Ubuntu 18.04 上的 JModelica

Question

你好 JModelica 社区。我已经设法在 CentOS 上编译 JModelica，但在 Ubuntu 18.04 上仍然失败。编译本身是成功的，但是正在运行

from pyjmi.examples import cstr_casadi
cstr_casadi.run_demo()

失败

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-4-33de447ae4ee> in <module>()
----> 1 cstr_casadi.run_demo()

/opt/JModelica.org/Python/pyjmi/examples/cstr_casadi.pyc in run_demo(with_plots)
    179 
    180     # Solve the optimal control problem
--> 181     res = op.optimize(options=opt_opts)
    182 
    183     # Extract variable profiles

/opt/JModelica.org/Python/pyjmi/casadi_interface.pyc in optimize(self, algorithm, options)
    446                              "algorithm.")
    447         return self._exec_algorithm('pyjmi.jmi_algorithm_drivers',
--> 448                                     algorithm, options)
    449 
    450     # Make solve synonymous with optimize

/opt/JModelica.org/Python/pyjmi/common/core.pyc in _exec_algorithm(self, module, algorithm, options)
    166         alg = algorithm(self, options)
    167         # solve optimization problem/initialize
--> 168         alg.solve()
    169         # get and return result
    170         return alg.get_result()

/opt/JModelica.org/Python/pyjmi/jmi_algorithm_drivers.pyc in solve(self)
    351         Solve the optimization problem using ipopt solver.
    352         """
--> 353         self.nlp.solve_and_write_result()
    354 
    355     def get_result(self):

/opt/JModelica.org/Python/pyjmi/optimization/casadi_collocation.pyc in solve_and_write_result(self)
    995         t0 = time.clock()
    996         # todo: account for preprocessing time within solve_nlp separately?
--> 997         self.times['sol'] = self.solve_nlp()
    998         self.result_file_name = self.export_result_dymola(self.result_file_name)
    999         self.times['post_processing'] = time.clock() - t0 - self.times['sol'] - self.extra_update

/opt/JModelica.org/Python/pyjmi/optimization/casadi_collocation.pyc in solve_nlp(self)
    610 
    611         # Get the result
--> 612         primal_opt = N.array(self.solver_object.output(casadi.NLP_SOLVER_X))
    613         self.primal_opt = primal_opt.reshape(-1)
    614         if self.order != "default":

/opt/JModelica.org/Python/casadi/casadi_core.pyc in <lambda>(self, name)
  30560     for _s in [Function]:
  30561         __swig_getmethods__.update(getattr(_s, '__swig_getmethods__', {}))
> 30562     __getattr__ = lambda self, name: _swig_getattr(self, NlpSolver, name)
  30563     __repr__ = _swig_repr
  30564 

/opt/JModelica.org/Python/casadi/casadi_core.pyc in _swig_getattr(self, class_type, name)
     78     if method:
     79         return method(self)
---> 80     raise AttributeError("'%s' object has no attribute '%s'" % (class_type.__name__, name))
     81 
     82 

AttributeError: 'NlpSolver' object has no attribute 'output'

我尝试降级几个软件包，但问题仍然存在。顺便说一句，我在 Arch Linux 上遇到了同样的错误。

我知道 Ubuntu 18.04 不受官方支持，但我希望有人已经找到解决此问题的方法。

编辑： 来自make install 和make casadi_interface 的日志：

https://pastebin.com/ADRyE7XV

https://pastebin.com/dL4SCWdb

Answer 1

我正在回答我自己的问题以将其标记为已解决。解决方案来自@Vital（参见原始问题的 cmets）。

为了让 JModelica 在 Ubuntu 18.04 或 Arch Linux 上运行，solver_object.output 需要在所有受影响的 Python 文件中替换为 solver_object.getOutput。

Answer 2

您似乎加载了错误的 casadi 版本。 你能检查一下输出吗

from casadi import __version__
print(__version__)

编辑

可能问题与 swig 版本有关。

将solver_object.output 替换为solver_object.getOutput 应该会有所帮助！ 例如

primal_opt = N.array(self.solver_object.output(casadi.NLP_SOLVER_X)) 

与

primal_opt = N.array(self.solver_object.getOutput(casadi.NLP_SOLVER_X))

也许是 changset：https://trac.jmodelica.org/changeset/8074 有助于概览文件和位置。

Answer 3

我根据 Stackoverflow 上的贡献和原始 JModelica 用户指南创建了一个 Linux 安装脚本。此安装脚本托管在 https://gitlab.com/christiankral/install_jmodelica/。