On 7/11/07, Charles Acknin <charlesacknin@gmail.com> wrote:
> It surely sounds like a good idea, but in my opinion not feasible. It
> would imply including every single object state we're manipulating
> *before* and *after* the operation. So for instance, in a remove
> directory operation, at the time of creating the patch, instead of
> just writing "remove dir foo", we would have to somehow *backup* the
> whole content of foo/, which could happen to be tremendously huge (in
> size).
>
> /usr/bin/patch does this, but since it only operates on
> contextual-diffs, that's way easier task to achieve. Reversing
> directories and binary-files is a luxury, and as expected, expensive.
The old version is sitting there in version control. Does it make any
sense to have the patch hold the URL/revision information of the file
being deleted? The you could conceive of reversing the patch
deterministically and efficiently.
Of course, this information only exists for a patch that has been
committed and is only useful for repositories that you have access to.
Just a thought...
-Dan C (king of feature creep)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Jul 11 18:44:22 2007