Colorbars change plot dimensions

Thanks so much for adding colorbars to Bokeh 0.12.2!

If I take the colorbar example from http://bokeh.pydata.org/en/latest/docs/user_guide/annotations.html and comment out the colorbar, I get a nice square plot:

But if I enable the colorbar, it’s all squished:

To me, this is very surprising behavior – why would I want my plot to get squished when I add a colorbar? A colorbar is something outside my plot. It seems to be putting the entire plot into a box and then adjusting the internal plot size, which makes sense for making it easier to lay out the rest of the page, but it’s definitely not what I would have wanted. And I don’t see any obvious way to restore my plot to its rightful dimensions, which would appear to require finding out how big the colorbar is somehow.

The code is below; my only change was to specify the size as 200x200, but the default size has the same issue.

import numpy as np
from matplotlib.mlab import bivariate_normal

from bokeh.plotting import figure, output_file, show
from bokeh.models import LogColorMapper, LogTicker, ColorBar

output_file('color_bar.html')

N = 100
X, Y = np.mgrid[-3:3:complex(0, N), -2:2:complex(0, N)]
Z1 = bivariate_normal(X, Y, 0.1, 0.2, 1.0, 1.0) +  \
    0.1 * bivariate_normal(X, Y, 1.0, 1.0, 0.0, 0.0)
image = Z1 * 1e6

color_mapper = LogColorMapper(palette="Viridis256", low=1, high=1e7)

plot = figure(x_range=(0,1), y_range=(0,1), height=200, width=200, toolbar_location=None)
plot.image(image=[image], color_mapper=color_mapper,
           dh=[1.0], dw=[1.0], x=[0], y=[0])

color_bar = ColorBar(color_mapper=color_mapper, ticker=LogTicker(),
                     label_standoff=12, border_line_color=None, location=(0,0))

plot.add_layout(color_bar, 'right')

show(plot)

[cchwala]

Not a real solution, but at least a workaround that is a bit more pleasant to use than the one from @jbednar because only the width of the dummy figure has to be set to a meaningful value. The “trick” is to specify the figure height (and width) and use this height for all figures. Unfortunately one cannot get the height of an existing figure via figure.height and use it for the dummy figure. That would be easiest. Alternatively one could get the width of the bounding box of the colorbar, but I have not found out how to get that…

Here is my “solution” (a modification of the above on from jbednar, incorporating ideas from this SO answer

import numpy as np
from bokeh.plotting import figure, output_file, show, output_notebook
from bokeh.models import LogColorMapper, LogTicker, ColorBar
from bokeh.palettes import viridis
from bokeh.layouts import row, column

output_notebook()

fig_height = 400
fig_width = 300

Z = np.random.rand(100, 100)

color_mapper = LogColorMapper(palette=viridis(256), 
                              low=0, 
                              high=10)

plot = figure(x_range=(0,1), 
              y_range=(0,1), 
              height=fig_height, 
              width=fig_width, 
              toolbar_location=None)
plot.image(image=[Z], 
           color_mapper=color_mapper,
           dh=[1.0], 
           dw=[1.0], 
           x=[0], 
           y=[0])

color_bar = ColorBar(color_mapper=color_mapper, 
                     border_line_color=None, 
                     location=(0,0))

dummy = figure(height=fig_height, 
               width=120, 
               toolbar_location=None, 
               min_border=0, 
               outline_line_color=None,
               title='colorbar label',
               title_location='right')
dummy.add_layout(color_bar, 'right')
dummy.title.align = 'center'
dummy.title.text_font_size = '10pt'

r = row(plot,dummy)

show(r)gful values

[mattpap]

By default nothing will change, however, PR #8085 adds support for configuring frame dimensions (instead of plot dimensions), so with some configuration this issue will be solvable.