Learning Data Science

Lecture 5
Visualisation Craft

Project Scaffolding

Starting projects with uv
Adding your local project to GitHub
Installing external packages

uv init -p 3.13

uv add numpy

Python Notebooks

Ways to interact with Python code

REPL

✅ Interactive
❌ Sharable
❌ Version Control friendly
❌ Reproducible
❌ Mix code, text, plots

Scripts

❌ Interactive

✅ Sharable

✅ Version Control friendly

✅ Reproducible
❌ Mix code, text, plots

Notebooks

✅ Interactive

✅ Sharable

⚠️ Version Control friendly

⚠️ Reproducible

✅ Mix code, text, plots

Linear Algebra in a Nutshell

The math of vectors and matrices

How to combine and transform them

Everything is a tensor

Graphics from: St. Lawrence U CS140 and Montesinos-López et al (2022)

Image width (4px)

Image height (4px)

Image "depth"
(3 color channels)

NumPy

an array library that powers all of scientific Python

Fancy lists

Vectors

Matrices

Tensors

x = np.array([1,2,3,4,5])
# OR
l = [1, 2, 3, 4, 5]
x = np.array(l)

x = np.ones(5)
x = np.zeros(5)

x = np.array([1,2,3,4,5])
y = np.ones_like(x)

x = np.arange(6)
x = np.arange(1, 6, 2)
x = np.linspace(0, 1, 11)

rng = np.random.default_rng()
# 5 ints between [0,100)
rng.integers(0, 100, 5)

# 10 ints between [0, 1)
rng.random(10)

# 6 samples from a gaussian
# mean=5, std=3
rng.normal(5, 3, 6)

x = np.arange(1, 6)
# [1, 2, 3, 4, 5]
x[1] # = 2
x[2:4] # = [3, 4]
# everything from index -2 and onwards
x[-2:] # = [4, 5]
# Every 2 indices (step=2)
x[::2] # = [1, 3, 5]
# specifically indicies 1, 3, 4
x[[1, 3, 4]] # = [2, 4, 5]

x = np.array([1, 2])
y = np.array([3, 6])
x + y # = [4, 8]
x * y # = [3, 12]
x + 2 # = [3, 4]
np.sqrt(x) 
np.sin(x)

x.max()
x.sum()
x.mean()
# etc

np.array([1, 2, 3, 4])

Normal 1D Vector (4,)

4 rows, 0 columns

np.array([[1, 2, 3, 4]])

Row 2D Vector (1, 4)

1 row, 4 columns

np.array([[1], [2], [3], [4]])

Column 2D Vector (4, 1)

4 rows, 1 column

x.T

x.T

x.reshape(1, -1)

x.flatten()

x.reshape(-1, 1)

x.flatten()

Lecture 5

Recap
Why visualization
Matplotlib essentials
Seaborn quick tour
Important file formats

Why Visualize Anything?

Sight is often regarded as the dominant sense
Sight is often claimed to be the most "objective" sense

Debatable!

Clearly not true for everyone!

However, we use our sight often to better understand scientific results:

Reading journals and textbooks
Looking at slides
Looking at graphs

Listening to data

"When we synchronize our different ways of perceiving the world, our sensitivity to events that are masked to the eye ... increases exponentially."

More here

Listening to data

While today we're going to talk about visual ways to communicate data, always remember there are other ways too!

Why Visualization?

