log: add %S option (like --source) to log --format

PATCH In review [email protected] 4 participants 7 messages Dec 19, 2018 – Dec 31, 2018 Documentation

lore

Share & tools

lore

[email protected]

author

2018-12-19 08:33 UTC #0 lore raw

PATCH 1/1 log: add %S option (like --source) to log --format

From: Issac Trotts <[email protected]>

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.

Signed-off-by Issac Trotts <[email protected]>

---

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                         | 15 ++++++++++
 pretty.h                         |  1 +
 t/t4205-log-pretty-formats.sh    | 50 ++++++++++++++++++++++++++++++++
 6 files changed, 70 insertions(+), 1 deletion(-)

Documentation/pretty-formats.txt +2 -0

index 417b638cd..de6953108 100644
@@ -134,6 +134,8 @@ The placeholders are:
 - '%cI': committer date, strict ISO 8601 format
 - '%d': ref names, like the --decorate option of linkgit:git-log[1]
 - '%D': ref names without the " (", ")" wrapping.
+- '%S': ref name given on the command line by which the commit was reached
+  (like `git log --source`), only works with `git log`
 - '%e': encoding
 - '%s': subject
 - '%f': sanitized subject line, suitable for a filename

builtin/log.c +1 -1

index e8e51068b..be3025657 100644
@@ -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) && w.source)) {
 		init_revision_sources(&revision_sources);
 		rev->sources = &revision_sources;
 	}

log-tree.c +1 -0

index 10680c139..3cb14256e 100644
@@ -700,6 +700,7 @@ void show_log(struct rev_info *opt)
 	ctx.color = opt->diffopt.use_color;
 	ctx.expand_tabs_in_log = opt->expand_tabs_in_log;
 	ctx.output_encoding = get_log_output_encoding();
+	ctx.rev = opt;
 	if (opt->from_ident.mail_begin && opt->from_ident.name_begin)
 		ctx.from_ident = &opt->from_ident;
 	if (opt->graph)

pretty.c +15 -0

index b83a3ecd2..258e86e9c 100644
@@ -1084,6 +1084,7 @@ static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
 	struct commit_list *p;
 	const char *arg;
 	int ch;
+	char **slot;
 
 	/* these are independent of the commit */
 	switch (placeholder[0]) {

@@ -1194,6 +1195,17 @@ static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
 		load_ref_decorations(NULL, DECORATE_SHORT_REFS);
 		format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
 		return 1;
+	case 'S':		/* tag/branch like --source */
+		if (c->pretty_ctx->rev->sources == NULL) {
+			return 0;
+		}
+		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"));
+		}
 	case 'g':		/* reflog info */
 		switch(placeholder[1]) {
 		case 'd':	/* reflog selector */

@@ -1498,6 +1510,9 @@ static size_t userformat_want_item(struct strbuf *sb, const char *placeholder,
 	case 'N':
 		w->notes = 1;
 		break;
+	case 'S':
+		w->source = 1;
+		break;
 	}
 	return 0;
 }

pretty.h +1 -0

index 7359d318a..87ca5dfcb 100644
@@ -60,6 +60,7 @@ static inline int cmit_fmt_is_mail(enum cmit_fmt fmt)
 
 struct userformat_want {
 	unsigned notes:1;
+	unsigned source:1;
 };
 
 /* Set the flag "w->notes" if there is placeholder %N in "fmt". */

t/t4205-log-pretty-formats.sh +50 -0

54 lines not on this page · raw message

Signature · 1 line

2.19.1

Jeff King

2018-12-22 22:22 UTC #2 lore raw

Quoted 1 line · Issac Trotts

Hi all, friendly ping. Is there still interest in merging this patch?

Yes, but I think people are largely absent for holidays at this point. If there hasn't been any movement on it, I'll try to review after Jan 5th (it also wouldn't hurt to re-post around then).

-Peff

Issac Trotts

author

2018-12-25 02:12 UTC #3 lore raw

Got it, sounds good.

