[PATCH 0/3] "log --stdin" updates

[PATCH 0/3] "log --stdin" updates

[Junio C Hamano]

These three come on top of 61ab67a (Teach --stdin option to "log" family,
2009-11-03), which gave "--stdin" to the log family of commands (e.g. log,
rev-list).  The earlier patch allowed you to feed only the revs (which was
what gitk originally wanted), but after zero or more revs (one rev per
line), you can now feed a line that consists of "--" and then pathspecs
(one path per line).

The whole series probably need to be updated to learn -z option to read
NUL terminated input.

Junio C Hamano (3):
  read_revision_from_stdin(): use strbuf
  setup_revisions(): do not call get_pathspec() too early
  Make --stdin option to "log" family read also pathspecs

 revision.c |   95 +++++++++++++++++++++++++++++++++++++++++++++++++++--------
 1 files changed, 82 insertions(+), 13 deletions(-)

[PATCH 1/3] read_revision_from_stdin(): use strbuf

[Junio C Hamano]

It is so 2005 (and Linus ;-) to have a fixed 1000-byte buffer that
reads from the user.  Let's use strbuf to unlimit the input length.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 revision.c |   18 ++++++++++--------
 1 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/revision.c b/revision.c
index 45c5de8..f5b735f 100644
--- a/revision.c
+++ b/revision.c
@@ -955,19 +955,21 @@ int handle_revision_arg(const char *arg, struct rev_info *revs,
 
 static void read_revisions_from_stdin(struct rev_info *revs)
 {
-	char line[1000];
+	struct strbuf sb;
 
-	while (fgets(line, sizeof(line), stdin) != NULL) {
-		int len = strlen(line);
-		if (len && line[len - 1] == '\n')
-			line[--len] = '\0';
+	strbuf_init(&sb, 1000);
+	while (strbuf_getwholeline(&sb, stdin, '\n') != EOF) {
+		int len = sb.len;
+		if (len && sb.buf[len - 1] == '\n')
+			sb.buf[--len] = '\0';
 		if (!len)
 			break;
-		if (line[0] == '-')
+		if (sb.buf[0] == '-')
 			die("options not supported in --stdin mode");
-		if (handle_revision_arg(line, revs, 0, 1))
-			die("bad revision '%s'", line);
+		if (handle_revision_arg(sb.buf, revs, 0, 1))
+			die("bad revision '%s'", sb.buf);
 	}
+	strbuf_release(&sb);
 }
 
 static void add_grep(struct rev_info *revs, const char *ptn, enum grep_pat_token what)
-- 
1.6.5.3.342.g14bb9

[PATCH 2/3] setup_revisions(): do not call get_pathspec() too early

[Junio C Hamano]

This is necessary because in the next patch I'll allow pathspecs to be fed
from the standard input, and pathspecs taken from the command line (and
converted via get_pathspec() already) in revs->prune_data too early gets
in the way when we want to append from the standard input.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 revision.c |    9 ++++++---
 1 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/revision.c b/revision.c
index f5b735f..4410a45 100644
--- a/revision.c
+++ b/revision.c
@@ -1230,6 +1230,7 @@ void parse_revision_opt(struct rev_info *revs, struct parse_opt_ctx_t *ctx,
 int setup_revisions(int argc, const char **argv, struct rev_info *revs, const char *def)
 {
 	int i, flags, left, seen_dashdash, read_from_stdin;
+	const char **prune_data = NULL;
 
 	/* First, search for "--" */
 	seen_dashdash = 0;
@@ -1240,7 +1241,7 @@ int setup_revisions(int argc, const char **argv, struct rev_info *revs, const ch
 		argv[i] = NULL;
 		argc = i;
 		if (argv[i + 1])
-			revs->prune_data = get_pathspec(revs->prefix, argv + i + 1);
+			prune_data = argv + i + 1;
 		seen_dashdash = 1;
 		break;
 	}
@@ -1321,12 +1322,14 @@ int setup_revisions(int argc, const char **argv, struct rev_info *revs, const ch
 			for (j = i; j < argc; j++)
 				verify_filename(revs->prefix, argv[j]);
 
-			revs->prune_data = get_pathspec(revs->prefix,
-							argv + i);
+			prune_data = argv + i;
 			break;
 		}
 	}
 
+	if (prune_data)
+		revs->prune_data = get_pathspec(revs->prefix, prune_data);
+
 	if (revs->def == NULL)
 		revs->def = def;
 	if (revs->show_merge)
-- 
1.6.5.3.342.g14bb9

[PATCH 3/3] Make --stdin option to "log" family read also pathspecs

[Junio C Hamano]

Similar to the command line arguments, after giving zero or more revs, you can
feed a line "--" and then feed pathspecs one at a time.

