Plotting in Python

Matplotlib, Seaborn, Bokeh, and APLpy

Kevin Gullikson

Outline

• Matplotlib
  • Basics
  • Latex
  • legends
  • Annotation
  • Multiple axes
  • Multiple subplots
• Seaborn: Pretty colors, easy configuration
• Bokeh
  • hover tooltips
  • linked brushing
• APLpy: Astronomy plotting library

All images and source for this slideshow are available here

Aims to make easy things easy, and difficult things possible.

• Main plotting library
• Almost everyone has it
• Extensive gallery
• Jet == sad panda

# Imports
import numpy as np
%pylab inline

Populating the interactive namespace from numpy and matplotlib

Basic Plotting

x = np.arange(0, 2.0*np.pi, 0.01)   # I think the IDL equivalent is findgen...
y = np.sin(x)

fig, ax = pylab.subplots()
_ = ax.plot(x,y, lw=2)

Axis Labels

x = np.arange(0, 2.0*np.pi, 0.01)
y = np.sin(x)

fig, ax = pylab.subplots()
_ = ax.plot(x,y, lw=2)
_ = ax.set_xlabel('Time', fontsize=15)
_ = ax.set_ylabel('Value', fontsize=15)
plt.savefig('gullikson_gsps_python_plotting/Sin_curve_with_axes.svg')

ax.set_xlabel('Time', fontsize=15)
ax.set_ylabel('Value', fontsize=15)

Legends + LaTex

x = np.arange(0, 2.0*np.pi, 0.01)
y1 = np.sin(x)
y2 = np.sin(2*x)

fig, ax = pylab.subplots()
_ = ax.plot(x,y1, lw=2, label=r'$\nu = 1 s^{-1}$')
_ = ax.plot(x,y2, lw=2, label=r'$\nu = 2 s^{-1}$')
_ = ax.set_xlabel('Time', fontsize=15)
_ = ax.set_ylabel('Value', fontsize=15)
leg = ax.legend(loc='best', fancybox=True)
leg.get_frame().set_alpha(0.5)
plt.savefig('gullikson_gsps_python_plotting/sinplot_with_legend.svg')

_ = ax.plot(x,y1, lw=2, label=r'$\nu = 1 s^{-1}$')
_ = ax.plot(x,y2, lw=2, label=r'$\nu = 2 s^{-1}$')
leg = ax.legend(loc='best', fancybox=True)
leg.get_frame().set_alpha(0.5)

Annotation

x = np.arange(0, 2.0*np.pi, 0.01)
y1 = np.sin(x)
y2 = np.sin(2*x)

# Plot
fig, ax = pylab.subplots()
_ = ax.plot(x,y1, lw=2, label=r'$\nu = 1 s^{-1}$')
_ = ax.plot(x,y2, lw=2, label=r'$\nu = 2 s^{-1}$')
_ = ax.set_xlabel('Time', fontsize=15)
_ = ax.set_ylabel('Value', fontsize=15)
leg = ax.legend(loc='best', fancybox=True)
leg.get_frame().set_alpha(0.5)

# Adjust axes
ax.set_ylim((-1.2, 1.3))
ax.set_xlim((0, 2.0*np.pi))

_ = ax.annotate('Peak', xy=(np.pi/2, 1), xytext=(np.pi, 1.1), 
            arrowprops=dict(facecolor='black', # Arrow color
                            shrink=0.1,        # Distance between arrow and text
                            width=2),          # Width of the arrow, in points
            fontsize=15)
pylab.savefig('gullikson_gsps_python_plotting/Annotated.svg')

# Adjust axes
ax.set_ylim((-1.2, 1.3))
ax.set_xlim((0, 2.0*np.pi))

ax.annotate('Peak', xy=(np.pi/2, 1), xytext=(np.pi, 1.1), 
            arrowprops=dict(facecolor='black', # Arrow color
                            shrink=0.1,        # Distance between arrow and text
                            width=2),          # Width of the arrow, in points
            fontsize=15)

Adding axes

• For a full example, see what I did here (It uses data on my local computer, so you won't be able to run it as-is)
• Here is the relevent code

def add_top_axis(axis, spt_values=('M5', 'M0', 'K5', 'K0', 'G5', 'G0')):
    # Find the temperatures at each spectral type
    temp_values = MS.Interpolate('Temperature', spt_values)
    
    # make the axis
    top = axis.twiny()
    
    # Set the full range to be the same as the data axis
    xlim = axis.get_xlim()
    top.set_xlim(xlim)
    
    # Set the ticks at the temperatures corresponding to the right spectral type
    top.set_xticks(temp_values)
    top.set_xticklabels(spt_values)
    top.set_xlabel('Spectral Type')
    return top

