negative_indices=False seems to be ignored in some cases #2539
Description
From enhancements buffer:
negative_indices-- Boolean, defaults to True. If set to False, the buffer will not support the usual wrap-around for negative indices, i.e. all negative indices are out-of-bounds. This can make it more convenient to write code that is optimal when bounds-checking is turned off (i.e. no casting to unsigned ints is necesarry to get optimial access without any if-tests).
It seems like this argument is ignored if using a negative index to access a row of the array:
# bufferstruct.pyx
from cython cimport Py_ssize_t
import numpy as np
from numpy cimport ndarray, float64_t, int64_t
cimport numpy as cnp
cnp.import_array()
def negind_read_one():
cdef ndarray[dtype=float64_t, ndim=2, negative_indices=False] arr
arr = np.arange(6, dtype=np.float64).reshape(2, 3)
return arr[-1]
def negind_read_both():
cdef ndarray[dtype=float64_t, ndim=2, negative_indices=False] arr
arr = np.arange(6, dtype=np.float64).reshape(2, 3)
return arr[-1, -1]
def negind_write():
cdef ndarray[dtype=float64_t, ndim=2, negative_indices=False] arr
arr = np.arange(6, dtype=np.float64).reshape(2, 3)
arr[-1, -1] = np.nan
return arr
def normal_write():
cdef ndarray[dtype=float64_t, ndim=2] arr
arr = np.arange(6, dtype=np.float64).reshape(2, 3)
arr[-1, -1] = np.nan
return arrThat is, shouldn't negind_read_one() raise the same IndexError as the rest?
>>> negind_read_one()
array([3., 4., 5.])
>>> negind_read_both()
Traceback (most recent call last):
File "<ipython-input-4-1dbb30ed3ea9>", line 1, in <module>
negind_read_both()
File "bufferstruct.pyx", line 17, in bufferstruct.negind_read_both
return arr[-1, -1]
IndexError: Out of bounds on buffer access (axis 1)
>>> negind_write()
Traceback (most recent call last):
File "<ipython-input-5-36ef52623779>", line 1, in <module>
negind_write()
File "bufferstruct.pyx", line 22, in bufferstruct.negind_write
arr[-1, -1] = np.nan
IndexError: Out of bounds on buffer access (axis 1)
>>> normal_write()
array([[ 0., 1., 2.],
[ 3., 4., nan]])