From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-08-17 08:37:53
At 07:25 PM 8/15/2001, Daryle Walker wrote:
>on 8/14/01 4:28 PM, Beman Dawes at bdawes_at_[hidden] wrote:
>> Daryle Walker's Base-from-Member library is accepted by Boost subject
>> two caveats:
>> * Issues raised during the review be addressed, particularly Ed Brey's
>> http://groups.yahoo.com/group/boost/message/15406. (Note that Daryle
>> already dealt with some issues in an update in the files section.)
>I think I covered every issue. Do you have any specific concerns?
No, I just hadn't checked to see if any issues were still outstanding.
>> * The documentation needs to be modified to more clearly indicate that
>> some applications the best use of the library is not to #include the
>> directly, but to use it as a demonstration of how to hand-code the
>> base-from-member idiom.
>When can't the code be used directly (besides compiler problems,
>constructor cases we don't cover, and the limitations of C++ syntax)?
The point is more that some reviewers felt they would prefer to view
Base-from-Member as an idiom to be followed, rather than a library to be
used as is. Note that it was explicitly stated that it wasn't a question
of your code, which was fine.
The message we need to get across is that this library is fine as it
stands, but it is also understood that users may wish to use it as a
pattern for their own code.
>> This last point is the key issue with Base-from-Member; some users will
>> wish to view the library as a demonstration of the idiom or technique
>> rather than as code they will always reuse.
>> This seems to me to require some explanation. Some of the points are
>> specific to Base-from-Member, such as "When is it better to hand
>> and so should be dealt with in the Base-from-Member docs.
>Should this be done soon? When is the next release, anyway? (We've have
>had three formal reviews, including the current Bind, since that last
>release. We also have the results of the Tuples and Pre-processing
>outstanding from before the last release.)
I'm aiming for a release within the next few days. Vacation plans get in
the way after that.
Tuples are ready, but the preprocessor stuff isn't. Whether or not
Base-from-Member goes in this release or the next one depends a bit on
unrelated factors, like how smooth the rest of the release work goes. And
whether the power stays up here; it has been off and on the last two days.
>> But it also seems to me that Boost should have a page titled something
>> "Demonstration Libraries" which Base-from-Member and similar libraries
>> link to. It would describe the concept that we reuse code when there
>> good library code solution, but other times there is no applicable
>> code, so instead we reuse code idioms and techniques (for small-scale
>> problems), or we reuse design patterns (for large-scale
>> to non-boost resources might also be appropriate.
>This is separate from cleaning up Base-from-Member, right?
Yes, it isn't something you need to worry about. It is on my "do list".
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk