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 <[email protected]>
Signed-off-by: Junio C Hamano <[email protected]>

Author: pks-t

Documentation/git-history.adoc

@@ -9,6 +9,7 @@ SYNOPSIS
 --------
 [synopsis]
 git history reword <commit>
+git history split <commit> [--] [<pathspec>...]
 
 DESCRIPTION
 -----------
@@ -37,11 +38,72 @@ Several 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.
+
 CONFIGURATION
 -------------
 
 include::includes/cmd-config-section-all.adoc[]
 
+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,e,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,e,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

builtin/history.c

@@ -1,24 +1,30 @@
 #define USE_THE_REPOSITORY_VARIABLE
 
 #include "builtin.h"
+#include "cache-tree.h"
 #include "commit-reach.h"
 #include "commit.h"
 #include "config.h"
 #include "editor.h"
 #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,
@@ -364,18 +370,186 @@ static int cmd_history_reword(int argc,
 	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;
+	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;
+	char original_commit_oid[GIT_MAX_HEXSZ + 1];
+	struct commit_list *parents = NULL;
+	struct commit *first_commit;
+	struct tree *split_tree;
+	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, 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);
+
+	/*
+	* 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(repo, "split-out", original_commit,
+					      &split_tree->object.oid,
+					      original_commit->parents, &parent_tree_oid, &out[0]);
+	if (ret < 0) {
+		ret = error(_("failed writing split-out 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.
+	*/
+	first_commit = lookup_commit_reference(repo, &out[0]);
+	commit_list_append(first_commit, &parents);
+
+	ret = commit_tree_with_edited_message(repo, "split-out", original_commit,
+					      &original_commit_tree_oid,
+					      parents, get_commit_tree_oid(first_commit), &out[1]);
+	if (ret < 0) {
+		ret = error(_("failed writing split-out commit"));
+		goto out;
+	}
+
+	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,
+	};
+	struct option options[] = {
+		OPT_END(),
+	};
+	struct commit *original_commit, *parent, *head;
+	struct strvec commits = STRVEC_INIT;
+	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);
+
+	parse_pathspec(&pathspec, 0,
+		PATHSPEC_PREFER_FULL | PATHSPEC_SYMLINK_LEADING_PATH | PATHSPEC_PREFIX_ORIGIN,
+		prefix, argv + 1);
+
+	ret = gather_commits_between_head_and_revision(repo, argv[0], &original_commit,
+						       &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:
+	clear_pathspec(&pathspec);
+	strvec_clear(&commits);
+	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(),
 	};
 

t/meson.build

@@ -388,6 +388,7 @@ 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',