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

Michael G Schwern schwern at pobox.com
Thu Mar 31 10:38:18 UTC 2005

On Thu, Mar 31, 2005 at 09:16:15AM +0200, Karel Gardas wrote:
> > * "darcs diff some/file" should only report on changes to affect some/file,
> >   not the entire repository.  This was the source of 90% of my frustration.
> >   --last 1, for example, showing me the last change to the repository instead
> >   of the last change to some/file.
> Could you be so kind and enter appropriate ticket in darcs bug database?


> > * The lack of simple revision numbers really complicates what would be a
> >   very simple task in most other version control systems.
> Is adding complicated revision numbers to the darcs just for helping with

How did the revision numbers become complicated?

> diffing worth the effort? I personally consider --to-* and --from-*
> options enough, perhaps there should be --to-date and --from-date added,
> but this probably depends on next item.

I don't know if its worth the effort to add them because I don't know what
the effort would be.  

I mention it because, well... its the truth.  Its a lot easier to remember 
and type in "3214" than "a TODO test which confirms that RT#266 is a 
regression" or "Sat Mar 12 14:08:46 PST 2005".  If I was communicating with 
a human the dates and names might be easier because I can be inexact, though 
I've worked in shops where we casually said things like "who's working on 39?".
In order to usefully ask darcs "show me the diff since patch X" I have to
remember the name or date pretty exactly.

Not to mention what happens when you have two patches with the same name such
as "Bug fix" or "Typo".  They're bad names but it happens.  A lot.

There's also less to remember in the short term.  When I'm looking for a
particular change I first search the log of change descriptions for the
one in question then I run a diff using the identifier of that change.
If the identifier is 2391 that's easy to just look at and type in.  If its
"Allow parens to be optional when calling darcs() test command in some cases."
well I'm going to have to highlight, copy and paste that.  Its a bit more
work but enough to be annoying if you're used to doing less work.

Finally, I also mention it because its not only diff that would find this
useful.  annotate is also crippled because of a lack of a simple revision
number, there's a discussion going on about the annotate format right now.

One thing the annotate discussion has revealed, its not easy to stick a
number on each change in darcs.

> > * "fancy dates" such as "last month" and "last year" should work if
> >   not-so-fancy dates like "last week" work.
> IIRC there is a ticket in darcs bug db asking for writting more advanced
> fancy date parser.

Can't find one.

> > This was most surprising.  Why is diff giving me a change log entry and no
> > diff?
> It is diffing against the last revision (not last revision in which
> appropriate file was changed) in which probably your file has not been
> changed, so it just show change log for this revision. In this case it
> would be good if darcs shows nothing when there is no diff for the output.


More information about the darcs-users mailing list