5.2 KiB
title |
---|
Sparo profiles |
Background
Git's sparse checkout feature normally relies on a collection of glob patterns that are stored in the .git/info/sparse-checkout
config file. The Git maintainers found that regular glob syntax was too inefficient, so they introduced a "cone mode" glob interpretation that ignores file-matching patterns and only matches directories.
The syntax looks something like this:
.git/info/sparse-checkout example
/*
!/*/
/apps/
!/apps/*/
/apps/my-app/
!/apps/my-app/*/
/apps/my-app/_/
To simplify management, Git also provides a git sparse-checkout
command that simplifies the syntax for adding/removing patterns from this file. However, in a large monorepo with hundreds of projects, managing these globs would nonetheless be confusing and error-prone.
Sparo improves sparse checkout
Sparo makes life easier by generating the .git/info/sparse-checkout
configuration automatically from config files called profiles. This offers many benefits:
-
Sparo profiles are defined using project selectors, for example: "Give me app1, app2, and all the projects needed to build them." This is more concise and maintainable than specifying globs.
-
Profiles are stored in a config file and committed to Git. This makes it easy to share them with your teammates.
-
Profiles are automatically updated when switching between branches, which ensures deterministic results. For example, when checking out a very old branch, you want the old profile definition, not today's version of it.
-
You can combine multiple profiles together (
sparo checkout --profile team1 --profile team2
), which selects the union of their projects. This is useful for example when modifying a library project that is consumed by projects belonging to several other teams. You could check out their projects using--from the-library
of course, but it's likely those other teams will have included other relevant projects in their profiles. -
Sparo avoids common mistakes by imposing additional restrictions beyond
git sparse-checkout
. This avoids mistakes such as trying to switch to a profile that is missing a project folder containing files that are locally modified. It is better for users to stash or commit such modifications first.
Best practices for profiles
You an add JSON comments to your profile config files. In a large shared codebase, we recommend adding a standardized header to the top of your files indicating their ownership and purpose. Something like this:
common/sparo-profiles/example-profile.json
/**
* OWNER: Customer service team
* PURPOSE: Use this profile when working on the customer service apps.
*/
{
"$schema": "https://tiktok.github.io/sparo/schemas/sparo-profile.schema.json",
/**
* A list of Rush project selectors indicating the project folders to be
* included for sparse checkout. The selectors will be combined to make
* the union superset of projects. See the Rush selector docs for details:
* https://rushjs.io/pages/developer/selecting_subsets/
*/
"selections": [
{
"selector": "--to",
"argument": "tag:cs-dashboard"
},
{
"selector": "--to",
"argument": "tag:cs-tools"
}
]
}
Combining profiles
The simple way to combine profiles is to specify --profile
multiple times. For example:
# Check out the union of profiles team-a.json, team-b.json, team-c.json
# NOTE: This will replace whatever profile selection was already checked out.
sparo checkout --profile team-a --profile team-b --profile team-c
You can also use --add-profile
to incrementally combine them. For example:
# These three commands are equivalent to the above command.
sparo checkout --profile team-a
sparo checkout --add-profile team-b
sparo checkout --add-profile team-c
How to checkout no profile at all? That is, how to return to the initial state of a clean sparo clone
that only includes the skeleton folders? The answer is to use the --no-profile
parameter:
# NOT IMPLEMENTED YET - check out just the skeleton folders
# without applying any profiles
sparo checkout --no-profile
If sparo checkout
without --profile
or --add-profile
or --no-profile
, then the existing profile selection is preserved. In other words, your profile choices are generally "sticky" across commands.
Querying profiles
Users can discover available profiles in the current branch by invoking the sparo list-profiles command. The --project
parameter enables you to query relevant profiles for a given project. For example:
# Suppose you need to make a fix for the "example-app" project.
# Which sparse checkout profiles include the "example-app" project?
sparo list-profiles --project example-app
# Great, let's add the "example-profile" result to our current checkout
# (combining it with the existing profile).
sparo checkout --add-profile example-profile
See also
- <profile-name>.json config file