builtin/history: implement "split" subcommand

pks-t · gitster · commit d563ecec2845 · 2026-03-03T15:09:37.000-08:00
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 &lt;ps@pks.im&gt;
Signed-off-by: Junio C Hamano &lt;gitster@pobox.com&gt;
diff --git a/Documentation/git-history.adoc b/Documentation/git-history.adoc
@@ -9,6 +9,7 @@ SYNOPSIS
 --------
 [synopsis]
 git history reword <commit> [--dry-run] [--update-refs=(branches|head)]
+git history split <commit> [--dry-run] [--update-refs=(branches|head)] [--] [<pathspec>...]
 
 DESCRIPTION
 -----------
@@ -57,6 +58,26 @@ The following commands are available to rewrite history in different ways:
 	details of this commit remain unchanged. This command will spawn an
 	editor with the current message of that commit.
 
+`split <commit> [--] [<pathspec>...]`::
+	Interactively split up <commit> into two commits by choosing
+	hunks introduced by it that will be moved into the new split-out
+	commit. These hunks will then be written into a new commit that
+	becomes the parent of the previous commit. The original commit
+	stays intact, except that its parent will be the newly split-out
+	commit.
++
+The commit messages of the split-up commits will be asked for by launching
+the configured editor. Authorship of the commit will be the same as for the
+original commit.
++
+If passed, _<pathspec>_ can be used to limit which changes shall be split out
+of the original commit. Files not matching any of the pathspecs will remain
+part of the original commit. For more details, see the 'pathspec' entry in
+linkgit:gitglossary[7].
++
+It is invalid to select either all or no hunks, as that would lead to
+one of the commits becoming empty.
+
 OPTIONS
 -------
 
@@ -72,6 +93,47 @@ OPTIONS
 	descendants of the original commit will be rewritten. With `head`, only
 	the current `HEAD` reference will be rewritten. Defaults to `branches`.
 
+EXAMPLES
+--------
+
+Split a commit
+~~~~~~~~~~~~~~
+
+----------
+$ git log --stat --oneline
+3f81232 (HEAD -> main) original
+ bar | 1 +
+ foo | 1 +
+ 2 files changed, 2 insertions(+)
+
+$ git history split HEAD
+diff --git a/bar b/bar
+new file mode 100644
+index 0000000..5716ca5
+--- /dev/null
++++ b/bar
+@@ -0,0 +1 @@
++bar
+(1/1) Stage addition [y,n,q,a,d,p,?]? y
+
+diff --git a/foo b/foo
+new file mode 100644
+index 0000000..257cc56
+--- /dev/null
++++ b/foo
+@@ -0,0 +1 @@
++foo
+(1/1) Stage addition [y,n,q,a,d,p,?]? n
+
+$ git log --stat --oneline
+7cebe64 (HEAD -> main) original
+ foo | 1 +
+ 1 file changed, 1 insertion(+)
+d1582f3 split-out commit
+ bar | 1 +
+ 1 file changed, 1 insertion(+)
+----------
+
 GIT
 ---
 Part of the linkgit:git[1] suite
diff --git a/builtin/history.c b/builtin/history.c
@@ -1,24 +1,32 @@
 #define USE_THE_REPOSITORY_VARIABLE
 
 #include "builtin.h"
+#include "cache-tree.h"
 #include "commit.h"
 #include "commit-reach.h"
 #include "config.h"
 #include "editor.h"
 #include "environment.h"
 #include "gettext.h"
 #include "hex.h"
+#include "lockfile.h"
+#include "oidmap.h"
 #include "parse-options.h"
+#include "path.h"
+#include "read-cache.h"
 #include "refs.h"
 #include "replay.h"
 #include "revision.h"
 #include "sequencer.h"
 #include "strvec.h"
 #include "tree.h"
+#include "unpack-trees.h"
 #include "wt-status.h"
 