[[-0.99582463  0.99785717]
 [-0.98329854  0.9975254 ]
 [-0.97077244  0.99820004]
 [-0.95824635  0.99651706]
 [-0.94572025  0.99664487]
 [-0.93319415  0.99538239]
 [-0.92066806  0.995524  ]
 [-0.90814196  0.99469056]
 [-0.89561587  0.99384034]
 [-0.88308977  0.99323822]
 [-0.87056367  0.99233328]
 [-0.85803758  0.99210751]
 [-0.84551148  0.99107832]
 [-0.83298539  0.99010279]
 [-0.82045929  0.98942019]
 [-0.80793319  0.98934844]
 [-0.7954071   0.98927283]
 [-0.782881    0.98808468]
 [-0.77035491  0.98737426]
 [-0.75782881  0.98696232]
 [-0.74530271  0.98682547]
 [-0.73277662  0.98587808]
 [-0.72025052  0.98514856]
 [-0.70772443  0.98531636]
 [-0.69519833  0.98464728]
 [-0.68267223  0.9847355 ]
 [-0.67014614  0.98388559]
 [-0.65762004  0.98306121]
 [-0.64509395  0.98279164]
 [-0.63256785  0.98383771]
 [-0.62004175  0.98375677]
 [-0.60751566  0.98293767]
 [-0.59498956  0.98383009]
 [-0.58246347  0.98452698]
 [-0.56993737  0.98345948]
 [-0.55741127  0.98350519]
 [-0.54488518  0.98408205]
 [-0.53235908  0.9836294 ]
 [-0.51983299  0.98374858]
 [-0.50730689  0.98364951]
 [-0.49478079  0.98381505]
 [-0.4822547   0.98420902]
 [-0.4697286   0.98364914]
 [-0.45720251  0.98383823]
 [-0.44467641  0.9827292 ]
 [-0.43215031  0.98349927]
 [-0.41962422  0.98315183]
 [-0.40709812  0.98297283]
 [-0.39457203  0.9830812 ]
 [-0.38204593  0.98327179]
 [-0.36951983  0.98374565]
 [-0.35699374  0.98282346]
 [-0.34446764  0.98334186]
 [-0.33194154  0.9831124 ]
 [-0.31941545  0.98315234]
 [-0.30688935  0.98366272]
 [-0.29436326  0.98345594]
 [-0.28183716  0.9837556 ]
 [-0.26931106  0.98313967]
 [-0.25678497  0.98290479]
 [-0.24425887  0.98307619]
 [-0.23173278  0.98324617]
 [-0.21920668  0.98320411]
 [-0.20668058  0.9826849 ]
 [-0.19415449  0.98314091]
 [-0.18162839  0.98318293]
 [-0.1691023   0.98408642]
 [-0.1565762   0.98381884]
 [-0.1440501   0.98271634]
 [-0.13152401  0.9829331 ]
 [-0.11899791  0.98245396]
 [-0.10647182  0.98279521]
 [-0.09394572  0.98315076]
 [-0.08141962  0.98350076]
 [-0.06889353  0.98272155]
 [-0.05636743  0.98274717]
 [-0.04384134  0.98214642]
 [-0.03131524  0.98293766]
 [-0.01878914  0.98257601]
 [-0.00626305  0.98278833]
 [ 0.00626305  0.98264936]
 [ 0.01878914  0.98296328]
 [ 0.03131524  0.98229963]
 [ 0.04384134  0.98241852]
 [ 0.05636743  0.98386052]
 [ 0.06889353  0.98249822]
 [ 0.08141962  0.9825797 ]
 [ 0.09394572  0.98375928]
 [ 0.10647182  0.98419352]
 [ 0.11899791  0.98257071]
 [ 0.13152401  0.98290075]
 [ 0.1440501   0.98319426]
 [ 0.1565762   0.98375958]
 [ 0.1691023   0.98268469]
 [ 0.18162839  0.98299352]
 [ 0.19415449  0.98341565]
 [ 0.20668058  0.98329256]
 [ 0.21920668  0.98298301]
 [ 0.23173278  0.98309564]
 [ 0.24425887  0.98261577]
 [ 0.25678497  0.98308789]
 [ 0.26931106  0.98309491]
 [ 0.28183716  0.98331351]
 [ 0.29436326  0.98301856]
 [ 0.30688935  0.98345023]
 [ 0.31941545  0.98368849]
 [ 0.33194154  0.98336824]
 [ 0.34446764  0.98347031]
 [ 0.35699374  0.98337527]
 [ 0.36951983  0.98337933]
 [ 0.38204593  0.98311682]
 [ 0.39457203  0.98355906]
 [ 0.40709812  0.98342144]
 [ 0.41962422  0.98432639]
 [ 0.43215031  0.9841481 ]
 [ 0.44467641  0.98370361]
 [ 0.45720251  0.98327543]
 [ 0.4697286   0.98316794]
 [ 0.4822547   0.98412248]
 [ 0.49478079  0.98378512]
 [ 0.50730689  0.98390695]
 [ 0.51983299  0.98305279]
 [ 0.53235908  0.98415821]
 [ 0.54488518  0.98400649]
 [ 0.55741127  0.9834189 ]
 [ 0.56993737  0.98371369]
 [ 0.58246347  0.98404789]
 [ 0.59498956  0.98400435]
 [ 0.60751566  0.98289261]
 [ 0.62004175  0.98302591]
 [ 0.63256785  0.98307655]
 [ 0.64509395  0.98339993]
 [ 0.65762004  0.98413959]
 [ 0.67014614  0.98278563]
 [ 0.68267223  0.98402645]
 [ 0.69519833  0.98454638]
 [ 0.70772443  0.98499224]
 [ 0.72025052  0.98516417]
 [ 0.73277662  0.98508764]
 [ 0.74530271  0.98643427]
 [ 0.75782881  0.98753893]
 [ 0.77035491  0.98680817]
 [ 0.782881    0.98836621]
 [ 0.7954071   0.98850978]
 [ 0.80793319  0.98925487]
 [ 0.82045929  0.9895102 ]
 [ 0.83298539  0.99045925]
 [ 0.84551148  0.99178033]
 [ 0.85803758  0.99216214]
 [ 0.87056367  0.99328812]
 [ 0.88308977  0.99372549]
 [ 0.89561587  0.99407863]
 [ 0.90814196  0.99523416]
 [ 0.92066806  0.99532697]
 [ 0.93319415  0.99607476]
 [ 0.94572025  0.99619085]
 [ 0.95824635  0.99687223]
 [ 0.97077244  0.99706969]
 [ 0.98329854  0.99821132]
 [ 0.99582463  0.9977719 ]]

