def as_array(obj):
"""Create a numpy array from a ctypes array. The numpy array
shares the memory with the ctypes object."""
tp = type(obj)
try: tp.__array_struct__
except AttributeError: prep_array(tp)
return multi_array(obj, copy=False)
def as_array(obj):
"""Create a numpy array from a ctypes array. The numpy array
shares the memory with the ctypes object."""
tp = type(obj)
try:
tp.__array_struct__
except AttributeError:
prep_array(tp)
return multi_array(obj, copy=False)
def convert(obj,dims=None,order="C",out=None):
"""Converts ctypes array to numpy array and vice versa
convert determines the input type (ctypes or numpy)
internally and returns an object of the opposite type.
NOTE: do NOT do the following:
n1 = numpy_array; c1 = convert(n1); n1 = convert(n1)
^^ ^^
nasty things will happen! it is ok to do
n1 = numpy_array; c1 = convert(n1); n2 = convert(n1)
^^ ^^
A 1D ctypes array ca can be converted to a (m,n,k,...)
numpy array na in C order with convert(ca,(m,n,k,...),"C")
and to na in F order with convert(ca,(m,n,k,...),"F").
(m,n,k,...) is reversed internally and the order of
matrix-matrix or matrix-vector nultiplication must be
inverted, when comparing with C oder results.
This code is based on similar code posted by Thomas Heller"""
# for obj in simple c_types (e.g.c_float(1.))
if isSimpleType(obj):
"""convert simple ctype to numpy array"""
obj = obj.value,
return multi_array(obj,copy=False)
# for obj in scalars(e.g. 1.), lists and tuples
if not (isCtypesFamily(obj) or isNumpyArray(obj)):
"""convert Python scalar, list or tuple to numpy array"""
obj = tuple(obj)
return multi_array(obj,copy=False)
# numpy ==> ctypes
if isNumpyArray(obj):
"""convert numpy array to ctypes array"""
do_copy = False
# if obj is C order and return object should be
# Fortran order, transpose obj for Fortran order
if not isfortran(obj) and order == "F":
obj = obj.T
ai = obj.__array_interface__
if ai["strides"]:
pass
# do something sensible
# obj = obj.T
# ai = obj.__array_interface__
addr,readonly = ai["data"]
if readonly: # make a copy
do_copy = True
## code below should consider strides
i_size = obj.itemsize
if out is None:
# print "SIZE",eval(PRODUCT % "obj.shape")
t = n_Dict[ai["typestr"]]
for dim in ai["shape"][::-1]:
t = t*dim
if do_copy:
out = t()
memmove(out,addr,obj.size*i_size)
else:
out = t.from_address(addr)
out.__array_interface__ = ai
out.__keep = ai
return out
else:
size1 = obj.size
size2 = len(out)
size = min(size1,size2)*i_size
memmove(out,addr,size)
out.__array_interface__ = ai
out.__keep = ai
return out
# ctypes ==> numpy
else:
"""convert ctypes array to numpy array"""
typestr = c_Dict[typeName(obj)]
strides = None
if dims is None:
shape = []
o = obj
while isArrayType(o):
shape.append(o._length_)
o = o._type_
shape = tuple(shape)
else:
shape = tuple(dims)
p = sizeof(eval(typeName(obj)))
products = [p]
for d in dims[:-1]:
p *= d
products.append(p)
if order == "F":
strides = tuple(products)
ao = addressof(obj)
ai = \
{
'descr' : [('',typestr)],
'__ref' : ao,
'strides': strides,
'shape' : shape,
'version': 3,
'typestr': typestr,
'data' : (ao,False)
}
obj.__array_interface__ = ai
return multi_array(obj,copy=False)
def convert(obj, dims=None, order="C", out=None):
"""Converts ctypes array to numpy array and vice versa
convert determines the input type (ctypes or numpy)
internally and returns an object of the opposite type.
NOTE: do NOT do the following:
n1 = numpy_array; c1 = convert(n1); n1 = convert(n1)
^^ ^^
nasty things will happen! it is ok to do
n1 = numpy_array; c1 = convert(n1); n2 = convert(n1)
^^ ^^
A 1D ctypes array ca can be converted to a (m,n,k,...)
numpy array na in C order with convert(ca,(m,n,k,...),"C")
and to na in F order with convert(ca,(m,n,k,...),"F").
(m,n,k,...) is reversed internally and the order of
matrix-matrix or matrix-vector nultiplication must be
inverted, when comparing with C oder results.
This code is based on similar code posted by Thomas Heller"""
# for obj in simple c_types (e.g.c_float(1.))
if isSimpleType(obj):
"""convert simple ctype to numpy array"""
obj = obj.value,
return multi_array(obj, copy=False)
# for obj in scalars(e.g. 1.), lists and tuples
if not (isCtypesFamily(obj) or isNumpyArray(obj)):
"""convert Python scalar, list or tuple to numpy array"""
obj = tuple(obj)
return multi_array(obj, copy=False)
# numpy ==> ctypes
if isNumpyArray(obj):
"""convert numpy array to ctypes array"""
do_copy = False
# if obj is C order and return object should be
# Fortran order, transpose obj for Fortran order
if not isfortran(obj) and order == "F":
obj = obj.T
ai = obj.__array_interface__
if ai["strides"]:
pass
# do something sensible
# obj = obj.T
# ai = obj.__array_interface__
addr, readonly = ai["data"]
if readonly: # make a copy
do_copy = True
## code below should consider strides
i_size = obj.itemsize
if out is None:
# print "SIZE",eval(PRODUCT % "obj.shape")
t = n_Dict[ai["typestr"]]
for dim in ai["shape"][::-1]:
t = t * dim
if do_copy:
out = t()
memmove(out, addr, obj.size * i_size)
else:
out = t.from_address(addr)
out.__array_interface__ = ai
out.__keep = ai
return out
else:
size1 = obj.size
size2 = len(out)
size = min(size1, size2) * i_size
memmove(out, addr, size)
out.__array_interface__ = ai
out.__keep = ai
return out
# ctypes ==> numpy
else:
"""convert ctypes array to numpy array"""
typestr = c_Dict[typeName(obj)]
strides = None
if dims is None:
shape = []
o = obj
while isArrayType(o):
shape.append(o._length_)
o = o._type_
shape = tuple(shape)
else:
shape = tuple(dims)
p = sizeof(eval(typeName(obj)))
products = [p]
for d in dims[:-1]:
p *= d
products.append(p)
if order == "F":
strides = tuple(products)
ao = addressof(obj)
ai = \
{
'descr' : [('',typestr)],
'__ref' : ao,
'strides': strides,
'shape' : shape,
'version': 3,
'typestr': typestr,
'data' : (ao,False)
}
obj.__array_interface__ = ai
return multi_array(obj, copy=False)