Closed
Description
Bug (?) report
Bug summary
I do not know if it was ever intended to combine the bbox_to_anchor and the loc kwargs in legend. But it results in a behavior that I find rather surprising.
Code for reproduction
import matplotlib.pyplot as plt
def do_plot(ax, loc='best', bbox_to_anchor=(0.5, 0.5), color='black'):
ax.plot([0, 1], color=color, label='A label')
ax.set_ylim(top=3)
ax.legend(loc=loc, bbox_to_anchor=bbox_to_anchor)
# Cosmeticks
ax.xaxis.set_major_locator(plt.NullLocator())
ax.yaxis.set_major_locator(plt.NullLocator())
ax.set_title("loc={l}\nbbox_to_anchor={b}".format(l=loc, b=bbox_to_anchor))
return ax
fig, ((ax0, ax1), (ax2, ax3), (ax4, ax5)) = plt.subplots(nrows=3, ncols=2)
# First row
do_plot(ax0, color='r', loc='upper right', bbox_to_anchor=(-0.001, 1))
do_plot(ax1, color='g', loc='best', bbox_to_anchor=(-0.001, 1))
# Second row
do_plot(ax2, color='b', loc='upper left', bbox_to_anchor=(0.001, 1))
do_plot(ax3, color='y', loc='best', bbox_to_anchor=(0.001, 1))
# Third row
do_plot(ax4, color='m', loc='upper left', bbox_to_anchor=(0.1, 1))
do_
5B26
plot(ax5, color='c', loc='best', bbox_to_anchor=(0.1, 1))
fig.subplots_adjust(left=0.2, hspace=0.5, wspace=0.75, bottom=0.05, top=0.90)
plt.show()Actual outcome
Matplotlib 2.0.0:
Matplotlib 1.5.1:
Expected outcome
- First row is fine for me: 'best' <- 'upper right' because else the legend frame would have overlapped with the left spine.
- Second and third rows seem weird to me: why would the best location not be something like 'upper left'?
If I am not the only one to find the behavior of the left column a bit surprising, then the real problem is that it is now the default legend behavior in 2.0 if the user only provide a bbox_to_anchor but no loc, isn't it?
Matplotlib version
Tested on Linux (CentOS 7), with Python 2.7 and mpl 2.0.0 and 1.5.1 (installed from conda).
Metadata
Metadata
Assignees
Labels
No labels