With this

	(
		echo ^maint
		echo --
		echo Documentation
	) | git log --stat --oneline --stdin master -- t

lists commits that touch Documentation/ or t/ between maint and master.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 revision.c |   72 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 files changed, 68 insertions(+), 4 deletions(-)

diff --git a/revision.c b/revision.c
index 4410a45..8750c20 100644
--- a/revision.c
+++ b/revision.c
@@ -953,9 +953,38 @@ int handle_revision_arg(const char *arg, struct rev_info *revs,
 	return 0;
 }
 
-static void read_revisions_from_stdin(struct rev_info *revs)
+static void read_pathspec_from_stdin(struct rev_info *revs, struct strbuf *sb, const char ***prune_data)
+{
+	const char **prune = *prune_data;
+	int prune_nr;
+	int prune_alloc;
+
+	/* count existing ones */
+	if (!prune)
+		prune_nr = 0;
+	else
+		for (prune_nr = 0; prune[prune_nr]; prune_nr++)
+			;
+	prune_alloc = prune_nr; /* not really, but we do not know */
+
+	while (strbuf_getwholeline(sb, stdin, '\n') != EOF) {
+		int len = sb->len;
+		if (len && sb->buf[len - 1] == '\n')
+			sb->buf[--len] = '\0';
+		ALLOC_GROW(prune, prune_nr+1, prune_alloc);
+		prune[prune_nr++] = xstrdup(sb->buf);
+	}
+	if (prune) {
+		ALLOC_GROW(prune, prune_nr+1, prune_alloc);
+		prune[prune_nr] = NULL;
+	}
+	*prune_data = prune;
+}
+
+static void read_revisions_from_stdin(struct rev_info *revs, const char ***prune)
 {
 	struct strbuf sb;
+	int seen_dashdash = 0;
 
 	strbuf_init(&sb, 1000);
 	while (strbuf_getwholeline(&sb, stdin, '\n') != EOF) {
@@ -964,11 +993,18 @@ static void read_revisions_from_stdin(struct rev_info *revs)
 			sb.buf[--len] = '\0';
 		if (!len)
 			break;
-		if (sb.buf[0] == '-')
+		if (sb.buf[0] == '-') {
+			if (len == 2 && sb.buf[1] == '-') {
+				seen_dashdash = 1;
+				break;
+			}
 			die("options not supported in --stdin mode");
+		}
 		if (handle_revision_arg(sb.buf, revs, 0, 1))
 			die("bad revision '%s'", sb.buf);
 	}
+	if (seen_dashdash)
+		read_pathspec_from_stdin(revs, &sb, prune);
 	strbuf_release(&sb);
 }
 
@@ -1220,6 +1256,34 @@ void parse_revision_opt(struct rev_info *revs, struct parse_opt_ctx_t *ctx,
 	ctx->argc -= n;
 }
 
+static void append_prune_data(const char ***prune_data, const char **av)
+{
+	const char **prune = *prune_data;
+	int prune_nr;
+	int prune_alloc;
+
+	if (!prune) {
+		*prune_data = av;
+		return;
+	}
+
+	/* count existing ones */
+	for (prune_nr = 0; prune[prune_nr]; prune_nr++)
+		;
+	prune_alloc = prune_nr; /* not really, but we do not know */
+
+	while (*av) {
+		ALLOC_GROW(prune, prune_nr+1, prune_alloc);
+		prune[prune_nr++] = *av;
+		av++;
+	}
+	if (prune) {
+		ALLOC_GROW(prune, prune_nr+1, prune_alloc);
+		prune[prune_nr] = NULL;
+	}
+	*prune_data = prune;
+}
+
 /*
  * Parse revision information, filling in the "rev_info" structure,
  * and removing the used arguments from the argument list.
@@ -1294,7 +1358,7 @@ int setup_revisions(int argc, const char **argv, struct rev_info *revs, const ch
 				}
 				if (read_from_stdin++)
 					die("--stdin given twice?");
-				read_revisions_from_stdin(revs);
+				read_revisions_from_stdin(revs, &prune_data);
 				continue;
 			}
 
