filter: introduce GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT

Provide a mechanism to filter using attribute data from a specific
commit (making use of `GIT_ATTR_CHECK_INCLUDE_COMMIT`).

Author: ethomson

include/git2/blob.h

@@ -114,6 +114,12 @@ typedef enum {
 	 * in the HEAD commit.
 	 */
 	GIT_BLOB_FILTER_ATTRIBUTES_FROM_HEAD = (1 << 2),
+
+	/**
+	 * When set, filters will be loaded from a `.gitattributes` file
+	 * in the specified commit.
+	 */
+	GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT = (1 << 3),
 } git_blob_filter_flag_t;
 
 /**
@@ -128,6 +134,12 @@ typedef struct {
 
 	/** Flags to control the filtering process, see `git_blob_filter_flag_t` above */
 	uint32_t flags;
+
+	/**
+	 * The commit to load attributes from, when
+	 * `GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT` is specified.
+	 */
+	git_oid *commit_id;
 } git_blob_filter_options;
 
 #define GIT_BLOB_FILTER_OPTIONS_VERSION 1

include/git2/filter.h

@@ -49,6 +49,12 @@ typedef enum {
 
 	/** Load attributes from `.gitattributes` in the root of HEAD */
 	GIT_FILTER_ATTRIBUTES_FROM_HEAD = (1u << 2),
+
+	/**
+	 * Load attributes from `.gitattributes` in a given commit.
+	 * This can only be specified in a `git_filter_options`.
+	 */
+	GIT_FILTER_ATTRIBUTES_FROM_COMMIT = (1u << 3),
 } git_filter_flag_t;
 
 /**
@@ -59,6 +65,12 @@ typedef struct {
 
 	/** See `git_filter_flag_t` above */
 	uint32_t flags;
+
+	/**
+	 * The commit to load attributes from, when
+	 * `GIT_FILTER_ATTRIBUTES_FROM_COMMIT` is specified.
+	 */
+	git_oid *commit_id;
 } git_filter_options;
 
  #define GIT_FILTER_OPTIONS_VERSION 1

src/blob.c

@@ -421,7 +421,7 @@ int git_blob_filter(
 	int error = 0;
 	git_filter_list *fl = NULL;
 	git_blob_filter_options opts = GIT_BLOB_FILTER_OPTIONS_INIT;
-	git_filter_flag_t flags = GIT_FILTER_DEFAULT;
+	git_filter_options filter_opts = GIT_FILTER_OPTIONS_INIT;
 
 	GIT_ASSERT_ARG(blob);
 	GIT_ASSERT_ARG(path);
@@ -441,14 +441,19 @@ int git_blob_filter(
 		return 0;
 
 	if ((opts.flags & GIT_BLOB_FILTER_NO_SYSTEM_ATTRIBUTES) != 0)
-		flags |= GIT_FILTER_NO_SYSTEM_ATTRIBUTES;
+		filter_opts.flags |= GIT_FILTER_NO_SYSTEM_ATTRIBUTES;
 
 	if ((opts.flags & GIT_BLOB_FILTER_ATTRIBUTES_FROM_HEAD) != 0)
-		flags |= GIT_FILTER_ATTRIBUTES_FROM_HEAD;
+		filter_opts.flags |= GIT_FILTER_ATTRIBUTES_FROM_HEAD;
 
-	if (!(error = git_filter_list_load(
+	if ((opts.flags & GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT) != 0) {
+		filter_opts.flags |= GIT_FILTER_ATTRIBUTES_FROM_COMMIT;
+		filter_opts.commit_id = opts.commit_id;
+	}
+
+	if (!(error = git_filter_list_load_ext(
 			&fl, git_blob_owner(blob), blob, path,
-			GIT_FILTER_TO_WORKTREE, flags))) {
+			GIT_FILTER_TO_WORKTREE, &filter_opts))) {
 
 		error = git_filter_list_apply_to_blob(out, fl, blob);
 

src/filter.c

@@ -443,6 +443,11 @@ static int filter_list_check_attributes(
 	if ((src->options.flags & GIT_FILTER_ATTRIBUTES_FROM_HEAD) != 0)
 		attr_opts.flags |= GIT_ATTR_CHECK_INCLUDE_HEAD;
 
