Control angle of arrow in matplotlib.pyplot.annotate

Question

it is driving me nuts that the annotate arrow in the screenshot below is not straight. What am I missing? I'm suspecting it has to do with the way I specify the xy and xytext position, yet they both have the same y.

This is how I'm specifying the annotate:

plt.annotate('Head Motion Cutoff',
             xy=[data.shape[0], motion_cutoff],
             xytext=[data.shape[0]+12, motion_cutoff],
             fontsize=14,
             arrowprops=dict(fc='#A9A9A9',
                             ec='#A9A9A9',
                             arrowstyle='simple',
                             shrinkA=4,
                             shrinkB=4))

Can you provide the code that allows to reproduce the issue? See minimal reproducible example. — ImportanceOfBeingErnest
– ImportanceOfBeingErnest, Commented Jan 11, 2018 at 16:09

ImportanceOfBeingErnest · Accepted Answer · 2018-01-12 12:11:30Z

4

The problem is that the arrow is by default connected to the center of the text.

You may however change the position to which the arrow connects using the relpos argument of the arrowproperties.

plt.annotate(...,  arrowprops=dict(..., relpos=(0, 0)) )

The relative position is specified in coordinates of the text's bounding box.

For bottom aligned text, one would choose relpos=(0,0).
For center aligned text, one would choose relpos=(0,0.5).
For top aligned text, one would choose relpos=(0,1).
A problem is, if the text does not contain any character (like the "g" here), which goes down to the bottom, then a relpos=(0,0.2) might make sense.

Example:

import matplotlib.pyplot as plt
fig, ax = plt.subplots(1, dpi=200)
ax.grid(color='k', alpha=0.5, ls=':')

plt.annotate('Head Motion Cutoff',
         xy=[0.1, 0.8],
         xytext=[60, 0], 
         verticalalignment = "baseline",
         arrowprops=dict(arrowstyle='simple',
                         fc='#A9A9A9', ec='#A9A9A9',
                         shrinkA=4, shrinkB=4,
                         relpos=(0, 0.2)),             
         fontsize=11, 
         textcoords="offset points")



ap = dict(fc='#A9A9A9', ec='#A9A9A9', arrowstyle='simple',
          shrinkA=4, shrinkB=4)

fontsize = 11
aligns = ["bottom", "top", "center"]
positions =  [dict(relpos=(0, 0.)),dict(relpos=(0, 1)),dict(relpos=(0, 0.5))]
kw = dict(fontsize=fontsize, textcoords="offset points")


for i, (align,pos) in enumerate(zip(aligns,positions)):
    ap.update(pos)
    kw.update(dict(arrowprops=ap)) 
    plt.annotate('Head Motion Cutoff (va={})'.format(align),
             xy=[0.1, i*0.2+0.2],
             xytext=[60, 0],
             va=align, ha="left", **kw)

plt.show()

edited Jan 12, 2018 at 12:11

answered Jan 11, 2018 at 22:00

ImportanceOfBeingErnest

342k61 gold badges739 silver badges771 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

Maarten Over a year ago

Great thanks! Accepted this as the answer for showing the combined effects of 'relpos' and 'va'. Thanks again, my OCD is now slightly alleviated :)

tmdavison Over a year ago

why do you have horizontalalignment = "bottom", in the first annotate? I presume that's a typo?

tmdavison Over a year ago

OK, but that's not actually doing anything. Try changing that to top and you'll see it makes no difference. horizontalalignment can take [ 'center' | 'right' | 'left' ] as its value, not 'bottom'. Are you sure you didn't mean verticalalignment there?

ImportanceOfBeingErnest Over a year ago

Oh, now I see what you mean. Sorry for that. I updated the answer.

tmdavison · Accepted Answer · 2018-01-11 17:33:54Z

It's due to the vertical alignment of the text (in this case 'Head Motion Cutoff'). The arrow is aligned to the center of the text, but by default, the baseline of the text will be set to the coordinates you give in xytext.

The simplest fix to this is to set the verticalalignment (or va) of the text to 'center', so everything lines up nicely.

For example:

import matplotlib.pyplot as plt
fig, ax = plt.subplots(1)
ax.grid(color='k', alpha=0.5, ls=':')

plt.annotate('Head Motion Cutoff (va=bottom)',
             xy=[0.1, 0.2],
             xytext=[0.4, 0.2],
             fontsize=14,
             arrowprops=dict(fc='#A9A9A9',
                             ec='#A9A9A9',
                             arrowstyle='simple',
                             shrinkA=4,
                             shrinkB=4),
             va='bottom')

plt.annotate('Head Motion Cutoff (va=baseline)',
             xy=[0.1, 0.4],
             xytext=[0.4, 0.4],
             fontsize=14,
             arrowprops=dict(fc='#A9A9A9',
                             ec='#A9A9A9',
                             arrowstyle='simple',
                             shrinkA=4,
                             shrinkB=4),
             va='baseline')

plt.annotate('Head Motion Cutoff (va=top)',
             xy=[0.1, 0.6],
             xytext=[0.4, 0.6],
             fontsize=14,
             arrowprops=dict(fc='#A9A9A9',
                             ec='#A9A9A9',
                             arrowstyle='simple',
                             shrinkA=4,
                             shrinkB=4),
             va='top')

plt.annotate('Head Motion Cutoff (va=center)',
             xy=[0.1, 0.8],
             xytext=[0.4, 0.8],
             fontsize=14,
             arrowprops=dict(fc='#A9A9A9',
                             ec='#A9A9A9',
                             arrowstyle='simple',
                             shrinkA=4,
                             shrinkB=4),
             va='center')

plt.show()

excellent, thanks for pointing me to the 'va' option. Very useful! I would like to have accepted your answer too, but I can only accept one :(. Accepted the other one for bringing 'relpos' in the mix too and showing that I could have chosen another number than 0, 0.5 or 1 (go figure :D)

Collectives™ on Stack Overflow

Control angle of arrow in matplotlib.pyplot.annotate

2 Answers 2

4 Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

4 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related