I find myself working this way almost all the time. In the middle of a refactoring or other change, I try as much as possible to get it commited by the end of

I find myself working this way almost all the time. In the middle of a refactoring or other change, I try as much as possible to get it commited by the end of the day, in a runnable state, even if accessing the affected unfinished area will crash. Better that than to have to walk in the next morning and try to remember what portion of the code change is actually finished. In a way this is like looking at your changes as atomic operations: If a change is so big you can't get yourself back into a runnable state by the end of one day coding, it probably should be broken up into smaller stable steps.