🔒 Locked

std::vector clear zeroing capacity

Started by Aface Jun 10, 2006 at 9:29 PM 4 replies 3.9k views

Original Post

Jun 10, 2006 at 9:29 PM

Hi All, Have searched this, with no luck. I was under the impression that, having filled my vector with objects, i could .clear() it and the underlying capacity should not be changed? I have searched some online vector resources, and almost all of them report that .clear() simply calls .erase(begin(), end()), however, I have done a test and this is not the case. Is my version of the STL (default that is installed with vc7.1) not conforming, or is this expected behaviour? Alex The test I performed is below



	std::vector<int> testVector;

	int iCapacity = testVector.capacity(); //returns 0

	testVector.push_back(5);

	iCapacity = testVector.capacity(); //returns 1

	testVector.resize(10);

	iCapacity = testVector.capacity(); //returns 10

	testVector.pop_back();

	iCapacity = testVector.capacity(); //returns 10

	testVector.erase(testVector.begin(), testVector.end());

	iCapacity = testVector.capacity(); //returns 10

	testVector.resize(10);

	iCapacity = testVector.capacity(); //returns 10

	testVector.clear();

	iCapacity = testVector.capacity(); //RETURNS 0!!!!!

JBourrie

Jun 10, 2006 at 9:31 PM

Yep, pain in the ass, huh? :)

The secret is calling resize(0) instead.

Aface

Jun 10, 2006 at 10:15 PM

Thanks for the info.

Any idea whether this is the correct behaviour, or is my implementation of the STL buggy?

Nice game by the way.

Alex

smart_idiot

Jun 10, 2006 at 10:36 PM

I don't know if it's wrong, but it isn't something I would have expected.

libstdc++ and stlport both give me 10 if you wanted something to compare with it to.

deathkrush

Jun 10, 2006 at 10:37 PM

I get the same behaviour with Visual Studio .NET 2003.

jflanglois

Jun 10, 2006 at 11:05 PM

Quote:
Original post by deathkrush
I get the same behaviour with Visual Studio .NET 2003.

Indeed you do, since that is 7.1.

In VC2005, however, the last call to vector::capacity() returns 10 (which is what I and I am sure many would expect), so it must be something with previous implementations. The standard draft of 1998 (and the working draft of 2005) makes no mention (as far as I can tell) regarding capacity after a call to clear(), the only requirements being that it is equivalent to erase( begin(), end() ) and post size() == 0, so capacity could be 0, I guess.

[edit] For those who are interested, I believe the correct way to ensure a capacity of 0 is the empty-temporary/swap idiom:
std::vector< int > v; // ... std::vector< int >().swap( v );

jfl.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.