In article <935.363.uupcb@ftl.atl.ga.us> chris.smith@ftl.atl.ga.us (Chris Smith)  writes:
>In article <CJnr3H.BxD@ucc.su.OZ.AU>
>maxtal@physics.su.OZ.AU (John Max Skaller) writes:
>
>JMS>There are many better alternatives.
>JMS>And that is the problem. My solution: use your own string class :-)
>JMS>That, being written in Standard C++, will be portable between
>JMS>systems, even if it does not interface to other string classes.
>
>Oh Nooooo!  I could have gone the rest of the year without hearing
>_you_ say that.  I am aware that you have an important voice
>in the current ANSI proceeedings.
>This (use your own <insert name here> class) is
>killing me, and I had real hopes that the committee would do
>something about it.

 We are. But, there is no Standard C++ yet.

>I don't disagree with your advice.
>For the time being, I'm doing _exactly_ as you recommend.  But so
>are all of the vendors of the libraries I use.

 Yes. And one vendor I spoke to said they' have to
continue to supply their own string class anyhow.

 The difference in the final Standard, however is that
there will be a namespace mechanism to control this.

>So my code is
>littered with places where I have to convert from my string class
>to zApp's string class, and Borland's string class.  OK, converting
>from one string class to another is somewhat trivial, but
>converting from TArrayAsVector<float> to floatDynaArray is not as
>trivial and a real waste of effort.

 I think restraint is called for. I intend to use
the new string class supplied by Borland for a while to see
how it goes. I hope not to convert ANY of my code: I'll just pull out:

 #include <mystring.hpp>

and put in

 #include <borlandstring.h> // or whatever

and see what happens. Probably, may of the more conservative
codes will just work, where I used weird concepts of my own,
maybe I'll have a bit more work to do.

>I have often resorted to
>avoiding container classes in my own code, because I know I'll
>have to convert to somebody elses container class anyway.
>I beleive that I'm not alone in this practice of writing
>significant portions of code that should be OO in standard C,
>just because of the lack of definition of some important
>container classes.  This lack of definition of standard
>class libraries is IMO a serious detriment to the
>implementation of the results of OOD in C++.

 I agree.
>
>ANSI and POSIX have defined the standard C libraries.  Container
>classes are as important to OOP as <string.h> is to standard C.
>I feel that one of the very most important things ANSI could
>do would be to define a minimal set of standard class libraries.
>And tear down this temple of babble.

 Again, I agree. But it takes time.

--
        JOHN (MAX) SKALLER,         INTERNET:maxtal@suphys.physics.su.oz.au
 Maxtal Pty Ltd,      CSERVE:10236.1703
        6 MacKay St ASHFIELD,     Mem: SA IT/9/22,SC22/WG21
        NSW 2131, AUSTRALIA

In article <2hk7et$jjv@Tandem.CAM.ORG>
vr@CAM.ORG (Alain Lauzon) writes:

AL>Chris Smith (chris.smith@ftl.atl.ga.us) wrote:
AL>: In article <CJnr3H.BxD@ucc.su.OZ.AU>
AL>: maxtal@physics.su.OZ.AU (John Max Skaller) writes:

AL>: JMS>And that is the problem. My solution: use your own string class :-)
AL>: JMS>That, being written in Standard C++, will be portable between
AL>: JMS>systems, even if it does not interface to other string classes.

AL>: Oh Nooooo!  I could have gone the rest of the year without hearing
AL>: _you_ say that.  I am aware that you have an important voice

[my whining about the lack of standard containers elided]

AL>: ANSI and POSIX have defined the standard C libraries.  Container
AL>: classes are as important to OOP as <string.h> is to standard C.
AL>: I feel that one of the very most important things ANSI could
AL>: do would be to define a minimal set of standard class libraries.

AL>The containers defined in the ODMG-93 book could be proposed for a
AL>standard, they are very well constructed...

