Boost logo

Boost :

From: Andy Little (andy_at_[hidden])
Date: 2006-01-30 05:26:39

"Reece Dunn" wrote
> Andy Little wrote:
>>A Fixed Strings library but this one? . What interests me is that the
>>fixed_string library is way ahead as the top download in the Vault download
>>list, but few reviewers have resulted. I feel the mystery can be explained
>>by the reviews so far. looking through them we have all said pretty much
>>same thing. Surprise
>>that there is no overflow policy.
> I intend to fix this for the next review :).

The other surprise for me is that its possible to modify the capacity of a
fixed_string. That means the class name fixed_string is misleading. Unless fixed
means 'not broken' string implying that std::string is flawed. Its that sense
in which it doesnt promise what it delivers. I would expect a fixed_string to
have a fixed length. This fits best with the rationale of fixed_string to be a
char array replacement. I suspect that is what potential users expected, but
nothing like what they got so far. The solution now is to give users what they
expected in the first place.

>>Concern at the lack of detail in the
>>documentation and the way the documentation and the class itself are laid
> I hope to improve the documentation so that users can follow it better.

Ok but there is also a lack of information. Class function signatures are
documented but there is no information on what they do. As a potential user I
get no sense that you as the author are keen about the class and are interested
in trying to help me to get to grips with it. It looks like the documentation
was sketched in and then just never worked on after that. The good thing though
is that you have a potential user-base of 400 to get this right for next time.
It would be worth going through the documentation of other (boost) libraries
that you like to see how the documentation is laid out. There are also some
examples of poor documentation in boost, so use those as examples of what to

> I can only fix issues that are being raised.

Whatever... You need to figure why no issues are being raised. Answer I think is
that, because there is little information provided , then it is difficult to
raise any points about it by telepathy so to speak.

> That said, I intend to move the basic_string_impl class out of the detail
> namespace as this is more useful in other contexts. That class is based on
> the flex_string class written by Andrei Alexandrescu. However, this class -
> like the basic_string class - is *way* too large.
> I have experimented with splitting this class into several smaller ones
> based on the grouping in the standard documentation (iterators, capacity,
> etc.) so you can combine them. This way, it will be like the iterator
> classes. These have had some limited success and I intend on working on this
> after fixing the issues raised in the review.
> With this change, I intend to move the fixed_string class into boost/string
> so that other string classes like const_string can be placed there as well.

OK. Bear in mind that you have 400 potential users who wanted a fixed_string.
Its probably worth trying to find out what they thought a 'fixed_string' is and
how it differs from what you provided. OTOH It sounds now like you have opted
to design something completely different again, which seems a bit of a shame as
you will lose many of the potential fixed_string users.

Andy Little

Boost list run by bdawes at, gregod at, cpdaniel at, john at