Understanding the entity_lookup and entity_generate process plugins from Migrate Tools

In recent posts we have explored the Migrate Plus and Migrate Tools modules. They extend the Migrate API to provide migrations defined as configuration entities, groups to share configuration among migrations, a user interface to execute migrations, among other things. Yet another benefit of using Migrate Plus is the option to leverage the many process plugins it provides. Today, we are going to learn about two of them: `entity_lookup` and `entity_generate`. We are going to compare them with the `migration_lookup` plugin, show how to configure them, and explain their compromises and limitations. Let’s get started.

What is the difference among the migration_lookup, entity_lookup, entity_generate plugins?

In the article about migration dependencies we covered the `migration_lookup` plugin provided by the core Migrate API. It lets you maintain relationships among entities that are being imported. For example, if you are migrating a node that has associated users, taxonomy terms, images, paragraphs, etc. This plugin has a very important restriction: the related entities must come from another migration. But what can you do if you need to reference entities that already exists system? You might already have users in Drupal that you want to assign as node authors. In that case, the `migration_lookup` plugin cannot be used, but `entity_lookup` can do the job.

The `entity_lookup` plugin is provided by the Migrate Plus module. You can use it to query any entity in the system and get its unique identifier. This is often used to populate entity reference fields, but it can be used to set any field or property in the destination. For example, you can query existing users and assign the `uid` node property which indicates who created the node. If no entity is found, the module returns a `NULL` value which you can use in combination of other plugins to provide a fallback behavior. The advantage of this plugin is that it does not require another migration. You can query any entity in the entire system.

The `entity_generate` plugin, also provided by the Migrate Plus module, is an extension of `entity_lookup`. If no entity is found, this plugin will automatically create one. For example, you might have a list of taxonomy terms to associate with a node. If some of the terms do not exist, you would like to create and relate them to the node.

Note: The `migration_lookup` offers a feature called stubbing that neither `entity_lookup` nor `entity_generate` provides. It allows you to create a placeholder entity that will be updated later in the migration process. For example, in a hierarchical taxonomy terms migration, it is possible that a term is migrated before its parent. In that case, a stub for the parent will be created and later updated with the real data.

Getting the example code

You can get the full code example at https://github.com/dinarcon/ud_migrations The module to enable is `UD Config entity_lookup and entity_generate examples` whose machine name is `ud_migrations_config_entity_lookup_entity_generate`. It comes with one JSON migrations: `udm_config_entity_lookup_entity_generate_node`. Read this article for details on migrating from JSON files. The following snippet shows a sample of the file:

{
  "data": {
    "udm_nodes": [
      {
        "unique_id": 1,
        "thoughtful_title": "Amazing recipe",
        "creative_author": "udm_user",
        "fruit_list": "Apple, Pear, Banana"
      },
      {...},
      {...},
      {...}
    ]
  }
}

Additionally, the example module creates three users upon installation: ‘udm_user’, ‘udm_usuario’, and ‘udm_utilisateur’. They are deleted automatically when the module is uninstalled. They will be used to assign the node authors. The example will create nodes of types “Article” from the standard installation profile. You can execute the migration from the interface provided by Migrate Tools at `/admin/structure/migrate/manage/default/migrations`.

Using the entity_lookup to assign the node author

Let’s start by assigning the node author. The following snippet shows how to configure the `entity_lookup` plugin to assign the node author:

uid:
  - plugin: entity_lookup
    entity_type: user
    value_key: name
    source: src_creative_author
  - plugin: default_value
    default_value: 1

The `uid` node property is used to assign the node author. It expects an integer value representing a user ID (`uid`). The source data contains usernames so we need to query the database to get the corresponding user IDs. The users that will be referenced were not imported using the Migrate API. They were already in the system. Therefore, `migration_lookup` cannot be used, but `entity_lookup` can.

The plugin is configured using three keys. `entity_type` is set to machine name of the entity to query: `user` in this case. `value_key` is the name of the entity property to lookup. In Drupal, the usernames are stored in a property called `name`. Finally, `source` specifies which field from the source contains the lookup value for the `name` entity property. For example, the first record has a `src_creative_author` value of `udm_user`. So, this plugin will instruct Drupal to search among all the users in the system one whose `name` (username) is `udm_user`. If a value if found, the plugin will return the user ID. Because the `uid` node property expects a user ID, the return value of this plugin can be used directly to assign its value.

What happens if the plugin does not find an entity matching the conditions? It returns a `NULL` value. Then it is up to you to decide what to do. If you let the `NULL` value pass through, Drupal will take some default behavior. In the case of the `uid` property, if the received value is not valid, the node creation will be attributed to the anonymous user (uid: 0). Alternatively, you can detect if `NULL` is returned and take some action. In the example, the second record specifies the “udm_not_found” user which does not exists. To accommodate for this, a process pipeline is defined to manually specify a user if `entity_lookup` did not find one. The `default_value` plugin is used to return `1` in that case. The number represents a user ID, not a username. Particularly, this is the user ID of “super user” created when Drupal was first installed. If you need to assign a different user, but the user ID is unknown, you can create a pseudofield and use the `entity_lookup` plugin again to finds its user ID. Then, use that pseudofield as the default value.

Important: User entities do not have bundles. Do not set the `bundle_key` nor `bundle` configuration options of the `entity_lookup`. Otherwise, you will get the following error: “The entity_lookup plugin found no bundle but destination entity requires one.” Files do not have bundles either. For entities that have bundles like nodes and taxonomy terms, those options need to be set in the `entity_lookup` plugin.