Subplots

from matplotlib import gridspec

xfull = np.arange(0, 2.0*np.pi, 0.01)
xdata = np.random.random(size=20)* 2.0 * np.pi
yfull = np.sin(x)
ydata = np.sin(xdata) + np.random.normal(scale=0.2, size=xdata.size)

fig = pylab.figure()
gs = gridspec.GridSpec(2, 1, height_ratios=[3, 1.5]) 
top = plt.subplot(gs[0])
top.plot(xdata, ydata, 'rx')
top.plot(xfull, yfull, 'k--', lw=2)
bottom = plt.subplot(gs[1])
bottom.plot(xdata, ydata-np.sin(xdata), 'rx', lw=2)
bottom.plot(xfull, np.zeros(xfull.size), 'k--')

bottom.set_xlabel('Time')
bottom.set_ylabel('Residuals')
top.set_ylabel('Value')

# Save the figure before making it look prettier
plt.savefig('gullikson_gsps_python_plotting/sinplot_resid_basic.svg')

# Now, Make it look prettier
top.set_xticklabels([])  # Remove tick labels on the x axis
fig.subplots_adjust(hspace=0)  # Remove space between subplots
top.set_xlim(bottom.get_xlim())  # Make sure the x-limits are the same
bottom.set_yticks(bottom.get_yticks()[:-1])  # Remove the top y tick, which overlaps with the bottom one of the top plot
top.set_ylim((-1.6, 1.6))

plt.savefig('gullikson_gsps_python_plotting/sinplot_resid_nice.svg')

from matplotlib import gridspec

fig = pylab.figure()
gs = gridspec.GridSpec(2, 1, height_ratios=[3, 1.5]) 
top = plt.subplot(gs[0])
top.plot(xdata, ydata, 'rx')
top.plot(xfull, yfull, 'k--', lw=2)
bottom = plt.subplot(gs[1])
bottom.plot(xdata, ydata-np.sin(xdata), 'rx', lw=2)
bottom.plot(xfull, np.zeros(xfull.size), 'k--')

Setting Defaults

• ~/.matplotlib/matplotlibrc
  • backend : TkAgg
  • lines.linewidth : 1.0
  • Search github for matplotlibrc files
  • See here for documentation
• On top of the script

import matplotlib
matplotlib.rc('lines', linewidth=2, color='r')

# Must edit the rc params before this (or pylab import)
import matplotlib.pyplot as plt

Seaborn

• Aims to make plots prettier by default (jet not even available!)
• As easy as 'import seaborn as sns' on top of script
• Easy colormap changes
• Plays well with pandas
• Also has its own built-in functions for:
  • Linear fitting and plotting the result
  • Visualizing matrices
• Has its own example gallery

"If matplotlib “tries to make easy things easy and hard things possible”, seaborn aims to make a well-defined set of hard things easy too.

import seaborn as sns
sns.set_style('darkgrid')
x = np.arange(0, 2.0*np.pi, 0.01)
y1 = np.sin(x)
y2 = np.sin(2*x)

# Plot
fig, ax = pylab.subplots()
_ = ax.plot(x,y1, lw=2, label=r'$\nu = 1 s^{-1}$')
_ = ax.plot(x,y2, lw=2, label=r'$\nu = 2 s^{-1}$')
_ = ax.set_xlabel('Time', fontsize=15)
_ = ax.set_ylabel('Value', fontsize=15)
leg = ax.legend(loc='best', fancybox=True)
leg.get_frame().set_alpha(0.5)

# Adjust axes
ax.set_ylim((-1.2, 1.3))
ax.set_xlim((0, 2.0*np.pi))

_ = ax.annotate('Peak', xy=(np.pi/2, 1), xytext=(np.pi, 1.1), 
            arrowprops=dict(facecolor='black', # Arrow color
                            shrink=0.1,        # Distance between arror and text (also arrow and the xy coordinates)
                            width=2),          # Width of the arrow, in points
            fontsize=15)
ax.set_title('Seaborn Default', fontsize=20)
_ = pylab.savefig('gullikson_gsps_python_plotting/Annotated_Seaborn_default.svg')

Bokeh

• Make interactive plots via web presentation
• Similar to D3.js but you don't need to know any javascript.
• Hover tooltips
• Streaming large data (eventually...)
• Linked brushing
• Also has a gallery
• Takes a bit more work than seaborn
• Documentation is lacking and/or out of date, but the examples are really useful
• works in ipython notebook with %bokeh "magic"

APLpy: The Astronomical Plotting Library in Python

• Made specifically for astronomy plotting
• Reads in fits files
• Knows about RA/DEC coordinates and how we write them
• Has a (small) gallery

Examples

Credit: Jacob Hummel. Source

Credit: Jacob Hummel. Source

Credit: Aaron Jaurez.

Credit: Aaron Jaurez.

Conclusions

• For line plots and scatter plots, use matplotlib. import seaborn for aesthetics
• For images, consider APLpy
• For interactive data vizualation, consider bokeh

Thanks!