python：阻止导入的模块解析命令行参数

Question

我正在使用一个外部 python 模块，它不是由我编写的，因此无法更改。这个名为 magnum (http://micromagnum.informatik.uni-hamburg.de) 的模块处理所有可选的命令行参数。这里有一个例子来说明行为：

script1.py

#!/usr/bin/env python
import magnum

执行脚本产生：

>>> ./script1.py -h
[WARNING] - Python Imaging Library not found!
[WARNING] - -> This means that the ImageShapeCreator and related classes are not available!
[   INFO] - Imported FFTW wisdom from file
Usage: scipt1.py [options]

Options:
  --version             show program's version number and exit
  -h, --help            show this help message and exit

  Hardware options:
    Options that control which hardware is used.

    -g GPU_ID           enable GPU processing (using 32-bit accuracy) on cuda
                    device GPU_ID. The simulator will fall back to CPU
                    mode if it was not compiled with CUDA support or when
                    no CUDA capable graphics cards were detected.
    -G GPU_ID           enable GPU processing (using 64-bit accuracy) on cuda
                    device GPU_ID. TODO: Describe fallback behaviour.
    -t NUM_THREADS, --threads=NUM_THREADS
                    enable CPU multithreading with NUM_THREADS (1..64)
                    threads. This parameter instructs the fftw library to
                    use NUM_THREADS threads for computing FFTs.

  Logging options:
    Options related to logging and benchmarking.

    -l LEVEL, --loglevel=LEVEL
                    set log level (0:Debug, 1:Info, 2:Warn, 4:Error,
                    5:Critical), default is Debug (0).
    --prof              Log profiling info at program exit.

  Parameter sweep options:
    These options have only an effect when the simulation script uses a
    Controller object to sweep through a parameter range.

    -p RANGE, --param-range=RANGE
                    select parameter set to run, e.g. --param-range=0,64
                    to run sets 0 to 63.
    --print-num-params  print number of sweep parameters to stdout and exit.
    --print-all-params  print all sweep parameters to stdout and exit.

  Miscellanous options:
    --on_io_error=MODE  Specifies what to do when an i/o error occurs when
                    writing an .omf/.vtk file. 0: Abort (default), 1:
                    Retry a few times, then abort, 2: Retry a few times,
                    then pause and ask for user intervention

现在我想编写一个小脚本，它接受自己的命令行参数，然后使用 magnum 模块执行一些小计算。我想使用 argparse 来解析参数。但是，argparse 的优先级似乎比这个外部模块的参数处理低，并且我自己的参数无法被识别：

script2.py

#!/usr/bin/env python
import argparse
import magnum
parser = argparse.ArgumentParser(
        description='TEST')
parser.add_argument('--x',
        help='test_arg')
args = parser.parse_args()
print args.x

调用它：

>>>./scrip2.py --x 3
[WARNING] - Python Imaging Library not found!
[WARNING] - -> This means that the ImageShapeCreator and related classes are not available!
[   INFO] - Imported FFTW wisdom from file
Usage: test.py [options]

test.py: error: no such option: --x

我在 magnum 之前或之后导入 argparse 都没有关系。如果我不导入 magnum，则 argparse 有效：

script3.py

#!/usr/bin/env python
import argparse
parser = argparse.ArgumentParser(
        description='TEST')
parser.add_argument('--x',
        help='test_arg')
args = parser.parse_args()
print args.x

执行它会产生：

>>>./scrip2.py --x 3
3

我的问题是：如何在不编辑 magnum 的情况下阻止 magnum 处理我的命令行参数？

Answer 1

虽然我不认为有任何“好的”解决方案，但您可以通过猴子补丁 argparse 使其变为 nop：

class EmptyParser():
    def parse_args():
        return
    ... (more redirects for add_argument)
argparse.ArgumentParser = EmptyParser
import magnum

或者，您可以将导入 magnum 包装在一个函数中，并通过该函数创建一个接口，并在 magnum 解析它们之前创建参数。

def magnum(param1, param2):
    sys.argv = [param1, '--x', param2]
    import magnum

不过，这两种方法都非常 hacky。

Answer 2

为了将来参考，这里是基于上述 cmets / 答案的工作解决方案：

script3.py

#!/usr/bin/env python
import argparse
import sys
parser = argparse.ArgumentParser(
    description='TEST')
parser.add_argument('--x',
    help='test_arg')
args = parser.parse_args()
sys.argv = [sys.argv[0]]
import magnum
print args.x

执行脚本表明 magnum 被正确导入，而且参数已被解析并存储在 args 变量中：

>>>./script3.py --x 3
[WARNING] - Python Imaging Library not found!
[WARNING] - -> This means that the ImageShapeCreator and related classes are not available!
[   INFO] - Imported FFTW wisdom from file
[   INFO] - ----------------------------------------------------------------------
[   INFO] - MicroMagnum 0.2rc3
[   INFO] - Copyright (C) 2012 by the MicroMagnum team.
[   INFO] - This program comes with ABSOLUTELY NO WARRANTY.
[   INFO] - This is free software, and you are welcome to redistribute it under
[   INFO] - certain conditions; see the file COPYING in the distribution package.
[   INFO] - ----------------------------------------------------------------------
[   INFO] - FFTW using 1 threads from now on
[   INFO] - CUDA GPU support: no
3

此解决方案不会将任何参数转发给 magnum，这是我想要实现的。

Answer 3

看github源码，https://github.com/MicroMagnum

import magnum

导入一个包 (magnum/__init__.py)。 init 中的关键操作是

config.cfg.initialize(sys.argv)

这定义并创建了一个cfg = MagnumConfig()。那检查环境。它还定义并运行optparse 解析器。

因此，没有明显的方法可以绕过其解析器并仍然设置计算环境。在import magnum 之前进行自己的解析和调整sys.argv 似乎是唯一的解决方案。