[python] multiple axis in matplotlib with different scales

How can multiple scales can be implemented in Matplotlib? I am not talking about the primary and secondary axis plotted against the same x-axis, but something like many trends which have different scales plotted in same y-axis and that can be identified by their colors.

For example, if I have trend1 ([0,1,2,3,4]) and trend2 ([5000,6000,7000,8000,9000]) to be plotted against time and want the two trends to be of different colors and in Y-axis, different scales, how can I accomplish this with Matplotlib?

When I looked into Matplotlib, they say that they don't have this for now though it is definitely on their wishlist, Is there a way around to make this happen?

Are there any other plotting tools for python that can make this happen?

This question is related to python matplotlib

The answer is


Bootstrapping something fast to chart multiple y-axes sharing an x-axis using @joe-kington's answer: enter image description here

# d = Pandas Dataframe, 
# ys = [ [cols in the same y], [cols in the same y], [cols in the same y], .. ] 
def chart(d,ys):

    from itertools import cycle
    fig, ax = plt.subplots()

    axes = [ax]
    for y in ys[1:]:
        # Twin the x-axis twice to make independent y-axes.
        axes.append(ax.twinx())

    extra_ys =  len(axes[2:])

    # Make some space on the right side for the extra y-axes.
    if extra_ys>0:
        temp = 0.85
        if extra_ys<=2:
            temp = 0.75
        elif extra_ys<=4:
            temp = 0.6
        if extra_ys>5:
            print 'you are being ridiculous'
        fig.subplots_adjust(right=temp)
        right_additive = (0.98-temp)/float(extra_ys)
    # Move the last y-axis spine over to the right by x% of the width of the axes
    i = 1.
    for ax in axes[2:]:
        ax.spines['right'].set_position(('axes', 1.+right_additive*i))
        ax.set_frame_on(True)
        ax.patch.set_visible(False)
        ax.yaxis.set_major_formatter(matplotlib.ticker.OldScalarFormatter())
        i +=1.
    # To make the border of the right-most axis visible, we need to turn the frame
    # on. This hides the other plots, however, so we need to turn its fill off.

    cols = []
    lines = []
    line_styles = cycle(['-','-','-', '--', '-.', ':', '.', ',', 'o', 'v', '^', '<', '>',
               '1', '2', '3', '4', 's', 'p', '*', 'h', 'H', '+', 'x', 'D', 'd', '|', '_'])
    colors = cycle(matplotlib.rcParams['axes.color_cycle'])
    for ax,y in zip(axes,ys):
        ls=line_styles.next()
        if len(y)==1:
            col = y[0]
            cols.append(col)
            color = colors.next()
            lines.append(ax.plot(d[col],linestyle =ls,label = col,color=color))
            ax.set_ylabel(col,color=color)
            #ax.tick_params(axis='y', colors=color)
            ax.spines['right'].set_color(color)
        else:
            for col in y:
                color = colors.next()
                lines.append(ax.plot(d[col],linestyle =ls,label = col,color=color))
                cols.append(col)
            ax.set_ylabel(', '.join(y))
            #ax.tick_params(axis='y')
    axes[0].set_xlabel(d.index.name)
    lns = lines[0]
    for l in lines[1:]:
        lns +=l
    labs = [l.get_label() for l in lns]
    axes[0].legend(lns, labs, loc=0)

    plt.show()

Since Steve Tjoa's answer always pops up first and mostly lonely when I search for multiple y-axes at Google, I decided to add a slightly modified version of his answer. This is the approach from this matplotlib example.

Reasons:

  • His modules sometimes fail for me in unknown circumstances and cryptic intern errors.
  • I don't like to load exotic modules I don't know (mpl_toolkits.axisartist, mpl_toolkits.axes_grid1).
  • The code below contains more explicit commands of problems people often stumble over (like single legend for multiple axes, using viridis, ...) rather than implicit behavior.

Plot

import matplotlib.pyplot as plt 

# Create figure and subplot manually
# fig = plt.figure()
# host = fig.add_subplot(111)

# More versatile wrapper
fig, host = plt.subplots(figsize=(8,5)) # (width, height) in inches
# (see https://matplotlib.org/3.3.3/api/_as_gen/matplotlib.pyplot.subplots.html)
    
par1 = host.twinx()
par2 = host.twinx()
    
host.set_xlim(0, 2)
host.set_ylim(0, 2)
par1.set_ylim(0, 4)
par2.set_ylim(1, 65)
    
host.set_xlabel("Distance")
host.set_ylabel("Density")
par1.set_ylabel("Temperature")
par2.set_ylabel("Velocity")

color1 = plt.cm.viridis(0)
color2 = plt.cm.viridis(0.5)
color3 = plt.cm.viridis(.9)

p1, = host.plot([0, 1, 2], [0, 1, 2],    color=color1, label="Density")
p2, = par1.plot([0, 1, 2], [0, 3, 2],    color=color2, label="Temperature")
p3, = par2.plot([0, 1, 2], [50, 30, 15], color=color3, label="Velocity")

lns = [p1, p2, p3]
host.legend(handles=lns, loc='best')

# right, left, top, bottom
par2.spines['right'].set_position(('outward', 60))

# no x-ticks                 
par2.xaxis.set_ticks([])

# Sometimes handy, same for xaxis
#par2.yaxis.set_ticks_position('right')

# Move "Velocity"-axis to the left
# par2.spines['left'].set_position(('outward', 60))
# par2.spines['left'].set_visible(True)
# par2.yaxis.set_label_position('left')
# par2.yaxis.set_ticks_position('left')

host.yaxis.label.set_color(p1.get_color())
par1.yaxis.label.set_color(p2.get_color())
par2.yaxis.label.set_color(p3.get_color())

# Adjust spacings w.r.t. figsize
fig.tight_layout()
# Alternatively: bbox_inches='tight' within the plt.savefig function 
#                (overwrites figsize)

# Best for professional typesetting, e.g. LaTeX
plt.savefig("pyplot_multiple_y-axis.pdf")
# For raster graphics use the dpi argument. E.g. '[...].png", dpi=200)'

if you want to do very quick plots with secondary Y-Axis then there is much easier way using Pandas wrapper function and just 2 lines of code. Just plot your first column then plot the second but with parameter secondary_y=True, like this:

df.A.plot(label="Points", legend=True)
df.B.plot(secondary_y=True, label="Comments", legend=True)

This would look something like below:

enter image description here

You can do few more things as well. Take a look at Pandas plotting doc.