Re: fix null assignment

[Jaroslav Hajek]

On Mon, Sep 22, 2008 at 1:42 PM, Jaroslav Hajek <address@hidden> wrote:
> On Sat, Sep 20, 2008 at 4:32 PM, John W. Eaton <address@hidden> wrote:
>> On 20-Sep-2008, John W. Eaton wrote:
>>
>> | Please take a look at the two sets of diffs below.
>> |
>> | The first set of changes seem to have the desired effect, but I'm not
>> | entirely happy with them.
>>
>> If you are interested in experimenting with these changes, please use
>> the following version of the first patch instead.
>>
>> Thanks,
>>
>> jwe
>>
>>
>
> Close enough for government work, I guess, but still looks a bit
> dirty. The special check in assign is not immediatelly obvious, and it
> does not catch the more obscure cases like `b{1} = []; a = ones(3);
> a(1:2,:) = b{1}'. The problem is that not all octave_value assignments
> happen through octave_value:assign.
>
> The first attached patch (on top of your first one) hacks two more
> places to prevent a null matrix from being stored anywhere,
> encapsulating the hack in a member function.
> (empty
>
> You may also consider the second patch. It avoids using the extra bool
> flag in Array.h by introducing a special static member for null
> matrices.
>
> I can't say I like this solution. I like my idea with op_nul_asn_eq
> better. Still, this one is here and is working, it's a solution, while
> my idea is just an idea. I guess I can attend to it in the future.
>

Ooops. It seems that your first patch inadvertently removes matrix =
float_matrix assignment. Attached is a fix (should probably be folded
into your original patch).



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz

empty-assign-diffs-1.x.diff
Description: Text Data