Why Visualization?

Alberto Cairo’s "Datasaurus Dozen"

Distributions with the exact same statistics
But visually completely different!

Looking at the numbers alone is not enough!

Challenge #1

Find what is off, weird, confusing, misleading, or 'too much' in each of the following visualizations:

Previous president of Colombia claiming that collective homicides went down during his time in office (2018-2020)

Open AI's Official
ChatGPT 5 Release Video
(Aug 2025)

Lecture 5

Recap
Why visualization
Matplotlib essentials
Seaborn quick tour
Important file formats

matplotlib

Create visualizations:

static
animated
interactive

Generally flexible enough to do whatever you need

Installing matplotlib

uv add matplotlib

import matplotlib.pyplot as plt

Importing matplotlib

Everyone nicknames this package to plt!

Plotting Functions with line plots

Making Plots

x = np.array([-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5])

y = [i**2 for i in x]

plt.plot(x, y)
plt.show()

Making Plots

x = np.linspace(-10, 10, 1000)
y = [i**2 for i in x]

plt.plot(x, y)
plt.show()

Make the plot even smoothing by plotting 1000 points inside our range
Recall linspace is similar to the builtin range function

Add title and axis labels

x = np.linspace(-10, 10, 1000)
y = [i**2 for i in x]

plt.plot(x, y)

plt.title("A plot of $f(x) = x^2$")
plt.xlabel("x")
plt.ylabel("$x^2$")

plt.show()

Tip: if you know how to write latex math, this also works inside plots!

Scatter Plot

rng = np.random.default_rng()

x = rng.normal(0, 1, 20)
y = rng.normal(0, 1, 20)

plt.scatter(x, y)
plt.show()

Good for unordered, 2D data

Histograms

x = rng.normal(171, 20, 2000)

plt.hist(x)
plt.xlabel("Height (cm)")
plt.ylabel("Frequency")
plt.title("Height Distribution of 2000 People")
plt.show()

Show the frequency distribution of 1D data

Axes and Labels

rng = np.random.default_rng()

x = rng.normal(0, 1, 20)
y = rng.normal(0, 1, 20)

plt.scatter(x, y)
plt.show()

Same scatter as before

Axes and Labels

rng = np.random.default_rng()

x = rng.normal(0, 1, 200)
y = rng.normal(0, 1, 200)

plt.scatter(x, y)
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("My Great Scatter Plot")
plt.show()

Can set x and y axes, as well as title

Overplotting and Legends

x = np.linspace(0, 10, 100)
y1 = np.sin(x)
y2 = np.cos(x)

plt.plot(x, y1)
plt.plot(x, y2)
plt.show()

Anything you do before 'show()' will all show up on the same axes!

