Review results for "format"

From: Jens Maurer (Jens.Maurer_at_[hidden])
Date: 2002-01-24 17:05:01

The review period for Samuel Krempp's "format" library has
ended.  The library is accepted into boost.
Thanks to Samuel for his ongoing efforts, and to all the
reviewers for their comments.

The polished candidate should be posted to the files
section and announced on the mailing list for an
informal mini-review.  Should no vetoes be cast in the
week following the announcement, the candidate will go in.

I'd like to highlight a few issues come up during review that
I feel should be addressed.

 - Documentation rework (there were quite a few ideas during the review,
and good documentation is very important for boost libraries)
 - boost::format (no sub-namespace) appears to be fine, seldom-used
helpers with commonly used names ("manip"?) might go in boost::io
(or a boost::format sub-namespace if it weren't taken by the main
stuff) implementation details should go in boost::io::detail.
 - It seems to me that % won over [] by a small margin.  But
judge for yourself, Samuel.
 - Try to improve portability.
 - Have only ONE interface for a given feature (see below),
unless there are *very* compelling reasons to do otherwise.

One issue that I would like to re-emphasize is "minimality of
interface":  The full Open Group syntax for argument reordering
("%1$d, %2$d" if I read my Linux man page correctly) does not
appear to me as significantly more clumsy than "%1 %2", and of
course, the full Open Group syntax has a lot more features.  Similarly,
I see no compelling reason to provide e.g. both   format("xxx", a, b)
and format("xxx") % a % b  syntax variants:  Make an informed
decision, document the rationale, say what the other option was,
and why it wasn't chosen.

Jens Maurer