I am working with scikitlearn and would like to override the build method of the treebuilder class for regression trees implemented using cython. In order to do so I figured I would need access to the cython code, so I have added scikitlearn as a git submodule.
My project structure thus looks as follows:
.
|-- setup.py
|-- MyNewTree
| |-- __init__.py
| |-- MyNewTree.pyx
| `-- scikitlearn
| `-- sklearn
| `-- tree
| |-- __init__.py
| |-- _tree.pxd
| |-- _tree.pyx
| |-- setup.py
| `-- tree.py
In my setup.py I am doing the following:
from setuptools import setup, find_packages
from setuptools.extension import Extension
from Cython.Build import cythonize
import numpy
extensions = [
Extension(
"newtree.MyNewTree",
["newtree/MyNewTree.pyx"],
include_dirs=['modulenetwork/scikitlearn/sklearn/tree', numpy.get_include()]
)
]
setup(
name = 'MyNewTree',
version = '0.0.1',
packages = find_packages(),
ext_modules = cythonize(extensions)
)
Finally MyNewTree.pyx
# cython: cdivision=True
# cython: boundscheck=False
# cython: wraparound=False
import numpy as np
cimport numpy as np
np.import_array()
from .scikitlearn.sklearn.tree._tree cimport BestFirstTreeBuilder
cdef class TreeBuilder(BestFirstTreeBuilder):
cpdef build(self):
print('This is an overridden build method!')
What I would like this to produce is a TreeBuilder class that has a different build method from the original scikitlearn implementation, but has everything else the same.
To compile I run python setup.py build_ext --inplace
However I get the following error:
Error compiling Cython file:
------------------------------------------------------------
...
import numpy as np
cimport numpy as np
np.import_array()
from .scikitlearn.sklearn.tree._tree cimport BestFirstTreeBuilder
^
------------------------------------------------------------
newtree/MyNewTree.pyx:9:0: 'newtree/scikitlearn/sklearn/tree/_tree.pxd' not found
Error compiling Cython file:
------------------------------------------------------------
...
import numpy as np
cimport numpy as np
np.import_array()
from .scikitlearn.sklearn.tree._tree cimport BestFirstTreeBuilder
^
------------------------------------------------------------
newtree/MyNewTree.pyx:9:0: 'newtree/scikitlearn/sklearn/tree/_tree/BestFirstTreeBuilder.pxd' not found
Error compiling Cython file:
------------------------------------------------------------
...
cimport numpy as np
np.import_array()
from .scikitlearn.sklearn.tree._tree cimport BestFirstTreeBuilder
cdef class TreeBuilder(BestFirstTreeBuilder):
^
------------------------------------------------------------
newtree/MyNewTree.pyx:11:5: 'BestFirstTreeBuilder' is not a type name
Traceback (most recent call last):
File "setup.py", line 18, in <module>
ext_modules = cythonize(extensions)
File "/Users/__/miniconda3/lib/python3.6/site-packages/Cython/Build/Dependencies.py", line 1039, in cythonize
cythonize_one(*args)
File "/Users/__/miniconda3/lib/python3.6/site-packages/Cython/Build/Dependencies.py", line 1161, in cythonize_one
raise CompileError(None, pyx_file)
Cython.Compiler.Errors.CompileError: newtree/MyNewTree.pyx
Clearly the files that are reported as not existing actually do exist. Is this a problem with my setup script? How do I properly cimport scikitlearn classes into my code?
