matplotlib-users Mailing List for matplotlib

Nathann Cohen, on 2010-12-26 22:27,  wrote:
> Hello everybody !!!
> 
> When adding some text to a plot, is there a way to know the actual
> size of the letters as they appear, in such a way that I could, for
> instance, draw a circle around a 'A', so that the A perfectly fits
> inside ("the smallest circle containing the letter"), regardless of
> the actual size of the picture I'm drawing ?

Hi Nathann,

Here's a quick and dirty way of getting what you want:

  import matplotlib.pyplot as plt
  import numpy as np
  ax = plt.subplot(111)
  t = plt.text(0.5,0.5,'A', ha='center', va='center')
  plt.draw()
  b = t.get_window_extent() # bounding box in pixel coordinates
  r = np.sqrt((b.bounds[-2]/2)**2 + (b.bounds[-1]/2)**2)
  plt.scatter(0.5,0.5, s=np.pi*r**2, marker='o', edgecolor='k', facecolor='w', lw=1,zorder=-1)

I don't think there's a super simple way of doing this by hand -
because text keeps its size constant regardless of how you
manipulate the axes. Here's an example that does what you want,
if you only need the circle in one particular view (i.e. if you
won't be rescaling/zooming in or out the axes:

  ax = plt.subplot(111)
  plt.plot([0,1,1],[0,0,1]) # keep the axes from resizing when we draw our circle
  t = plt.text(0.5,0.5,'A')
  plt.axis('equal')
  plt.draw()
  b = t.get_window_extent() # bounding box in pixel coordinates
  bbox = b.inverse_transformed(ax.transData)
  xc,yc = bbox.get_points().mean(0)
  r = np.sqrt((bbox.bounds[-2]/2)**2 + (bbox.bounds[-1]/2)**2)
  theta = np.linspace(0,2*np.pi,200)
  x,y = r*(np.cos(theta)), r*np.sin(theta)
  l = plt.plot(x+xc, y+yc)

This does exactly what you want, but now, anytime you resize the
axes, the A will stay the same size, but that circle will get
resized.

  ax = plt.subplot(111)
  plt.plot([0,1,1],[0,0,1]) # keep the axes from resizing when we draw our circle
  t = plt.text(0.5,0.5,'A')
  plt.axis('equal')
  plt.draw()
  b = t.get_window_extent() # bounding box in pixel coordinates
  bbox = b.inverse_transformed(ax.transAxes)
  xc,yc = bbox.get_points().mean(0)
  r = np.sqrt((bbox.bounds[-2]/2)**2 + (bbox.bounds[-1]/2)**2)
  theta = np.linspace(0,2*np.pi,200)
  x,y = r*(np.cos(theta)), r*np.sin(theta)
  l = plt.plot(x+xc, y+yc, transform=ax.transAxes)

The above will keep the circle from resizing when you move - but
now it prevents the circle from following 'A' as you pan around.

I see that matplotlib.collections (which is what plt.scatter
creates in the quick-and-dirty example) uses offset and
transOffset to get the job done, but I couldn't figure out a way
to get my last two examples to do something similar by just
manipulating the transforms. 

Hopefully someone chimes in with a better solution. For more on
transformations see:
http://matplotlib.sourceforge.net/users/transforms_tutorial.html

And you can wrap my hand-rolled solution nicely using something
like:
http://matplotlib.sourceforge.net/examples/api/line_with_text.html

best, 
-- 
Paul Ivanov
314 address only used for lists,  off-list direct email at:
http://pirsquared.org | GPG/PGP key id: 0x0F3E28F7 

Philippe Piot, on 2010-12-27 13:58,  wrote:
> 1/  When executing the above script, I also display the histograms
> generated when invoking "hist" (I tried to make this line tranparaent
> by using alpha=0 but it did not work).

Hi Philippe,

welcome to matplotlib - I think what you really want to do is
just use numpy's histogram function, and then plot the result.
This is actually the function matplotlib's hist uses behind the
scenes.

> 2/ can I directly manipulate the data within an histogram to
> arbitrarily offset the histogram?

Once you get the histogram using numpy, you can plot it in in any
way you want. Here's a small example:

  import numpy as np
  import matplotlib.pyplot as plt
  h,edges = np.histogram(np.random.randn(1000))
  plt.bar(edges[:-1],h, width=edges[1]-edges[0])
  plt.bar(edges[:-1]-10,h, width=edges[1]-edges[0]) #offset
  plt.step(edges[1:],h) # plot as step lines instead of bars
  plt.step(edges[1:]+10,h) # same as above, with an offset

best,
-- 
Paul Ivanov
314 address only used for lists,  off-list direct email at:
http://pirsquared.org | GPG/PGP key id: 0x0F3E28F7