[email protected]
Unlocked Packages
All the details about unlocked package

There is a huge amount of documentation on unlocked packages. Rather than repeating all the information here, a curated list of links are provided which every [email protected] practitioner should be well versed with
The Basics
Advanced Materials
The following sections deal with items that are particular to [email protected] or more emphasis is required in large scale programs.

  • If you have the following scenarios, it may be a good idea to put them in domain specific packages.
    • A group of related code and customization
    • Independent from other components and can be called from other packages
    • Standalone and released independently
  • Don’t package the metadata merely because it is supported by Unlocked Packaging. Some components (e.g. profiles) will need to be managed outside of unlocked packages and it may be preferable to deploy other related metadata using the same mechanism. In other cases, metadata like non-critical reports, email templates etc, may be better to manage outside of source control entirely to allow these to be changed dynamically by end users.
  • Metadata acting on components in a package should be scoped where possible to operate within the confines of the package so these can be included too. Examples include:
    • Permission Sets, which should be granular and specific to a function, for example ability to create customers (create/edit accounts and contacts) or ability to administrate customers (delete/view all accounts and contacts). It's also important that permission sets are composable to reduce the impact of changes - for example including field security for all non-sensitive fields of an object/package in a single permission set simplifies the process of creating a new field, as permissions for this field can be added to just this permission set
    • Flows, which should be small and modular
    • Triggers and Trigger Handler Classes, which should be implemented through a dependency injection framework, and scoped to be lightweight, calling service classes from common packages where necessary
  • Objects should ideally be "owned" by one domain specific package which includes the majority of fields, validation rules and other metadata related to this object. This allows for simpler governance around core object changes and ensures that when metadata is added to an object this is pulled by SFDX into the owning package by default.
  • Workflows and Process Builders need to be placed in the same unlocked package that contains the parent object definition. This is a restriction of this particular metadata component. In these scenarios rather than building automation using workflows and process builders, it is better to use apex or flow
  • Custom Labels: Group and manage custom labels for each package separately to ensure they don't cause deployment errors. sfpowerkit provides some tooling around to work with maintaining custom labels. Check the command here and here. Ensure you use sfpowerkit label create command to create these labels. Custom labels should be used for their intended purpose to support translation and renaming of standard components. Do not use them to store constants and other hard coded values.
  • Usually some metadata will be cross-cutting, and these components should generally be moved to global packages such as src-access-mgmt and src-ui:
    • Permission Set Groups, which should collate granular permission sets from multiple packages into a specific access level
    • Apps, which include tabs from multiple packages
    • Layouts, which can include related lists, buttons and fields referencing metadata outside of the object's package
    • Flexipages, which may include components, actions and lists from across packages

  • Don’t package the metadata that is not supported by Metadata API. Always check the latest Metadata Coverage. Ensure you run the following command during Pull Request Validation / Locally using the following command.
sfdx sfpowerkit:package:valid -n <name_of_package>
  • Profiles should not be included in unlocked packages and need special attention.
  • Reports can only be put in an unlocked if
    • They serve a specific purpose within the package, for example: providing behavioural data of some components
    • They serve as a template that admin can clone

Unlocked Packages, excluding Org-Dependent unlocked packages have mandatory test coverage requirements. Each package should have minimum of 75% coverage requirement. A validated build (or build command in sfpowerscripts) validates the coverage of package during the build phase. To enable the feedback earlier in the process, sfpowerscripts provide you functionality to validate test coverage of a package during the Pull Request Validation process.
Please note that during pull request validation, all apex tests in a package are triggered in parallel and test cases must be written to ensure the guidelines mentioned in the Locking Records developer guide notes.

For unlocked packages, we ask our practitioners to follow a semantic versioning of packages.
Please note Salesforce packages do not support the concept of PreRelease/BuildMetadata. The last segment of a version number is a build number. We recommend to utilize the auto increment functionality provided by Salesforce rather than rolling out your own build number substitution ( Use 'NEXT' while describing the build version of the package and 'LATEST' to the build number where the package is used as a dependency)
Note that an unlocked package must be promoted before it can be installed to a production org, and either the major, minor or patch (not build) version must be higher than the last version of this package which was promoted. These version number changes should be made in the sfdx-project.json file before the final package build and promotion.

