BUG: boxplot does not set limits of axis

[aseyboldt]

DataFrame.boxplot fails to properly set the limits of the y-axis is this example:

df = pd.DataFrame({'X': [1, 0, 1, 0],
                   'Y': [1, 0, 1, 0],
                   'Z': [1, 0, 1, 0],
                   #'M': [1, 0, 1, 0],
                   'B': list("XXYY")}, 
                  index=list("UUVV"))

print df
"""
   B  X  Y  Z
U  X  1  1  1
U  X  0  0  0
V  Y  1  1  1
V  Y  0  0  0
"""

df.boxplot(by='B')

I would expect the y-axis to go from 0 to 1, not -0.06 to 0.06.
If I uncomment the M it works.

I use pandas 0.12.0 and matplotlib 1.3.1

[aseyboldt]

Attaching the image did not work as I expected. The image is here:
http://nbviewer.ipython.org/7475835

[jorisvandenbossche]

Note that it also works when you comment both M and Z eg (so you only have two columns). So it seems it has something to do with the fourth empty ax.

EDIT: It's not really a pandas bug (but could maybe solved in pandas), but it also happens if you do it yourself manually in matplotlib. Due to the shared axes, there goes something wrong with the limits when there is an empty subplot.

[jorisvandenbossche]

It seems this should be fixed in matplotlib master: matplotlib/matplotlib#2356 and matplotlib/matplotlib#2357

[aseyboldt]

I can confirm that this is fixed in matplotlib master.

[jreback]

is this still a bug? or just in matplotlib? anything to do?

[jorisvandenbossche]

It's a bug in matplotlib, which it is fixed in matplotlib master, and not a real pandas bug. But it would be possible to workaround it in pandas (as it produces a wrong boxplot by only using pandas code). But I do not know how easy and/or necessary this would be.

[jreback]

ok....prob by the time 0.14 comes out, will just close this