I am not aware of that book, and would like to know more about
it.  Are you saying that it is a possibility that such a thing
may be proposed?  ANY STANDARD container class, even a poorly
constructed one would be better than NO STANDARD. (Not shouting,
just emphasizing).  I don't like the 'standard' string class
that has been (??) adopted, it's too large and complex, but
I like it better than no standard at all.

Might there be a proposal forthcoming with regards to container
classes?

regards,
chris
---
. OLX 2.2 . What's another word for Thesaurus?

In article <CJnr3H.BxD@ucc.su.OZ.AU>
maxtal@physics.su.OZ.AU (John Max Skaller) writes:

JMS>There are many better alternatives.
JMS>And that is the problem. My solution: use your own string class :-)
JMS>That, being written in Standard C++, will be portable between
JMS>systems, even if it does not interface to other string classes.

Oh Nooooo!  I could have gone the rest of the year without hearing
_you_ say that.  I am aware that you have an important voice
in the current ANSI proceeedings.
This (use your own <insert name here> class) is
killing me, and I had real hopes that the committee would do
something about it.  I don't disagree with your advice.
For the time being, I'm doing _exactly_ as you recommend.  But so
are all of the vendors of the libraries I use.  So my code is
littered with places where I have to convert from my string class
to zApp's string class, and Borland's string class.  OK, converting
from one string class to another is somewhat trivial, but
converting from TArrayAsVector<float> to floatDynaArray is not as
trivial and a real waste of effort.  I have often resorted to
avoiding container classes in my own code, because I know I'll
have to convert to somebody elses container class anyway.
I beleive that I'm not alone in this practice of writing
significant portions of code that should be OO in standard C,
just because of the lack of definition of some important
container classes.  This lack of definition of standard
class libraries is IMO a serious detriment to the
implementation of the results of OOD in C++.

ANSI and POSIX have defined the standard C libraries.  Container
classes are as important to OOP as <string.h> is to standard C.
I feel that one of the very most important things ANSI could
do would be to define a minimal set of standard class libraries.
And tear down this temple of babble.
regards,
chris
---
. OLX 2.2 . I hear radio waves in my head.

Chris Smith (chris.smith@ftl.atl.ga.us) wrote:
: In article <CJnr3H.BxD@ucc.su.OZ.AU>
: maxtal@physics.su.OZ.AU (John Max Skaller) writes:

: JMS>There are many better alternatives.
: JMS>And that is the problem. My solution: use your own string class :-)
: JMS>That, being written in Standard C++, will be portable between
: JMS>systems, even if it does not interface to other string classes.

: Oh Nooooo!  I could have gone the rest of the year without hearing
: _you_ say that.  I am aware that you have an important voice
: in the current ANSI proceeedings.
: This (use your own <insert name here> class) is
: killing me, and I had real hopes that the committee would do
: something about it.  I don't disagree with your advice.
: For the time being, I'm doing _exactly_ as you recommend.  But so
: are all of the vendors of the libraries I use.  So my code is
: littered with places where I have to convert from my string class
: to zApp's string class, and Borland's string class.  OK, converting
: from one string class to another is somewhat trivial, but
: converting from TArrayAsVector<float> to floatDynaArray is not as
: trivial and a real waste of effort.  I have often resorted to
: avoiding container classes in my own code, because I know I'll
: have to convert to somebody elses container class anyway.
: I beleive that I'm not alone in this practice of writing
: significant portions of code that should be OO in standard C,
: just because of the lack of definition of some important
: container classes.  This lack of definition of standard
: class libraries is IMO a serious detriment to the
: implementation of the results of OOD in C++.

: ANSI and POSIX have defined the standard C libraries.  Container
: classes are as important to OOP as <string.h> is to standard C.
: I feel that one of the very most important things ANSI could
: do would be to define a minimal set of standard class libraries.
: And tear down this temple of babble.
: regards,
: chris
: ---
: . OLX 2.2 . I hear radio waves in my head.

The containers defined in the ODMG-93 book could be proposed for a
standard, they are very well constructed...


Alain Lauzon