Quoted 10 lines · Jeff King

Quoted 1 line · Issac Trotts
Hi all, friendly ping. Is there still interest in merging this patch?
Yes, but I think people are largely absent for holidays at this point. If there hasn't been any movement on it, I'll try to review after Jan 5th (it also wouldn't hurt to re-post around then).
-Peff

Derrick Stolee

2018-12-27 13:20 UTC #4 lore raw

In reply to [email protected]

Quoted 32 lines · On 12/19/2018 3:33 AM, [email protected] wrote:

From: Issac Trotts <[email protected]>
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.
Signed-off-by Issac Trotts <[email protected]>
---
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                         | 15 ++++++++++
  pretty.h                         |  1 +
  t/t4205-log-pretty-formats.sh    | 50 ++++++++++++++++++++++++++++++++
  6 files changed, 70 insertions(+), 1 deletion(-)
Documentation/pretty-formats.txt +0 -0
index 417b638cd..de6953108 100644

Documentation/pretty-formats.txt The placeholders are: · line 138

Quoted 5 lines

Documentation/pretty-formats.txt +2 -0

@@ -134,6 +134,8 @@ The placeholders are:
  - '%cI': committer date, strict ISO 8601 format
  - '%d': ref names, like the --decorate option of linkgit:git-log[1]
  - '%D': ref names without the " (", ")" wrapping.
+- '%S': ref name given on the command line by which the commit was reached
+  (like `git log --source`), only works with `git log`

This "only works with git log" made me think about what would happen with git rev-list --pretty=format:"%h %S" and the answer (on my machine) was a segfault.

Quoted 8 lines

Documentation/pretty-formats.txt +0 -0

  - '%e': encoding
  - '%s': subject
  - '%f': sanitized subject line, suitable for a filename

builtin/log.c +0 -0

index e8e51068b..be3025657 100644

builtin/log.c static void cmd_log_init_finish(int argc, const char **argv, const char *prefix, · line 209

Quoted 8 lines

builtin/log.c +1 -1

@@ -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) && w.source)) {
  		init_revision_sources(&revision_sources);
  		rev->sources = &revision_sources;
  	}

Likely, you'll want to duplicate this initialization in the revision machinery. Keep this one in builtin/log.c as it was before, but add something like this initialization to setup_revisions(). Add a test for rev-list.

[snip]

builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1199

Quoted 6 lines

builtin/log.c +2 -0

@@ -1194,6 +1195,17 @@ static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
  		load_ref_decorations(NULL, DECORATE_SHORT_REFS);
  		format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
  		return 1;
+	case 'S':		/* tag/branch like --source */
+		if (c->pretty_ctx->rev->sources == NULL) {

Use "if (!c->pretty_ctx->rev->sources".

builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1203

Quoted 4 lines

builtin/log.c +4 -0

+			return 0;
+		}
+		slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
+		if (slot && *slot) {

I'm not sure this check for 'slot' being non-null is necessary, as we would already get a failure in the commit-slab code (for revision_sources_at()) if the slab is not initialized.

builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1207

Quoted 4 lines

builtin/log.c +4 -0

+			strbuf_addstr(sb, *slot);
+			return 1;
+		} else {
+			die(_("failed to get info for %%S"));

Here, you die() when you fail to get a slot but above you return 0 when the sources are not initialized.

I don't see another use of die() in this method. Is that the right way to handle failure here? (I'm legitimately asking because I have over-used 'die()' in the past and am still unclear on when it is appropriate.)

Quoted 43 lines

builtin/log.c +41 -0

  '
 
+test_expect_success 'set up %S tests' '
+	git checkout --orphan source-a &&
+	test_commit one &&
+	test_commit two &&
+	git checkout -b source-b HEAD^ &&
+	test_commit three
+'
+
+test_expect_success 'log --format=%S paints branch names' '
+	cat >expect <<-\EOF &&
+	source-b
+	source-a
+	source-b
+	EOF
+	git log --format=%S source-a source-b >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'log --format=%S paints tag names' '
+	git tag -m tagged source-tag &&
+	cat >expect <<-\EOF &&
+	source-tag
+	source-a
+	source-tag
+	EOF
+	git log --format=%S source-tag source-a >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'log --format=%S paints symmetric ranges' '
+	cat >expect <<-\EOF &&
+	source-b
+	source-a
+	EOF
+	git log --format=%S source-a...source-b >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success '%S in git log --format works with other placeholders (part 1)' '
+	git log --format="source-b %h" source-b >expect &&
+	git log --format="%S %h" source-b >actual &&

builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */

Quoted 10 lines

builtin/log.c +9 -0

+	test_cmp expect actual
+'
+
+test_expect_success '%S in git log --format works with other placeholders (part 2)' '
+	git log --format="%h source-b" source-b >expect &&
+	git log --format="%h %S" source-b >actual &&
+	test_cmp expect actual
+'
+
  test_done

