builtin/history: implement "split" subcommand

It is quite a common use case that one wants to split up one commit into
multiple commits by moving parts of the changes of the original commit
out into a separate commit. This is quite an involved operation though:

  1. Identify the commit in question that is to be dropped.

  2. Perform an interactive rebase on top of that commit's parent.

  3. Modify the instruction sheet to "edit" the commit that is to be
     split up.

  4. Drop the commit via "git reset HEAD~".

  5. Stage changes that should go into the first commit and commit it.

  6. Stage changes that should go into the second commit and commit it.

  7. Finalize the rebase.

This is quite complex, and overall I would claim that most people who
are not experts in Git would struggle with this flow.

Introduce a new "split" subcommand for git-history(1) to make this way
easier. All the user needs to do is to say `git history split $COMMIT`.
From hereon, Git asks the user which parts of the commit shall be moved
out into a separate commit and, once done, asks the user for the commit
message. Git then creates that split-out commit and applies the original
commit on top of it.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

1 files changed, 218 insertions, 0 deletions

diff --git a/builtin/history.c b/builtin/history.c
index cb251ae2e0..cae841707d 100644
--- a/builtin/history.c
+++ b/builtin/history.c
@@ -1,6 +1,7 @@
 #define USE_THE_REPOSITORY_VARIABLE
 
 #include "builtin.h"
+#include "cache-tree.h"
 #include "commit-reach.h"
 #include "commit.h"
 #include "config.h"
@@ -8,17 +9,22 @@
 #include "environment.h"
 #include "gettext.h"
 #include "hex.h"
+#include "oidmap.h"
 #include "parse-options.h"
+#include "path.h"
+#include "read-cache.h"
 #include "refs.h"
 #include "replay.h"
 #include "reset.h"
 #include "revision.h"
+#include "run-command.h"
 #include "sequencer.h"
 #include "strvec.h"
 #include "tree.h"
 #include "wt-status.h"
 
 #define GIT_HISTORY_REWORD_USAGE N_("git history reword <commit>")
+#define GIT_HISTORY_SPLIT_USAGE  N_("git history split <commit> [--] [<pathspec>...]")
 
 static int collect_commits(struct repository *repo,
 			   struct commit *old_commit,
@@ -323,6 +329,216 @@ out:
 	return ret;
 }
 
