[darcs-users] Re: Frustrations diffing against the last change to a file

[Tommy Pettersson]

On Sat, Apr 02, 2005 at 01:44:26AM -0800, Michael G Schwern wrote:
> On Sat, Apr 02, 2005 at 10:08:32AM +0200, Thomas Zander wrote:
> > Adding revision numbers to a system that is designed in such a way that 
> > revision numbers are inherently unneeded an counter productive is not an 
> > improvement, I'd call it a step back.
> > This has been said and argued; and all you keep doing is saying you don't 
> > want to copy paste and you want to keep naming your patches "fix". 
> 
> If I was the only one arguing for revision numbers then I might say, "Ok,
> you've had this discussion before and concluded its not worth while."  I'm
> new around here and would be willing to accept that.  Maybe if there was
> some sort of FAQ on the subject.
> 
> However, so far you're the only one saying revision numbers are unneeded.
> In fact a lot of the replies have been people saying that revision numbers
> might be helpful.  That they, too, are having trouble with the current system
> of patch naming.  Also, over in the annotate format thread, it was realized
> that having a short, unique way to identify a patch would allow for a much
> more concise format.
> 
> Also, Thomas, I took a look at you.  Your role so far in these discussions
> has been largely negative.  "If you used darcs more you'd realize you don't 
> need that" has been the tone of your replies and generally not constructive.  
> I had assumed you've been around for a while and were an important 
> contributor.  You've been on darcs-users since December 2004, four months.  
> You have contributed six patches to darcs, same number as I have.  So I'm
> not going to stop the discussion on your say so.  If David or Mark said,
> "enough", that would be different.
> 
> Finally, another distributed version control system which I use, SVK, does 
> have revision numbers.  And they work pretty well over there.
> 
> So I'm going to continue to poke at this issue.  While patch names do work,
> there is value in continuing to optimize and improve the UI.  I don't know
> if revision numbers are doable for darcs, but I certainly won't know if I
> don't try to discuss it.

First, I'm not opposed to make it easier to refer to patches
in a repo.  But I agree with Thomas that "revision numbers"
is wrong.  They will make you think the wrong way.

I came from CVS to darcs, and I remember I was very
uncomfortable with history browsing at first.  I thought it
was awkward to find the "right place" in the changes output
and then refer to it with the name of a patch.

Now I'm a bit dissatisfied the other way around.  --from-patch
and --to-patch works on ranges in the quite arbitrary order
of patches that may or may not contain the patches I'm
interested in.

A diff from one point to another doesn't say half as much as
the patches that make the changes, but it takes a while to
were off the habit of always trying to get a diff to examine.


Anyway...

It would mostly be querying commands that need a shorthand
patch reference, since the other commands let you select
patches interactively.  Why not number the patches backwards
so that the last applied patch (the first one in a normal
changes listing) is 1.

Pros:
  you will usually type rather short references
  you will be quite aware that references are temporary

Cons:
  it won't work in repos that others modify simultaneously




-- 
Tommy Pettersson <ptp at lysator.liu.se>