Replaying git history
Work that was never watched, such as a branch an agent pushed overnight or a pull request, can still be typed out from git. Replay Git Commits… (in the panel’s ... menu) builds the buffer from a range of commits, one change per file per commit, oldest first. Pick one of:
- This branch: the current branch’s commits that are not on the default branch (
origin/HEAD, elseorigin/main,origin/master,mainormaster). - A branch…: another local or remote-tracking branch’s commits that are not on the default branch. A pull request you fetched but did not check out is replayed this way.
- Last commit, or Enter a commit or range… in git’s syntax:
main..feature,HEAD~5..HEAD, or a single commit.
Each change’s before and after text come from the commits, not from the files on disk, so uncommitted edits do not show up. Merge commits are skipped, as git log -p does, and so are files the watch filters leave out, binary files, files over maxFileSizeKB, Git LFS files, symbolic links and submodules. Renames git detects replay as renames. Nothing is downloaded: in a partial clone, file versions git does not have locally are skipped (this needs git 2.44 or later; older git fetches them), and in a shallow clone the oldest commit, whose parent is missing, is skipped.
The replay replaces the buffer (after asking) and starts playing at the first change; pause it from the panel or the replay editor. It also stops watching, so live edits do not mix into the history; start watching again when you are done. The repository is only read: nothing is checked out or fetched, and HEAD, the index and your files stay as they are.