Re: [boost] Re: Re: string_algo Formal Review

From: Peter Dimov (pdimov_at_[hidden])
Date: 2003-10-22 15:55:17

Pavel Vozenilek wrote:
> "Peter Dimov" <pdimov_at_[hidden]> wrote
>
>> Part of the problem is that the library isn't really a string
>> algorithm library, it's a container algorithm library with
>> string-like operations. There isn't anything wrong with that but the
>> focus is lost.
>>
>> For example,
>>
>>     bool isspace(std::basic_string<T> const & t, std::locale const &
>> l);
>>
>> is much less controversial than the generic 'all'.
>>
> What to do with people who use Alexandrescu's flex_string<> or other?

Nothing? With all due respect, I estimate the number of people using
flex_string<> and needing 'all' as close to zero. A line needs to be drawn
somewhere; it is this line that separates ivory tower "might be a good idea
if" code from useful libraries that have a clear goal.

All-encompassing "algorithm" libraries (of MPL scope, say) are a fairly
special case and as such aren't handled very well by the Boost review
process. A simplified "std::basic_string convenience functions" component
will be more likely to pass review without much trouble - except, of course,
the functional vs in place debate.

For the "but what about people that use other character sequences" case, a
separate library of supporting algorithms may be more appropriate, where the
focus would be seamless STL integration, perhaps with the algorithms being
submitted one by one and reviewed on a case by case basis (the goal being to
keep a minimal but complete set of algorithms.)