chart

Charts with pure python

Description

A zero-dependency python package that prints basic charts to a Jupyter output

Programming language: Python

License: MIT License

Tags: Data Visualization     Scientific     Engineering     Utilities     Visualization     Multimedia     Printing     Graphics     Artistic Software    

README

A zero-dependency python package that prints basic charts to a Jupyter output

Charts supported:

• Bar graphs
• Scatter plots
• Histograms
• 🍑📊👏

Examples

Bar graphs can be drawn quickly with the bar function:

from chart import bar

x = [500, 200, 900, 400]
y = ['marc', 'mummify', 'chart', 'sausagelink']

bar(x, y)

       marc: ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇             
    mummify: ▇▇▇▇▇▇▇                       
      chart: ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇
sausagelink: ▇▇▇▇▇▇▇▇▇▇▇▇▇                              

And the bar function can accept columns from a pd.DataFrame:

from chart import bar
import pandas as pd

df = pd.DataFrame({
    'artist': ['Tame Impala', 'Childish Gambino', 'The Knocks'],
    'listens': [8_456_831, 18_185_245, 2_556_448]
})
bar(df.listens, df.artist, width=20, label_width=11, mark='🔊')

Tame Impala: 🔊🔊🔊🔊🔊🔊🔊🔊🔊           
Childish Ga: 🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊🔊
 The Knocks: 🔊🔊🔊                                

Histograms are just as easy:

from chart import histogram

x = [1, 2, 4, 3, 3, 1, 7, 9, 9, 1, 3, 2, 1, 2]

histogram(x)

▇        
▇        
▇        
▇        
▇ ▇      
▇ ▇      
▇ ▇      
▇ ▇     ▇
▇ ▇     ▇
▇ ▇   ▇ ▇

And they can accept objects created by scipy:

from chart import histogram
import scipy.stats as stats
import numpy as np

np.random.seed(14)
n = stats.norm(loc=0, scale=10)

histogram(n.rvs(100), bins=14, height=7, mark='🍑')

            🍑              
            🍑   🍑          
            🍑 🍑 🍑          
            🍑 🍑 🍑          
        🍑   🍑 🍑 🍑          
      🍑 🍑 🍑 🍑 🍑 🍑 🍑 🍑 🍑    
      🍑 🍑 🍑 🍑 🍑 🍑 🍑 🍑 🍑   🍑

Scatter plots can be drawn with a simple scatter call:

from chart import scatter

x = range(0, 20)
y = range(0, 20)

scatter(x, y)

                                       •
                                   • •  
                                 •      
                             • •        
                         • •            
                       •                
                  •  •                  
                •                       
            • •                         
        • •                             
      •                                 
  • •                                   
•                                       

And at this point you gotta know it works with any np.array:

from chart import scatter
import numpy as np

np.random.seed(1)
N = 100
x = np.random.normal(100, 50, size=N)
y = x * -2 + 25 + np.random.normal(0, 25, size=N)

scatter(x, y, width=20, height=9, mark='^')

^^                  
 ^                  
    ^^^             
    ^^^^^^^         
       ^^^^^^       
        ^^^^^^^     
            ^^^^    
             ^^^^^ ^
                ^^ ^

In fact, all chart functions work with pandas, numpy, scipy and regular python objects.

Preprocessors

In order to create the simple outputs generated by bar, histogram, and scatter I had to create a couple of preprocessors, namely: NumberBinarizer and RangeScaler.

I tried to adhere to the scikit-learn API in their construction. Although you won't need them to use chart here they are for your tinkering:

from chart.preprocessing import NumberBinarizer

nb = NumberBinarizer(bins=4)
x = range(10)
nb.fit(x)
nb.transform(x)

[0, 0, 0, 1, 1, 2, 2, 3, 3, 3]

from chart.preprocessing import RangeScaler

rs = RangeScaler(out_range=(0, 10), round=False)
x = range(50, 59)
rs.fit_transform(x)

[0.0, 1.25, 2.5, 3.75, 5.0, 6.25, 7.5, 8.75, 10.0]

Installation

pip install chart

Contribute

For feature requests or bug reports, please use Github Issues

Inspiration

I wanted a super-light-weight library that would allow me to quickly grok data. Matplotlib had too many dependencies, and Altair seemed overkill. Though I really like the idea of termgraph, it didn't really fit well or integrate with my Jupyter workflow. Here's to chart 🥂 (still can't believe I got it on PyPI)