 #define GIT_HISTORY_REWORD_USAGE \
 	N_("git history reword <commit> [--dry-run] [--update-refs=(branches|head)]")
+#define GIT_HISTORY_SPLIT_USAGE \
+	N_("git history split <commit> [--dry-run] [--update-refs=(branches|head)] [--] [<pathspec>...]")
 
 static void change_data_free(void *util, const char *str UNUSED)
 {
@@ -484,18 +492,260 @@ static int cmd_history_reword(int argc,
 	return ret;
 }
 
+static int write_ondisk_index(struct repository *repo,
+			      struct object_id *oid,
+			      const char *path)
+{
+	struct unpack_trees_options opts = { 0 };
+	struct lock_file lock = LOCK_INIT;
+	struct tree_desc tree_desc;
+	struct index_state index;
+	struct tree *tree;
+	int ret;
+
+	index_state_init(&index, repo);
+
+	opts.head_idx = -1;
+	opts.src_index = &index;
+	opts.dst_index = &index;
+
+	tree = repo_parse_tree_indirect(repo, oid);
+	init_tree_desc(&tree_desc, &tree->object.oid, tree->buffer, tree->size);
+
+	if (unpack_trees(1, &tree_desc, &opts)) {
+		ret = error(_("unable to populate index with tree"));
+		goto out;
+	}
+
+	prime_cache_tree(repo, &index, tree);
+
+	if (hold_lock_file_for_update(&lock, path, 0) < 0) {
+		ret = error_errno(_("unable to acquire index lock"));
+		goto out;
+	}
+
+	if (write_locked_index(&index, &lock, COMMIT_LOCK)) {
+		ret = error(_("unable to write new index file"));
+		goto out;
+	}
+
+	ret = 0;
+
+out:
+	rollback_lock_file(&lock);
+	release_index(&index);
+	return ret;
+}
+
+static int split_commit(struct repository *repo,
+			struct commit *original,
+			struct pathspec *pathspec,
+			struct commit **out)
+{
+	struct interactive_options interactive_opts = INTERACTIVE_OPTIONS_INIT;
+	struct strbuf index_file = STRBUF_INIT;
+	struct index_state index = INDEX_STATE_INIT(repo);
+	const struct object_id *original_commit_tree_oid;
+	const struct object_id *old_tree_oid, *new_tree_oid;
+	struct object_id parent_tree_oid;
+	char original_commit_oid[GIT_MAX_HEXSZ + 1];
+	struct commit *first_commit, *second_commit;
+	struct commit_list *parents = NULL;
+	struct tree *split_tree;
+	int ret;
+
+	if (original->parents) {
+		if (repo_parse_commit(repo, original->parents->item)) {
+			ret = error(_("unable to parse parent commit %s"),
+				    oid_to_hex(&original->parents->item->object.oid));
+			goto out;
+		}
+
+		parent_tree_oid = *get_commit_tree_oid(original->parents->item);
+	} else {
+		oidcpy(&parent_tree_oid, repo->hash_algo->empty_tree);
+	}
+	original_commit_tree_oid = get_commit_tree_oid(original);
+
+	/*
+	 * 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");
+
+	ret = write_ondisk_index(repo, &parent_tree_oid, index_file.buf);
+	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->object.oid);
+	ret = run_add_p_index(repo, &index, index_file.buf, &interactive_opts,
+			      original_commit_oid, pathspec, ADD_P_DISALLOW_EDIT);
+	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);
+	strbuf_release(&index_file);
+
+	/*
+	 * 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;
+	}
+
+	/*
+	 * The first commit is constructed from the split-out tree. The base
+	 * that shall be diffed against is the parent of the original commit.
+	 */
+	ret = commit_tree_with_edited_message_ext(repo, "split-out", original,
+						  original->parents, &parent_tree_oid,
