matplotlib label subplots of different sizes the exact same distance from corner of each subplot

Question

I am searching for an elegant solution to placing a figure label (A, B, C) above the corner of each subplot using matplotlib- but have each be the exact same distance from the subplot axes corner. The problem I am having is that typical labeling solutions take advantage of the axes fraction- so it is easy to place A,B,C in the same place relative to each plot.

import matplotlib as mpl
import matplotlib.pyplot as plt
fig, ax = plt.subplots(2,2, figsize = (10,10))

texts = ['A', 'B', 'C', 'D']
axes = fig.get_axes()
for a,l in zip(axes, texts):
    a.annotate(l, xy=(-0.1, 1.1), xycoords="axes fraction", fontsize=15, weight = 'bold')
plt.suptitle('Label_Distance Consistent', fontsize = 20)

however, if the plots are different sizes you will get labels that are variably far from the corners of the plots (due to aspect ratio). See Label A and C for example. I am looking for a good way to ensure proportional distance of labels from axes corners for panels containing multiple sizes/aspect ratio subplots, and/or to explicitly set text a specific distance (in inches or maybe figure coordinate units) from axes corners.

In the past I have placed same sized square axes at the corner of each subplot axes in the panel, made those invisible, and scaled text to these, but it is a convoluted approach.

fig, ax = plt.subplots(2,2, figsize = (10,10))

ax1 = plt.subplot2grid((3, 3), (0, 0), colspan=3)
ax2 = plt.subplot2grid((3, 3), (1, 0), colspan=2)
ax3 = plt.subplot2grid((3, 3), (1, 2), rowspan=2)
ax4 = plt.subplot2grid((3, 3), (2, 0))
ax5 = plt.subplot2grid((3, 3), (2, 1))

axes = fig.get_axes()
texts = ['A', 'B', 'C', 'D', 'E']
for a,l in zip(axes, texts):
    a.annotate(l, xy=(-0.1, 1.1), xycoords="axes fraction", fontsize=15, weight = 'bold')
plt.suptitle('Label Distance Inconsistent', fontsize = 20)

score 3 · Answer 1 · answered Sep 12 '18 at 07:08

3

You could use axes pixels istead of axes fraction as reference and add the individual heights to compensate the reference origin lower left corner:

Here for brevity only the changed line within your loop, rest of your code untouched:

    a.annotate(l, xy=(-40, 10 + a.bbox.height), xycoords="axes pixels", fontsize=15, weight = 'bold')

answered Sep 12 '18 at 07:08

SpghttCd

10,510
2
20
25

Great answer, this almost does what I was hoping. I am finding that this doesn't work consistently when the dpi is not 80, is there a way to account for this? So if I scale up my dpi to 100 for example, labels will not be consistent from axes again- C will be much further (height wise) from the corner of ax3 (bottom right). Is there a way perhaps to set the text a specified distance in inches (x inches, y inches) in relation to axes top corners so that this could be robust to dpi changes/figure dimensions? Thanks so much for the help! – djakubosky Sep 12 '18 at 19:17
On second thought, this doesn't seem to quite work at any DPI I try, it could be matplotlib version specific too because your image looks consistent (I am 2.2.3). It seems that a.bbox.height doesn't return a value that is true to the heights of the plots for me. – djakubosky Sep 12 '18 at 19:34
Just tested it with matplotlib 2.2.3 and it still works with `a.bbox.height` and `axes pixels`. – SpghttCd Sep 13 '18 at 08:58
it definitely makes sense that it would work, but for whatever reason, my lettering still scales with the dimensions of the plots. In fact, even if I plot them with xy= (0,bbox.height) (theoretically top corner) the letters aren't placed in the corners. strange. I have observed other strange results in this ax.annotate method, namely that plotting on xycoords='figure points' or 'figure pixels' doesn't seem to place text where its expected- this is why in my answer below I converted from display back to axes fraction. – djakubosky Sep 13 '18 at 19:49
Confirming this works for others I work with, must be something specific to my software config. – djakubosky Sep 13 '18 at 20:13

score 1 · Answer 2 · answered Sep 13 '18 at 07:42

1

What about simply adding the title of the subplots?

(Again only the line to be replaced within your loop:)

a.set_title(l, size=15, weight='bold', loc='left')

Result:

answered Sep 13 '18 at 07:42

SpghttCd

10,510
2
20
25

I never thought of this! Highly efficient if they don't need to be offset. – djakubosky Sep 13 '18 at 19:33

score 1 · Accepted Answer · answered Sep 13 '18 at 08:29

1

You can get the position of each subplot and add text on figure-level, so that offsets are in fractions of the figure size - which is equal for all subplots:

X = a.get_position().x0
Y = a.get_position().y1    
fig.text(X - .04, Y + .015, l, size=15, weight='bold')

Result:

answered Sep 13 '18 at 08:29

SpghttCd

10,510
2
20
25

accepting this answer as it solves most requirements of my question and will probably serve most peoples needs. – djakubosky Sep 13 '18 at 19:39

score 0 · Answer 4 · answered Sep 13 '18 at 19:27

While both of the above solutions work (and are more simple/meet most peoples needs), I thought I'd post a solution I pieced together allowing you to place text/objects with exact inches/points of offset from a particular axes point. This means that regardless of figure panel size/dpi, the text could be placed the same distance from corners- useful if making multiple figures of different dimensions and want them to be consistent (eg for publication). Apparently the matplotlib.transforms.offset_copy() function was designed for this purpose allowing specification of inches, points (1/72 inch) or dots as offset.

https://matplotlib.org/examples/pylab_examples/transoffset.html

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker

fig, ax = plt.subplots(2,2, figsize = (10,7), dpi = 500)

ax1 = plt.subplot2grid((3, 3), (0, 0), colspan=3)
ax2 = plt.subplot2grid((3, 3), (1, 0), colspan=2)
ax3 = plt.subplot2grid((3, 3), (1, 2), rowspan=2)
ax4 = plt.subplot2grid((3, 3), (2, 0))
ax5 = plt.subplot2grid((3, 3), (2, 1))

axes = fig.get_axes()
texts = ['A', 'B', 'C', 'D', 'E']
for a,l in zip(axes, texts):
    a.set_xlim(0,10)
    inv = a.transData.inverted()

    # specify the number of points or inches or dots to transform by
    # places an axes transform here for the axes on which you want to pick an anchor point on
    offsetter = mpl.transforms.offset_copy(a.transAxes, fig=fig, x= -2, y= 6, units = 'points')

    # offset an anchor point - this will return display coordinates 
    corner_offset = offsetter.transform([0,1])

    # convert display coordinate to axes fraction
    axes_frac = inv.transform(corner_offset)
    a.annotate(l, xy=(axes_frac[0],axes_frac[1]), xycoords="axes fraction", fontsize=15, weight = 'bold', color='blue')

or with figsize/dpi changed

matplotlib label subplots of different sizes the exact same distance from corner of each subplot

4 Answers4