Boost logo

Boost :

Subject: Re: [boost] [review][constrained_value] Review of Constrained Value Library begins today
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2008-12-03 07:22:16


> -----Original Message-----
> From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]]
On
> Behalf Of Jeff Garland
> Sent: 01 December 2008 13:00
> To: boost_at_[hidden]
> Subject: [boost] [review][constrained_value] Review of Constrained Value
Library
> begins today
>
> The review of the Robert Kawulak's Constrained Value library begins today
> December 1, 2008, and will end on December 10th -- I will be the review
> manager. Please post reviews to the developer list.
> - What is your evaluation of the design?

OK.

The floating point issue has been ducked for now, wisely I think, and with
sufficient explanation - for now.

Though the concepts could (can) also be applied very usefully to
floating-point types.

Before constrained FP can be unleashed on unsuspecting users, I believe we
need to establish separately a clear package of functions for handling
approximate comparisons, something using the ideas in BOOST_CHECK_CLOSE.

I would very much like to see this as a future Boost library component.

With sensible defaults that handle the few epsilon uncertainties that arise
from conversion from FP registers and memory,
and round off from calculations, it would be useful in practice without
exposing users unfamiliar with floating-point to too much risk of nasty
surprises.

> - What is your evaluation of the implementation?

Looks OK.

BUT a suite of tests (promised) is essential before release.

These should not only include the simple features (for which tests may seem
a bit over the top but give a warm feeling),
but also some of the 'fancier' features that are much more likely to cause
trouble, and for which tests are more likely to highlight problems.

> - What is your evaluation of the documentation?

Very good. (I found no spelling mistakes ;-) )
And Robert did not fall into the trap of thinking that using Doxygen meant
that you didn't need to write any other documentation.
The usage and rationale and design compromises were discussed sufficiently.

> - What is your evaluation of the potential usefulness of the library?

Very useful. An essential building block for reliable code.

> - Did you try to use the library? Used OK previously.

> - How much effort did you put into your evaluation?

An hour or so (re-)reading the docs.

> - Are you knowledgeable about the problem domain?

Slightly.

Paul

PS In the absence of a decent numbers of reviewers, it would very helpful to
judge the value of software (and of reviews) to know the size of the 'user
base' of packages submitted for reviews. Is there any way we can get this
information?

---
Paul A. Bristow
Prizet Farmhouse
Kendal, UK   LA8 8AB
+44 1539 561830, mobile +44 7714330204
pbristow_at_[hidden]

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk