This chapter begins with an introduction to Matplotlib, including the architecture of Matplotlib and the elements of a figure, followed by the recipes. The following are the recipes that will be covered in this chapter:

• Working in interactive mode
• Working in non-interactive mode
• Reading from external files and plotting
• How to change and reset default environment variables

Matplotlib is a cross-platform Python library for plotting two-dimensional graphs (also calledplots). It can be used in a variety of user interfaces such as Python scripts, IPython shells, Jupyter Notebooks, web applications, and GUI toolkits. It can be used to develop professional reporting applications, interactive analytical applications,complexdashboard applications or embed into web/GUI applications. It supports saving figures into various hard-copy formats as well. It also has limited support for three-dimensional figures. It also supports many third-party toolkits to extend its functionality.

Matplotlib has a three-layer architecture:backend,artist, andscripting, organized logically as a stack. Scripting is an API that developers use to create the graphs. Artist does the actual job of creating the graph internally. Backend is where the graph is displayed.

This is the middle layer of the stack. Matplotlib uses theartist object to draw various elements of the graph. So, every element (see elements of a figure) we see in the graph is an artist. This layer provides anobject-oriented API for plotting graphs with maximum flexibility. This interface is meant for seasoned Python programmers, who can create complex dashboard applications.

This is the topmost layer of the stack. This layer provides a simple interface for creating graphs. This is meant for use by end users who don't have much programming expertise. This is called apyplot API.

The high-level Matplotlib object that contains all the elements of the output graph is called afigure. Multiple graphs can be arranged in different ways to form a figure. Each of the figure's elements is customizable.

The following diagram is the anatomy of afigure, containing all its elements:

axes is a sub-section of the figure, where a graph is plotted.axes has atitle, anx-label and ay-label. Afigure can have many suchaxes, each representing one or more graphs. In the preceding figure, there is only oneaxes, two line graphs in blue and red colors.

These are number lines representing the scale of the graphs being plotted. Two-dimensional graphs have anx axis and ay axis, and three-dimensional graphs have anx axis, ay axis, and az axis.

This is the name given to various elements of the figure, for example,xaxis label,yaxis label, graph label (blue signal/red signal in the preceding figureAnatomy of a figure), and so on.

When there are multiple graphs in theaxes (as in the preceding figureAnatomy of a figure), each of them has its own label, and all these labels are represented as a legend. In the preceding figure, the legend is placed at the top-right corner of the figure.

It is the name given to each of theaxes. Thefigure also can have its own title, when the figure has multiple axes with their own titles. The preceding figure has only one axes, so there is only one title for the axes as well as the figure.

Each axis (x,y, orz) will have a range of values that are divided into many equal bins. Bins are chosen at two levels. In the preceding figureAnatomy of a figure, thexaxis scale ranges from 0 to 4, divided into four major bins (0-1, 1-2, 2-3, and 3-4) and each of the major bins is further divided into four minor bins (0-0.25, 0.25-0.5, and 0.5-0.75). Ticks on both sides of major bins are calledmajor ticks and minor bins are calledminor ticks, and the names given to them aremajor ticklabels andminor ticklabels.

Boundaries of the figure are calledspines. There are four spines for each axes(top, bottom, left, and right).

For easier readability of the coordinates of various points on the graph, the area of the graph is divided into a grid. Usually, this grid is drawn along major ticks of thex andy axis. In the preceding figure, the grid is shown in dashed lines.

Matplotlib can be used in aninteractive ornon-interactive modes.In the interactive mode, the graph display gets updated after each statement. In the non-interactive mode, the graph does not get displayed until explicitly asked to do so.

You need working installations of Python, NumPy, and Matplotlib packages.

Using the following commands, interactive mode can be set on or off, and also checked for current mode at any point in time:

• matplotlib.pyplot.ion() to set the interactive modeON
• matplotlib.pyplot.ioff() to switchOFF the interactive mode
• matplotlib.is_interactive() to check whether the interactive mode isON (True) orOFF (False)

Let's see how simple it is to work in interactive mode:

1. Set the screen output as the backend:

%matplotlib inline

2. Import thematplotlib andpyplot libraries. It is common practice in Python to import libraries with crisp synonyms. Noteplt is the synonym for thematplotlib.pyplot package:

import matplotlib as mpl
import matplotlib.pyplot as plt

3. Set the interactive mode to ON:

plt.ion()

4. Check the status of interactive mode:

mpl.is_interactive()

5. You should get the output asTrue.
6. Plot a line graph:

plt.plot([1.5, 3.0])

You should see the following graph as the output:

