[issue10181] Problems with Py_buffer management in memoryobject.c (and elsewhere?)

[Stefan Krah]

Stefan Krah <stefan-usenet at bytereef.org> added the comment:

Nick Coghlan <report at bugs.python.org> wrote:
> Radical suggestion: make it public as collections.simple_ndarray?

Heh, that's a nice one. :)


While I think that the code in _testbuffer.c is sound and very well tested,
the API is low-level and optimized for testing all sorts of quirks.

Examples: _testbuffer.ndarray has the questionable capability of changing
views while buffers are exported. This is to test the pathological case
that came up in the discussion.

Then, similar to NumPy's ndarray, _testbuffer.ndarray constructs arrays
from a flat list. This is a bit clumsy but quite suitable for testing.

NumPy's ndarray is also the low level API. I think most people use
the high level array:

>>> a = array([[1,2], [3,4]], dtype="L")
>>> a
array([[1, 2],
       [3, 4]], dtype=uint64)

So it would take some effort to polish the API.


Meanwhile, to eliminate the use of _testbuffer in the documentation, I think
the following might work:

I've just added complete support for displaying multi-dimensional lists and
multi-dimensional comparisons to memoryobject.c in my private repo.

The code is quite succinct and follows exactly the same pattern as
copy_base/copy_rec.


Then, in the documentation we can use cast() and memoryview.tolist(),
but add a comment that cast() is just used for demonstration purposes.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue10181>
_______________________________________