@@ -1322,7 +1386,7 @@ int setup_revisions(int argc, const char **argv, struct rev_info *revs, const ch
 			for (j = i; j < argc; j++)
 				verify_filename(revs->prefix, argv[j]);
 
-			prune_data = argv + i;
+			append_prune_data(&prune_data, argv + i);
 			break;
 		}
 	}
-- 
1.6.5.3.342.g14bb9

Re: [PATCH 0/3] "log --stdin" updates

[Jeff King]

On Fri, Nov 20, 2009 at 03:25:12AM -0800, Junio C Hamano wrote:

> These three come on top of 61ab67a (Teach --stdin option to "log" family,
> 2009-11-03), which gave "--stdin" to the log family of commands (e.g. log,
> rev-list).  The earlier patch allowed you to feed only the revs (which was
> what gitk originally wanted), but after zero or more revs (one rev per
> line), you can now feed a line that consists of "--" and then pathspecs
> (one path per line).

This looks like a good API feature, though I am curious about the
missing "4/3" that was the motivation (of course, with a feature
like this, it may be for your out-of-git script, but as I said, I am
curious :) ). Is it a response to

  http://article.gmane.org/gmane.comp.version-control.git/133268

?

The implementation looks good from my cursory reading. Tests and docs
are of course missing. We don't seem to have any rev-list --stdin tests
already. Maybe even something as simple as:

  git rev-list HEAD -- file >expect &&
  (echo HEAD; echo --; echo file) | git rev-list --stdin >actual &&
  test_cmp expect actual

For docs, maybe squash this in (it mentions the pathspecs, but also
loosens --stdin to be available for "git log" and friends, which should
have been part of 61ab67a).

diff --git a/Documentation/rev-list-options.txt b/Documentation/rev-list-options.txt
index bf66116..b44fdd9 100644
--- a/Documentation/rev-list-options.txt
+++ b/Documentation/rev-list-options.txt
@@ -243,12 +243,14 @@ endif::git-rev-list[]
 	Pretend as if all the refs in `$GIT_DIR/refs/remotes` are listed
 	on the command line as '<commit>'.
 
-ifdef::git-rev-list[]
 --stdin::
 
 	In addition to the '<commit>' listed on the command
-	line, read them from the standard input.
+	line, read them from the standard input. If a '--' separator is
+	seen, stop reading commits and start reading paths to limit the
+	result.
 
+ifdef::git-rev-list[]
 --quiet::
 
 	Don't print anything to standard output.  This form

Re: [PATCH 0/3] "log --stdin" updates

[Junio C Hamano]

Jeff King <peff@peff.net> writes:

> On Fri, Nov 20, 2009 at 03:25:12AM -0800, Junio C Hamano wrote:
>
>> These three come on top of 61ab67a (Teach --stdin option to "log" family,
>> 2009-11-03), which gave "--stdin" to the log family of commands (e.g. log,
>> rev-list).  The earlier patch allowed you to feed only the revs (which was
>> what gitk originally wanted), but after zero or more revs (one rev per
>> line), you can now feed a line that consists of "--" and then pathspecs
>> (one path per line).
>
> This looks like a good API feature, though I am curious about the
> missing "4/3" that was the motivation (of course, with a feature
> like this, it may be for your out-of-git script, but as I said, I am
> curious :) ). Is it a response to
>
>   http://article.gmane.org/gmane.comp.version-control.git/133268
>
> ?

No.  This series is a follow-up to $gmane/131971; I was suspicious that I
missed something for the API feature to be complete back then (and the
patch was not signed-off partly for that reason), and I found one thing
that was lacking.