[endian] Updated version available in the file vault
There's an updated version available in the file vault. Changes include: * Use of namespaces to simulate C++0x scoped enums. * Addition of operator= for improved efficiency. * Updated docs, including new FAQ entries. * Jamfile.v2 added. See http://tinyurl.com/4bswsh Does this look ready for formal review? Comments appreciated! --Beman
On 2008-05-26 05:22, Beman Dawes wrote:
There's an updated version available in the file vault. Changes include:
* Use of namespaces to simulate C++0x scoped enums. * Addition of operator= for improved efficiency. * Updated docs, including new FAQ entries. * Jamfile.v2 added.
Does this look ready for formal review? Comments appreciated!
The 0.8 version looks good to me, except for two things: 1. As I commented earlier about the name:
... there is already an endian.hpp elsewhere in boost and I think it would be easy for people to miss this endian, and never discover its functionality.
When I was searching to see if boost had something like this, the great majority of the hits were for the existing boost/detail/endian.hpp, which wasn't what I was hoping for. It doesn't seem good to me to have two headers in boost with the same file name, even if they are in different directories. The only good solution I can see is to rename detail/endian.hpp to detail/endianness.hpp, which is more descriptive, IMHO, but would probably break a lot of existing client code. Although maybe clients include detail/ headers at their own risk? 2. As mentioned in other posts, the implementation of operators using <boost/operators.hpp> needs to be reconsidered. Thanks, Beman, for your work in getting this moving again. --Neil
Neil Mayhew wrote:
When I was searching to see if boost had something like this, the great majority of the hits were for the existing boost/detail/endian.hpp, ... IMHO, but would probably break a lot of existing client code.
What about the name packed.hpp, since this is what endian.hpp really does? Just a random idea. -- _________________________________________ _ _ | Roland Schwarz |_)(_ | aka. speedsnail | \__) | mailto:roland.schwarz@chello.at ________| http://www.blackspace.at
On 5/29/08 11:41 PM, Roland Schwarz wrote:
What about the name packed.hpp, since this is what endian.hpp really does?
Good suggestion. In my mind, however, what this code really does is allow me to handle other-endian values, and the packing just comes along with it as a bonus. Also, if I was searching for something to handle other-endian data, I wouldn't be searching for "packed". Even so, I think this name is definitely worth considering. --Neil
----- Original Message ----- From: "Neil Mayhew" <neil_mayhew@users.sourceforge.net> To: <boost@lists.boost.org> Sent: Friday, May 30, 2008 3:13 PM Subject: Re: [boost] [endian] Updated version available in the file vault
On 5/29/08 11:41 PM, Roland Schwarz wrote:
What about the name packed.hpp, since this is what endian.hpp really does?
Good suggestion. In my mind, however, what this code really does is allow me to handle other-endian values, and the packing just comes along with it as a bonus. Also, if I was searching for something to handle other-endian data, I wouldn't be searching for "packed". Even so, I think this name is definitely worth considering.
What about endiannes.hpp? Vicente
Neil Mayhew wrote:
On 2008-05-26 05:22, Beman Dawes wrote:
There's an updated version available in the file vault. Changes include:
* Use of namespaces to simulate C++0x scoped enums. * Addition of operator= for improved efficiency. * Updated docs, including new FAQ entries. * Jamfile.v2 added.
Does this look ready for formal review? Comments appreciated!
The 0.8 version looks good to me, except for two things:
1. As I commented earlier about the name:
... there is already an endian.hpp elsewhere in boost and I think it would be easy for people to miss this endian, and never discover its functionality.
When I was searching to see if boost had something like this, the great majority of the hits were for the existing boost/detail/endian.hpp, which wasn't what I was hoping for. It doesn't seem good to me to have two headers in boost with the same file name, even if they are in different directories. The only good solution I can see is to rename detail/endian.hpp to detail/endianness.hpp, which is more descriptive, IMHO, but would probably break a lot of existing client code. Although maybe clients include detail/ headers at their own risk?
Use of detail/ headers is certainly "at-your-own-risk". My quick take on detail/endian.hpp is that it should be moved to boost/config/platform, but John Maddock is the config maintainer so that would be his call. As far as boost/integer/endian.hpp goes, I plan to stick with that name.
2. As mentioned in other posts, the implementation of operators using <boost/operators.hpp> needs to be reconsidered.
Yes. On my do list.
Thanks, Beman, for your work in getting this moving again.
Thank you and everyone else responding to this thread for all your ideas and suggestions! --Beman
participants (4)
-
Beman Dawes -
Neil Mayhew -
Roland Schwarz -
vicente.botet