我正在尝试使用 ctypes 从 Python 调用 Fortran 函数。我试图从子例程和函数(都具有相同的功能)中获取结果,但是我无法从函数中获得预期的输出,而子例程运行良好。问题是我有很多带有 Fortran 函数而不是子例程的库。Fortran 函数和 ctypes 有什么问题吗?
一段 Fortran 代码:
MODULE Vector
! Public types
TYPE VectorType
PRIVATE
DOUBLE PRECISION, DIMENSION(3):: components = 0.0d0
END TYPE VectorType
!---------------------------------------------------------------------
CONTAINS
!---------------------------------------------------------------------
SUBROUTINE newVect(this,vectorIn)
TYPE (VectorType), INTENT(OUT):: this
DOUBLE PRECISION, DIMENSION(3), INTENT(IN)::vectorIn
this%components = (/vectorIn(1), vectorIn(2), vectorIn(3)/)
END SUBROUTINE newVect
!---------------------------------------------------------------------
SUBROUTINE subVect(this,vectorOut)
TYPE(VectorType), INTENT (OUT):: vectorOut
TYPE(VectorType), INTENT (IN) :: this
vectorOut%components = this%components
END SUBROUTINE subVect
!----------------------------------------------------------------------
TYPE(VectorType) FUNCTION getVect(this) RESULT(vectorOut)
TYPE(VectorType), INTENT (IN) :: this
vectorOut%components = this%components
END FUNCTION getVect
!--------------------------------------------------------------------
END MODULE Vector
我正在使用的 Python 代码是:
import ctypes
import numpy as np
class _VectorType(ctypes.Structure):
_fields_ = [('components', ctypes.c_double*3)]
lib_gen_ctypes = '/local/scratch/jfreixa/lib/lib_ctypes_vector.so'
try_ctypes = ctypes.CDLL(lib_gen_ctypes,ctypes.RTLD_GLOBAL)
class vector(object):
_ctypes_newVect = try_ctypes['Vector.newVect_']
_ctypes_subVect = try_ctypes['Vector._subVect_']
_ctypes_getVect = try_ctypes['Vector.getVect_']
vector_pointer = ctypes.POINTER(_VectorType)
_ctypes_getVect.argtypes = [vector_pointer,]
_ctypes_getVect.restype = _VectorType
def __init__(self,*args):
self._vector = _VectorType()
self._newVect(*args)
def _newVect(self,vectIn):
pdb.set_trace()
c_vect = (ctypes.c_double*3)(*vectIn)
self._ctypes_newVect(self._vector,c_vect)
def subVect(self):
pdb.set_trace()
c_vect = _VectorType()
self._ctypes_subVect(ctypes.byref(self._vector),ctypes.byref(c_vect))
print c_vect.components[:]
return np.array(c_vect.components[:])
def getVect(self):
pdb.set_trace()
c_vect = self._ctypes_getVect(ctypes.byref(self._vector))
vect = self._ctypes_getVect(self.vector_pointer.from_address(ctypes.addressof(c_vect)))
print vect.components[:]
return np.array(vect.components[:])
对于这个功能,我尝试了很多东西,但我从来没有得到正确的结果。要运行该程序,请尝试:
import pyctp.vector
newVect = pyctp.vector.vector((1.0,2.0,3.0))
newVect.subVect()
newVect.getVect()
子例程调用返回预期的向量,而函数调用返回空向量或充满垃圾的向量。