如何将numpy和cythonize-脚本包括在一起？

Question

最近，我开始研究Cython，Anaconda发行版有一个方便的脚本cythonize，允许这样做。

 >cythonize -i foo.pyx

将文件编译到位。

我的问题是如何使用这个脚本来包含numpy，这样我的cimport numpy才能正常工作？

我试过了

>cythonize -X include_path=C:\Anaconda3\lib\site-packages\numpy\core\include  -i foo.pyx
>cythonize -s include_path=C:\Anaconda3\lib\site-packages\numpy\core\include  -i foo.pyx

还有include_dir=...，include_dirs=...，-X和-s

如果我使用-X，我会得到

Traceback (most recent call last):
  File "C:\Anaconda3\Scripts\cythonize-script.py", line 5, in <module>
    sys.exit(Cython.Build.Cythonize.main())
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Cythonize.py", line 185, in main
    options, paths = parse_args(args)
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Cythonize.py", line 172, in parse_args
    options, args = parser.parse_args(args)
  File "C:\Anaconda3\lib\optparse.py", line 1387, in parse_args
    stop = self._process_args(largs, rargs, values)
  File "C:\Anaconda3\lib\optparse.py", line 1431, in _process_args
    self._process_short_opts(rargs, values)
  File "C:\Anaconda3\lib\optparse.py", line 1536, in _process_short_opts
    option.process(opt, value, values, self)
  File "C:\Anaconda3\lib\optparse.py", line 785, in process
    self.action, self.dest, opt, value, values, parser)
  File "C:\Anaconda3\lib\optparse.py", line 805, in take_action
    self.callback(self, opt, value, parser, *args, **kwargs)
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Cythonize.py", line 38, in parse_directives
    value, relaxed_bool=True, current_settings=old_directives)
  File "C:\Anaconda3\lib\site-packages\Cython\Compiler\Options.py", line 424, in parse_directive_list
    raise ValueError('Unknown option: "%s"' % name)
ValueError: Unknown option: "include_path"

用-s我得到

Traceback (most recent call last):
  File "C:\Anaconda3\Scripts\cythonize-script.py", line 5, in <module>
    sys.exit(Cython.Build.Cythonize.main())
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Cythonize.py", line 196, in main
    cython_compile(path, options)
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Cythonize.py", line 90, in cython_compile
    **options.options)
  File "C:\Anaconda3\lib\site-packages\Cython\Build\Dependencies.py", line 809, in cythonize
    ctx = c_options.create_context()
  File "C:\Anaconda3\lib\site-packages\Cython\Compiler\Main.py", line 581, in create_context
    self.cplus, self.language_level, options=self)
  File "C:\Anaconda3\lib\site-packages\Cython\Compiler\Main.py", line 90, in __init__
    self.include_directories = include_directories + [standard_include_path]
TypeError: unsupported operand type(s) for +: 'bool' and 'list'

是的，我可以制作一个setup.py，但是我发现这很烦人，它会把构建文件夹留在那里，而cythonize则不会这样做。

Answer 1

看起来不可能通过命令行版本的cythonize传递包含dirs。

问题是，在cython源代码中，它希望选项是布尔值：

https://github.com/cython/cython/blob/master/Cython/Build/Cythonize.py#L48

因此，当您包含一个路径时，它将被转换为True

cythonize -i -s include_path=path_to_numpy_include test.pyx

这些选项被解析为：

{'directives': {}, 'options': {'include_path': True}, 'python3_mode': None, 'annotate': None, 'excludes': [], 'build': True, 'build_inplace': True, 'parallel': 12, 'force': None, 'quiet': None, 'lenient': None, 'keep_going': None}

这将导致您报告的类型错误：

TypeError: unsupported operand type(s) for +: 'bool' and 'list'

我认为你最好的选择就是使用一个合适的setup.py。

Answer 2

您可以通过设置由C编译器解释的环境变量(它将运行作为子进程来运行)来实现这一点。例如，使用GCC或嘎吱声

C_INCLUDE_PATH=$(python -c 'import numpy; print(numpy.get_include())') cythonize -i foo.pyx