Re: behaviour of vector<T>::operator[]

"Jason Winnebeck" <gillius-ng@xxxxxxxxxxxxxxxxxx> wrote in message
news:%237ivQtxsFHA.2592@xxxxxxxxxxxxxxxxxxxxxxx
>I would think that the point of the question is to see what you can do
>simply. I would imagine that writing a function to do this and the
>technical ability to do this is not out of the question for the original
>poster.
>
> One of the points of using [] over .at is that it is more intuitive and
> (IMO) readable than .at. Also, using [] might be important for generic
> programming. That said, I'm not sure a non-standard (to other programmers
> reading the code ) macro or a templated function is a good solution. To
> me, "mydata[5]" is OK, but "At( mydata, 5 )" looks too unintuitive when
> the only benefit of the non-cannonical code is bounds checking in debug
> mode.

Of course, that's not the only solution. You can always use some form of
assert to check the precondition. In fact, you probably wouldn't like error
message at() function is throwing anyway. It varies from implementation to
implementation, but generally what() function I've seen returned just
"vector" with no more information. So even for this reason most people will
avoid at(), resorting to their own checks. Moreover for debug purposes, you
don't want the exception that at() throws, you need some form of assert that
aborts your program right where the bug is and doesn't unwind the stack.

>
> I'm guessing the right solution is what Plauger said was going into
> Whidbey. I applaud MS and/or the library writer(s) as appropriate for
> doing things because I have been wishing for that feature for quite awhile
> myself.

I am not that sure. Firstly, as I mentioned above, debug checks should not
be based on exceptions. Secondly, I for example, rarely use debug build for
debugging because it's impractically slow. I'd rather do a release build
without optimizations or select optimizations in some places. In many cases
I need to run the code at full throttle, having only few places checked.
What I think we need is a checked implementation of the standard library,
that would allow a degree of flexibility of turning on and off certain
checks in certain places using some sort of asserts, not exceptions.

- gene


.

Relevant Pages

  • Re: Exception handling of filesys.exe and gwes.exe
    ... Most errors in this modules will hit an assert that tells you what you're doing wrong in the debug output when using a debug build. ... I have wrapped each thread I have created with __try and __exception blocks, but that does not seem to catch these types of exceptions. ...
    (microsoft.public.windowsce.app.development)
  • Re: Null exception error and general tidying of code..
    ... > beginning programming courses and books don't teach how to ... is far shorter than the time wasted in attempting to debug code ... HTML anchors to the major sections ... nice to see you posting to the groups again, ...
    (comp.lang.java.help)
  • Re: Weird error during programme startup
    ... Microsoft Visual C++ Debug Library ... I've stepped through the innards of _initterm and it ... it won't assert in release builds. ... You said that you moved all strings to string tables. ...
    (microsoft.public.vc.language)
  • beginner with programming, how to learn to debug and few C general questions
    ... I am first an accountant and decided to take on programming a few ... That's why I'd like to learn how to debug. ... I have mainly been programming under Windows Xp DJGPP Dolorie... ... int symbol; ...
    (comp.lang.c)
  • Re: Debugging designer bug
    ... that's not the Designer panel. ... Load any debug project into this instance (for ... exception) so you'll have source code when you enter the debugger. ... Then under Debug choose Exceptions ... ...
    (microsoft.public.dotnet.languages.csharp)
Loading