python-quantities-developers team mailing list archive

["Darren Dale" <dsdale24@xxxxxxxxx>]

I experimented with that approach for most of the morning yesterday. The
problem is that lots of operations that create new Quantity objects, like
multiplication, yield an object that you would think owns its own data, but
in fact does not. I asked about it on the numpy mailing list, and Robert
Kern suggested that it would probably be best to simply disallow in-place
modification of units.

Robert said it might be possible to check if q.base is an instance of
Quantity, and I had been trying that approach, but the arithmetic operations
yield an object whose base is an ndarray, when we would be counting on the
base being None or a Quantity. If we can find a way to yield objects that
own their own data, without additional overhead of creating additional
copies of the data, then I think we could allow in-place operations.
otherwise, we choose between forbidding them, or allowing them but warning
people that in-place operations are potentially dangerous.

Darren

P.S. please respond to the mailing list so we can capture these
considerations in the mailing list archives

On Sun, Jan 18, 2009 at 10:29 PM, John Salvatier
<jsalvati@xxxxxxxxxxxxxxxx>wrote:

> That would be the only case when it's not ambiguous, really.
>
>
> On Sun, Jan 18, 2009 at 7:28 PM, John Salvatier <jsalvati@xxxxxxxxxxxxxxxx
> > wrote:
>
>> Would it work to support in place arithmetic only for cases when the array
>> is the owner of the data (basically is not a view)?
>>
>>
>> On Sun, Jan 18, 2009 at 7:22 PM, Darren Dale <dsdale24@xxxxxxxxx> wrote:
>>
>>> I looked at it for quite a while today, and I dont think there is a safe
>>> way to support it. I could consider adding it back in if there is a big
>>> demand for it. I think I remember you suggested not allowing in-place
>>> modification of units in one of our first discussions, and I've finally come
>>> around to that point of view.
>>>
>>> We can still allow in-place arithmetic that doesnt modify the units.
>>>
>>> I just pushed some big changes in the constants subpackage up to
>>> launchpad.
>>>
>>> Darren
>>>
>>> On Sun, Jan 18, 2009 at 9:53 PM, John Salvatier <
>>> jsalvati@xxxxxxxxxxxxxxxx> wrote:
>>>
>>>> I see your latest update about not allowing in-place arithmetic. Do you
>>>> think this is something we should support eventually? It's not something I
>>>> generally use, but it seems like something that is important for some
>>>> people.
>>>>
>>>> _______________________________________________
>>>> Mailing list: https://launchpad.net/~python-quantities-developers<https://launchpad.net/%7Epython-quantities-developers>
>>>> Post to     : python-quantities-developers@xxxxxxxxxxxxxxxxxxx
>>>> Unsubscribe : https://launchpad.net/~python-quantities-developers<https://launchpad.net/%7Epython-quantities-developers>
>>>> More help   : https://help.launchpad.net/ListHelp
>>>>
>>>>
>>>
>>> _______________________________________________
>>> Mailing list: https://launchpad.net/~python-quantities-developers<https://launchpad.net/%7Epython-quantities-developers>
>>> Post to     : python-quantities-developers@xxxxxxxxxxxxxxxxxxx
>>> Unsubscribe : https://launchpad.net/~python-quantities-developers<https://launchpad.net/%7Epython-quantities-developers>
>>> More help   : https://help.launchpad.net/ListHelp
>>>
>>>
>>
>