|
|
Blitz Bugs : |
From: Julian C. Cummings (cummings_at_[hidden])
Date: 2004-09-16 12:40:33
Hi Patrick,
I wrote a little test program using the STL vector class and compiled with
gcc 3.3.
It indicates that you can use a default constructor, call the constructor
with size 0 or call the reserve() method with the value 0, and you always
get begin()==end(). I think it would make sense to modify the blitz Array
class to follow this behavior. That would mean allowing resize(0) and
correcting the default constructor so that the begin and end iterators are
equal.
Regards, Julian C.
Dr. Julian C. Cummings
Staff Scientist, CACR/Caltech
(626) 395-2543
cummings_at_[hidden]
> -----Original Message-----
> From: blitz-bugs-bounces_at_[hidden]
> [mailto:blitz-bugs-bounces_at_[hidden]] On Behalf Of Patrick Guio
> Sent: Thursday, September 16, 2004 1:22 AM
> To: blitz-bugs_at_[hidden]
> Subject: Re: [Blitz-bugs] Zero size arrays bug: begin() != end()
>
>
> On Tue, 14 Sep 2004, N Smethurst wrote:
>
> I am not sure what is the expected behaviour of a standard
> (STL style) begin and end iterator for an object of dimension
> 0 but what I can say is that resize(0) is not valid in blitz
> You can check in blitz/array/resize.cc line 37 here there is
> a precondition BZPRECONDITION(length0 > 0); so if you compile
> with BZ_DEBUG you will get an error. Now there might be a
> need for a precondition in the construtor as well? line 166
> in blitz/array-impl.h?
>
> Array(GeneralArrayStorage<N_rank> storage =
> GeneralArrayStorage<N_rank>())
> : storage_(storage)
> {
> length_ = 0;
> stride_ = 0;
> zeroOffset_ = 0;
> }
>
> What do you think Julian?
>
> Cheers, Patrick
>
> > Hello there
> >
> > I discovered a bug in Array. When an array is initialised to zero
> > size, the
> > begin() iterator does not equal the end() interator.
> >
> > For example:
> > Array<float, 1> vec(0);
> > Array<float, 1>::iterator begin = vec.begin();
> > Array<float, 1>::iterator end = vec.end();
> >
> > cout << "begin == end : " << (begin == end) << endl;
> > cout << "begin != end : " << (begin != end) << endl;
> >
> > This gives:
> > begin == end : 0
> > begin != end : 1
> >
> > However, when an array is created then resized, there is no bug:
> > Array<float, 1> vec;
> > vec.resize(0);
> > Array<float, 1>::iterator begin = vec.begin();
> > Array<float, 1>::iterator end = vec.end();
> > cout << "begin == end : " << (begin == end) << endl;
> > cout << "begin != end : " << (begin != end) << endl;
> >
> > This gives:
> > begin == end : 1
> > begin != end : 0
> >
> > The behaviour is the same with higher dimension arrays:
> > Array<float, 2> mat(0, 0);
> > // begin != end
> > Array<float, 2>::iterator begin = mat.begin();
> > Array<float, 2>::iterator end = mat.end();
> >
> > Array<float, 2> mat1;
> > mat1.resize(0, 0);
> > // begin1 == end1
> > Array<float, 2>::iterator begin1 = mat1.begin();
> > Array<float, 2>::iterator end1 = mat1.end();
> >
> > Regards
> >
> > Nicholas
> _______________________________________________
> Blitz-bugs mailing list
> Blitz-bugs_at_[hidden]
> http://www.oonumerics.org/mailman/listinfo.cgi/blitz-bugs
>