+static int split_commit(struct repository *repo,
+			struct commit *original_commit,
+			struct pathspec *pathspec,
+			struct object_id *out)
+{
+	struct interactive_options interactive_opts = INTERACTIVE_OPTIONS_INIT;
+	struct strbuf index_file = STRBUF_INIT, split_message = STRBUF_INIT;
+	struct child_process read_tree_cmd = CHILD_PROCESS_INIT;
+	struct index_state index = INDEX_STATE_INIT(repo);
+	struct object_id original_commit_tree_oid, parent_tree_oid;
+	const char *original_message, *original_body, *ptr;
+	char original_commit_oid[GIT_MAX_HEXSZ + 1];
+	char *original_author = NULL;
+	struct commit_list *parents = NULL;
+	struct commit *first_commit;
+	struct tree *split_tree;
+	size_t len;
+	int ret;
+
+	if (original_commit->parents)
+		parent_tree_oid = *get_commit_tree_oid(original_commit->parents->item);
+	else
+		oidcpy(&parent_tree_oid, repo->hash_algo->empty_tree);
+	original_commit_tree_oid = *get_commit_tree_oid(original_commit);
+
+	/*
+	* Construct the first commit. This is done by taking the original
+	* commit parent's tree and selectively patching changes from the diff
+	* between that parent and its child.
+	*/
+	repo_git_path_replace(repo, &index_file, "%s", "history-split.index");
+
+	read_tree_cmd.git_cmd = 1;
+	strvec_pushf(&read_tree_cmd.env, "GIT_INDEX_FILE=%s", index_file.buf);
+	strvec_push(&read_tree_cmd.args, "read-tree");
+	strvec_push(&read_tree_cmd.args, oid_to_hex(&parent_tree_oid));
+	ret = run_command(&read_tree_cmd);
+	if (ret < 0)
+		goto out;
+
+	ret = read_index_from(&index, index_file.buf, repo->gitdir);
+	if (ret < 0) {
+		ret = error(_("failed reading temporary index"));
+		goto out;
+	}
+
+	oid_to_hex_r(original_commit_oid, &original_commit->object.oid);
+	ret = run_add_p_index(repo, &index, index_file.buf, &interactive_opts,
+			      original_commit_oid, pathspec);
+	if (ret < 0)
+		goto out;
+
+	split_tree = write_in_core_index_as_tree(repo, &index);
+	if (!split_tree) {
+		ret = error(_("failed split tree"));
+		goto out;
+	}
+
+	unlink(index_file.buf);
+
+	/*
+	* We disallow the cases where either the split-out commit or the
+	* original commit would become empty. Consequently, if we see that the
+	* new tree ID matches either of those trees we abort.
+	*/
+	if (oideq(&split_tree->object.oid, &parent_tree_oid)) {
+		ret = error(_("split commit is empty"));
+		goto out;
+	} else if (oideq(&split_tree->object.oid, &original_commit_tree_oid)) {
+		ret = error(_("split commit tree matches original commit"));
+		goto out;
+	}
+
+	/* We retain authorship of the original commit. */
+	original_message = repo_logmsg_reencode(repo, original_commit, NULL, NULL);
+	ptr = find_commit_header(original_message, "author", &len);
+	if (ptr)
+		original_author = xmemdupz(ptr, len);
+
+	ret = fill_commit_message(repo, &parent_tree_oid, &split_tree->object.oid,
+				  "", "split-out", &split_message);
+	if (ret < 0)
+		goto out;
+
+	ret = commit_tree(split_message.buf, split_message.len, &split_tree->object.oid,
+			  original_commit->parents, &out[0], original_author, NULL);
+	if (ret < 0) {
+		ret = error(_("failed writing split-out commit"));
+		goto out;
+	}
+
+	/*
+	* The second commit is much simpler to construct, as we can simply use
+	* the original commit details, except that we adjust its parent to be
+	* the newly split-out commit.
+	*/
+	find_commit_subject(original_message, &original_body);
+	first_commit = lookup_commit_reference(repo, &out[0]);
+	commit_list_append(first_commit, &parents);
+
+	ret = commit_tree(original_body, strlen(original_body), &original_commit_tree_oid,
+			  parents, &out[1], original_author, NULL);
+	if (ret < 0) {
+		ret = error(_("failed writing second commit"));
+		goto out;
+	}
+
+	ret = 0;
+
+out:
+	if (index_file.len)
+		unlink(index_file.buf);
+	strbuf_release(&split_message);
+	strbuf_release(&index_file);
+	free_commit_list(parents);
+	free(original_author);
+	release_index(&index);
+	return ret;
+}
+
+static int cmd_history_split(int argc,
+			     const char **argv,
+			     const char *prefix,
+			     struct repository *repo)
+{
+	const char * const usage[] = {
+		GIT_HISTORY_SPLIT_USAGE,
+		NULL,
+	};
+	struct option options[] = {
+		OPT_END(),
+	};
+	struct oidmap rewritten_commits = OIDMAP_INIT;
+	struct commit *original_commit, *parent, *head;
+	struct strvec commits = STRVEC_INIT;
+	struct commit_list *from_list = NULL;
+	struct object_id split_commits[2];
+	struct pathspec pathspec = { 0 };
+	int ret;
+
+	argc = parse_options(argc, argv, prefix, options, usage, 0);
+	if (argc < 1) {
+		ret = error(_("command expects a revision"));
+		goto out;
+	}
+	repo_config(repo, git_default_config, NULL);
+
+	original_commit = lookup_commit_reference_by_name(argv[0]);
+	if (!original_commit) {
+		ret = error(_("commit to be split cannot be found: %s"), argv[0]);
+		goto out;
+	}
+
+	parent = original_commit->parents ? original_commit->parents->item : NULL;
+	if (parent && repo_parse_commit(repo, parent)) {
+		ret = error(_("unable to parse commit %s"),
+			    oid_to_hex(&parent->object.oid));
+		goto out;
+	}
+
+	head = lookup_commit_reference_by_name("HEAD");
+	if (!head) {
+		ret = error(_("could not resolve HEAD to a commit"));
+		goto out;
+	}
+
+	commit_list_append(original_commit, &from_list);
+	if (!repo_is_descendant_of(repo, head, from_list)) {
+		ret = error(_("split commit must be reachable from current HEAD commit"));
+		goto out;
+	}
+
+	parse_pathspec(&pathspec, 0,
+		PATHSPEC_PREFER_FULL | PATHSPEC_SYMLINK_LEADING_PATH | PATHSPEC_PREFIX_ORIGIN,
+		prefix, argv + 1);
+
+	/*
+	 * Collect the list of commits that we'll have to reapply now already.
+	 * This ensures that we'll abort early on in case the range of commits
+	 * contains merges, which we do not yet handle.
+	 */
+	ret = collect_commits(repo, parent, head, &commits);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * Then we split up the commit and replace the original commit with the
+	 * new ones.
+	 */
+	ret = split_commit(repo, original_commit, &pathspec, split_commits);
+	if (ret < 0)
+		goto out;
+
+	replace_commits(&commits, &original_commit->object.oid,
+			split_commits, ARRAY_SIZE(split_commits));
+
+	ret = apply_commits(repo, &commits, parent, head, "split");
+	if (ret < 0)
+		goto out;
+
+	ret = 0;
+
+out:
+	oidmap_clear(&rewritten_commits, 0);
+	free_commit_list(from_list);
+	clear_pathspec(&pathspec);
+	strvec_clear(&commits);
+	return ret;
+}
+
 int cmd_history(int argc,
 		const char **argv,
 		const char *prefix,
@@ -330,11 +546,13 @@ int cmd_history(int argc,
 {
 	const char * const usage[] = {
 		GIT_HISTORY_REWORD_USAGE,
+		GIT_HISTORY_SPLIT_USAGE,
 		NULL,
 	};
 	parse_opt_subcommand_fn *fn = NULL;
 	struct option options[] = {
 		OPT_SUBCOMMAND("reword", &fn, cmd_history_reword),
+		OPT_SUBCOMMAND("split", &fn, cmd_history_split),
 		OPT_END(),
 	};