[PATCH] Implement a textconv filter for "git diff"

[Matthieu Moy <Matthieu.Moy@imag.fr>]

That feature is similar to the custom diff driver, but the user only has
to provide a text filter (a command to convert a file into a plain-text
representation). Git takes care of diffing, mode change, ...

For example, with

[textconv "odt2txt"]
          command=odt2txt

*.ods textconv=odt2txt
*.odp textconv=odt2txt
*.odt textconv=odt2txt

One can use "git diff" on OpenOffice files (including "git diff --color"
and friends).

This could be extended so that "git blame" can use it also.
---
 diff.c |   68 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 68 insertions(+), 0 deletions(-)

diff --git a/diff.c b/diff.c
index 6917981..7cbc42b 100644
--- a/diff.c
+++ b/diff.c
@@ -181,6 +181,12 @@ int git_diff_ui_config(const char *var, const char *value, void *cb)
 		if (ep != var + 4 && !strcmp(ep, ".command"))
 			return parse_ll_command(var, var + 5, ep, value, DIFF_DRIVER);
 	}
+	if (!prefixcmp(var, "textconv.")) {
+		const char *ep = strrchr(var, '.');
+
+		if (ep != var + 8 && !strcmp(ep, ".command"))
+			return parse_ll_command(var, var + 9, ep, value, TEXTCONV_DRIVER);
+	}
 
 	return git_diff_basic_config(var, value, cb);
 }
@@ -262,6 +268,10 @@ static struct diff_tempfile {
 
 /* Forward declarations */
 static int run_command_to_buf(const char **argv, char **buf, size_t * size);
+static const char *external_cmd_attr(const char *name, int driver_index);
+static void prepare_temp_file(const char *name,
+			      struct diff_tempfile *temp,
+			      struct diff_filespec *one);
 /* End forward declarations */
 
 static int count_lines(const char *data, int size)
@@ -1511,6 +1521,52 @@ static int run_command_to_buf(const char **argv, char **buf, size_t * size)
 	return 0;
 }
 
+static int textconv_two_files(const char *textconv,
+			      const char *name_a,
+			      const char *name_b,
+			      mmfile_t *mf1,
+			      mmfile_t *mf2,
+			      struct diff_filespec *one,
+			      struct diff_filespec *two)
+{
+	const char *spawn_arg[3];
+	const char **arg = &spawn_arg[0];
+	struct diff_tempfile *temp = diff_temp;
+	char *buf_a, *buf_b;
+	size_t size_a, size_b;
+	prepare_temp_file(name_a, &temp[0], one);
+	*arg++ = textconv;
+	*arg++ = temp[0].name;
+	*arg++ = NULL;
+	/*
+	 * Run both commands before touching arguments to make sure we
+	 * do all or nothing.
+	 */
+	if(run_command_to_buf(spawn_arg, &buf_a, &size_a))
+		return -1;
+
+	prepare_temp_file(name_b, &temp[1], two);
+	spawn_arg[1] = temp[1].name;
+	if(run_command_to_buf(spawn_arg, &buf_b, &size_b))
+		return -1;
+
+	mf1->ptr = buf_a;
+	mf1->size = (long)size_a;
+	one->data = mf1->ptr;
+	one->size = mf1->size;
+	one->should_free = 1;
+	one->should_munmap = 1;
+
+	mf2->ptr = buf_b;
+	mf2->size = (long)size_b;
+	two->data = mf2->ptr;
+	two->size = mf2->size;
+	two->should_free = 1;
+	two->should_munmap = 1;
+	return 0;
+}
+
+
 static void builtin_diff(const char *name_a,
 			 const char *name_b,
 			 struct diff_filespec *one,
@@ -1574,6 +1630,18 @@ static void builtin_diff(const char *name_a,
 	if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
 		die("unable to read files to diff");
 
+	const char * textconv = external_cmd_attr(name_a, TEXTCONV_DRIVER);
+	const char * textconvb = external_cmd_attr(name_b, TEXTCONV_DRIVER);
+
+	/*
+	 * Only use the textconv driver if it is set, and is the same
+	 * for source and destination file.
+	 */
+	if (textconv && textconvb && !strcmp(textconv, textconvb))
+		if(textconv_two_files(textconv, name_a, name_b, &mf1, &mf2, one, two))
+			fprintf(stderr, "warning: textconv filter failed, "
+				"falling back to plain diff\n");
+
 	if (!DIFF_OPT_TST(o, TEXT) &&
 	    (diff_filespec_is_binary(one) || diff_filespec_is_binary(two))) {
 		/* Quite common confusing case */
-- 
1.6.0.2.312.g1ef81a