【已解决】OSError: could not find or load spatialindex

我通过minconda安装spyder5.0后，每次启动都用报错

虽说不影响正常使用，但是每次跳出来提醒你“你有错”还是很不爽的，于是我想了办法搞定了它，接下来我先介绍方法，在讲明原理。

1. 在python安装路径里搜索spatialindex_c-64.dll（为什么后面会讲到）
   
2. 然后点击文件所在位置，你会发现有两个dll文件应该是相辅相成的，一起复制（不要剪切）。
   
3. 接下来找到python第三方库安装路径，我的是D:ProgramDataMiniconda3Libsite-packages，如果你是原生python你可以在python安装路径里搜索site-packages文件夹，在其中找到rtree的文件夹，我的如下图所示：
   
   
4. 在rtree文件夹下粘贴ctrl + v，然后就可以随便使用spyder了。

通过解析rtree源码可以得到上面的解决办法。
首先引用rtree，得到如下报错：

>>> import rtree
Traceback (most recent call last):
  File "", line 1, in 
  File "D:ProgramDataMiniconda3libsite-packagesrtree__init__.py", line 9, in 
    from .index import Rtree, Index  # noqa
  File "D:ProgramDataMiniconda3libsite-packagesrtreeindex.py", line 6, in 
    from . import core
  File "D:ProgramDataMiniconda3libsite-packagesrtreecore.py", line 75, in 
    rt = finder.load()
  File "D:ProgramDataMiniconda3libsite-packagesrtreefinder.py", line 67, in load
    raise OSError("could not find or load {}".format(lib_name))
OSError: could not find or load spatialindex_c-64.dll

错误在倒数的那个报错代码上，生成了OSError，因为找不到spatialindex_c-64.dll，那么让python能找到就可以了，就是说要搞清楚

  File "D:ProgramDataMiniconda3libsite-packagesrtreefinder.py", line 67, in load
    raise OSError("could not find or load {}".format(lib_name))

finder.py的运行逻辑，注意到其上层报错

 File "D:ProgramDataMiniconda3libsite-packagesrtreecore.py", line 75, in 
    rt = finder.load()

因此着重看load()函数

"""
finder.py
------------

Locate `libspatialindex` shared library by any means necessary.
"""
import os
import sys
import ctypes
import platform
from ctypes.util import find_library

# the current working directory of this file
_cwd = os.path.abspath(os.path.expanduser(
    os.path.dirname(__file__)))

# generate a bunch of candidate locations where the
# libspatialindex shared library *might* be hanging out
_candidates = [
    os.environ.get('SPATIALINDEX_C_LIBRARY', None),
    os.path.join(_cwd, 'lib'),
    _cwd,
    '']


def load():
    """
    Load the `libspatialindex` shared library.

    Returns
    -----------
    rt : ctypes object
      Loaded shared library
    """
    if os.name == 'nt':
        # check the platform architecture
        if '64' in platform.architecture()[0]:
            arch = '64'
        else:
            arch = '32'
        lib_name = 'spatialindex_c-{}.dll'.format(arch)

        # add search paths for conda installs
        if 'conda' in sys.version:
            _candidates.append(
                os.path.join(sys.prefix, "Library", "bin"))

        # get the current PATH
        oldenv = os.environ.get('PATH', '').strip().rstrip(';')
        # run through our list of candidate locations
        for path in _candidates:  # 从这里开始就是遍历路径，查找dll
            if not path or not os.path.exists(path):
                continue
            # temporarily add the path to the PATH environment variable
            # so Windows can find additional DLL dependencies.
            os.environ['PATH'] = ';'.join([path, oldenv])
            try:
                rt = ctypes.cdll.LoadLibrary(os.path.join(path, lib_name))
                if rt is not None:
                    return rt
            except (WindowsError, OSError):
                pass
            except baseException as E:
                print('rtree.finder unexpected error: {}'.format(str(E)))
            finally:
                os.environ['PATH'] = oldenv
        raise OSError("could not find or load {}".format(lib_name))

    elif os.name == 'posix':

        # posix includes both mac and linux
        # use the extension for the specific platform
        if platform.system() == 'Darwin':
            # macos shared libraries are `.dylib`
            lib_name = "libspatialindex_c.dylib"
        else:
            # linux shared libraries are `.so`
            lib_name = 'libspatialindex_c.so'

        # get the starting working directory
        cwd = os.getcwd()
        for cand in _candidates:
            if cand is None:
                continue
            elif os.path.isdir(cand):
                # if our candidate is a directory use best guess
                path = cand
                target = os.path.join(cand, lib_name)
            elif os.path.isfile(cand):
                # if candidate is just a file use that
                path = os.path.split(cand)[0]
                target = cand
            else:
                continue

            if not os.path.exists(target):
                continue

            try:
                # move to the location we're checking
                os.chdir(path)
                # try loading the target file candidate
                rt = ctypes.cdll.LoadLibrary(target)
                if rt is not None:
                    return rt
            except baseException as E:
                print('rtree.finder ({}) unexpected error: {}'.format(
                    target, str(E)))
            finally:
                os.chdir(cwd)

    try:
        # try loading library using LD path search
        rt = ctypes.cdll.LoadLibrary(
            find_library('spatialindex_c'))
        if rt is not None:
            return rt
    except baseException:
        pass

    raise OSError("Could not load libspatialindex_c library")

注意上述代码中的中文注释，是我加上的，这说明我们需要搞清楚_candidates里有哪些路径，将rtree要找的dll复制过去不就ok了！

# the current working directory of this file
_cwd = os.path.abspath(os.path.expanduser(
    os.path.dirname(__file__)))

# generate a bunch of candidate locations where the
# libspatialindex shared library *might* be hanging out
_candidates = [
    os.environ.get('SPATIALINDEX_C_LIBRARY', None),
    os.path.join(_cwd, 'lib'),
    _cwd,
    '']

_cwd在_candidates列表中，因此这里只找_cwd的位置就行了，观其注释与函数调用，应该是finder.py所在的目录，即rtree的安装目录，对我来说就是D:ProgramDataMiniconda3Libsite-packagesrtree。
这就是我解决办法的由来，只要将要找的dll文件复制到安装目录就能解决问题。