Manage Unpackaged Configuration#

Not everything that’s part of an application can be part of a package.

CumulusCI implements the Product Delivery Model by offering support for complex applications – applications that may include multiple managed packages as well as unpackaged metadata, and setup automation that configures org settings or makes precise changes to existing configuration.

The tools used to implement that support are unpackaged metadata and Metadata ETL.

Unpackaged metadata refers to metadata that is not delivered as part of a package, and can include both support metadata delivered to users as well as metadata that operationally configures orgs used by the product.

Metadata ETL is a suite of tasks that supports surgically altering existing metadata in an org. It’s a powerful technique that alters the unpackaged configuration in an org without risking damage to existing customizations by overwriting them with incoming metadata. Metadata ETL is relevant for delivering applications to customers safely, and is often a superior alternative to unpackaged metadata.

To learn more, see Metadata ETL.

Roles of Unpackaged Metadata#

unpackaged/pre: Prepare an Org#

Some projects require that unpackaged metadata be deployed to finish the customization of an org before the package’s own code and metadata are deployed.

For example, the Nonprofit Success Pack (NPSP) must deploy unpackaged Record Types prior to installing its own packages. unpackaged/pre is the location designed for such metadata, which is stored in subdirectories such as unpackaged/pre/first.

CumulusCI’s standard flows that build orgs, such as dev_org and install_prod, always deploy metadata bundles found in unpackaged/pre before proceeding to the deployment of the application. It’s also easy to include unpackaged/pre metadata in customer-facing installers run via MetaDeploy.

The deploy_pre task, which is part of the dependencies flow, is responsible for deploying these bundles.

Important

Do not include metadata in unpackaged/pre unless it is intended to be delivered to all installations of the product.

unpackaged/post: Configuration After Package Install#

Projects often include metadata that is genuinely part of the application, but cannot be delivered as part of a managed package for operational reasons. This metadata must be deployed after the package’s own code and metadata are deployed first and the org is configured.

For example, a product can’t deliver TopicsForObjects metadata as part of a managed package because that type of metadata isn’t packageable. unpackaged/post is the home for this kind of metadata, which is stored in subdirectories such as unpackaged/post/first.

Note

To learn more about which components are packageable, see the Metadata Coverage Report.

CumulusCI’s standard flows that build orgs, such as dev_org and install_prod, always deploy metadata bundles found in unpackaged/post, making it a full-fledged part of the application. It’s also easy to include unpackaged/post metadata in customer-facing installers run via MetaDeploy.

The deploy_post task, which is part of the config_dev, config_qa, and config_managed flows, is responsible for deploying these bundles.

Important

Do not include metadata in unpackaged/post unless it is intended to be delivered to all environments (both managed installations and unmanaged deployments). It’s also critical for managed package projects that this metadata include namespace tokens (see namespace injection).

unpackaged/config: Tailor an Org#

Projects can come with more than one supported configuration in their CumulusCI automation. For example, projects often support distinct, tailored dev_org, qa_org, and install_prod flows, each of which performs a unique setup for their specific use case.

Unpackaged metadata stored in unpackaged/config is a tool to support operational needs that tailor orgs to different configurations. For instance, a testing-oriented scratch org may need to deploy a customized set of Page Layouts to help testers easily visualize data under test. Such page layouts are stored in unpackaged/config/qa.

Unpackaged Metadata Folder Structure#

All unpackaged metadata is stored in the unpackaged directory tree, which contains these top-level directories.

  • unpackaged/pre

  • unpackaged/post

  • unpackaged/config

These trees contain metadata bundles in Metadata API or Salesforce DX format. CumulusCI automatically converts Salesforce DX-format unpackaged bundles to Metadata API format before deploying them.

Namespace Injection#

Projects that build managed packages often construct their unpackaged metadata to be deployable in multiple contexts, such as:

  • Unmanaged deployments, such as developer orgs.

  • Unmanaged namespaced scratch orgs.

