'exception' question

[mclark]

Can someone quickly tell me what type of exception is thrown when an STL container is full...?


imagine...

int main(void) {
std::vector<int> myVec;

// loop only exited via exception (which exception???)
for ( ; ; ) {
myVec.push_back(42); // the answer
}

return 0;
}

what exception is generated???

[Graham]

There is no specific exception for "container full". In your case, you would probably get bad_alloc when you run out of memory.

[Kheun]

I dont' think there is any size limit for all containers in the STL. Even if the number of reserved element has been exceeded, the container is capable of resizing itself by allocating more memory.

[mclark]

Originally posted by Kheun
I dont' think there is any size limit for all containers in the STL. Even if the number of reserved element has been exceeded, the container is capable of resizing itself by allocating more memory.

I think you missed the infinite loop. Regardless of container resizing - it will run out of resources and at very least a bad_alloc exception will occur.

Graham is on the right track...

Graham - assuming the std::allocator is the exception guarenteed to be 'bad_alloc' (this can probably be inferred from the std::allocator spec in the standard) or is that implementation dependent?

Now down to the tricky part of my question, intrestingly reserve throws length_error if the size request is larger than max_size. If the std::vector<>:: push_back (or any other modifier) uses reserve internally and their has been enough free memory that the new requested size (based on the growth strategy) is both larger than max_size and unsatisifiable by the allocator - do I get a 'bad_alloc', 'length_error' or is it all just implementation dependent.

[Graham]

Originally posted by mclark
Now down to the tricky part of my question, intrestingly reserve throws length_error if the size request is larger than max_size. If the std::vector<>:: push_back (or any other modifier) uses reserve internally and their has been enough free memory that the new requested size (based on the growth strategy) is both larger than max_size and unsatisifiable by the allocator - do I get a 'bad_alloc', 'length_error' or is it all just implementation dependent.

I've had a quick look at the standard and I can't find the answer. As you say, it boils down to whether internal reallocation uses reserve(), and, if so, does it check for a possible length_error before it does it? I couldn't find anything covering that. All it says about the allocator is "reserve() uses Allocator::allocate() which may throw an appropriate exception". I did try this out once but, of course, being VC++, it doesn't throw bad_alloc, it returns NULL, so the test was inconclusive, and I didn't have time to try to persuade VC++ to throw the exception like it's supposed to.

I suppose to some extent, the question may be moot, if max_size() is moderately huge, then you're likely to hit bad_alloc before length_error anyway.

[sleeping]

Searching on the SGI website, I found a bunch of exceptions in stdexcept. I'm too lazy to check them all (waiting for an infinite loop to run out of ressource is not my idea of a good time), but I have a good feeling on std::runtime_error

Would be interresting to check though.

F.