Question

I want to use single annotation text to annotate several data points with several arrows. I made a simple workaround:

ax = plt.gca()
ax.plot([1,2,3,4],[1,4,2,6])
an1 = ax.annotate('Test',
  xy=(2,4), xycoords='data',
  xytext=(30,-80), textcoords='offset points',
  arrowprops=dict(arrowstyle="-|>",
                  connectionstyle="arc3,rad=0.2",
                  fc="w"))
an2 = ax.annotate('Test',
  xy=(3,2), xycoords='data',
  xytext=(0,0), textcoords=an1,
  arrowprops=dict(arrowstyle="-|>",
                  connectionstyle="arc3,rad=0.2",
                  fc="w"))
plt.show()

Producing following result:

But I don’t really like this solution because it is… well, an ugly dirty hack.

Besides that, it affects the appearance of annotation (mainly if using semi-transparent bboxes etc).

So, if anyone got an actual solution or at least an idea how to implement it, please share.

Answer 1

I guess the proper solution will require too much effort – subclassing _AnnotateBase and adding support for multiple arrows all by yourself. But I managed to eliminate that issue with second annotate affecting visual appearance simply by adding alpha=0.0. So the updated solution here if no one will provide anything better:

def my_annotate(ax, s, xy_arr=[], *args, **kwargs):
  ans = []
  an = ax.annotate(s, xy_arr[0], *args, **kwargs)
  ans.append(an)
  d = {}
  try:
    d['xycoords'] = kwargs['xycoords']
  except KeyError:
    pass
  try:
    d['arrowprops'] = kwargs['arrowprops']
  except KeyError:
    pass
  for xy in xy_arr[1:]:
    an = ax.annotate(s, xy, alpha=0.0, xytext=(0,0), textcoords=an, **d)
    ans.append(an)
  return ans

ax = plt.gca()
ax.plot([1,2,3,4],[1,4,2,6])
my_annotate(ax,
            'Test',
            xy_arr=[(2,4), (3,2), (4,6)], xycoords='data',
            xytext=(30, -80), textcoords='offset points',
            bbox=dict(boxstyle='round,pad=0.2', fc='yellow', alpha=0.3),
            arrowprops=dict(arrowstyle="-|>",
                            connectionstyle="arc3,rad=0.2",
                            fc="w"))
plt.show()

Resulting picture: