python-quantities-developers team mailing list archive

["Darren Dale" <dsdale24@xxxxxxxxx>]

On Tue, Jan 6, 2009 at 4:48 AM, John Salvatier <jsalvati@xxxxxxxxxxxxxxxx>wrote:
>
> On Tue, Jan 6, 2009 at 1:32 AM, John Salvatier <jsalvati@xxxxxxxxxxxxxxxx>wrote:
>
>> I have noticed that while many members functions work as we would like
>> them to work after I overload them, the static member functions in the numpy
>> namespace are unaffected. For example:
>>
>> for the sum function
>> In [13]: y
>> Out[13]:
>> [[ 1.  3.  4.  5.]
>>  [ 1.  2.  3.  6.]]*inch
>>
>> In [14]: y.sum()
>> Out[14]: 25.0*inch
>>
>> In [15]: numpy.sum(y)      # this actually throws an error
>> ... error stuff...
>>
>> for the ravel function
>> In [19]: y
>> Out[19]:
>> [[ 1.  3.  4.  5.]
>>  [ 1.  2.  3.  6.]]*inch
>>
>> In [20]: y.ravel()
>> Out[20]: [ 1.  3.  4.  5.  1.  2.  3.  6.]*inch
>>
>> In [21]: numpy.ravel(y)
>> Out[21]: array([ 1.,  3.,  4.,  5.,  1.,  2.,  3.,  6.])
>>
>> I am not sure what --if anything-- we want to do about this, and I am not
>> sure there is a whole lot we *can* do.
>> Thoughts?
>>
>
Can't we implement our own versions of these functions, like numpy's masked
array package does?


>
> Additionally:
>
> In [10]: x
> Out[10]: [ 1.  2.  3.  4.]*Btu
>
> In [11]: math.e ** x             #this works correctly
> ValueError: exponent is not dimensionless
>
> In [12]: numpy.exp(x)         # but this does not
> Out[12]: [  2.71828183   7.3890561   20.08553692  54.59815003]*Btu


We need to implement exp()