27 April 2009

Revert by Commit

Here is a sequence that happens to me regularly:
  1. Have a theory about how to fix something
  2. Try it out
  3. Find out it was a stupid theory
  4. Realize that there is a kernel of learning in what you tried
  5. Realize that you don't want to throw away that kernel of learning quite yet
  6. But want to get the bad changes out of the way
  7. But there are some good uncommitted changes you don't want to throw away
  8. Ack, what should I do now?
Here is a recipe that makes it really easy to save the bad experiment *and* get it out of the way in ONE step.
  • git checkout -b experiment-that-stunk
  • git add -i (pick the one or two files that contain the changes you want to get rid of)
  • git commit -m "tried to do something stupid, but it didn't work because ..."
  • git checkout stuff-i-was-working-on
Revert by commit. Now the changes are gone, saved, explained, pickled for fermentation for more ideas or conversation.

That is something that was NOT part of my experience before git.

Actually, I take that back, I did this once or twice before, but I wasted lots of time because I resisted give up on the experiment at the first sign of stupidity -- I subconciously knew it would be painful to save a whole patch, partially revert, hope I got it right, forget what I was doing, mess it up, revert all the way, re-apply the patch, repeat. So it took longer to abort the bad experiment because of the hesitation because of the perceived expense of partial revert.


I just realized this must be how "git stash" is implemented, and that it's definitely not a new idea.

No comments:

Post a Comment