Revert implmentation and how to undo a bzr remove

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Benno wrote:
> On Wed Apr 27, 2005 at 19:52:35 +1000, Aaron Bentley wrote:
>>I should mention that the merge code can do revert (or switch, update,
>>etc).  It can be restricted to operate only on particular files.
> 
> 
> In your opinion is it better piggy backing on that (which I could understand
> to actually be quite complicated), or use a simple implementation to just
> do revert, which is actually quite simple, since you don't need to handle
> any actual merging.


My preference is for the smallest possible set of core primitives to
achieve the maximum functionality.  So using merge for revert, even
though it's way too much tool for the job, appeals to my sensabilities.
 Given that we need merge code already, it reduces the amount of code to
maintain.  It's not hard to do-- at the command line, it's just "merge
./@ ."  And it already handles IDs properly.

On the other hand, the merge code *is* quite complicated.  I agree that
there are virtues to a simple revert command.  For one thing, it's more
readable.  For another, it's easier for Martin to verify that it's
implemented properly and merge it.

Anyhow, I'm not the guy who makes the call on this.

Aaron