pyplot: change ncols within a legend

Question

I'm plotting a large data set and some regressions in pyplot. The data is colored according to an additional value. I decided to set the number of columns in the legend to 2.

It looks nice for the data points, but for the regressions, I'd like to go back to ncols=1. It is possible to do this within one legend?

(I know, I could declare two legends but I'd like to avoid this...)

Not sure I understand what you are asking. You want a legend that is part 2 columns, part 1 column? — tacaswell, Jul 28 '13 at 17:54
Exactly. Some of the labels are very short, i.e. just a single digit, some are much longer. — Tyrax, Jul 29 '13 at 09:04
Alright, I'll try to set up two legends that look like one. Still have to figure out how... — Tyrax, Jul 29 '13 at 15:37
Might be worth emailing the mpl mailing lists, the core devs (who would be able to answer this for sure) don't watch SO. — tacaswell, Jul 29 '13 at 15:40

score 8 · Accepted Answer · answered Aug 06 '13 at 14:10

Legend in matplotlib is a collection of OffsetBox. In principle, you can change "ncol" within a legend by manually rearranging these offsetboxes. Of course, this needs some knowledge of matplotlib internals and will be difficult for normal users. Not sure what would be the best way to expose this. Here is a simple code that creates two legends and combine them into one.

import matplotlib.pyplot as plt
import matplotlib.legend as mlegend

ax = plt.subplot(111)
l1, = ax.plot([1,2,3])

leg1 = ax.legend([l1], ["long lable"], ncol=1)

leg2 = mlegend.Legend(ax, [l1, l1, l1, l1], tuple("1234"), ncol=2)

leg1._legend_box._children.append(leg2._legend_box._children[1])
leg1._legend_box.align="left" # the default layout is 'center'

plt.show()

Excellent, this is pretty close to what I had in mind. Thanks a lot. — Tyrax, Aug 07 '13 at 13:54

pyplot: change ncols within a legend

1 Answers1

Linked