[Numpy-discussion] Is this expected behavior?
Russel Howe
russel@appliedminds....
Mon Jan 18 14:43:23 CST 2010
Oh, of course. I can reverse it myself. Thanks, I did not think of that.
Russel
Warren Weckesser wrote:
> Russel Howe wrote:
>> Since they are iterators, is it possible to check for the second
>> condition and reverse both of them so the behavior I expect happens or
>> does this break something else?
>>
>
> You may already know this, but just in case...
>
> In the second case, you can accomplish the shift by using reversed slices:
>
> a[:, -1:0:-1] = a[:, -2::-1]
>
>
> Warren
>
>> Russel
>> Robert Kern wrote:
>>
>>> On Mon, Jan 18, 2010 at 13:41, Russel Howe <russel@appliedminds.com> wrote:
>>>
>>>> This looks like the difference between memmove and memcpy to me, but I
>>>> am not sure what the expected behavior of numpy should be. The first
>>>> shift behaves the way I expect, the second is surprising.
>>>>
>>> memmove() and memcpy() are not used for these operations (and in
>>> general, they can't be). Rather, iterators are created and looped over
>>> to do the assignments. Because you are not making copies on the
>>> right-hand-side, you are modifying the RHS as the iterators assign to
>>> the LHS.
>>>
>>>
>>>> In [3]: a[:, :-1] = a[:, 1:]
>>>>
>>>> In [4]: a
>>>> Out[4]:
>>>> array([[0, 5, 4, 8, 2, 7, 8, 7, 6, 6],
>>>> [6, 3, 3, 9, 8, 0, 8, 9, 5, 5],
>>>> [0, 1, 1, 2, 5, 8, 2, 5, 3, 3],
>>>> [0, 0, 2, 8, 2, 0, 7, 7, 0, 0],
>>>> [8, 6, 9, 6, 3, 9, 4, 4, 5, 5],
>>>> [7, 6, 9, 3, 8, 9, 9, 6, 9, 9],
>>>> [8, 8, 4, 0, 3, 7, 6, 7, 6, 6],
>>>> [4, 9, 2, 4, 7, 3, 6, 7, 4, 4],
>>>> [2, 0, 7, 0, 7, 6, 6, 1, 6, 6],
>>>> [3, 8, 8, 9, 6, 7, 2, 5, 0, 0]], dtype=uint8)
>>>>
>>> The first one works because the RHS pointer is always one step ahead
>>> of the LHS pointer, thus it always reads pristine data.
>>>
>>>
>>>> In [5]: a[:, 1:] = a[:, :-1]
>>>>
>>>> In [6]: a
>>>> Out[6]:
>>>> array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
>>>> [6, 6, 6, 6, 6, 6, 6, 6, 6, 6],
>>>> [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
>>>> [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
>>>> [8, 8, 8, 8, 8, 8, 8, 8, 8, 8],
>>>> [7, 7, 7, 7, 7, 7, 7, 7, 7, 7],
>>>> [8, 8, 8, 8, 8, 8, 8, 8, 8, 8],
>>>> [4, 4, 4, 4, 4, 4, 4, 4, 4, 4],
>>>> [2, 2, 2, 2, 2, 2, 2, 2, 2, 2],
>>>> [3, 3, 3, 3, 3, 3, 3, 3, 3, 3]], dtype=uint8)
>>>>
>>> The second one fails to work as you expect because the RHS pointer is
>>> always one step behind the LHS pointer, thus it always reads the data
>>> that just got modified in the previous step. The data you expected it
>>> to read has already been wiped out.
>>>
>>>
>> _______________________________________________
>> NumPy-Discussion mailing list
>> NumPy-Discussion@scipy.org
>> http://mail.scipy.org/mailman/listinfo/numpy-discussion
>>
>
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion@scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion
More information about the NumPy-Discussion
mailing list