7. Now add the axis labels and a title to the graph with the help of the following code:

# Add labels and title
plt.title("Interactive Plot") #Prints the title on top of graph
plt.xlabel("X-axis")         # Prints X axis label as "X-axis"
plt.ylabel("Y-axis")         # Prints Y axis label as "Y-axis"

After executing the preceding three statements, your graph should look as follows:

So, this is how the explanation goes:

• plt.plot([1.5, 3.0]) plots a line graph connecting two points (0, 1.5) and (1.0, 3.0).
• Theplot command expects two arguments (Python list, NumPyarray or pandasDataFrame) for thex andy axis respectively.
• If only one argument is passed, it takes it asy axis co-ordinates and forx axis co-ordinates it takes the length of the argument provided.
• In this example, we are passing only one list of two points, which will be taken asyaxis coordinates.
• For thex axis, it takes the default values in the range of 0 to 1, since the length of the list[1.5, 3.0] is 2.
• If we had three coordinates in the list fory, then forx,it would take the range of 0 to 2.
• You should see the graph like the one shown instep 6.
• plt.title("Interactive Plot"), prints the title on top of the graph asInteractive Plot.
• plt.xlabel("X-axis"), prints thex axis label asX-axis.
• plt.ylabel("Y-axis"), prints they axis label asY-axis.
• After executing preceding three statements, you should see the graph as shown instep 7.

If you are using Python shell, after executing each of the code statements, you should see the graph getting updated with title first, then thex axis label, and finally they axis label.

If you are using Jupyter Notebook, you can see the output only after all the statements in a given cell are executed, so you have to put each of these three statements in separate cells and execute one after the other, to see the graph getting updated after each code statement.

You can add one more line graph to the same plot and go on until you complete your interactive session:

1. Plot a line graph:

plt.plot([1.5, 3.0])

2. Add labels and title:

plt.title("Interactive Plot")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")

3. Add one more line graph:

plt.plot([3.5, 2.5])

The following graph is the output obtained after executing the code:

Hence, we have now worked in interactive mode.

In the interactive mode, we have seen the graph getting built step by step with each instruction. In non-interactive mode, you give all instructions to build the graph and then display the graph with a command explicitly.

Working on non-interactive mode won't be difficult either:

1. Start the kernel afresh, and import thematplotlibandpyplot libraries:

import matplotlib
import matplotlib.pyplot as plt

2. Set the interactive mode to OFF:

plt.ioff()

3. Check the status of interactive mode:

matplotlib.is_interactive()

4. You should get the outputFalse.
5. Execute the following code; you will not see the plot on your screen:

# Plot a line graph
plt.plot([1.5, 3.0])

# Plot the title, X and Y axis labels
plt.title("Non Interactive Mode")
plt.xlabel("X-axis")
plt.ylabel("Y-axis")

6. Execute the following statement, and then you will see the plot on your screen:

# Display the graph on the screen
plt.show()

Each of the preceding code statements is self-explanatory. The important thing to note is in non-interactive mode, you write complete code for the graph you want to display, and callplt.show() explicitly to display the graph on the screen.

The following is the output obtained:

By default, Matplotlib accepts input data as a Python list, NumPy array, or pandas DataFrame. So all external data needs to be read and converted to one of these formats before feeding it to Matplotlib for plotting the graph. From a performance perspective, NumPy format is more efficient, but for default labels, pandas format is convenient.

If the data is a.txt file, you can use NumPy function to read the data and put it in NumPy arrays. If the data is in.csv or.xlsx formats, you can use pandas to read the data. Here we will demonstrate how to read.txt,.csv, and.xlsx formats and then plot the graph.

Import thematplotlib.pyplot,numpy , andpandas packages that are required to read the input files:

1. Import thepyplot library with theplt synonym:

import matplotlib.pyplot as plt

2. Import thenumpy library with thenp synonym. Thenumpy library can manage n-dimensional arrays, supporting all mathematical operations on these arrays:

import numpy as np

3. Import thepandas package withpd as a synonym:

import pandas as pd

We will follow the order of.txt,.csv, and.xlsx files, in three separate sections.

Here are some steps to follow:

1. Read the text file into thetxt variable:

txt = np.loadtxt('test.txt', delimiter = ',')
txt

Here is the explanation for the preceding code block:

• Thetest.txttext file has 10 numbers separated by a comma, representing thex andy coordinates of five points (1, 1), (2, 4), (3, 9), (4, 16), and (5, 25) in a two-dimensional space.
• Theloadtxt()function loads text data into a NumPy array.

You should get the following output:

array([ 1., 1., 2., 4., 3., 9., 4., 16., 5., 25.])

