[libcxx-commits] [PATCH] D74290: [libcxx] Qualify make_move_iterator in vector::insert for input iterators

[Logan Smith via Phabricator via libcxx-commits]

logan-5 added a comment.

In D74290#1865949 <https://reviews.llvm.org/D74290#1865949>, @Quuxplusone wrote:

> I don't understand the actual algorithm being used in `insert`-with-input-iterators


Seems reasonable enough to me:

1. insert up to capacity at the end of the vector
2. construct any additional elements that would overflow capacity in a scratch buffer, then see how big it is (because input iterators), and reserve new capacity (if exception is thrown, destroy items from step 1--easy since they're still sitting at the end)
3. rotate items from step 1 into place, and then insert() the scratch buffer after those

I'm not sure why you'd insert() the scratch buffer instead of just moving its elements onto the vector's end and then rotating all the elements into place at once, but maybe the insert() overload called for the scratch buffer does things more efficiently somehow.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74290/new/

https://reviews.llvm.org/D74290