By Alex Hagen
pyg started as a simple wrapper around matplotlib to help me keep my style the same in plotting, but now it's expanded to a full graphics suite. If you get bored reading through the first two examples, skip to the bottom. Those examples are a bit cooler.
For pyg, we need quite a few requirements. Installation right now is pretty manual, but this should do the trick on unix systems:
pip install numpy scipy matplotlib colours
mkdir ~/util
cd ~/util
git clone https://github.com/alexhagen/pyg -b master pyg
sudo echo "export PYTHONPATH=${PYTHONPATH}:~/util" >> ~/.bashrc
source ~/.bashrcand then we can just import pyg whenever with
from pyg import twod as pyg2dpyg has one main class, a twod plot type, and it has several other classes. The table module has some table printing help for Jupyter notebooks and some LaTeX publication helper functions. The threed module has some matplotlib three dimensional plotting (this is good for surface plotting, if you're doing geometric visualization, use my pyb class, which I'll include into pyg soon), three2twod is a class for annotating three dimensional plotting (if you have the transformation matrix from 3-d to 2-d). I've created some informative examples of these below. I've put interesting examples first, but they're a little complex. If you want to get started, skip to the "Boring Examples" section.
The following shows an example for 3d plotting using pyg, which is generally simple except for the conversion from matrix data into a form that can be plotted. Below shows a simple example for a power fit, but an API is soon coming for converting data into the right formats.
from pyg import threed as pyg3d
import numpy as np
plot = pyg3d.pyg3d()
x = np.linspace(0.0, 5.0)
y = np.linspace(0.0, 5.0)
X, Y = np.meshgrid(x, y)
z = np.power(X, 2.0) - np.power(Y, 3.0)
plot.surf(x, y, z)
plot.export('_static/threed_surf')
plot.show(caption='An arbitrary three dimensional surface')/Users/hage581/code/pyg/pyg/threed.py:100: UserWarning: Requested projection is different from current axis projection, creating new axis with requested projection.
self.ax = self.fig.gca(projection='3d')
<div class='pygfigure' name='['Anarbitrarythreedimensionalsurface']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/threed_surf.svg?794197054' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 1:</b> An arbitrary three dimensional surface</div>
</div>
The description of 3 dimensional geometry by annotation is difficult in the best circumstances, but very few things are "best circumstances". Most of the field of visualization relies on software specifically designed for visualization, such as VTK and its derivatives. This is very powerful, but for someone analyzing data or writing simulations, the last thing they want to do is write an interface to these programs. So, I've written a quick and easy API to keep all the data in Python and visualize the geometry in Blender. Then, I have an interface for which the code can extract the camera parameters of an exported render, and then the user can plot two dimensional annotations overtop of the three-dimensional geometry, in place. A more advanced example is coming, but a rudimentary example is shown below.
from pyg.pyb import pyb
from pyg import three2twod as pyg32d
scene = pyb.pyb()
scene.rpp(c=(0., 0., 0.), l=(100., 100., 100.), name='cube')
scene.flat(color='#fc8d82', name='newgold')
scene.set_matl(obj='cube', matl='newgold')
scene.rpp(c=(0., 0., -65.), l=(500., 500., 30.), name='floor')
scene.flat(color='#888888', name='gray')
scene.set_matl(obj='floor', matl='gray')
scene.run('_static/blenderrender.png')
plot = pyg32d.ann_im('_static/blenderrender.png')
plot.add_data_pointer(0., 0., 0., string=r'$\vec{c} = \left( 0, 0, 0 \right)$',
place=(-500., 200.))
plot.add_legend_entry(color='#fc8d82', name='Cube')\
.add_legend_entry(color='#888888', name='Floor')
plot.legend(loc=2)
plot.export('_static/ann_im', ratio='golden')
plot.show('Using two dimensional annotations on a three dimensional geometric plot') <div class='pygfigure' name='['Usingtwodimensionalannotationsonathreedimensionalgeometricplot']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/ann_im.svg?1938495231' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 2:</b> Using two dimensional annotations on a three dimensional geometric plot</div>
</div>
One thing I always hated about most plotting programs is how hard it is to add "measurements". These are so useful in calling out visual information that they're near universal in CAD, but in most plotting and other visualization, they're nowhere to be found. So, for the most part, I've included measurements in pyg. The following example shows some measurements of grade distributions at IU's School of Medicine versus their nursing department. It shows how the distributions are clearly not normally distributed, but it also shows the grade inflation for the nursing department, with the overwhelming majority of classes giving A's, whereas the medical school fails a large proportion of students.
from scipy.stats import gaussian_kde as gkde
grades = [4.0, 4.0, 3.7, 3.3, 3.0, 2.7, 2.3, 2.0, 1.7, 1.3, 1.0, 0.7, 0.0]
med = [51, 188, 84, 74, 141, 69, 54, 84, 45, 30, 51, 19, 53]
nur = [228, 160, 89, 58, 77, 38, 17, 10, 1, 0, 0, 0, 0]
_med = []
for m, g in zip(med, grades):
_med += [g] * m
_nur = []
for n, g in zip(nur, grades):
_nur += [g] * n
m_dist = gkde(_med)
n_dist = gkde(_nur)
sigma_m = np.std(_med)
mu_m = np.mean(_med)
sigma_n = np.std(_nur)
mu_n = np.mean(_nur)
_grades = np.linspace(0., 4.0)
plot = pyg2d.pyg2d()
plot.add_line(_grades, m_dist(_grades), linestyle='-', linecolor='#285668')
plot.add_line(_grades, n_dist(_grades), linestyle='-', linecolor='#fc8d82')
plot.add_hmeasure(mu_m + sigma_m, mu_m - sigma_m, 0.35, 'middle $2\sigma$')
plot.add_hmeasure(mu_n + sigma_n, mu_n - sigma_n, 1.5, 'middle $2\sigma$')
plot.xlabel(r'Grade ($g$) [$\text{GPA Points}$]')
plot.ylabel(r'Likelihood ($P$) [ ]')
plot.lines_on()
plot.markers_off()
plot.export('_static/measure', ratio='silver')
plot.show(caption='Depiction of useful measurements on a two-d plot') <div class='pygfigure' name='['Depictionofusefulmeasurementsonatwodplot']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/measure.svg?442736828' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 3:</b> Depiction of useful measurements on a two-d plot</div>
</div>
The simplest plotting in pyg is line plotting, and the following two figures show the api for plotting a line with its associated uncertainty.
