Logarithmic scale "raises" small values into range limit instead of cropping them

[eisenlohr]

lin = np.linspace(10,30,21)
log = np.logspace(-10,-1,21)

canvas = toyplot.Canvas(width=400,height=200)
linlog  = canvas.cartesian(grid=(1,2,0),ymin=1e-6,ymax=1e-4,yscale='log')
loglin = canvas.cartesian(grid=(1,2,1),xmin=1e-6,xmax=1e-4,xscale='log')
linlog.scatterplot(lin,log)
loglin.scatterplot(log,lin)
canvas

results in "raised" values at the low end of the logarithmic range:

linlog_raising.pdf

[tshead2]

@eisenlohr, this is definitely wrong, thanks for the heads up.

Cheers,
Tim