matplotlib legend: Including markers and lines from two different graphs in one line

Question

I\'ve been doing some linear regression and want to plot the markers (original data) and the lines (regression) on the same line in the legend. For simplicity, here\'s a fak

Answer 1

You just need to use legend a bit more directly. See Matplotlib - How to make the marker face color transparent without making the line transparent and user guide.

ax.legend([(p1, p3), (p2, p4)], ['set 1', 'set 2'])
plt.draw()

Answer 2

You can just try with

from pylab import *
ax = subplot(1,1,1)
p1, = ax.plot([1,2,3,4,5,6],'r-')
p2, = ax.plot([6,5,4,3,2,1],'b-')

p3, = ax.plot([1.2,1.8,3.1,4.1,4.8,5.9],'r-o', label="dots 1")
p4, = ax.plot([6.1,5.1,3.8,3.1,1.9,0.9],'b-o', label="dots 2")

ax.legend(loc='center right',numpoints=1)
show()

or if you want a poor man's solution, you can plot something outside your plotting range and label only that plot. For instance

p5 = ax.plot(ones(2)*1e6,ones(2)*1e6,'r-o', label="dots 1")

do the same for the other label and then force your plot not to include p5, for example, like this

ax.set_xlim(0,10);ax.set_ylim(0,10)

Answer 3

I usually solve this problem by creating dummy lines with the plot properties that I am interested in showing. However, I think @tcaswell's solution is better.

from matplotlib.lines import Line2D

def create_dummy_line(**kwds):
    return Line2D([], [], **kwds)

# your code here

# Create the legend
lines = [
    ('name A', {'color': 'red', 'linestyle': '-', 'marker': 'o'}),
    ('name B', {'color': 'blue', 'linestyle': '-', 'marker': 'o'}),
]
ax.legend(
    # Line handles
    [create_dummy_line(**l[1]) for l in lines],
    # Line titles
    [l[0] for l in lines],
    loc='center right'
)