Stacking Overview - Git Productivity Toolkit

Stacked branches (or stacked PRs) are a workflow where you build a chain of dependent branches, each one based on the previous. This lets you break large changes into small, reviewable pieces without waiting for each PR to merge before starting the next.

The Problem

Without stacking, you either:

Wait: Finish PR 1, wait for review and merge, then start PR 2. Slow.
Branch from main: Start PR 2 from main, duplicate or conflict with PR 1’s changes. Messy.
Stack manually: Create PR 2 on top of PR 1’s branch, but then juggle rebasing when PR 1 changes. Error-prone.

How gx Solves It

gx tracks parent-child relationships between branches and automates the rebasing, pushing, and bookkeeping:

main
  <- feature/auth          PR #1 (3 commits)
    <- feature/tests       PR #2 (2 commits, depends on auth)
      <- feature/dashboard PR #3 (4 commits, depends on tests)

When feature/auth changes (new commits, or after code review updates), gx sync --stack rebases the entire chain in order and pushes all branches.

Workflow

1. Create the stack

gx stack feature/auth main
# ... make commits ...
gx stack feature/tests feature/auth
# ... make commits ...
gx stack feature/dashboard feature/tests

2. Visualize

gx graph

3. Sync after changes

When you update a branch (e.g., address PR review feedback on feature/auth):

gx sync --stack

This rebases feature/tests and feature/dashboard in sequence, then pushes all three.

4. Navigate

Move through the stack without remembering branch names:

gx up       # Move to child
gx down     # Move to parent
gx top      # Jump to leaf
gx bottom   # Jump to first branch above trunk

5. Retarget after merge

When feature/auth is merged into main, retarget feature/tests:

gx retarget feature/tests main

This rebases feature/tests onto main, pushes, updates the config, and retargets the GitHub PR.

6. Clean up

gx sweep         # Delete merged branches
gx nuke --orphans # Delete orphaned branches

The stack.json Format

gx stores stack relationships in .git/gx/stack.json:

{
  "branches": {
    "feature/auth": {
      "parent": "main",
      "parent_head": "a1b2c3d4e5f6..."
    },
    "feature/tests": {
      "parent": "feature/auth",
      "parent_head": "b2c3d4e5f6g7..."
    },
    "feature/dashboard": {
      "parent": "feature/tests",
      "parent_head": "c3d4e5f6g7h8..."
    }
  },
  "metadata": {
    "main_branch": "main",
    "last_updated": "2025-04-15T10:30:00Z"
  }
}

Fields

Field	Description
`branches.<name>.parent`	The parent branch name
`branches.<name>.parent_head`	The parent’s HEAD SHA when the relationship was last synced
`branches.<name>.pr_number`	PR number (optional, for future use)
`metadata.main_branch`	The trunk branch name
`metadata.last_updated`	Timestamp of last config update

Key Properties

Stored in .git/gx/: Not tracked by git, local to each clone
Auto-created: First gx stack call creates it if missing
Read-only tree building: BuildTree reads stack.json but never writes to it. Only explicit commands (gx stack, gx sync, gx retarget) modify the config
Explicit only: Only branches explicitly created with gx stack appear in the graph. There is no auto-discovery of relationships
parent_head for precise rebasing: The parent_head field stores the parent’s HEAD SHA at the time of the last sync, enabling precise --onto rebasing

Command	Purpose
`gx init`	Initialize stacking config
`gx stack`	Create a stacked branch
`gx sync`	Rebase and push the stack
`gx retarget`	Move a branch to a new parent
`gx graph`	Visualize the stack tree
`gx up/down/top/bottom`	Navigate the stack
`gx parent`	Print parent branch (for scripting)
`gx view`	Stack overview with PR status
`gx handoff`	Generate branch summary

​The Problem

​How gx Solves It

​Workflow

​1. Create the stack

​2. Visualize

​3. Sync after changes

​4. Navigate

​5. Retarget after merge

​6. Clean up

​The stack.json Format

​Fields

​Key Properties

​Related Commands