Overplotting and Legends

x = np.linspace(0, 10, 100)
y1 = np.sin(x)
y2 = np.cos(x)

plt.plot(x, y1, label="sin(x)")
plt.plot(x, y2, label="cos(x)")
plt.legend()
plt.show()

Use the 'label' kwarg, and add plt.legend() to automatically make a Legend!

Overplotting and Legends

x = np.linspace(0, 10, 100)
y1 = np.sin(x)
y2 = np.cos(x)

plt.plot(x, y1, label="sin(x)")
plt.plot(x, y2, label="cos(x)")
plt.legend(loc="lower right")
plt.show()

Use the loc kwarg to select the position of the legend!

Overplotting and Legends

Style

Line style
Marker Style
Color
Width
Alpha

Line Style

y1 = np.sin(x)
y2 = np.cos(x)

plt.plot(x, y1, label="sin(x)", linestyle="dashed")
plt.plot(x, y2, label="cos(x)", linestyle="dotted")
plt.legend()
plt.show()

Use the loc kwarg to select the position of the legend!

Line Style

plt.plot(x, y1, label="sin(x)", linewidth=5, linestyle="dashed")
plt.plot(x, y2, label="cos(x)", linewidth=10, linestyle="dotted")
plt.legend()
plt.show()

Change the width of your lines

Marker Style

plt.scatter(x, y, marker="v", label="People who do", )
plt.scatter(x2, y2, marker="x", label="People who do not", )
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("My Great Scatter Plot")
plt.legend()
plt.show()

Change the shape of points:
full list here

Marker Style

plt.scatter(x, y, marker="v", label="People who do", )
plt.scatter(x2, y2, marker="x", label="People who do not", )
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("My Great Scatter Plot")
plt.legend()
plt.show()

Change the size of the markers

Color

New overplots automatically cycle through a list of colors

Color

You can also set the colors you want specifically.

plt.scatter(x, y, color="red", label="People who do", marker="v")
plt.scatter(x2, y2, color="blue", label="People who do not", marker="x")
plt.scatter(x3, y3, color="green", label="People who might", marker="D")
plt.scatter(x4, y4, color="#6f42c1", label="People who would never", marker="+")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("My Great Scatter Plot")
plt.legend()
plt.show()

Marker Size

Use the 's' variable,

represented in area^2 of the plot

plt.scatter(x, y, s=5, label="People who do", marker="v")
plt.scatter(x2, y2, s=100, label="People who do not", marker="x")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("My Great Scatter Plot")
plt.legend()
plt.show()

Alpha (Transparency)

Change the opacity of curves/markers. Alpha is in the range [0,1]

plt.plot(x, y1, alpha=0.9, color="#5a2094")
plt.plot(x, y2, alpha=0.8, color="#5a2094")
plt.plot(x, y3, alpha=0.5, color="#5a2094")
plt.plot(x, y4, alpha=0.3, color="#5a2094")
plt.plot(x, y5, alpha=0.1, color="#5a2094")

Log Scale

x = np.linspace(0, 100, 1000)
y = np.exp(x)

plt.plot(x, y)
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("Normal Axes")
plt.show()

plt.plot(x, y)
plt.yscale("log")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("Logarithmic Axes")
plt.show()

Log Scale

x = np.linspace(0, 100, 1000)
y = np.exp(x)

plt.plot(x, y)
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("Normal Axes")
plt.show()

plt.plot(x, y)
plt.yscale("log")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")
plt.title("Logarithmic Axes")
plt.show()

Use plt.xscale or plt.yscale

Saving Plots

Can always drag out of a notebook

Saving Plots

Can also save the file from code

Just add the plt.savefig() name
Can export to many different file types
Just change the suffix at the end of "my_plot.png" to match the filet ype you wanted

x = np.linspace(-10, 10, 1000)
y = [i**2 for i in x]

plt.plot(x, y)

plt.title("A plot of $f(x) = x^2$")
plt.xlabel("x")
plt.ylabel("$x^2$")
plt.savefig("my_plot.png")
plt.show()

Challenge #1

Try to recreate this figure as exact as possible!

