In workflows that a single working tree, like co-located branches, sometimes you want to switch while you have uncommitted changes. By default, switch will apply your uncommitted changes to the new branch that you switch to. But often you don’t want that. You just want to do some work in the other branch, and eventually return to this branch and work some more.
You could run bzr shelve --all before switching, to store the changes safely. So you have to know that there are uncommitted changes present, and you have to remember to run bzr shelve --all. Then when you switch back to the branch, you need to remember to unshelve the changes, and you need to know what their shelf-id was.
Using switch --store takes care of all of this for you. If there are any uncommitted changes in your tree, it stores them in your branch. It then restores any uncommitted changes that were stored in the branch of your target tree. It’s almost like having two working trees and using cd to switch between them.
To take an example, first we’d set up a co-located branch:
$ bzr init foo Created a standalone tree (format: 2a) $ cd foo $ bzr switch -b foo
Now create committed and uncommitted changes:
$ touch committed $ bzr add adding committed $ bzr commit -m "Add committed" Committing to: /home/abentley/sandbox/foo/ added committed Committed revision 1. $ touch uncommitted $ bzr add adding uncommitted $ ls committed uncommitted
Now create a new branch using --store. The uncommitted changes are stored in “foo”, but the committed changes are retained.
$ bzr switch -b --store bar Uncommitted changes stored in branch "foo". Tree is up to date at revision 1. Switched to branch: /home/abentley/sandbox/foo/ abentley@speedy:~/sandbox/foo$ ls committed
Now, create uncommitted changes in “bar”:
$ touch uncommitted-bar $ bzr add adding uncommitted-bar
Finally, switch back to “foo”:
$ bzr switch --store foo Uncommitted changes stored in branch "bar". Tree is up to date at revision 1. Switched to branch: /home/abentley/sandbox/foo/ $ ls committed uncommitted
Each branch holds only one set of stored changes. If you try to store a second set, you get an error. If you use --store all the time, this can’t happen. But if you use plain switch, then it won’t restore the uncommitted changes already present:
$ bzr switch bar Tree is up to date at revision 1. Switched to branch: /home/abentley/sandbox/foo/ $ bzr switch --store foo bzr: ERROR: Cannot store uncommitted changes because this branch already stores uncommitted changes.
If you’re working in a branch that already has stored changes, you can restore them with bzr switch . --store:
$ bzr shelve --all -m "Uncommitted changes from foo" Selected changes: -D uncommitted Changes shelved with id "1". $ bzr switch . --store Tree is up to date at revision 1. Switched to branch: /home/abentley/sandbox/foo/ $ ls committed uncommitted-bar