Skip to content

NumPy Array vs Python List

Run the NumPy Array vs Python List MicroSim Fullscreen
Edit in the p5.js Editor

About This MicroSim

A Python list and a NumPy array can hold the same five numbers, but they store them in very different ways. This MicroSim draws both.

  • Python list (left): the list object is a row of five pointers. Each pointer leads to a separate int object somewhere else in memory.
  • NumPy array (right): a small header records the dtype and the shape, and the five values sit side by side in one block, 8 bytes each.

Pressing Next doubles the numbers. The list needs one pass through the Python interpreter for every element: follow the pointer, check the type, multiply, store a pointer to the result. The array is finished after one step, because arr * 2 is a single call whose loop runs in compiled C code along the block.

The bottom panel scales the comparison up. Memory is computed from the sizes that CPython reports on a 64-bit machine: a list costs 56 bytes plus 8 bytes per pointer and 28 bytes per int object, and an int64 array costs 8 bytes per value. For one million numbers that is 36.0 MB against 8.0 MB, the same figures the chapter's sys.getsizeof code prints. The speed statement is a typical range, not a measurement, because real timings depend on the operation and the computer. The addresses on screen are made up for illustration.

How to Use

  1. Read the two panels at step 0. Find the five pointers in the list and the five values in the array.
  2. Press Next and watch which cells turn gold. Read the sentence under each panel.
  3. Keep pressing Next until the list is done. Count how many steps each side needed.
  4. Untick Show addresses, then tick it again. Compare the scattered addresses of the int objects with the evenly spaced addresses of the array values.
  5. Drag the Array size n slider and watch the memory bars and the number of interpreter passes. Press Reset to start again.

Iframe Embed Code

You can add this MicroSim to any web page by adding this to your HTML:

1
2
3
4
<iframe src="https://dmccreary.github.io/data-science-course/sims/numpy-array-vs-python-list/main.html"
        height="616"
        width="100%"
        scrolling="no"></iframe>

Lesson Plan

Grade Level

High School (11-12) and College Freshman

Duration

10-15 minutes

Prerequisites

Python lists and for loops, and the idea that a program's data lives in the computer's memory.

Bloom's Taxonomy Level

Understand (L2)

Learning Objective

Students will be able to explain how a Python list and a NumPy array store the same numbers and why that difference makes whole-array NumPy operations use less memory and run faster.

Activities

  1. Predict (3 min): Before pressing Next, students predict how many steps the list and the array will each need to double five numbers, and say why.
  2. Step through (6 min): Students step from 0 to 5, describing in their own words what the interpreter does on each pass and what NumPy did at step 1.
  3. Scale up (5 min): Students set the slider to 1,000 and to 1,000,000 and record the memory for each structure, then work out where the 36 bytes per list element come from.

Assessment

Students draw their own picture of the list [10, 20, 30] and the array np.array([10, 20, 30]) in memory and write two sentences explaining why arr * 2 is faster than [x * 2 for x in nums] on a large array, and why the gain disappears on a tiny one.

References

  1. NumPy documentation. What is NumPy?.
  2. VanderPlas, Jake. Python Data Science Handbook, 2nd ed. O'Reilly Media, 2022. Part II, Introduction to NumPy. See the chapter Understanding Data Types in Python.
  3. Wikipedia. NumPy.