Generation of sample points for log-scale plots

Blair Mason

unread,

Apr 16, 2020, 4:21:42 AM4/16/20

to sage-devel

Hello all,

Should generate_plot_points() be modified to log-sample on plots where the x scale is logarithmic? I'd be willing to write the patch up but I'd like to know that it would be desired behavior before I do so.

I've been using sage to graph the frequency responses of a couple of filters. While I was generating the plots, the results sage generated by default were not correct - sage had picked too few sample points. Increasing the number of points by an order of magnitude worked, but it significantly slowed generation of the figures, which was undesirable as I was looking to make them interactive. Restating the function in terms of x <- 2^x and changing the axes appropriately solved the problem.

I think plot() should 'just work' the way one would intuitively expect. However, I don't know - there may be (/probably are) use cases where this would be far from expected behavior.

Code, to illustrate:

def mag2db(x):
    return 20*log(x, 10)

def plot_tf(tf, cutoff, var, domain=(1, 1e6), thresh=1e-2, **kwargs):
    tf = tf.substitute({var:var*i})
    return plot_semilogx(mag2db(abs(tf)), (var, domain[0], domain[1]), base=2, **kwargs)

def plot_tf_log(tf, cutoff, var, domain=(1, 1e6), thresh=1e-2, **kwargs):
    tf = tf.substitute({var:2^(var)*i})
    return plot(mag2db(abs(tf)), (var, log(domain[0], 2), log(domain[1], 2)), **kwargs)
    
@interact
def _(w=slider(0, 20000, step_size=500, default=2000, label='$\\omega_c$'), R=slider(0, 5, default=1, step_size=0.2), C=slider(0, 10, default=1, step_size=0.2)):
    l = L.substitute(omega_c=w, r=R, c=C)
    b = B.substitute(omega_c=w, r=R, c=C)
    h = H.substitute(omega_c=w, r=R, c=C)

    p = plot_tf(l, w, s, color='blue')
    p += plot_tf(b, w, s, color='red')
    p += plot_tf(h, w, s, color='green')
    p.show(ymin=-30, ymax=5)

    p2 = plot_tf_log(l, w, s, color='blue')
    p2 += plot_tf_log(b, w, s, color='red')
    p2 += plot_tf_log(h, w, s, color='green')
    p2.show(ymin=-30, ymax=5)

    print(l) # 4000000/(s^2 + 4000*s + 4000000)
    print(b) # 2000*s/(s^2 + 4000*s + 4000000)
    print(h) # s^2/(s^2 + 4000.00000000000*s + 4000000)

Thanks,

Blair Mason

Eric Gourgoulhon

unread,

Apr 16, 2020, 9:12:24 AM4/16/20

to sage-devel

Hi,

Le jeudi 16 avril 2020 06:21:42 UTC+2, Blair Mason a écrit :

Hello all,

Should generate_plot_points() be modified to log-sample on plots where the x scale is logarithmic? I'd be willing to write the patch up but I'd like to know that it would be desired behavior before I do so.

Yes please! Thanks for your initiative!

The current sampling is definitely annoying for log plots.

Best wishes,

Eric.