Re: containers tentative design summary

[Jonathan Underwood <jonathan.underwood@gmail.com>]

2009/11/15 Tuomo Keskitalo <Tuomo.Keskitalo@iki.fi>:
> Hello,
>
> On 11/14/2009 04:42 PM, Brian Gough wrote:
>
>> At Mon, 09 Nov 2009 16:07:43 -0700,
>> Gerard Jungman wrote:
>>>
>>> On Fri, 2009-11-06 at 14:42 +0000, Brian Gough wrote:
>>>>
>>>> Ok, I have read the paper now.  I do think the practice of casting
>>>> described there is rather dated.  When people had no viable
>>>> alternative to C, they had to resort to such tricks.  It is not
>>>> something that should be encouraged today -- programs should either be
>>>> written safely, following the rules of type-checking in C, or be
>>>> written in another language.
>>
>> From 1.3 million lines of code they describe only one method which
>> does not use casts, which is the one we use.  I don't think we are
>> going to find anything that is better than the current method for
>> views.
>
> Apparently this is a fundamental question.
>
> Currently GSL uses C in a type-safe manner which forces somewhat complicated
> APIs for everyone but enables users to find some lethal bugs. More
> user-friendly APIs would allow people to silently break their programs if
> they are not careful. And there is no compromise. Does this summarize the
> situation?
>
> I am sure there are users for both approaches. I don't know.. Maybe GSL
> should be the safe, strict library, and there should be another scientific
> library in C which aims towards interoperability between data, libraries and
> languages? This would split forces, but both projects would also benefit
> from each other.
>


I think it would be better if GSL, being written in C,  were to
present the most functional interface at the expense of sacrificing
type safety, and higher level language bindings could then wrap that
API and do type checking as needed. As gerard says, C isn't about type
safety.

Jonathan