Re: [PATCH v2] diff: add --compact-summary option to complement --stat

[Jeff King <peff@peff.net>]

On Fri, Jan 19, 2018 at 07:26:28AM +0700, Duy Nguyen wrote:

> > (I know this is a bikeshed, so I'm perfectly willing to take "yuck, I
> > don't like that as well" as a response).
> 
> The position of A+x column is exactly where gerrit put it. Though web
> pages have more flexibility than our terminal console so its position
> does not have to be the same. I'm just throwing options out there
> 
> For many years I have this instead
> 
>  t/t5573-pull-verify-signatures.sh (new +x) | 81 ++++++++++++++++++++
> 
> Another option is just wrap the code in [] to make it look like check
> boxes. But that wastes two more columns
> 
>        builtin/merge.c                    |  2 +
>  [A+x] t/t5573-pull-verify-signatures.sh  | 81 ++++++++++++++++++++++++
>        t/t7612-merge-verify-signatures.sh | 45 +++++++++++++

Yeah, I almost suggested brackets, but wasn't sure if people would balk
at the extra 2 columns. But they do help it stand out more. Colors would
help, too, as you noted. Though they would not transfer over email, and
I wonder if people would want to use this for format-patch.

> Back to your suggestion, I kinda like the closeness between the +/-
> count and "|" though. The output on 10c78a162f is like this, which
> makes "A" looks a bit umm.. disconnected from the path name?
> 
>   Documentation/RelNotes/2.14.0.txt | A  97 +++++++++++++++++++++++++++
>   GIT-VERSION-GEN                   |     2 +-
>   RelNotes                          |     2 +-

Yeah, I was trying to get it away from the pathname, since it doesn't
stand out as much. I guess it depends how you think of the "A". To me it
is sensible as a modifier for the line-count change. I.e., My read on
the output above is "here is a path; it was added with 97 lines".

> > One thing I noticed is that --compact-summary by itself does nothing.
> > Should it imply --stat?
> 
> It might go with --numstat or --dirstat in future too. Didn't really
> think hard about this yet. But I probably will go with Eric suggestion
> and keep this in --stat=.... unless it really makes sense to have
> something like this in --numstat or --dirstat.

I'd think that most consumers of --numstat are not human, and would
just use "--numstat --raw" to get all the information. But I also have
not thought hard about it.

Anyway, thanks for listening. :)

-Peff