[RFC] _show_log supporting unmerged revisions (first step towards lp #4663)

[Wouter van Heyst]

On Sun, Jun 18, 2006 at 01:04:13PM -0500, John Arbash Meinel wrote:

...

> > I often make use of that, will try not to in bzr code. The function in
> > question is dead code in this instance of the patch, I removed
> > find_merging and inlined last_mainline_ancestor now that there is only
> > one user.
> 
> I'm not specifically against it, just didn't know about it before.

Ah.

> >> In general, the above seems to be a little convoluted and hard to
> >> follow. It would be nice if we could get a simpler definition, so that
> >> it can be maintained.
> > 
> > I agree the above functions are convoluted, I tried getting rid of them.
> > How bad is the rest of the patch? I dislike the cutting out of revisions
> > with the sorted_merge loop, nor the zero_depths calculation.
> 
> The zero_depth is because log formatters 'short' and 'line' only display
> the mainline. I think we need to keep something like that in, though if
> you look at my message to Aaron, he might be doing it differently.

Oh quite, I meant the calculation of the amount of them, so as to be
able to correctly label with revision numbers non-mainline revisions in
the reversed case. Forwards we could just increase revno in the loop,
but alas.

    zerodepths = sum(1 for pack in merge_sorted_revisions if pack[2] == 0)

> 
> I didn't really look at all of your steps, so I'll wait for your new
> patch and review again.

Thanks for reviewing by the way, I also still need to do cmd_log and
switch to Aarons log work as a basis, so you can wait for that if you
want.

> Just be aware that Aaron is also working on some log improvements. So
> you should work closely with him so that you don't step on eachother's toes.

I didn't spot an obvious branch containing his work, I'm now looking at
http://panoramicfeedback.com/opensource/bzr/repo/log.perf.submit/

Wouter van Heyst