Re: [ublas] Function 'equals' in matrix_assign.hpp

Subject: Re: [ublas] Function 'equals' in matrix_assign.hpp
From: nasos (nasos_i_at_[hidden])
Date: 2016-03-06 13:26:55

It is not by design, it is just that originally it was developed with 
floating point arithmetic in mind. Please change it to <= in the branch 
and we will propagate it. Make a separate commit just for this so we 
keep track of the change in case something goes  bad. Also run all the 
tests before submitting a pull request. If you don't plan to sumbit a 
pull request please create a new issue so that we deal with it.

-Nasos

On 03/04/2016 12:46 AM, Shangtong Zhang wrote:
> Hello,
>
> I’m trying to contribute to toeplitz matrix, here is my prototype
> https://github.com/ShangtongZhang/ublas/blob/toeplitz/include/boost/numeric/ublas/toeplitz.hpp
>
> However I have a problem.
> When I execute code
> matrix<int> m_new(m)
> where m is a toeplitz matrix, it will raise exception.
> When matrix_assign is executed, it calls ‘equals' in matrix_assign.hpp
> https://github.com/ShangtongZhang/ublas/blob/toeplitz/include/boost/numeric/ublas/detail/matrix_assign.hpp line 
> 32
> However the parameter epsilon is 0, norm_inf (e1 - e2 ) < 0 will never 
> succeed.
> So it will return false and raise exception.
> Same things happen even m is a banded matrix.
> Why is it ‘<' rather than '<= '? Is it by design?
>
> Thanks,
>
> Shangtong Zhang,
> Senior Student,
> School of Computer Science,
> Fudan University, PRC.
>
>
>
> _______________________________________________
> ublas mailing list
> ublas_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/ublas
> Sent to: athanasios.iliopoulos.ctr.gr_at_[hidden]