英文:
The value in cython array is changed randomly
问题
以下是您提供的代码的中文翻译：
这是我的Cython代码。
self.point_index_array 属性是一个数组，其值从0开始递增。
> [0, 1, 2, 3, 4, 5, .....]
当首次声明时，这些值是正常的，但当从另一个函数中访问该数组时，前四个元素的值会自动更改。
Python 和 Cython 版本
- Python 3.9.13
- Cython 0.29.34
这是我的伪代码
cdef class KdTree:
cdef uint32_t* point_index_array
def __cinit__(KdTree self):
self.point_index_array = NULL
def __init__(KdTree self,
np.ndarray points not None,
int num_of_leaf=16):
cdef np.ndarray[uint32_t, ndim=1] point_index_list
point_index = np.arange(0, self.num_of_points, dtype=np.uint32)
point_index_list = np.ascontiguousarray(point_index.ravel(), dtype=np.uint32)
self.point_index_array = <uint32_t*>point_index_list.data
# kdtree...
# 这是正常的
self.print_index_array()
def print_index_array(self):
for index in range(10):
print(f"index {index} => {self.point_index_array[index]}")
def neighbor_search(self, float distance):
# self.point_index_array 中的值会改变
self.print_index_array()
正常输出（__init__）
> index 0 => 390
index 1 => 469
index 2 => 493
index 3 => 492
index 4 => 396
index 5 => 490
index 6 => 397
index 7 => 117
index 8 => 484
index 9 => 9
异常输出（neighbor_search）
> index 0 => 1171518192
index 1 => 504
index 2 => 1174758304
index 3 => 504
index 4 => 396
index 5 => 490
index 6 => 397
index 7 => 117
index 8 => 484
index 9 => 9
我不明白为什么会发生这种情况。您能告诉我我漏掉了什么吗？
英文:
Here is my cython code.
self.point_index_array attribute is array whose values increase by 1 from 0.
> [0, 1, 2, 3, 4, 5, .....]
The values are normal when first declared, but when the array is accessed from another function, the values of the first four elements are automatically changed.
python and cython version
- Python 3.9.13
- Cython 0.29.34
Here is my sudo code
cdef class KdTree:
cdef uint32_t* point_index_array
def __cinit__(KdTree self):
self.point_index_array = NULL
def __init__(KdTree self,
np.ndarray points not None,
int num_of_leaf=16):
cdef np.ndarray[uint32_t, ndim=1] point_index_list
point_index = np.arange(0, self.num_of_points, dtype=np.uint32)
point_index_list = np.ascontiguousarray(point_index.ravel(), dtype=np.uint32)
self.point_index_array = <uint32_t*>point_index_list.data
# kdtree...
# its normal
self.print_index_array()
def print_index_array(self):
for index in range(10):
print(f"index {index} => {self.point_index_array[index]}")
def neighbor_search(self, float distance):
# values in self.point_index_array are changed
self.print_index_array()
normal output is (__init__)
> index 0 => 390
index 1 => 469
index 2 => 493
index 3 => 492
index 4 => 396
index 5 => 490
index 6 => 397
index 7 => 117
index 8 => 484
index 9 => 9
abnormal output is (neighbor_search)
> index 0 => 1171518192
index 1 => 504
index 2 => 1174758304
index 3 => 504
index 4 => 396
index 5 => 490
index 6 => 397
index 7 => 117
index 8 => 484
index 9 => 9
I don't understand why this happens at all.
Could you let me know what I'm missing?
答案1
得分: 0
Here is the translated content:
@slothrop 感谢您的赏识！
"point_index_list" 变量是前面代码中的 Python 对象，因此被垃圾收集器移除了。
因此，我将此对象保存到类属性中。
cdef class KdTree:
cdef uint32_t* point_index_array
cdef readonly np.ndarray point_index_list
def __init__(KdTree self,
np.ndarray points 不为 None,
int num_of_leaf=16):
point_index = np.arange(0, self.num_of_points, dtype=np.uint32)
# 这段代码将防止 point_index_list 被 CG（垃圾收集）
self.point_index_list = np.ascontiguousarray(point_index.ravel(), dtype=np.uint32)
self.point_index_array = <uint32_t*>self.point_index_list.data
(Note: I've left the code parts as is, as per your request, and only translated the surrounding text.)
英文:
@slothrop thanks for your appreciate!
"point_index_list" variable is python object in previous code, so it was removed by GC.
so, i save this object to class attribute.
cdef class KdTree:
cdef uint32_t* point_index_array
cdef readonly np.ndarray point_index_list
def __init__(KdTree self,
np.ndarray points not None,
int num_of_leaf=16):
point_index = np.arange(0, self.num_of_points, dtype=np.uint32)
# this code will prevent CG of point_index_list
self.point_index_list = np.ascontiguousarray(point_index.ravel(), dtype=np.uint32)
self.point_index_array = <uint32_t*>self.point_index_list.data
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论