Figure Size

plt.figure(figsize=(10, 5))
plt.plot(x, y)
plt.show()

plt.figure(figsize=(3, 10))
plt.plot(x, y)
plt.show()

plt.figure(figsize=(X,Y))

💫 You are now a master plot maker!

If you're someone who likes design, you might have noticed they are not so pretty by default.

Let's look at a tool which can help 💅

Lecture 5

Recap
Why visualization
Matplotlib essentials
Seaborn quick tour
Important file formats

What is seaborn?

Seaborn is a library built on top of matplotlib

Attempts to make your plots

✨effortlessly pretty✨

just like you

ALSO: Provides a user-friendly high-level interface for making statistical plots

What is seaborn?

Installing seaborn

uv add seaborn

import seaborn as sns

Starting seaborn

import matplotlib.pyplot as plt
import seaborn as sns

sns.set_theme()

Using seaborn styles in matplotlib

Now even matplotlib plots will look a bit nicer

Challenge #2

Redo the plot from Challenge #1 but with seaborn active

How does the plot look now?
Which do you prefer?

rng = np.random.default_rng()
height = rng.normal(171, 20, 2000)

sns.displot(height, kde=True, label="Height distribution")
plt.xlabel("Height (cm)")
plt.ylabel("Frequency")
plt.title("Height Distribution of 2000 People")
plt.legend()
plt.show()


sns.displot(height, kind="ecdf", label="Cumulative height distribution")
plt.xlabel("Height (cm)")
plt.ylabel("Cumulative frequency")
plt.title("Height Distribution of 2000 People")
plt.legend()
plt.show()

Seaborn historgrams

Heatmaps

Often can be a 2D histogram
Frequency (or some other feature) is represented with a color scale

Heatmaps

Can also be used for categorical data

And many more!

Lecture 5

Recap
Why visualization
Matplotlib essentials
Seaborn quick tour
Important file formats

Important Data Formats

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

Fancier:

.hdf5
.parquet

.lmdb

.SQL

.excel

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

Today

Text-based data formats

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

Today

Why so many?

Different shapes of data havedifferent priorities
human-readability
schema strictness
tabular vs nested
tooling support

Text-based data formats

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

Plaintext:

.txt
.json
.csv
.xml
.yaml
.toml

key-value? Like a dict!

Mental model:

.txt → free-form notes/logs
.csv → rows & columns (tabular)
.json/.xml/.yaml/.toml → nested key–value structures (configs, APIs)

Text files - .txt

What: Unstructured plain text; the simplest possible file.
Origins: Since the beginning of time
Used for:
- Notes
- logs
- docs
Why use it: Open with any software ever
Caveat: No built-in structure
- you must define your own conventions.

Reading text files with Python

with open("lec5.txt", "r") as file:
    data = file.read()

print(data)

We use this special 'with' format
When we open a file, we need to remember to close it after reading the data from it
The 'with' statement automatically closes the file for us once we run all the code underneath it! Thanks 'with'!

Challenge #3

44 12 96 12 129 120 49 60 38 11 20 478 938 40 102 222 102 23 40 58 40 12 12 12 12 49 60 48 27 37 40 17 172 11 98

Copy the text above and save it to a file called "challenge3.txt"
Read in the the file and use your Python skills to turn the text into a numpy array of integers
Calculate the mean and median of the array

CSV

What: Comma-Separated Values one row per line
- Commas to separate columns
Origins: 1970s
Used for: Spreadsheets, databases, etc
Why use it: Universal, great for tabular data.
Caveats:
- Schema not embedded
- Slooooow

Reading CSV with Python

import csv

with open("lec5-data/bus.csv", newline="", encoding="utf-8") as f:
    rows = list(csv.DictReader(f))
    
    
print(rows)
print(rows[0])

