Maximum iteration depth exceeded

[ahkab]

Maximum iteration depth exceeded

[itdaniher]

I just checked with most recent ahkab and

numpy-1.7.1
scipy-0.13.0.dev-c31f167
sympy-0.7.3
matplotlib-1.3.0

and didn't get that error. I'll update my F19 system after exams and try and reproduce, but this may be an upstream problem. What version of matplotlib are you using?

It looks like numpy changed around a handful of things from 1.7 to 1.8, as per release notes, could be related.

[liubenyuan]

with
matplotlib-1.3.1-1

I think it relate to the problem of plot r['tran']['T'], the error occurs there.

[ggventurini]

Thanks for the bug report!

Also good to know matplotlib 1.3.0 is unaffected.

I can reproduce this with this setup:

>>> import numpy, scipy, matplotlib
>>> for i in numpy, scipy, matplotlib:
...     print i.__version__
... 
1.7.1
0.12.0
1.3.1

I'll be looking into this and I'll try to come up with a patch shortly.

I'll be looking at the changes in numpy too while I'm at it.. :)

Cheers!

[ggventurini]

It is indeed a bug upstream in matplotlib.

I can reproduce the same RuntimeError: maximum recursion depth exceeded with just:

import pylab as plt
import numpy as np
a = np.mat([1, 2, 3])
plt.plot(a)

This merge is said to fix the issue: matplotlib/matplotlib#2290

It was merged yesterday to the git matplotlib repo, it will take a bit for package mantainers to add the patch to their code bases I'd expect.

A quick work-around is to pass all your data through np.array() before plotting it. It requires modifying your files, sorry about that.

If need be, a temporary work-around targeted to matplotlib 1.3.1 can be introduced in the data interface. If you guys think it is a good idea, please let me know.

Best,

Giuseppe

[ggventurini]

Kinda off topic but numpy 1.8 seems relatively safe to me. I see no worrisome changes. It was a nice read too!

[liubenyuan]

I tried np.array, how ever the data become multi-dimensional and the legend is a mess.

BTW: I am new to python maybe I need to learn more about the coding style to catch up.

[ggventurini]

Here is what I was referring to:

import numpy as np

import ahkab
from ahkab import ahkab, circuit, printing, devices

mycircuit = circuit.circuit(title="Butterworth Example circuit", filename=None)

gnd = mycircuit.get_ground_node()

mycircuit.add_resistor(name="R1", ext_n1="n1", ext_n2="n2", R=600)
mycircuit.add_inductor(name="L1", ext_n1="n2", ext_n2="n3", L=15.24e-3)
mycircuit.add_capacitor(name="C1", ext_n1="n3", ext_n2=gnd, C=119.37e-9)
mycircuit.add_inductor(name="L2", ext_n1="n3", ext_n2="n4", L=61.86e-3)
mycircuit.add_capacitor(name="C2", ext_n1="n4", ext_n2=gnd, C=155.12e-9)
mycircuit.add_resistor(name="R2", ext_n1="n4", ext_n2=gnd, R=1.2e3)

voltage_step = devices.pulse(v1=0, v2=1, td=500e-9, tr=1e-12, pw=1, tf=1e-12, per=2)
mycircuit.add_vsource(name="V1", ext_n1="n1", ext_n2=gnd, vdc=5, vac=1, function=voltage_step)

printing.print_circuit(mycircuit)

op_analysis = ahkab.new_op()
ac_analysis = ahkab.new_ac(start=1e3, stop=1e5, points=100)
tran_analysis = ahkab.new_tran(tstart=0, tstop=1.2e-3, tstep=1e-6, x0=None)

r = ahkab.run(mycircuit, an_list=[op_analysis, ac_analysis, tran_analysis])

import pylab

fig = pylab.figure()
pylab.title(mycircuit.title + " - TRAN Simulation")
T = np.array(r['tran']['T'])
VN1 = np.array(r['tran']['VN1'])
VN4 = np.array(r['tran']['VN4'])
w = np.array(r['ac']['w'])
absVN4 = np.array(r['ac']['|Vn4|'])
argVN4 = np.array(r['ac']['arg(Vn4)'])
pylab.plot(T, VN1, label="Input voltage")
pylab.hold(True)
pylab.plot(T, VN4, label="output voltage")
pylab.legend()
pylab.hold(False)
pylab.grid(True)
pylab.ylim([0,1.2])
pylab.ylabel('Step response')
pylab.xlabel('Time [s]')
fig.savefig('tran_plot.png')

fig = pylab.figure()
pylab.subplot(211)
pylab.semilogx(w, absVN4, 'o-')
pylab.ylabel('abs(V(n4)) [V]')
pylab.title(mycircuit.title + " - AC Simulation")
pylab.subplot(212)
pylab.grid(True)
pylab.semilogx(w, argVN4, 'o-')
pylab.xlabel('Angular frequency [rad/s]')
pylab.ylabel('arg(V(n4)) [rad]')
fig.savefig('ac_plot.png')
pylab.show()

Hopefully that bug will be fixed downstream quickly enough. The above is a temporary solution (slightly more verbose...)

Best, GV

[liubenyuan]

It works! I for got to delete the transpose operator .T

[ggventurini]

This should be fixed in 4c2dbe6, independently from whether your matplotlib version is affected by this regression or not.

[ggventurini]

Also it should be transparent to you, whether you are passing your results through array() or not makes no difference. And no need for the ugly transpose operator .T anymore.