Using the entity_generate to assign and create taxonomy terms

Now, let’s migrate a comma separated list of taxonomy terms. An example value is `Apple, Pear, Banana`.  The following snippet shows how to configure the `entity_generate` plugin to look up taxonomy terms and create them on the fly if they do not exist:

field_tags:
  - plugin: skip_on_empty
    source: src_fruit_list
    method: process
    message: 'No src_fruit_list listed.'
  - plugin: explode
    delimiter: ','
  - plugin: callback
    callable: trim
  - plugin: entity_generate
    entity_type: taxonomy_term
    value_key: name
    bundle_key: vid
    bundle: tags

The terms will be assigned to the `field_tags` field using a process pipeline of four plugins:

• `skip_on_empty` will skip the processing of this field if the record does not have a `src_fruit_list` column.
• `explode` will break the string of comma separated files into individual elements.
• `callback` will use the `trim` PHP function to remove any whitespace from the start or end of the taxonomy term name.
• `entity_generate` takes care of finding the taxonomy terms in the system and creating the ones that do not exist.

For a detailed explanation of the `skip_on_empty` and `explode` plugins see this article. For the `callback` plugin see this article. Let’s focus on the `entity_generate` plugin for now. The `field_tags` field expects an array of taxonomy terms IDs (`tid`). The source data contains term names so we need to query the database to get the corresponding term IDs. The taxonomy terms that will be referenced were not imported using the Migrate API. And they might exist in the system yet. If that is the case, they should be created on the fly. Therefore, `migration_lookup` cannot be used, but `entity_generate` can.

The plugin is configured using five keys. `entity_type` is set to machine name of the entity to query: `taxonomy_term` in this case. `value_key` is the name of the entity property to lookup. In Drupal, the taxonomy term names are stored in a property called `name`. Usually, you would include a `source` that specifies which field from the source contains the lookup value for the `name` entity property. In this case it is not necessary to define this configuration option. The lookup value will be passed from the previous plugin in the process pipeline. In this case, the trimmed version of the taxonomy term name.

If, and only if, the entity type has bundles, you also must define two more configuration options: `bundle_key` and `bundle`. Similar to `value_key` and `source`, these extra options will become another condition in the query looking for the entities. `bundle_key` is the name of the entity property that stores which bundle the entity belongs to. `bundle` contains the value of the bundle used to restrict the search. The terminology is a bit confusing, but it boils down to the following. It is possible that the same value exists in multiple bundles of the same entity. So, you must pick one bundle where the lookup operation will be performed. In the case of the taxonomy term entity, the bundles are the vocabularies. Which vocabulary a term belongs to is associated in the `vid` entity property. In the example, that is `tags`. Let’s consider an example term of “Apple”. So, this plugin will instruct Drupal to search for a taxonomy term whose `name` (term name) is “Apple” that belongs to the “tags” `vid` (vocabulary).

What happens if the plugin does not find an entity matching the conditions? It will create one on the fly! It will use the value from the source configuration or from the process pipeline. This value will be used to assign the `value_key` entity property for the newly created entity. The entity will be created in the proper bundle as specified by the `bundle_key` and `bundle` configuration options. In the example, the terms will be created in the `tags` vocabulary. It is important to note that values are trimmed to remove whispaces at the start and end of the name. Otherwise, if your source contains spaces after the commas that separate elements, you might end up with terms that seem duplicated like “Apple” and ” Apple”.

More configuration options

Both `entity_lookup` and `entity_generate` share the previous configuration options. Additionally, the following options are only available:
`ignore_case` contains a boolean value to indicate if the query should be case sensitive or not. It defaults to true.
`access_check` contains a boolean value to indicate if the system should check whether the user has access to the entity. It defaults to true.
`values` and `default_values` apply only to the `entity_generate` plugin. You can use them to set fields that could exist in the destination entity. An example configuration is included in the code for the plugin.

One interesting fact about these plugins is that none of the configuration options is required. The `source` can be skipped if the value comes from the process pipeline. The rest of the configuration options can be inferred by code introspection. This has some restrictions and assumptions. For example, if you are migrating nodes, the code introspection requires the `type` node property defined in the process section. If you do not set one because you define a `default_bundle` in the destination section, an error will be produced. Similarly, for entity reference fields it is assumed they point to one bundle only. Otherwise, the system cannot guess which bundle to lookup and an error will be produced. Therefore, always set the `entity_type` and `value_key` configurations. And for entity types that have bundles, `bundle_key` and `bundle` must be set as well.

Note: There are various open issues contemplating changes to the configuration options. See this issue and the related ones to keep up to date with any future change.

Compromises and limitations

The `entity_lookup` and `entity_generate` plugins violate some ETL principles. For example, they query the destination system from the process section. And in the case of `entity_generate` it even creates entities from the process section. Ideally, each phase of the ETL process is self contained. That being said, there are valid uses cases to use these plugins and they can you save time when their functionality is needed.

An important limitation of the `entity_generate` plugin is that it is not able to clean after itself. That is, if you rollback the migration that calls this plugin, any created entity will remain in the system. This would leave data that is potentially invalid or otherwise never used in Drupal. Those values could leak into the user interface like in autocomplete fields. Ideally, rolling back a migration should delete any data that was created with it.

The recommended way to maintain relationships among entities in a migration project is to have multiple migrations. Then, you use the `migration_lookup` plugin to relate them. Throughout the series, several examples have been presented. For example, this article shows how to do taxonomy term migrations.

What did you learn in today’s blog post? Did you know how to configure these plugins for entities that do not have bundles? Did you know that reverting a migration does not delete entities created by the `entity_generate` plugin? Did you know you can assign fields in the generated entity? Share your answers in the comments. Also, I would be grateful if you shared this blog post with others.