x = np.linspace(0.0, 4.0 * np.pi, 1000)
y = np.sin(x)
u_y = 0.1
plot = pyg2d.pyg2d()
plot.add_line(x, y, linestyle='-', linecolor='#285668', yerr=u_y, error_fill=True,
name=r'$\sin \left( \theta \right)$')
plot.xlabel('x-coordinate ($x$) [$\unit{cm}$]')
plot.ylabel('y-coordinate ($y$) [$\unit{cm}$]')
plot.lines_on()
plot.markers_off()
plot.export('_static/line', ratio='silver')
plot.show(caption='A line drawing with uncertainty in y') <div class='pygfigure' name='['Alinedrawingwithuncertaintyiny']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/line.svg?1326500369' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 4:</b> A line drawing with uncertainty in y</div>
</div>
x = np.linspace(0.0, 4.0 * np.pi, 1000)
y = 5.0 * np.cos(x)
u_y = 1.0
x_sparse = np.linspace(0.0, 4.0 * np.pi, 10)
y_sparse = 5.0 * np.cos(x_sparse)
u_y_sparse = 1.0
plot = pyg2d.pyg2d()
plot.add_line(x, y, linestyle='-', linecolor='#fc8d82', yerr=u_y, error_fill=True,
name=r'$\sin \left( \theta \right)$')
plot.add_line(x_sparse, y_sparse, linecolor='#000000', yerr=u_y_sparse,
name=r'sparse')
plot.lines_on()
plot.markers_off()
plot.lines['sparse'].set_alpha(1.0)
plot.lines['sparse'].set_markersize(6)
plot.lines['sparse'].set_linewidth(0.0)
plot.export('_static/err', ratio='silver')
plot.show(caption='Sinusoid with uncertainty and a sparsely sampled sinusoid with uncertainty') <div class='pygfigure' name='['Sinusoidwithuncertaintyandasparselysampledsinusoidwithuncertainty']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/err.svg?1944181666' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 5:</b> Sinusoid with uncertainty and a sparsely sampled sinusoid with uncertainty</div>
</div>
The following figure shows the API for plotting data on concurrent axes. There are two different APIs to this: the first requires you to plot your data, and then define a function that converts one axis to another. The other API requires you to plot two different data sets on axes with different limits.
x = np.linspace(0., 4.0 * np.pi, 1000)
y1 = 1.0 * np.sin(x)
y2 = 5.0 * np.cos(x)
plot = pyg2d.pyg2d()
plot.add_line(x, y1, linecolor='#fc8d82', name='$y_{1}$')
plot.add_line_yy(x, y2, linecolor='#285668', name='$y_{2}$')
plot.markers_off()
plot.xlabel('x coordinate ($x$)')
plot.ylabel('y coordinate ($y_{1}$)')
plot.ylabel('y coordinate ($y_{2}$)', axes=plot.ax2)
plot.legend(loc=3)
plot.export('_static/dual', ratio='silver')
plot.show('Sinusoids with the same $x$ axis, on different $y$ axes') <div class='pygfigure' name='['Sinusoidswiththesamexaxisondifferentyaxes']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/dual.svg?1592495023' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 6:</b> Sinusoids with the same $x$ axis, on different $y$ axes</div>
</div>
The next figure shows how you can compare a single function against different ordinate axes. This would be useful if you are comparing different units, but I particularly use it when there is some electrical measurement that is calibrated non-linearly (for example, in gamma spectroscopy).
x = np.linspace(0., 4.0 * np.pi, 1000)
y = 1.0 * np.sin(x)
plot = pyg2d.pyg2d()
plot.add_line(x, y, linecolor='#285668', name='$y$')
plot.markers_off()
def pi_div(x):
return x / np.pi
plot.add_xx(pi_div)
plot.xlabel('x coordinate ($x$) [$\unit{cm}$]')
plot.xlabel('x coordinate in terms of $\pi$ ($x$) [$\unit{\pi}$]', axes=plot.ax2)
plot.ylabel('y coordinate ($y$) [$\unit{cm}$]')
plot.export('_static/dualx', ratio='silver')
plot.show('Sinusoid in terms of radians and in terms of $\pi$') <div class='pygfigure' name='['Sinusoidintermsofradiansandintermsofpi']' style='text-align: center; max-width: 800px; margin-left: auto; margin-right: auto;'>
<img style='margin: auto; max-width:100%; width:1250.000000px; height: auto;' src='_static/dualx.svg?244575124' />
<div style='margin: auto; text-align: center;' class='figurecaption'><b>Figure 7:</b> Sinusoid in terms of radians and in terms of $\pi$</div>
</div>
- SVG import for illustrating on charts
- SVG addition via post processing - only suitable for SVG export https://stackoverflow.com/questions/31452451/importing-an-svg-file-a-matplotlib-figure
- SVG conversion to matplotlib via regexing https://matplotlib.org/examples/showcase/firefox.html
- HTML interactivity for websites http://jakevdp.github.io/blog/2014/01/10/d3-plugins-truly-interactive/