+						  &split_tree->object.oid, &first_commit);
+	if (ret < 0) {
+		ret = error(_("failed writing first commit"));
+		goto out;
+	}
+
+	/*
+	 * The second commit is constructed from the original tree. The base to
+	 * diff against and the parent in this case is the first split-out
+	 * commit.
+	 */
+	commit_list_append(first_commit, &parents);
+
+	old_tree_oid = &repo_get_commit_tree(repo, first_commit)->object.oid;
+	new_tree_oid = &repo_get_commit_tree(repo, original)->object.oid;
+
+	ret = commit_tree_with_edited_message_ext(repo, "split-out", original,
+						  parents, old_tree_oid,
+						  new_tree_oid, &second_commit);
+	if (ret < 0) {
+		ret = error(_("failed writing second commit"));
+		goto out;
+	}
+
+	*out = second_commit;
+	ret = 0;
+
+out:
+	if (index_file.len)
+		unlink(index_file.buf);
+	strbuf_release(&index_file);
+	free_commit_list(parents);
+	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,
+	};
+	enum ref_action action = REF_ACTION_DEFAULT;
+	int dry_run = 0;
+	struct option options[] = {
+		OPT_CALLBACK_F(0, "update-refs", &action, N_("<refs>"),
+			       N_("control ref update behavior (branches|head|print)"),
+			       PARSE_OPT_NONEG, parse_ref_action),
+		OPT_BOOL('n', "dry-run", &dry_run,
+			 N_("perform a dry-run without updating any refs")),
+		OPT_END(),
+	};
+	struct commit *original, *rewritten = NULL;
+	struct strbuf reflog_msg = STRBUF_INIT;
+	struct pathspec pathspec = { 0 };
+	struct rev_info revs = { 0 };
+	int ret;
+
+	argc = parse_options(argc, argv, prefix, options, usage, 0);
+	if (argc < 1) {
+		ret = error(_("command expects a committish"));
+		goto out;
+	}
+	repo_config(repo, git_default_config, NULL);
+
+	if (action == REF_ACTION_DEFAULT)
+		action = REF_ACTION_BRANCHES;
+
+	parse_pathspec(&pathspec, 0,
+		       PATHSPEC_PREFER_FULL |
+		       PATHSPEC_SYMLINK_LEADING_PATH |
+		       PATHSPEC_PREFIX_ORIGIN,
+		       prefix, argv + 1);
+
+	original = lookup_commit_reference_by_name(argv[0]);
+	if (!original) {
+		ret = error(_("commit cannot be found: %s"), argv[0]);
+		goto out;
+	}
+
+	ret = setup_revwalk(repo, action, original, &revs);
+	if (ret < 0)
+		goto out;
+
+	if (original->parents && original->parents->next) {
+		ret = error(_("cannot split up merge commit"));
+		goto out;
+	}
+
+	ret = split_commit(repo, original, &pathspec, &rewritten);
+	if (ret < 0)
+		goto out;
+
+	strbuf_addf(&reflog_msg, "split: updating %s", argv[0]);
+
+	ret = handle_reference_updates(&revs, action, original, rewritten,
+				       reflog_msg.buf, dry_run);
+	if (ret < 0) {
+		ret = error(_("failed replaying descendants"));
+		goto out;
+	}
+
+	ret = 0;
+
+out:
+	strbuf_release(&reflog_msg);
+	clear_pathspec(&pathspec);
+	release_revisions(&revs);
+	return ret;
+}
+
 int cmd_history(int argc,
 		const char **argv,
 		const char *prefix,
 		struct repository *repo)
 {
 	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(),
 	};
 
diff --git a/t/meson.build b/t/meson.build
@@ -392,6 +392,7 @@ integration_tests = [
   't3438-rebase-broken-files.sh',
   't3450-history.sh',
   't3451-history-reword.sh',
+  't3452-history-split.sh',
   't3500-cherry.sh',
   't3501-revert-cherry-pick.sh',
   't3502-cherry-pick-merge.sh',
diff --git a/t/t3452-history-split.sh b/t/t3452-history-split.sh
