I agree that it's better to know how much space you need up front - even if it turns a one-pass algorithm into a two-pass one.

One of my happiest moments as a programmer came when reading a co-worker's code to render 2d parametric functions smoothly in screen space. It recursively divides down the parameter interval until it's small enough, then draws line segments. The code is written iteratively and manually manages a stack to simulate recursion. The stack is a fixed-size static C array instead of a vector. "Aren't you worried about overflow?", I asked. But he had computed the number of divisions in half to get from a FLT_MAX-length interval down to an FLT_MIN-length interval, and made the array big enough to hold that many steps. Goodbye malloc(), hello one not-really-that-big static array that stays warm in the cache.