Re: [PATCH] log: add %S option (like --source) to log --format

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

Issac Trotts <issac.trotts@gmail.com> writes:

> Make it possible to write for example
>
>         git log --format="%H,%S"
>
> where the %S at the end is a new placeholder that prints out the ref
> (tag/branch) for each commit.
>
> Using %d might seem like an alternative but it only shows the ref for
> the last commit in the branch.

Have your sign-off here (see Documentation/SubmittingPatches),

and then have a line that only has three-dashes on it,

and then any other additional info like this.

>
> This change is based on a question from Stack Overflow:
> https://stackoverflow.com/questions/12712775/git-get-source-information-in-format
> ---




>  Documentation/pretty-formats.txt |  2 ++
>  builtin/log.c                    |  2 +-
>  log-tree.c                       |  1 +
>  pretty.c                         | 10 +++++++
>  t/t4205-log-pretty-formats.sh    | 50 ++++++++++++++++++++++++++++++++
>  5 files changed, 64 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/pretty-formats.txt b/Documentation/pretty-formats.txt
> index 417b638cd..acfe7e0a4 100644
> --- a/Documentation/pretty-formats.txt
> +++ b/Documentation/pretty-formats.txt
> @@ -104,6 +104,8 @@ The placeholders are:
>
>  - '%H': commit hash
>  - '%h': abbreviated commit hash
> +- '%S': ref name given on the command line by which the commit was reached
> +  (like `git log --source`)

This looks entirely out of place, among the description of the basic
basic data like the object name of the commit itself, its tree, etc.

Describe this immediately after %d and %D are explained, perhaps.

> diff --git a/builtin/log.c b/builtin/log.c
> index e8e51068b..a314ea2b6 100644
> --- a/builtin/log.c
> +++ b/builtin/log.c
> @@ -203,7 +203,7 @@ static void cmd_log_init_finish(int argc, const
> char **argv, const char *prefix,
>          rev->diffopt.filter || rev->diffopt.flags.follow_renames)
>          rev->always_show_header = 0;
>
> -    if (source) {
> +    if (source || (rev->pretty_given && rev->commit_format ==
> CMIT_FMT_USERFORMAT)) {

Broken line (you let your MUA line-wrap and corrupt your patch???)

>          init_revision_sources(&revision_sources);
>          rev->sources = &revision_sources;
>      }

This means anybody who asks for say --format='%aN %s' pays the price
of keeping track of each commit's source, which is unreasonable.

Perhaps mimick the way how presence of %N is detected in
userformat_want_item() so that we do not pay the price for
init_display_notes() when a format that does not care about %N is
given?

> @@ -1149,6 +1150,15 @@ static size_t format_commit_one(struct strbuf
> *sb, /* in UTF-8 */
>          parse_object(the_repository, &commit->object.oid);
>
>      switch (placeholder[0]) {
> +    case 'S':        /* tag/branch like --source */
> +        slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
> +        if (slot && *slot) {
> +            strbuf_addstr(sb, *slot);
> +            return 1;
> +        } else {
> +            die(_("failed to get info for %%S"));
> +            return 0;
> +        }

Have this next to case arms that deal with 'd' and 'D".

>      case 'H':        /* commit hash */
>          strbuf_addstr(sb, diff_get_color(c->auto_color, DIFF_COMMIT));
>          strbuf_addstr(sb, oid_to_hex(&commit->object.oid));

This is a tangent, but I think this existing case arm should move to
the previous block before parsing the commit, as %H only needs the
object name of the commit itself (reword the comment before that
switch to read from "independent of the commit" to "computable
without parsing the commit")..