You can view the existing metadata coverage by release using METADATA_SUPPORT.md
- It can be updated by running
- The script runs daily via github actions. You only need to run it if you want to see the results of your registry changes.
yarn update-supported-metadata
Got questions?
- If you work for Salesforce, #platform-cli.
- If not, open an issue
NOTE: If you have a “Settings” metadata type, you do not need to follow these instructions. Support for your type is already built into SDR.
To simplify modifying the registry, there's a script.
Note: The script depends on being able to create an org that will have your type. If you have a working org, but your type isn't compatible with scratch org, you can use an existing org
Note: The script will use your default dev hub. If you don't have one, set one up. If you want to use a different hub, update your default.
The script will:
- Look for missing types (similar to the completeness test)
- For missing types, generate a project and scratch org that includes the Features/Settings
- Run
sf org list metadata-types
to get the describe - Modify the registry to include the newly found types
yarn update-registry YourTypeNameHere
You can set the environment variable SF_ORG_API_VERSION
if you want to specify an API version.
inFolderTypes and types with childXml in their describe are not supported. You want to explore the various strategies for those (see the SDR Handbook in this repo) and create tests to validate that your types are being handled correctly.
For types that contain multiple files (e.g., bundles such as LWC, Aura, ExperienceBundle) where deleting 1 file should not delete the entire component you should set supportsPartialDelete
on the type definition. For an example of this see the ExperienceBundle
definition in metadataRegistry.json
.
For those situations, refer to another existing type in the registry that you want yours to behave like.
If that's confusing, it's a great time to reach out to the CLI team.
You can do what the script does yourself. As you work, run ./node_modules/mocha/bin/mocha test/registry/registryValidation.test.ts
to check your entries
Run sf org list metadata-types --json
to get the describe. >
the output to a file or pipe it to jq (| jq
) to find your type.
Your describe will contain something like this
{
"directoryName": "yourTypes",
"inFolder": "false",
"metaFile": "false",
"suffix": "yourtype",
"xmlName": "YourNewType"
},
Entries in the describe.json
have the following shape:
Property | Description |
---|---|
childXmlNames | List of child sub-components for this component |
directoryName | The name of the directory in the .zip file that contains this component. |
inFolder | Indicates whether the component is in a folder (true) or not (false). For example, documents, email templates and reports are stored in folders. |
metaFile | Indicates whether the component requires an accompanying metadata file. For example, documents, classes, and s-controls are components that require an additional metadata file. |
suffix | The file suffix for this component. |
xmlName | The name of the root element in the metadata file for this component. This name also appears in the Packages > types > name field in the manifest file package.xml. |
Create an entry in the types
property:
this fictitious type COULD HAVE used the automated script
"yournewtype": { // the key is lowercased xmlName
"id": "yournewtype", // lowercased xmlName
"name": "YourNewType", // xmlName
"suffix": "yourtype", // suffix
"directoryName": "yourTypes", //directoryName
"inFolder": false, // inFolder
"strictDirectoryName": false // set to true when suffixes are not unique in the registry
},
add a matching entry in suffixes
"yourtype": "yournewtype" // "suffix": "id"
If your type uses a suffix already in use, tell SDR to use the directoryName
as well by setting "strictDirectoryName"
: "true"
Then also add an entry in the strictDirectoryNames
section.
"yourtype" : "yournewtype" // "directory" : "id"
To preserve existing behavior, use strictDirectoryName
on the new types, not the old ones.
SDR includes 2 registry-related tests to check your changes to the metdataRegistry.json
yarn mocha test/registry/registryValidation.test.ts
Test failures here could be types that exist in the types
section but don't have entries in suffixes
or strictDirectoryNames
.
It also checks that suffixes are unique OR only one type that shares a suffix isn't strictDirectoryName
.
yarn test:registry
The library uses the registry file to resolve how to process metadata types. This needs to be updated as new metadata types are added to the platform at major releases.
The completeness is checked by comparing the registry to the metadata coverage report, but excluding
- Types that aren't supported in the metadata API
- Types in the nonSupportedTypes file (think of it as a registry-ignore file). You can ignore the types themselves, or the feature/settings they depend on. Be sure to explain why you're choosing to ignore that type.
If you find your types (or the features they require) excluded by nonSupportedTypes.ts
but think they're ready to go, feel free to remove them from the list.
