubliPlots,
showing point estimates with error bars connected by lines. Point plots
are ideal for visualizing trends across categorical variables and
comparing groups over time or conditions.
"""

import publiplots as pp
import pandas as pd
import numpy as np

# %%
# Simple Point Plot
# -----------------
# Basic point plot showing mean values with confidence intervals.

# Create sample data
np.random.seed(42)
simple_data = pd.DataFrame({
    'time': np.repeat(['Day 1', 'Day 2', 'Day 3', 'Day 4'], 20),
    'measurement': np.concatenate([
        np.random.normal(50, 8, 20),
        np.random.normal(55, 9, 20),
        np.random.normal(62, 10, 20),
        np.random.normal(70, 12, 20),
    ])
})

# Create simple point plot
ax = pp.pointplot(
    data=simple_data,
    x='time',
    y='measurement',
    title='Simple Point Plot',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y")
pp.show()


# %%
# Point Plot with Hue Grouping
# -----------------------------
# Compare multiple groups with different colors.

# Create grouped data
np.random.seed(123)
hue_data = pd.DataFrame({
    'time': np.repeat(['Day 1', 'Day 2', 'Day 3', 'Day 4'], 20),
    'group': np.tile(np.repeat(['Control', 'Treated'], 10), 4),
    'measurement': np.concatenate([
        np.random.normal(50, 8, 10), np.random.normal(52, 8, 10),
        np.random.normal(52, 9, 10), np.random.normal(65, 10, 10),
        np.random.normal(54, 9, 10), np.random.normal(78, 12, 10),
        np.random.normal(55, 10, 10), np.random.normal(85, 14, 10),
    ])
})

# Create point plot with hue
ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    title='Point Plot with Hue Grouping',
    xlabel='Time Point',
    ylabel='Measurement',
    palette="RdGyBu_r"
)
ax.grid(axis="y")
pp.show()


# %%
# Point Plot with Custom Markers
# -------------------------------
# Use different marker shapes for each group.

ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    markers=["o", "D"],
    palette="RdGyBu_r",
    title='Point Plot with Custom Markers',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y")
pp.show()

# %%
# Point Plot with Custom Line Styles
# -----------------------------------
# Use different line styles to distinguish groups.

ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    linestyles=["-", ":"],
    palette="RdGyBu_r",
    title='Point Plot with Custom Line Styles',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y")
pp.show()

# %%
# Complete Customization
# ----------------------
# Combine custom markers, line styles, and palette with error bars.

ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    errorbar='se',
    markers=["o", "D"],
    linestyles=["-", ":"],
    palette="RdGyBu_r",
    title='Fully Customized Point Plot',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y")
pp.show()

# %%
# Point Plot with Standard Error
# -------------------------------
# Use standard error instead of confidence intervals.

ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    palette="RdGyBu_r",
    errorbar='se',
    title='Point Plot with Standard Error',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y")
pp.show()

# %%
# Point Plot with Standard Deviation
# -----------------------------------
# Show standard deviation as error bars.

ax = pp.pointplot(
    data=hue_data,
    x='time',
    y='measurement',
    hue='group',
    palette="RdGyBu_r",
    errorbar='sd',
    title='Point Plot with Standard Deviation',
    xlabel='Time Point',
    ylabel='Measurement',
)
ax.grid(axis="y", alpha=0.3)
pp.show()

# %%
# Forest Plot: Log2 Odds Ratios with Significance Coloring
# ---------------------------------------------------------
# Forest plot showing log2 odds ratios from a genetic association study.
# Points are colored based on statistical significance:
# - Blue (#5D83C3): Protective effect (upper CI < 1, i.e., log2(upper) < 0)
# - Gray (0.5): Non-significant (CI crosses 1, i.e., log2 CI crosses 0)
# - Red (#e67e7e): Risk effect (lower CI > 1, i.e., log2(lower) > 0)
#
# .. tip::
#    PubliPlots supports **custom precomputed error bars** via 
#    ``errorbar=('custom', (lower, upper))``, where ``lower`` and ``upper`` 
#    are column names in your DataFrame. This is a distinguishing feature 
#    compared to seaborn, which only accepts long-format data for automatic 
#    CI calculation.
#
# .. note::
#    For more information on error bars in statistical visualization, see the
#    excellent guide by the seaborn authors: 
#    https://seaborn.pydata.org/tutorial/error_bars.html

