[Python] Immutability of bytes is ignored

[asfimport]

Creating a pyarrow.Buffer from Python bytes allows in-place changes of immutable Python strings:

>>> import pyarrow as pa
>>> import numpy as np
>>> a = b'123456'
>>> a[0] = 77 # bytes are immutable, so TypeError is expected
TypeError: 'bytes' object does not support item assignment
>>> b = pa.py_buffer(a)  # but with pyarrow bytes can be changed in-place
>>> arr = np.frombuffer(b, dtype=np.uint8)
>>> arr[0] = 66 # change 'a' in-place, would expect error
>>> a
b'B23456'
>>> hash(a)
-4581532003987476523
>>> arr[0] = 77
>>> a
b'M23456'
>>> hash(a) # hash value stays constant while changing 'a'
-4581532003987476523

Notice that numpy.frombuffer respects immutability of bytes:
``

>>> arr2 = np.frombuffer(a, dtype=np.uint8)
>>> arr2
array([77, 50, 51, 52, 53, 54], dtype=uint8)
>>> arr2[0] = 88 # expecting error
ValueError: assignment destination is read-only

Reporter: Pearu Peterson / @pearu
Assignee: Wes McKinney / @wesm

PRs and other links:

• GitHub Pull Request #2567

_{Note: This issue was originally created as ARROW-3228. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
Apparently somewhere a mutability flag isn't being respected. Wasn't obvious at a glance where the problem lies, see:

• https://github.com/apache/arrow/blob/master/cpp/src/arrow/python/common.cc#L61
• https://github.com/apache/arrow/blob/master/python/pyarrow/io.pxi#L831

[asfimport]

Wes McKinney / @wesm:
It looks like the flags argument from PyObject_GetBuffer is being ignored. NumPy tries to get a mutable buffer, and if that fails (which it must if the buffer is readonly), then it sets WRITEABLE to false: https://github.com/numpy/numpy/blob/e8d177f74adbe5c7630721a4ab3f20f58839ac99/numpy/core/src/multiarray/ctors.c#L3711. Patch incoming

[asfimport]

Wes McKinney / @wesm:
Issue resolved by pull request 2567
#2567