Please also add a simple test to t6006-rev-list-format.sh.

Thanks,

-Stolee

Junio C Hamano

2018-12-28 18:14 UTC #5 lore raw

Quoted 11 lines · Derrick Stolee

Documentation/pretty-formats.txt The placeholders are: · line 138
Quoted 7 lines
+++ b/Documentation/pretty-formats.txt
Documentation/pretty-formats.txt +2 -0
@@ -134,6 +134,8 @@ The placeholders are:
  - '%cI': committer date, strict ISO 8601 format
  - '%d': ref names, like the --decorate option of linkgit:git-log[1]
  - '%D': ref names without the " (", ")" wrapping.
+- '%S': ref name given on the command line by which the commit was reached
+  (like `git log --source`), only works with `git log`
This "only works with git log" made me think about what would happen with git rev-list --pretty=format:"%h %S" and the answer (on my machine) was a segfault.

That's a bad one X-<.

Quoted 13 lines

pretty.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1203
Quoted 2 lines
pretty.c +2 -0
+		slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
+		if (slot && *slot) {
I'm not sure this check for 'slot' being non-null is necessary, as we would already get a failure in the commit-slab code (for revision_sources_at()) if the slab is not initialized.
pretty.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1207
Quoted 4 lines
pretty.c +4 -0
+			strbuf_addstr(sb, *slot);
+			return 1;
+		} else {
+			die(_("failed to get info for %%S"));
Here, you die() when you fail to get a slot but above you return 0 when the sources are not initialized.

Quoted 4 lines

I don't see another use of die() in this method. Is that the right way to handle failure here? (I'm legitimately asking because I have over-used 'die()' in the past and am still unclear on when it is appropriate.)

This is definitely a bad one, too. If '%d' cannot find decoration, it would not "die".

Thanks.

Issac Trotts

author

2018-12-31 04:35 UTC #6 lore raw

In reply to Derrick Stolee

Hi Derrick, thanks for your feedback.

I'll send out a new patch with the changes.

Quoted 29 lines · Derrick Stolee

Quoted 27 lines · On 12/19/2018 3:33 AM, [email protected] wrote:
From: Issac Trotts <[email protected]>
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.
Signed-off-by Issac Trotts <[email protected]>
---
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                         | 15 ++++++++++
  pretty.h                         |  1 +
  t/t4205-log-pretty-formats.sh    | 50 ++++++++++++++++++++++++++++++++
  6 files changed, 70 insertions(+), 1 deletion(-)

Quoted 14 lines

Documentation/pretty-formats.txt The placeholders are: · line 138
Quoted 10 lines
Documentation/pretty-formats.txt +2 -0
index 417b638cd..de6953108 100644
@@ -134,6 +134,8 @@ The placeholders are:
  - '%cI': committer date, strict ISO 8601 format
  - '%d': ref names, like the --decorate option of linkgit:git-log[1]
  - '%D': ref names without the " (", ")" wrapping.
+- '%S': ref name given on the command line by which the commit was reached
+  (like `git log --source`), only works with `git log`
This "only works with git log" made me think about what would happen with git rev-list --pretty=format:"%h %S" and the answer (on my machine) was a segfault.

Good find. Checking for c->pretty_ctx->rev == NULL prevents the seg fault.

