Plotting Fundamentals Skill

Overview

Master quick plotting and interactive visualization with hvPlot and HoloViews basics. This skill covers essential techniques for creating publication-quality plots with minimal code.

Dependencies

• hvplot >= 0.9.0
• holoviews >= 1.18.0
• pandas >= 1.0.0
• numpy >= 1.15.0
• bokeh >= 3.0.0

Core Capabilities

1. hvPlot Quick Plotting

hvPlot provides an intuitive, pandas-like API for rapid visualization:

import hvplot.pandas
import pandas as pd
import numpy as np

# Create sample data
df = pd.DataFrame({
    'date': pd.date_range('2024-01-01', periods=100),
    'sales': np.cumsum(np.random.randn(100)) + 100,
    'region': np.random.choice(['North', 'South', 'East', 'West'], 100)
})

# Simple line plot
df.hvplot.line(x='date', y='sales', title='Sales Over Time')

# Grouped plot
df.hvplot.line(x='date', y='sales', by='region', subplots=True)

# Scatter with size and color
df.hvplot.scatter(x='sales', y='date', c='region', size=50)

2. Common Plot Types

# Bar plot
df.hvplot.bar(x='region', y='sales', rot=45)

# Histogram
df['sales'].hvplot.hist(bins=30, title='Sales Distribution')

# Box plot
df.hvplot.box(y='sales', by='region')

# Area plot
df.hvplot.area(x='date', y='sales')

# KDE (Kernel Density Estimation)
df['sales'].hvplot.kde()

# Hexbin (for large datasets)
df.hvplot.hexbin(x='sales', y='date', gridsize=20)

3. Customization Options

# Apply consistent styling
plot = df.hvplot.line(
    x='date',
    y='sales',
    title='Sales Trend',
    xlabel='Date',
    ylabel='Sales ($)',
    color='#2E86DE',
    line_width=2,
    height=400,
    width=700,
    responsive=True,
    legend='top_left'
)

# Color mapping
df.hvplot.scatter(
    x='sales',
    y='date',
    c='sales',
    cmap='viridis',
    s=100
)

# Multiple series
df.hvplot.line(
    x='date',
    y=['sales'],
    title='Performance Metrics'
)

4. Interactive Features

# Hover information
df.hvplot.scatter(
    x='sales',
    y='date',
    hover_cols=['region'],
    tools=['hover', 'pan', 'wheel_zoom']
)

# Selection and linked views
import holoviews as hv
scatter = df.hvplot.scatter(x='sales', y='date')
scatter.opts(tools=['box_select'])

# Responsive sizing
plot = df.hvplot.line(
    x='date',
    y='sales',
    responsive=True,
    height=400
)

5. Geographic Plotting with hvPlot

import geopandas as gpd

# Quick geographic plot
gdf = gpd.read_file(gpd.datasets.get_path('naturalearth_lowres'))
gdf.hvplot(
    c='pop_est',
    cmap='viridis',
    geo=True,
    frame_width=600
)

# City points on map
cities = gpd.GeoDataFrame({
    'name': ['City A', 'City B'],
    'geometry': [Point(0, 0), Point(1, 1)],
    'population': [1000000, 500000]
})
cities.hvplot(
    geo=True,
    c='population',
    size='population',
    cmap='plasma'
)

HoloViews Fundamentals

1. Basic Element Types

import holoviews as hv
from holoviews import opts

# Curve
curve = hv.Curve(df, 'date', 'sales')

# Scatter
scatter = hv.Scatter(df, 'sales', 'date')

# Histogram
hist = hv.Histogram(df['sales'].values)

# Image (heatmap)
image = hv.Image(data)

# Bars
bars = hv.Bars(df, 'region', 'sales')

# Text annotations
text = hv.Text(0.5, 0.5, 'Hello HoloViews')

2. Styling and Options

# Using .opts() method
plot = hv.Curve(df, 'date', 'sales').opts(
    title='Sales Trend',
    xlabel='Date',
    ylabel='Sales',
    color='#2E86DE',
    line_width=2,
    height=400,
    width=700
)

# Using opts object
opts_obj = opts.Curve(
    title='Sales',
    color='navy',
    line_width=2
)
plot = hv.Curve(df, 'date', 'sales').opts(opts_obj)

