RE: [boost] Re: Invariants Library

From: Reece Dunn (msclrhd_at_[hidden])
Date: 2004-01-01 22:46:38

Dan W. wrote:

>Reece Dunn wrote:
>>Has there been any discussion regarding an invariants library (advanced 
>>assert), or is there one already implemented in boost?

>The compiler I'm using, Digital Mars version 8.38 already supports 
>invariant, pre- and post-conditions. See...
>http://www.digitalmars.com/ctg/designbycontract.html

cool! But these seem to work via assert which will terminate the program 
upon invariant failure.

With C++ having exception support, surely the exception mechanism is a 
better way to deal with invariants that do not hold? This allows the client 
code to react to the failure and report something readible to the user, 
like:
   invariant failure: cannot access a null pointer, in ...
instead of something like:
   abnormal program termination

Also, you need a compiler that supports the DBC extensions. The current 
thinking behind the new C++ standard is to adopt a minimal set of language 
extensions and a large number of library extensions (c.f. Bjarne Stroustrup, 
et. al.). Having a library-based invariant mechanism (integrating into the 
exception handling mechanism) allows support for older compilers (GCC 
2.95.x; VC6; etc) that do not implement C++0x features.

>and I also just saw a couple of boost invariant implementations while 
>browsing through the files section,
>http://groups.yahoo.com/group/boost/files/

I could only find http://groups.yahoo.com/group/boost/files/DbC which 
appears to support assert, not throw. Also, the files are old (2001).

>Design By Contract is such a wonderful concept, BTW, I used to program in 
>Eiffel, and there's nothing I miss more than its built-in DBC.

I agree, DBC is a powerful concept. What I am trying to do is open up the 
discussion on how it should be implemented in C++.

Regards,
Reece

_________________________________________________________________
Express yourself with cool new emoticons http://www.msn.co.uk/specials/myemo