def test_buffers(self):
b = b"xx"
ty = typeof(b)
self.assertEqual(ty, types.Bytes(types.uint8, 1, "C"))
self.assertFalse(ty.mutable)
ty = typeof(memoryview(b))
self.assertEqual(ty,
types.MemoryView(types.uint8, 1, "C", readonly=True))
self.assertFalse(ty.mutable)
ty = typeof(array.array('i', [0, 1, 2]))
self.assertEqual(ty, types.PyArray(types.int32, 1, "C"))
self.assertTrue(ty.mutable)
b = bytearray(10)
ty = typeof(b)
self.assertEqual(ty, types.ByteArray(types.uint8, 1, "C"))
self.assertTrue(ty.mutable)
from llvmlite import ir
from numba import types, cgutils, unicode
from numba.extending import (overload, intrinsic, overload_method, lower_cast,
register_jitable)
# bytes and str arrays items are of type CharSeq and UnicodeCharSeq,
# respectively. See numpy/types/npytypes.py for CharSeq,
# UnicodeCharSeq definitions. The corresponding data models are
# defined in numpy/datamodel/models.py. Boxing/unboxing of item types
# are defined in numpy/targets/boxing.py, see box_unicodecharseq,
# unbox_unicodecharseq, box_charseq, unbox_charseq.
s1_dtype = np.dtype('S1')
assert s1_dtype.itemsize == 1
bytes_type = types.Bytes(types.uint8, 1, "C", readonly=True)
# Currently, NumPy supports only UTF-32 arrays but this may change in
# future and the approach used here for supporting str arrays may need
# a revision depending on how NumPy will support UTF-8 and UTF-16
# arrays.
u1_dtype = np.dtype('U1')
unicode_byte_width = u1_dtype.itemsize
unicode_uint = {1: np.uint8, 2: np.uint16, 4: np.uint32}[unicode_byte_width]
unicode_kind = {
1: unicode.PY_UNICODE_1BYTE_KIND,
2: unicode.PY_UNICODE_2BYTE_KIND,
4: unicode.PY_UNICODE_4BYTE_KIND
}[unicode_byte_width]
