Alternatives for big data (Datashader, HoloViews) in Matplotlib - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the output of this Datashader aggregation code?

Given a large dataset of points, this code uses Datashader to aggregate points into an image. What is the type of the output variable agg?

Matplotlib

import datashader as ds
import pandas as pd
import numpy as np

points = pd.DataFrame({
    'x': np.random.uniform(0, 100, 100000),
    'y': np.random.uniform(0, 100, 100000)
})

canvas = ds.Canvas(plot_width=400, plot_height=400)
agg = canvas.points(points, 'x', 'y')

A A Pandas DataFrame containing aggregated counts

B A NumPy array of shape (400, 400) with counts

C A Datashader Image object representing the rasterized aggregation

D A dictionary with keys 'x' and 'y' holding aggregated values

💡 Hint

Datashader's Canvas.points returns an Image object, not a DataFrame or array.

The canvas.points method returns a Datashader Image object that contains the rasterized aggregation of points. It is not a DataFrame or NumPy array directly.

🧠 Conceptual

intermediate

Why use HoloViews with Datashader for big data visualization?

Which of the following best explains the advantage of combining HoloViews with Datashader for visualizing large datasets?

A HoloViews provides interactive plotting while Datashader efficiently rasterizes large data for fast rendering

B HoloViews replaces Datashader's aggregation with simpler plotting methods

C Datashader adds 3D plotting capabilities to HoloViews

D HoloViews automatically reduces data size before plotting without Datashader

💡 Hint

Think about how each library complements the other in handling big data.

HoloViews offers easy-to-use, interactive plotting interfaces, while Datashader handles the heavy lifting of rasterizing large datasets efficiently. Together, they enable fast, interactive visualizations of big data.

❓ data_output

advanced

What is the shape of the aggregated data from this Datashader code?

Consider this code that aggregates points into a 300x200 canvas. What is the shape of the agg.values array?

Matplotlib

import datashader as ds
import pandas as pd
import numpy as np

points = pd.DataFrame({
    'x': np.random.uniform(0, 50, 50000),
    'y': np.random.uniform(0, 100, 50000)
})

canvas = ds.Canvas(plot_width=300, plot_height=200)
agg = canvas.points(points, 'x', 'y')
result_shape = agg.values.shape

A (50000, 2)

B (300, 200)

C (300, 300)

D (200, 300)

💡 Hint

Remember that the first dimension is height (rows), second is width (columns).

Datashader's aggregation values array shape corresponds to (plot_height, plot_width), so (200, 300) in this case.

🔧 Debug

advanced

Why does this HoloViews + Datashader code raise an error?

This code tries to rasterize a scatter plot with Datashader but raises an error. What is the cause?

Matplotlib

import holoviews as hv
import datashader as ds
import pandas as pd
import numpy as np

hv.extension('bokeh')

points = pd.DataFrame({
    'x': np.random.randn(1000),
    'y': np.random.randn(1000)
})

scatter = hv.Points(points)
rasterized = ds.rasterize(scatter)

A Datashader rasterize expects a HoloViews element, but hv.Points is not recognized without hv.operation.datashader

B The DataFrame columns must be named 'X' and 'Y' with uppercase letters

C Datashader cannot rasterize data with negative values

D hv.extension('bokeh') must be called after rasterize

💡 Hint

Check how Datashader integrates with HoloViews for rasterization.

Datashader's rasterize function expects a HoloViews element but is usually accessed via hv.operation.datashader.rasterize. Calling ds.rasterize directly on hv.Points causes an error.

🚀 Application

expert

How to efficiently visualize 10 million points interactively?

You have 10 million (x, y) points and want an interactive plot that updates quickly when zooming or panning. Which approach is best?

A Plot all 10 million points directly with matplotlib scatter plot for full detail

B Use Datashader to rasterize points on a canvas and integrate with HoloViews for interactive zoom and pan

C Downsample the data to 1000 points and plot with seaborn scatterplot

D Convert data to a CSV and open in Excel for interactive filtering

💡 Hint

Think about how to handle large data efficiently with interactivity.

Datashader efficiently rasterizes large datasets into images, and combined with HoloViews, it supports interactive zooming and panning without plotting every point individually.