TITLE: std::vector base address

(Source: comp.lang.c++.moderated, 28 Nov 2000)

GALLUCCI: "David Gallucci" <dgall812@concentric_junk.net>

>std::vector<int> iVec;
>int* pI = &iVec[0]
>
>is it guaranteed that pI will point to the base address? ( assuming of
>course that no reallocation takes place )

SUTTER: "Herb Sutter" <hsutter@peerdirect.com>

Yes. All existing implementations do this, and the guarantee was
formally added last month as part of Technical Corrigendum #1 to the C++
Standard.

I also discussed this issue in detail in my column "Standard Library
News, Part 1: Vectors and Deques" (C++ Report, 11(7), July/August 1999).
In particular, the reason you give is in fact the biggest motivation for
having this guarantee:

GALLUCCI:

>Is this concept too dependent on vector's implementation?  I need a managed
>array, but I need to be able to pass the base address to a C function. I'd
>rather reuse something that already exists, but I need to make sure that
>this will work regardless of possible changes to vector's implementation.
>Am I violating the black box?

SUTTER:

The first question addressed in that article is: "Can you use a vector
interchangeably with a C-style array (say, with an existing function
that expects an array)?" We want to encourage people writing new code to
use vector instead of plain old arrays, because vectors are safer,
automatically manager, and so forth. That would be much more difficult
if a vector weren't usable with legacy code that expects C-style arrays,
such as a (T* ptr, size_t len) argument list, so it's important that
they be compatible.



_______________________________________________
cpptips mailing list
http://cpptips.hyperformix.com