Terminating integration based on maxiters results in RuntimeError

[laurent90git]

Hi,

I have issues using the library if I specifiy a maximmum number of step (maxiters) and this number of steps is actually reached during the integration.

Here is an example:

from julia.api import Julia
jl = Julia(compiled_modules=False)
from diffeqpy import de

def f(u,p,t):
    return -2*u

u0 = 0.5
tspan = (0., 1.)
tol = 1e-10

prob = de.ODEProblem(f, u0, tspan)
sol = de.solve(prob, alg=de.RadauIIA5(autodiff=False), maxiters=1000, abstol=tol/10, reltol=tol) # works
sol = de.solve(prob, alg=de.RadauIIA5(autodiff=False), maxiters=10, abstol=tol/10, reltol=tol) # fails

import matplotlib.pyplot as plt
plt.plot(sol.t,sol.u)
plt.show()

Here is the obtained error message:

Traceback (most recent call last):

  File "[...]/test_diffeqpy.py", line 24, in <module>
    sol = de.solve(prob, alg=de.RadauIIA5(autodiff=False), maxiters=10, abstol=tol/10, reltol=tol)

RuntimeError: <PyCall.jlwrap (in a Julia function called from Python)
JULIA: IOError: write: broken pipe (EPIPE)
Stacktrace:
 [1] uv_write(s::Base.PipeEndpoint, p::Ptr{UInt8}, n::UInt64)
   @ Base ./stream.jl:1064
 [2] unsafe_write(s::Base.PipeEndpoint, p::Ptr{UInt8}, n::UInt64)
   @ Base ./stream.jl:1118>

Do you have any idea on how to circumvent this issue or resolve it ?

I am using the Python package "julia" version 0.6.1 with Python 3.8.3 and IPython 7.16.1. Julia itself is version 1.8.2.

EDIT: it is possible that this error simply comes from Julia trying to print a warning (like "too many steps"). Trying the following also fails:

from julia import Main
Main.eval('println("Hello World")')

File "[...]/anaconda3/lib/python3.8/site-packages/julia-0.6.1-py3.8.egg/julia/core.py", line 609, in check_exception
    raise JuliaError(u'Exception \'{}\' occurred while calling julia code:\n{}'

JuliaError: Exception 'IOError: write: broken pipe (EPIPE)' occurred while calling julia code:
println("Hello World")

Thanks in advance for you help !

[laurent90git]

Turns out using verbose=False solves the issue. I guess the problem is Julia trying to print something through the Python console and somehow not being able to.

[ChrisRackauckas]

Yeah this looks like an upstream pyjulia issue.

[ChrisRackauckas]

Fixed by transition to PythonCall