iter_changes, delta consistency and revert

[Robert Collins]

Aaron, please take my prior mail with a grain of salt. I was feeling
unhappy because I've put a lot of work into this patch and I didn't like
the idea of having nontrivial alterations to make - but of course, it
was up to me to circulate the ideas I was considering earlier, and I
didn't, or didn't do so widely enough.

I want to improve our underlying API's but not in a way that will make
higher level code harder to write or understand.

I am proposing one key change to iter_changes:
 - always give enough data to be make consistent deltas from in the
direction of target

Do you have any thoughts about whether this will really cause issues for
things you're building (or may want to build) on top iter_changes? If
its going to cause problems, I'll have to do it differently.

Clearly revert is going to suffer some fallout in the current form, but
I think that that is a special case rather than generally expected
fallout. I've got no real data supporting this thought though :(

I am concerned that we have an overly sharp API in the current form of
iter_changes, not because the API is bad, but because the expansion that
is needed is not easy to accomplish outside the implementation of
iter_changes, and is nearly always needed. (There is for example a bug
open - which my branch fixes - that 'bzr a b; echo foo >> b/c; bzr st
b/c' doesn't report the a->b rename).

-Rob
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20090807/d3d6f098/attachment.pgp