Code review comment for lp:~abentley/bzr/fix_get_mtime

Creating a path<->id map is expensive on large trees; I've found significant benefits in commit etc when enough layers are pushed through to avoid this; not passing in the path will cause that map to be created. My understanding from Martin is that we want to be heading across the board to proving path data as much as possible - to avoid unnecessary work.

I'm not sure why its complete for TreeTransform to calculate the path, surely its just a walk of transaction parents ?