Matplotlib force figure size of output image

Question

I'm creating a matplotlib figure. And I want the size (or at least the aspect ratio) of the output image to be consistent. It seems like using figsize changes the contained figure but not the image.

fig, ax = plt.subplots(figsize=(9,9))
ax.barh([0,1,2],[1,5,3])
ax.set_yticks([0,1,2])
ax.set_yticklabels(['a','b','c'])

Creates the following:

While a change in the ticklabels enlarges the image

fig, ax = plt.subplots(figsize=(9,9))
ax.barh([0,1,2],[1,5,3])
ax.set_yticks([0,1,2])
ax.set_yticklabels(['a really really really really really really long name','b','c'])

I wish for the opposite - keep the output the same size and have the chart change as needed.

Can you try adding the following: `fig.canvas.layout.width = '9in'` and `fig.canvas.layout.height= '9in'` and see if it works — Sheldore, Jun 11 '20 at 12:25
`AttributeError: 'FigureCanvasAgg' object has no attribute 'layout'` — DeanLa, Jun 11 '20 at 12:32
I remove `layout` using only `fig.canvas.width = '9in' fig.canvas.height = '9in'` but it doesn't work — DeanLa, Jun 11 '20 at 12:33

score 1 · Answer 1 · answered Jun 11 '20 at 13:03

I was able to devise the following solution. It does the trick somewhat although I do not claim it to be the best solution. The idea is the following:

Get the width of the y-ticklabels in inches (I used this method)
Extract the maximum width i.e. for the longest tick-label
Reset the figure size at the end using width = 9-max(label width)

# Rest of your code

fig.canvas.draw()

# Get the labels' widths in inches
label_widths = [item.get_window_extent().transformed(fig.dpi_scale_trans.inverted()).width 
                for item in ax.get_yticklabels()]

Since the original size was 9 inches by 9 inches, rescale the size
fig.set_size_inches((9-max(label_widths)), 9)

Matplotlib force figure size of output image

1 Answers1