how to avoid some function in the legend? - matplotlib

Question

I need to include a line into a figure every time a button is clicked (I'm using pyqt4), this line has to be labeled and I also need to compare these lines with a constant function. Here is what I've tried:
labels = []
fig = plt.figure()
ax = fig.add_subplot(111, axisbg='white')
ax.hold(True)
def function(k):
x = np.linspace(0, 2, 100)
y = np.sin(k * np.pi * x) * np.exp(-5 * x)
labels.append('k = {}'.format(k))
ax.plot(x, y)
# reference line
plt.axhline(y=0.1, c='k', linestyle='--')
plt.legend(labels)
for i in range(0,5):
function(i)
plt.show()
The result:
There is a simple way to skip the constant line marker in the legend frame?

Answer 1

Maybe I'm not following but it doesn't look like your reference line axhline(y=0.1, ...) is included in the legend.
I would set this separately, no reason to redraw it every time you plot a new line. Also try passing the label inside the plot function
fig = plt.figure()
ax = fig.add_subplot(111, axisbg='white')
ax.hold(True)
# reference line - only draw this once
plt.axhline(y=0.1, c='k', linestyle='--')
def function(k):
x = np.linspace(0, 2, 100)
y = np.sin(k * np.pi * x) * np.exp(-5 * x)
ax.plot(x, y, linestyle='-', label='k = {}'.format(k)) # set label attribute for line
for i in range(0,5):
function(i)
plt.legend() # you only need to call this once, it will generate based on the label assigned to line objects
plt.show()
Note: If you want to do this interactively (i.e. draw on a button press) then you'll have to call plt.legend() upfront and call plt.draw() after each new line is added, that way it'll update the legend.

Answer 2

This is because there are actually 10 lines in your plot but your legend only shows 5 labels. If you check this by putting the label in the plot and axhline commands like this.
def function(k):
x = np.linspace(0, 2, 100)
y = np.sin(k * np.pi * x) * np.exp(-5 * x)
ax.plot(x, y, label='k = {}'.format(k))
# reference line
ax.axhline(y=0.1, c='k', linestyle='--', label='reference')
ax.legend()
print "number of lines in plot: {}".format(len(ax.lines))
Because you set the Axes.hold property to True, the Axes is not cleared, but a new line is added to Axes object every time you call these commands. This may be faster but you have to be careful to avoid adding duplicate artists. A simple solution would be to split the drawing in two functions: one to create an empty plot and one to add a line.
import matplotlib.pyplot as plt
import numpy as np
def init_plot(ax):
ax.hold(True)
ax.axhline(y=0.1, c='k', linestyle='--', label='reference')
ax.legend()
def add_line(ax, k):
x = np.linspace(0, 2, 100)
y = np.sin(k * np.pi * x) * np.exp(-5 * x)
ax.plot(x, y, label='k = {}'.format(k))
ax.legend()
def main():
fig = plt.figure()
ax = fig.add_subplot(111, axisbg='white')
init_plot(ax)
for i in range(0,5):
add_line(ax, i)
plt.show()
#raw_input('please press enter\n') # for OS-X
if __name__ == "__main__":
main()
I recommend to read the Artist tutorial and of course the Legend guide.