3. Composing Visualizations

# Overlaying multiple plots
overlay = hv.Curve(df, 'date', 'sales') * hv.Scatter(df_subset, 'date', 'sales')

# Side-by-side layouts
layout = hv.Curve(df1, 'date', 'sales') + hv.Scatter(df2, 'date', 'value')

# Grid layouts
grid = (
    (hv.Curve(data1) + hv.Scatter(data2)) /
    (hv.Histogram(data3) + hv.Image(data4))
)

# Faceted views
faceted = hv.Curve(df, 'date', 'sales').facet('region')

4. Interactive Selection and Linking

# Brush selection
curve_selectable = hv.Curve(df, 'date', 'sales').opts(
    tools=['box_select'],
    selection_fill_color='red',
    nonselection_fill_alpha=0.2
)

# Dynamic linking with streams
from holoviews import streams

# Hover information
hover = streams.Tap(source=scatter, transient=True)

@hv.transform
def get_info(data):
    if data.empty:
        return hv.Text(0, 0, 'Hover to select')
    return hv.Text(0, 0, f"Point: {data.iloc[0].values}")

Best Practices

1. Data Preparation

• Always check data types before plotting
• Handle missing values explicitly
• Normalize columns for better visualization
• Use appropriate data ranges

2. Visual Design

• Choose colors for accessibility (colorblind-friendly palettes)
• Use title and axis labels
• Include legends for multiple series
• Maintain consistent styling across related plots

3. Performance

• Use datashader for datasets with >100k points
• Downsample or aggregate before plotting
• Use responsive=True for web dashboards
• Cache expensive plot computations

4. Code Organization

# Create a plotting utility module
class PlotBuilder:
    COLORS = {'primary': '#2E86DE', 'secondary': '#A23B72'}
    DEFAULTS = {'height': 400, 'width': 700, 'responsive': True}

    @staticmethod
    def style_plot(plot, **kwargs):
        return plot.opts(**{**PlotBuilder.DEFAULTS, **kwargs})

# Usage
styled = PlotBuilder.style_plot(df.hvplot.line(x='date', y='sales'))

Common Patterns

Pattern 1: Dashboard with Multiple Plots

def create_sales_dashboard(df):
    return hv.Column(
        df.hvplot.line(x='date', y='sales', title='Trend'),
        df.hvplot.bar(x='region', y='sales', title='By Region'),
        df['sales'].hvplot.hist(bins=20, title='Distribution')
    )

Pattern 2: Conditional Visualization

def plot_data(df, plot_type='line'):
    if plot_type == 'line':
        return df.hvplot.line(x='date', y='sales')
    elif plot_type == 'scatter':
        return df.hvplot.scatter(x='date', y='sales')
    else:
        return df.hvplot.bar(x='region', y='sales')

Pattern 3: Multi-Series Plot with Legend

def plot_multiple_metrics(df, metrics):
    plots = [df.hvplot.line(x='date', y=m, label=m) for m in metrics]
    return hv.Overlay(plots)

Integration with Other HoloViz Tools

• Panel: Embed plots in dashboards
• HoloViews: Advanced composition and interactivity
• Datashader: Large dataset visualization
• Param: Dynamic plot updates based on parameters

Common Use Cases

1. Time Series Analysis: Trends, anomalies, forecasting
2. Comparative Analysis: Category comparisons, rankings
3. Distribution Analysis: Histograms, KDEs, box plots
4. Correlation Analysis: Scatter plots, hexbins
5. Geographic Analysis: Maps, regional data
6. Statistical Summaries: Summary statistics with plots

Troubleshooting

Issue: Plot Won't Display

• Ensure hvplot.pandas or hvplot.xarray is imported
• Check data is not empty
• Verify x and y columns exist in dataframe

Issue: Poor Performance with Large Data

• Use datashader for >100k points
• Implement aggregation or sampling
• Use hexbin or rasterization

Issue: Unclear or Overlapping Labels

• Rotate x-axis labels with rot=45
• Use subplots with by='column'
• Adjust figure size with height/width

Resources

• hvPlot Documentation
• HoloViews Documentation
• HoloViews Gallery
• Colorcet for Color Palettes