# Create realistic genetic association data (log2 odds ratios)
np.random.seed(101)
genetic_data = pd.DataFrame({
    'gene': [
        'APOE-ε4',
        'TREM2',
        'CLU',
        'CR1',
        'PICALM',
        'BIN1',
        'ABCA7',
        'MS4A6A',
        'CD33',
        'EPHA1',
    ],
    'log2_or': [0.85, 0.45, -0.25, 0.15, -0.35, 0.08, 0.38, -0.18, 0.05, -0.12],
    'log2_lower': [0.65, 0.25, -0.45, -0.05, -0.55, -0.12, 0.18, -0.38, -0.15, -0.32],
    'log2_upper': [1.05, 0.65, -0.05, 0.35, -0.15, 0.28, 0.58, 0.02, 0.25, 0.08],
})

# Determine significance category for coloring
def get_significance(row):
    if row['log2_upper'] < 0:
        return 'Protective'  # Protective (upper CI < 1)
    elif row['log2_lower'] > 0:
        return 'Risk'  # Risk (lower CI > 1)
    else:
        return 'Non-significant'  # Non-significant (CI crosses 1)

genetic_data['Significance'] = genetic_data.apply(get_significance, axis=1)
ax = pp.pointplot(
    data=genetic_data.sort_values("log2_or", ascending=False),
    x='log2_or',
    y='gene',
    hue='Significance',
    hue_order=("Risk", "Non-significant", "Protective"),
    palette="RdGyBu",
    linestyle="none",
    errorbar=('custom', ('log2_lower', 'log2_upper')),
    capsize=0.1,
    title="Genetic Association Study: Alzheimer's Disease Risk",
    xlabel="Log2 Odds Ratio (95% CI)",
    ylabel="Gene/Variant",
)
ax.axvline(x=0, color='black', linestyle='--', linewidth=1, alpha=0.7, 
           label='Null effect (OR = 1)')
ax.grid(axis='x')

# %%
# Forest Plot: Long-Format Data with Automatic CI Calculation
# ------------------------------------------------------------
# This example demonstrates using long-format data (typical in seaborn)
# where confidence intervals are automatically calculated from raw observations.
# This is useful when you have individual measurements rather than 
# pre-computed summary statistics.

# Create long-format data with multiple observations per treatment
np.random.seed(202)
n_obs = 30
long_format_data = pd.DataFrame({
    'treatment': np.repeat([
        'Drug A',
        'Drug B', 
        'Drug C',
        'Drug D',
        'Placebo',
    ], n_obs),
    'response': np.concatenate([
        np.random.normal(0.65, 0.15, n_obs),  # Drug A
        np.random.normal(0.48, 0.18, n_obs),  # Drug B
        np.random.normal(0.35, 0.12, n_obs),  # Drug C
        np.random.normal(0.52, 0.16, n_obs),  # Drug D
        np.random.normal(0.25, 0.10, n_obs),  # Placebo
    ])
})

# Create forest plot with automatic CI calculation
ax = pp.pointplot(
    data=long_format_data,
    x='response',
    y='treatment',
    color='#8E8EC1',
    linestyle="none",
    errorbar='ci',  # Automatically calculate 95% CI from data
    capsize=0.1,
    title='Clinical Trial: Treatment Response Rates',
    xlabel='Response Rate (95% CI)',
    ylabel='Treatment',
)

# Add vertical line at placebo mean for reference
placebo_mean = long_format_data[long_format_data['treatment'] == 'Placebo']['response'].mean()
ax.axvline(x=placebo_mean, color='red', linestyle=':', linewidth=1.5,
           alpha=0.7, label=f'Placebo mean ({placebo_mean:.2f})')
ax.grid(axis='x', alpha=0.3)
ax.legend(loc='lower right')

pp.show()

# %%
# Annotated points
# ----------------
# ``annotate=True`` labels each mean with its value, above the errorbar
# cap by default. Point-plot anchors are directional (``top``, ``bottom``,
# ``left``, ``right``, ``center``). See the dedicated
# :doc:`annotations gallery <plot_23_annotate>` for more.

ax = pp.pointplot(
    data=simple_data,
    x='time', y='measurement',
    errorbar='se',
    annotate={"fmt": ".1f"},
    title="annotate=True",
)
pp.show()