Re: [PATCH] branch: allow "-" as a short-hand for "previous branch"

[Junio C Hamano <gitster@pobox.com>]

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> This patch does that:
>
> -- snip --
> diff --git a/object-name.c b/object-name.c
> index 4d2746574cd..a2732be3b71 100644
> --- a/object-name.c
> +++ b/object-name.c
> @@ -1616,6 +1616,20 @@ int repo_interpret_branch_name(struct repository *r,
>  	if (!namelen)
>  		namelen = strlen(name);
>
> +	if (namelen == 1 && *name == '-') {
> +		struct grab_nth_branch_switch_cbdata cb = {
> +			.remaining = 1,
> +			.sb = buf
> +		};
> +
> +		if (refs_for_each_reflog_ent_reverse(get_main_ref_store(r),
> +						     "HEAD",
> +						     grab_nth_branch_switch,
> +						     &cb) <= 0)
> +			return -1;
> +		return namelen;
> +	}
> +

This is very reasonable.  

Anywhere we take '@{-1}', 'main', or 'js/dash-is-previous', nobody
would be surprised if we take '-' and interpreted as '@{-1}' in
addition.

However, as I said earlier, we have command line disambiguation that
wants to tell dashed options, revs, and paths apart.  We need to
find places that need adjusting and adjust them, *AND* then make
sure that such tweaks do not introduce unintended side effect.
These, especially the last one, take a careful work I would rather
not to see unexperienced developer perform alone and take the
finding by them blindly.

> What does not work with this patch is `git show -`. I am not sure whether
> we want to make that work, although I have to admit that I would use it.
> Often. And this patch would make it work, the test suite even passes with
> it:

Exactly my above point.  Nobody including our tests expect that a
single '-' to be taken as a rev when we disambiguate command line
arguments, so it is very unlikely that it is tested to ensure that a
single '-' ends the revs and is checked for its "path-ness".  It is
not surprising that the tests do not fail ;-).

> -- snip --
> diff --git a/revision.c b/revision.c
> index f4eee11cc8b..207b554aef1 100644
> --- a/revision.c
> +++ b/revision.c
> @@ -2802,7 +2802,7 @@ int setup_revisions(int argc, const char **argv, struct rev_info *revs, struct s
>  		revarg_opt |= REVARG_CANNOT_BE_FILENAME;
>  	for (left = i = 1; i < argc; i++) {
>  		const char *arg = argv[i];
> -		if (!seen_end_of_options && *arg == '-') {
> +		if (!seen_end_of_options && *arg == '-' && arg[1]) {
>  			int opts;
>
>  			opts = handle_revision_pseudo_opt(
> -- snap --

Thanks.  These two patch snippets in the message I am responding to
are promising and exciting.