I recently postedthis question on SO, about what I believe is an incorrect behavior of setting{x/y}scale('symlog') in combination with alinthres{x/y}. For example:

importnumpyasnpimportmatplotlib.pylabasplz=np.linspace(0,1,100)x=np.ones_like(z)x[0]=0pl.figure()pl.plot(x,z,'-')pl.yscale('symlog',linthresy=1e-2)pl.xlim(-0.1,1.1)

Produces the image shown below. I would expect a linear region from 0 to 1e-2, and a log region containing 1e-2, 1e-1, 1. Why isn't this happening? Is this a bug?