# can also convert to numpy array
data = np.array(rows)
print(data)

date,station,rides
2025-09-05,Central,120
2025-09-05,West,95
2025-09-06,Central,130
2025-09-06,West,105

[{'date': '2025-09-05', 'station': 'Central', 'rides': '120'}
 {'date': '2025-09-05', 'station': 'West', 'rides': '95'}
 {'date': '2025-09-06', 'station': 'Central', 'rides': '130'}
 {'date': '2025-09-06', 'station': 'West', 'rides': '105'}]

JSON

What: JavaScript Object Notation
Origins: 2000s
Used for: Web APIs
- Most data passed around from websites to you browser is communicated via JSON
Why use it: Human-readable, things are typed
Caveats: No comments allowed! :(

JSON

{
    "users": [
        {
            "name": "Lady Gaga",
            "email": "lady.gaga@mpp.mpg.de",
            "age": 36,
            "signed_in": true
        },
        {
            "name": "David Hasselhoff",
            "email": "david.hasselhoff@tum.de",
            "age": 25,
            "signed_in": false
        },
        {
            "name": "Johann Sebastian Bach",
            "email": "bach@db.de",
            "age": 300,
            "signed_in": true
        }
    ]
}

# json

import json

with open("userinfo.json", "r") as f:
    data = json.load(f)

print(type(data))
print(data["users"][0])

The type is a dictionary!

XML

What: eXtensible Markup Language
Origins: W3C 1998
Used for: document formats (e.g., Office .docx inside is XML!), config files, RSS, HTML?
Why use it: uhhhhh....
Caveats: Verbose, usually overkill

<?xml version="1.0" encoding="UTF-8"?>
<users>
    <user>
        <name>Lady Gaga</name>
        <email>lady.gaga@mpp.mpg.de</email>
        <age>36</age>
        <signed_in>true</signed_in>
    </user>
    <user>
        <name>David Hasselhoff</name>
        <email>david.hasselhoff@tum.de</email>
        <age>25</age>
        <signed_in>false</signed_in>
    </user>
    <user>
        <name>Johann Sebastian Bach</name>
        <email>bach@db.de</email>
        <age>300</age>
        <signed_in>true</signed_in>
    </user>
</users>

XML

this is the same address book as before

YAML

What: YAML Ain’t Markup Language —
Origins: ~2001
Used for: Python Configs, GitHub Actions, CI/CD
Why choose it: Very readable for humans!
Caveats: indentation sensitivity

YAML

users:
  - name: Lady Gaga
    email: lady.gaga@mpp.mpg.de
    age: 36
    signed_in: true
  - name: David Hasselhoff
    email: david.hasselhoff@tum.de
    age: 25
    signed_in: false
  - name: Johann Sebastian Bach
    email: bach@db.de
    age: 300
    signed_in: true

# uv add pyyaml
import yaml

with open("lec5-data/userinfo.yaml", "r") as f:
    data = yaml.load(f, Loader=yaml.FullLoader)
    
print(data)

Also loads into a dictionary!

TOML

What: TOML (Tom’s Obvious, Minimal Language)
Origins: 2013, by Tom Preston-Werner (a GitHub co-founder!)
Used for: Python packaging (pyproject.toml)
Why choose it: simple grammar, types
Caveats: Nested structures are simpler

Reading in TOML files

import tomllib

with open("lec5-data/userinfo.toml", "rb") as f:
    data = tomllib.load(f)

Again, loads into a dictionary!

tomllib included in python since v3.11!

Cheat Sheet

File Format	When to use
CSV	Tabular data, not to much data
JSON	Web/API
TOML	Human-readable config
YAML	Human-readable config
XML	Working with super old machine
TXT	Notes, logs, etc

Lecture 5

Recap
Why visualization
Matplotlib essentials
Seaborn quick tour
Important file formats

The End

Learning Data Science

Lecture 5 Visualisation Craft

Lecture 5

Lecture 5

Lecture 5

Lecture 5

Lecture 5

Lecture 5
Visualisation Craft