+	if ((src->options.flags & GIT_FILTER_ATTRIBUTES_FROM_COMMIT) != 0) {
+		attr_opts.flags |= GIT_ATTR_CHECK_INCLUDE_COMMIT;
+		attr_opts.commit_id = src->options.commit_id;
+	}
+
 	error = git_attr_get_many_with_session(
 		strs, repo, filter_session->attr_session, &attr_opts, src->path, fdef->nattrs, fdef->attrs);
 

tests/filter/bare.c

@@ -132,3 +132,63 @@ void test_filter_bare__sanitizes(void)
 	git_blob_free(blob);
 }
 
+void test_filter_bare__from_specific_commit_one(void)
+{
+	git_blob_filter_options opts = GIT_BLOB_FILTER_OPTIONS_INIT;
+	git_blob *blob;
+	git_buf buf = { 0 };
+	git_oid commit_id;
+
+	cl_git_pass(git_oid_fromstr(&commit_id, "b8986fec0f7bde90f78ac72706e782d82f24f2f0"));
+
+	opts.flags |= GIT_BLOB_FILTER_NO_SYSTEM_ATTRIBUTES;
+	opts.flags |= GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT;
+	opts.commit_id = &commit_id;
+
+	cl_git_pass(git_revparse_single(
+		(git_object **)&blob, g_repo, "055c872")); /* ident */
+
+	cl_assert_equal_s("$Id$\n", git_blob_rawcontent(blob));
+
+	cl_git_pass(git_blob_filter(&buf, blob, "ident.bin", &opts));
+	cl_assert_equal_s("$Id$\n", buf.ptr);
+
+	cl_git_pass(git_blob_filter(&buf, blob, "ident.identlf", &opts));
+	cl_assert_equal_s("$Id: 055c8729cdcc372500a08db659c045e16c4409fb $\n", buf.ptr);
+
+	git_buf_dispose(&buf);
+	git_blob_free(blob);
+}
+
+void test_filter_bare__from_specific_commit_with_no_attributes_file(void)
+{
+	git_blob_filter_options opts = GIT_BLOB_FILTER_OPTIONS_INIT;
+	git_blob *blob;
+	git_buf buf = { 0 };
+	git_oid commit_id;
+
+	cl_git_pass(git_oid_fromstr(&commit_id, "5afb6a14a864e30787857dd92af837e8cdd2cb1b"));
+
+	opts.flags |= GIT_BLOB_FILTER_NO_SYSTEM_ATTRIBUTES;
+	opts.flags |= GIT_BLOB_FILTER_ATTRIBUTES_FROM_COMMIT;
+	opts.commit_id = &commit_id;
+
+	cl_git_pass(git_revparse_single(
+		(git_object **)&blob, g_repo, "799770d")); /* all-lf */
+
+	cl_assert_equal_s(ALL_LF_TEXT_RAW, git_blob_rawcontent(blob));
+
+	cl_git_pass(git_blob_filter(&buf, blob, "file.bin", &opts));
+	cl_assert_equal_s(ALL_LF_TEXT_RAW, buf.ptr);
+
+	/* we never convert CRLF -> LF on platforms that have LF */
+	cl_git_pass(git_blob_filter(&buf, blob, "file.lf", &opts));
+	cl_assert_equal_s(ALL_LF_TEXT_RAW, buf.ptr);
+
+	/* we never convert CRLF -> LF on platforms that have LF */
+	cl_git_pass(git_blob_filter(&buf, blob, "file.crlf", &opts));
+	cl_assert_equal_s(ALL_LF_TEXT_RAW, buf.ptr);
+
+	git_buf_dispose(&buf);
+	git_blob_free(blob);
+}

tests/resources/crlf.git/objects/05/5c8729cdcc372500a08db659c045e16c4409fb

@@ -0,0 +1,3 @@
+x
��M
+�0�]�s
%3�Dt���$�b�6�F��Q�����2Mc�nSWU`�6*��C��ĽG�w��,�|S��,��\
�����#��Y��
8t�);q��ȳ��
+���5��V�G��-���~��ۥ2
=�`��ֲ���ml�?1�3�cn��H�]a�j�@�U