  • Managed contexts, such as a beta test org or a demo org created with install_prod.

For example, metadata located in unpackaged/post is deployed after the application code in both unmanaged and managed contexts. If that metadata contains references to the application components, it must be deployable when that metadata is namespaced (in a managed context or namespaced scratch org) and when it is not (in an unmanaged context).

CumulusCI uses a strategy called namespace injection to support this use case. Namespace injection is very powerful, and requires care from the project team to ensure that metadata remains deployable in all contexts.

Important

Projects that are building an org implementation or a non-namespaced package do not have a namespace, or a distinction between managed and unmanaged contexts. These projects typically don’t need to use namespace injection.

Metadata files where a namespace is conditionally applied to components for insertion into different contexts must replace the namespace with a token, which CumulusCI replaces with the appropriate value or with an empty string as appropriate to the context.

  • %%%NAMESPACE%%% is replaced with the package’s namespace in any context with a namespace (such as a namespaced org or managed org). Otherwise, it remains blank.

  • %%%NAMESPACED_ORG%%% is replaced with the package’s namespace in a namespaced org only, not in a managed installation. Otherwise, it remains blank.

Note

This token supports use cases where components in one unpackaged metadata bundle refer to components in another, and the dependency bundle acquires a namespace by being deployed into a namespaced org.

  • %%%NAMESPACE_OR_C%%% is replaced with the package’s namespace in any context with a namespace (such as a namespaced org or managed org). Otherwise, it is replaced with c, the generic namespace used in Lightning components.

  • %%%NAMESPACED_ORG_OR_C%%% is replaced with the package’s namespace in a namespaced org only, not in a managed installation. Otherwise, it is replaced with c, the generic namespace used in Lightning components.

  • %%%NAMESPACE_DOT%%% is replaced with the package’s namespace in any context with a namespace (such as a namespaced org or managed org) followed by a period (.) rather than two underscores.

Note

This token is used to construct references to packaged Record Types and Apex classes.

An example case for namespace injection can be found in Salesforce.org’s Nonprofit Success Pack (NPSP) managed package. A portion of metadata from NPSP is stored in a subdirectory under unpackaged/post, meaning it’s deployed after the application metadata. This metadata updates a Compact Layout on the Account object, and references packaged metadata from the application as well as from other managed packages. To deploy this as a managed context, this metadata requires the use of namespace tokens to represent the npsp namespace, letting CumulusCI automatically adapt the metadata to deploy into managed and unmanaged contexts.

<?xml version="1.0" encoding="UTF-8"?>
<CustomObject xmlns="http://soap.sforce.com/2006/04/metadata">
    <compactLayouts>
        <fullName>NPSP_Household_Account</fullName>
        <fields>Name</fields>
        <fields>npo02__TotalOppAmount__c</fields>
        <fields>%%%NAMESPACE%%%Number_of_Household_Members__c</fields>
        <label>NPSP Household Account</label>
    </compactLayouts>
</CustomObject>

Note that only the reference to the NPSP field Number_of_Household_Members__c is tokenized. (When installed as part of the managed package, this field appears as npsp__Number_of_Household_Members__c.) References to NPSP’s own managed package dependency, npo02, are not tokenized because this metadata is always namespaced when installed.

If this metadata isn’t tokenized, it fails to deploy into an org containing NPSP as a beta or released managed package (because in that context the field Number_of_Household_Members__c is namespaced as npsp__Number_of_Household_Members__c, and must be referred to as such).

Note

The resolution of component references in namespaced scratch orgs and in managed installations of the same metadata are not identical. Metadata that is tokenized and deploys cleanly in a namespaced scratch org can still fail in a managed context.

Configuration#

Most CumulusCI tasks can intelligently determine whether or not to inject the namespace based on the target org. For example, if tokenized metadata is being deployed into an org that contains the project installed as a managed package, CumulusCI knows to inject the namespace; otherwise, it replaces namespace tokens with an empty string for an unmanaged installation.