2. Convert the flat array into five points in 2D space:

txt = txt.reshape(5,2)
txt

After executing preceding code, you should see the following output:

array([[ 1., 1.], [ 2., 4.], [ 3., 9.], [ 4., 16.], [ 5., 25.]])

3. Split the.txt variable intox andy axis co-ordinates:

x = txt[:,0]
y = txt[:,1]
print(x, y)

Here is the explanation for the preceding code block:

• Separate thex andy axis points from thetxt variable.
• x is the first column intxt andy is the second column.
• The Python indexing starts from 0.

After executing the preceding code, you should see the following output:

[ 1. 2. 3. 4. 5.] [ 1. 4. 9. 16. 25.]

The.csv file has a relational database structure of rows and columns, and thetest.csv file hasx,y co-ordinates for five points in 2D space. Each point is a row in the file, with two columns:x andy. The same NumPyloadtxt() function is used to load data:

x, y = np.loadtxt ('test.csv', unpack = True, usecols = (0,1), delimiter = ',')
print(x)
print(y)

On execution of the preceding code, you should see the following output:

[ 1. 2. 3. 4. 5.] [ 1. 4. 9. 16. 25.]

Now let's read the same data from an.xlsx file and create thex andy NumPyarrays. The.xlsx file format is not supported by the NumPyloadtxt() function. A Python data processing package,pandas can be used:

1. Read the.xlsx file into pandas DataFrame. This file has the same five points in 2D space, each in a separate row withx,y columns:

df = pd.read_excel('test.xlsx', 'sheet', header=None)

2. Convert the pandas DataFrame to a NumPy array:

data_array = np.array(df)
print(data_array)

You should see the following output:

[[ 1 1] [ 2 4] [ 3 9] [ 4 16] [ 5 25]]

3. Now extract thex andy coordinates from the NumPy array:

x , y = data_array[:,0], data_array[:,1]
print(x,y)

You should see the following output:

[1 2 3 4 5] [ 1 4 9 16 25]

After reading the data from any of the three formats (.txt,.csv,.xlsx) and format it tox andy variables, then we plot the graph using these variables as follows:

plt.plot(x, y)

Display the graph on the screen:

plt.show()

The following is the output obtained:

Depending on the format and the structure of the data, we will have to use the Python, NumPy, or pandas functions to read the data and reformat it into an appropriate structure that can be fed into thematplotlib.pyplot function. After that, follow the usual plotting instructions to plot the graph that you want.

Matplotlib uses thematplotlibrc file to store default values for various environment and figure parameters used across matplotlib functionality. Hence, this file is very long. Thesedefault values are customizable to apply for all the plots within a session.

You can use theprint(matplotlib.rcParams) command to get all the default parameter settings from this file.

Thematplotlib.rcParams command is used to change these default values to any other supported values, one parameter at a time. Thematplotlib.rc command is used to set default values for multiple parameters within a specific group, for example, lines, font, text, and so on. Finally, thematplotlib.rcdefaults() command is used to restore default parameters.

The following code block provides the path to the file containing all configuration the parameters:

# Get the location of matplotlibrc file
import matplotlib
matplotlib.matplotlib_fname()

You should see the directory path like the one that follows. The exact directory path depends on your installation:

'C:\\Anaconda3\\envs\\keras35\\lib\\site-packages\\matplotlib\\mpl-
    data\\matplotlibrc'

The following block of code along with comments helps you to understand the process of changing and resetting default environment variables:

1. Import thematplotlib.pyplot package with theplt synonym:

import matplotlib.pyplot as plt

2. Loadx andy variables from sametest.csv file that we used in the preceding recipe:

x, y = np.loadtxt ('test.csv', unpack = True, usecols = (0,1),
                    delimiter = ',')

3. Change the default values for multiple parameters within the group'lines':

matplotlib.rc('lines', linewidth=4, linestyle='-', marker='*')

4. Change the default values for parameters individually:

matplotlib.rcParams['lines.markersize'] = 20
matplotlib.rcParams['font.size'] = '15.0'

5. Plot the graph:

plt.plot(x,y)

6. Display the graph:

plt.show()

The following is the output that will be obtained:

Thematplotlib.rc andmatplotlib.rcParams commands overwrite the default values for specified parameters as arguments in these commands. These new values will be used by thepyplot tool while plotting the graph.

You can reset all the parameters to their default values, using thersdefaults() command, as shown in the following block:

# To restore all default parameters
matplotlib.rcdefaults()
plt.plot(x,y)
plt.show()

The graph will look as follows:

.