There are 2 main ways this happens (or ways to make this work if it currently isn't)
- A feature is available to scratch orgs when it previously wasn't
- The metadata coverage report's "settings" were not sufficient to enable your type to appear in the
describe
call.
Fixing those problems not only makes it easier to automate your type's support in the library, but also makes your type usable by customers (features) and fixes your documentation (coverageReport).
Want to make sure your types are working as expected?
- Create a new project with
sf project generate -n registryTest
- Create a scratch org
sf org create scratch
- Open the org and create your types.
- Run
sf project deploy preview
and verify the remote add. - Run
sf project retrieve start
to pull the metadata and examine what is retrieved - Run
sf project deploy preview
and verify the changes were retrieved and no longer appear. - Delete the org
sf org delete scratch --no-prompt
- Create a new scratch org.
sf org create scratch
- Push the source
sf project deploy start
- Convert the source to mdapi format
sf project convert source -d mdapiOut
- Look in the resulting
mdapiOut
directory and thepackage.xml
to see that it looks as expected - Deploy it to the org using
sf project deploy start --metadata-dir mdapiOut --wait 30
and verify that it succeeds - Delete the source directory
rm -rf force-app/main/default/*
- Create a new scratch org and convert the source back
- Convert back from mdapi to source format
sf project convert mdapi -r mdapiOut -d force-app
sf project deploy start
Target types must be MDAPI addressable on the server. If they aren’t MDAPI addressable, special code is needed to support source tracking for these components. See the document Metadata API Types: End to End, Cradle to Grave (Salesforce internal only) for more details.
Reach out to the CLI team for help with unit tests.
metadataResolver.ts is an example of unit testing one behavior (resolving from source files) of a real metadata type.
If you're doing anything complex (you've used any of the following properties strategies
, folderType
, inFolder=true
, ignoreParsedFullName
, folderContentType
, ignoreParentName
), you'll want to add some NUTs (not-unit-tests) that verify the behavior or your types using real orgs and to prevent SDR changes from causing regressions on your types.
This NUT validates the behavior for a particularly bizarre metadataType, territory2.
NUTs live in plugin-deploy-retrieve but run on branches to SDR.
See testkit for examples and usage.
If you see a whole bunch of new unsupported types, you can "ignore" all the features and work through them in chunks (uncomment a feature at a time) using nonSupportedTypes.ts
If you want to update the registry for only a subset of the currently missing metadata types, add your types as arguments to the script.
# normal, update all missing types
yarn update-registry
# only update the 2 types listed
yarn update-registry AssessmentQuestion AssessmentQuestionSet
Some metadata types require features which require modifications to the DevHub (licenses, etc) and some may have to stay ignored (ex: a pilot feature you can't enable)
You can use an existing org for the metadata describe portion of the script by
- setting its alias to
registryBuilder
- setting the env
RB_EXISTING_ORG
ex:RB_EXISTING_ORG=true yarn update-registry
1. A sfdx project must exist in local.
`sf project generate --name <projectname> --default-package-dir <directory> -x`
2. An authorized devhub org must exists
`sf org login web -a <alias> -r <localhost url> -d`
3. A scratch org must exists with alias `registryBuilder`
1. Update `project-scratch-def.json` as per your requirements.
2. `sf org create scratch -f config/project-scratch-def.json -a registryBuilder -d`
1. Fork SourceDeployRetrieve github repo
(https://github.com/forcedotcom/source-deploy-retrieve)
2. Clone forked repo in local and checkout a new branch
3. Setup Yarn
1. Go to the repo directory
2. Run `yarn install && yarn build`
4. Setup an environment variable by executing command
`export RB_EXISTING_ORG=true`
5. Execute yarn update command for required metadata entities
`yarn update-registry <MetadataEntity1> <MetadataEntity2>`
6. Check if respective file(`src/registry/metadataRegistry.json`)
got updated. `git status`
Now changes are available in local, we have to link the registry with sfdx project
8. From SDR git repo directory, run `yarn build; yarn link`
9. Clone plugin repo (https://github.com/salesforcecli/plugin-deploy-retrieve)
10. Execute `yarn link @salesforce/source-deploy-retrieve` and `sfdx plugins:link .` and `yarn build` from cloned plugin repo directory
Registry has been set for your entities, now you can run (e.g.) `sf project deploy start` command for your entities:
Proceed to `Manual Testing` section above in this document.