You can also specify explicit configuration for namespace injection in circumstances where CumulusCI’s automatic functionality does not meet your needs, such as when deploying tokenized metadata from another project. If the metadata you are deploying has been tokenized, and you want to deploy metadata with a namespace, use the namespace_inject: <namespace> option to inject the namespace.

project:
    dependencies:
        - zip_url: https://github.com/SalesforceFoundation/EDA/archive/master.zip
          subfolder: EDA-master/dev_config/src/admin_config
          namespace_inject: hed

The metadata in the zip contains the string tokens %%%NAMESPACE%%% and ___NAMESPACE___ which is replaced with hed__ before the metadata is deployed.

To deploy tokenized metadata without any namespace references, specify both namespace_inject: <namespace> and unmanaged: True. In this example, we do just this for the EDA dependency.

project:
    dependencies:
        - zip_url: https://github.com/SalesforceFoundation/EDA/archive/master.zip
          subfolder: EDA-master/dev_config/src/admin_config
          namespace_inject: hed
          unmanaged: True

The namespace tokens are replaced with an empty string instead of the namespace, effectively stripping the tokens from the files and filenames.

Retrieve Unpackaged Metadata#

CumulusCI provides tasks to retrieve changes to unpackaged metadata, just as with packaged metadata.

When working with unpackaged metadata, it’s important to maintain awareness of key considerations related to retrieving metadata that is not part of the main application.

  • Take care to separate your development between the different bundles you wish to retrieve. For example, if you have changes to make in the application as well as in unpackaged metadata, complete the application changes first, retrieve them, and then make the unpackaged changes and retrieve those. If you conflate changes to components that live in separate elements of your project, it’s difficult to untangle them.

  • Whenever possible, build your unpackaged metadata in an org that contains a beta or released managed package. By doing so, the metadata contains namespaces when extracted, which CumulusCI easily replaces with tokens when retrieving metadata. It’s difficult to manually tokenize metadata that’s retrieved from an unmanaged org without namespaces.

After building changes to unpackaged metadata in a managed org, retrieve it using the retrieve_changes task with the additional namespace_tokenize option, and use the path option to direct the retrieved metadata to your desired unpackaged directory.

In the following example, we run the retrieve_changes task to retrieve metadata changes into the unpackaged/config/qa subdirectory, and replace references to the namespace npsp with the appropriate token.

$ cci task run retrieve_changes --path unpackaged/config/qa --namespace_tokenize npsp

Projects that use unpackaged metadata extensively define retrieve tasks to streamline this process.

For example, here is a custom task that retrieves changes to specific directory where metadata for QA configuration is kept.

retrieve_qa_config:
    description: Retrieves changes to QA configuration metadata
    class_path: cumulusci.tasks.salesforce.sourcetracking.RetrieveChanges
    options:
        path: unpackaged/config/qa
        namespace_tokenize: $project_config.project__package__namespace

The retrieve_changes task retrieves unpackaged metadata in a managed org, but in this case you must manually insert namespace tokens to deploy metadata in a managed or namespaced context.

Customize Config Flows#

Projects often customize new tasks that deploy unpackaged/config bundles, and harness these tasks in flows.

Projects that use unpackaged/config/qa often define a deploy_qa_config task.

deploy_qa_config:
    description: Deploys additional fields used for QA purposes only
    class_path: cumulusci.tasks.salesforce.Deploy
    options:
        path: unpackaged/config/qa

This task is then added to relevant flows, such as config_qa.

config_qa:
    steps:
        3:
            task: deploy_qa_config

In most cases, CumulusCI intelligently determines whether or not to inject the namespace. It’s rarely necessary to explicitly configure an injection mode. If you need to do so, use the unmanaged option:

config_regression:
    steps:
        3:
            task: deploy_qa_config
            options:
                unmanaged: False

For more details on customizing tasks and flows, see the Configure CumulusCI section.