Quoted 11 lines

Documentation/pretty-formats.txt +0 -0

  - '%e': encoding
  - '%s': subject
  - '%f': sanitized subject line, suitable for a filename

builtin/log.c +0 -0

index e8e51068b..be3025657 100644
@@ -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;

Quoted 10 lines

builtin/log.c static void cmd_log_init_finish(int argc, const char **argv, const char *prefix, · line 209
Quoted 5 lines
builtin/log.c +1 -1
-     if (source) {
+     if (source || (rev->pretty_given && (rev->commit_format == CMIT_FMT_USERFORMAT) && w.source)) {
              init_revision_sources(&revision_sources);
              rev->sources = &revision_sources;
      }
Likely, you'll want to duplicate this initialization in the revision machinery. Keep this one in builtin/log.c as it was before, but add something like this initialization to setup_revisions(). Add a test for rev-list.

Okay, I added a test to check that rev-list leaves %S alone and implicitly that it doesn't segfault.

I'd like to limit the scope of this change to what's already here, with the check preventing the segfault, if that's okay. Could support for %S in rev-list be in a later patch?

Quoted 22 lines

[snip]
builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1199
Quoted 6 lines
builtin/log.c +2 -0
@@ -1194,6 +1195,17 @@ static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */
              load_ref_decorations(NULL, DECORATE_SHORT_REFS);
              format_decorations_extended(sb, commit, c->auto_color, "", ", ", "");
              return 1;
+     case 'S':               /* tag/branch like --source */
+             if (c->pretty_ctx->rev->sources == NULL) {
Use "if (!c->pretty_ctx->rev->sources".
builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1203
Quoted 4 lines
builtin/log.c +4 -0
+                     return 0;
+             }
+             slot = revision_sources_at(c->pretty_ctx->rev->sources, commit);
+             if (slot && *slot) {
I'm not sure this check for 'slot' being non-null is necessary, as we would already get a failure in the commit-slab code (for revision_sources_at()) if the slab is not initialized.
builtin/log.c static size_t format_commit_one(struct strbuf *sb, /* in UTF-8 */ · line 1207
Quoted 4 lines
builtin/log.c +4 -0
+                     strbuf_addstr(sb, *slot);
+                     return 1;
+             } else {
+                     die(_("failed to get info for %%S"));

Quoted 7 lines

Here, you die() when you fail to get a slot but above you return 0 when the sources are not initialized.
I don't see another use of die() in this method. Is that the right way to handle failure here? (I'm legitimately asking because I have over-used 'die()' in the past and am still unclear on when it is appropriate.)

Fixed.

Quoted 48 lines

builtin/log.c +45 -0

  '
 
+test_expect_success 'set up %S tests' '
+     git checkout --orphan source-a &&
+     test_commit one &&
+     test_commit two &&
+     git checkout -b source-b HEAD^ &&
+     test_commit three
+'
+
+test_expect_success 'log --format=%S paints branch names' '
+     cat >expect <<-\EOF &&
+     source-b
+     source-a
+     source-b
+     EOF
+     git log --format=%S source-a source-b >actual &&
+     test_cmp expect actual
+'
+
+test_expect_success 'log --format=%S paints tag names' '
+     git tag -m tagged source-tag &&
+     cat >expect <<-\EOF &&
+     source-tag
+     source-a
+     source-tag
+     EOF
+     git log --format=%S source-tag source-a >actual &&
+     test_cmp expect actual
+'
+
+test_expect_success 'log --format=%S paints symmetric ranges' '
+     cat >expect <<-\EOF &&
+     source-b
+     source-a
+     EOF
+     git log --format=%S source-a...source-b >actual &&
+     test_cmp expect actual
+'
+
+test_expect_success '%S in git log --format works with other placeholders (part 1)' '
+     git log --format="source-b %h" source-b >expect &&
+     git log --format="%S %h" source-b >actual &&
+     test_cmp expect actual
+'
+
+test_expect_success '%S in git log --format works with other placeholders (part 2)' '

Quoted 8 lines

builtin/log.c
Quoted 6 lines
builtin/log.c +5 -0
+     git log --format="%h source-b" source-b >expect &&
+     git log --format="%h %S" source-b >actual &&
+     test_cmp expect actual
+'
+
  test_done
Please also add a simple test to t6006-rev-list-format.sh.

Done.

Quoted 5 lines

Thanks,
-Stolee