Unlocked packages provide traceability in the org by locking down the metadata components to the package that introduces it. This feature which is the main benefit of unlocked package can also create issues when you want to refactor components from one package to another. Let's look at some scenarios and common strategies that need to be applied
For a project that has two packages.
  • Package A and Package B
  • Package B is dependent on Package A.

  • Remove a component from Package A, provided the component has no dependency
  • Solution: Create a new version of Package A with the metadata component being removed and install the package.

  • Move a metadata component from Package A to Package B
  • Solution: This scenario is pretty straight forward, one can remove the metadata component from Package A and move that to Package B. When a new version of Package A gets installed, the following things happen:
    • If the deployment of the unlocked package is set to mixed, and no other metadata component is dependent on the component, the component gets deleted.
    • On the subsequent install of Package B, Package B restores the field and takes ownership of the component.

  • Move a metadata component from Package B to Package A, where the component currently has other dependencies in Package B
  • Solution: In this scenario, one can move the component to Package A and get the packages built. However during deployment to an org, Package A will fail with an error this component exists in Package B. To mitigate this one should do the following:
    • Deploy a version of Package B which removes the lock on the metadata component using deprecate mode. Some times this needs extensive refactoring to other components to break the dependencies. So evaluate whether the approach will work.
    • If not, you can go to the UI (Setup > Packaging > Installed Packages > <Name of Package> > View Components and Remove) and remove the lock for a package.

Package dependencies are defined in the sfdx-project.json. More information on defining package dependencies can be found in the Salesforce docs.
"packageDirectories": [
"path": "util",
"default": true,
"package": "Expense-Manager-Util",
"versionName": "Winter β€˜20",
"versionDescription": "Welcome to Winter 2020 Release of Expense Manager Util Package",
"versionNumber": "4.7.0.NEXT"
"path": "exp-core",
"default": false,
"package": "ExpenseManager",
"versionName": "v 3.2",
"versionDescription": "Winter 2020 Release",
"versionNumber": "3.2.0.NEXT",
"dependencies": [
"package": "ExpenseManager-Util",
"versionNumber": "4.7.0.LATEST"
"package": "TriggerFramework",
"versionNumber": "1.7.0.LATEST"
"package": "External Apex Library -"
"sourceApiVersion": "47.0",
"packageAliases": {
"TriggerFramework": "0HoB00000004RFpLAM",
"Expense Manager - Util": "0HoB00000004CFpKAM",
"External Apex [email protected]": "04tB0000000IB1EIAW",
"Expense Manager": "0HoB00000004CFuKAM"
Let's unpack the concepts utilizing the above example:
  • There are two unlocked packages
    • Expense Manager - Util is an unlocked package in your DevHub, identifiable by 0H in the packageAlias
    • Expense Manager - another unlocked package which is dependent on ' Expense Manager - Util', 'TriggerFramework' and 'External Apex Library -'
  • External Apex Library is an external dependency, it could be a managed package or any unlocked package released on a different Dev Hub. All external package dependencies must be defined with a 04t ID, which can be determined from the installation URL from AppExchange or by contacting your vendor.

Unlocked packages have two build modes, one with skip dependency check and one without. A package being built without skipping dependency check cant be deployed into production and can usually take a long time to build. sfpowerscripts tries to build packages in parallel understanding your dependency, however some of your packages could spend a significant time in validation.
During these situations, we ask you to consider whether the time taken to build all validated packages on an average is within your build budget, If not, here are your options
  • ​Move to org dependent package: Org-dependent unlocked packages are a variant of unlocked packages. Org-dependent packages do not validate the dependencies of a package and will be faster. However please note that all the org's where the earlier unlocked package was installed, had to be deprecated and the component locks removed, before the new org-dependent unlocked package is installed.
  • ​Move to source-package: Use it as the least resort, source packages have a fairly loose lifecycle management.

Create a source package and move the metadata and any associated dependencies over to that particular package.
Copy link
On this page
What to Package
What Not to Package
Unlocked Package and Test Coverage
Managing Version Numbers of Unlocked Package
Deprecating Components from an Unlocked Package
Scenario 1:
Scenario 2:
Scenario 3:
Managing Package Dependencies
Build Options with Unlocked Packages
Handling metadata that is not supported by unlocked packages