(New) Entity Transforms Framework

Background

During Import Process, entity transforms are required to make changes to the entity before it gets committed to the database. These modifications are necessary to make the entity conform to the environment it is going to reside. The Import Process provided a mechanism to do that.

Transformation Framework

A transformation framework allows a mechanism to selectively transform an entity or specific attributes of that entity.

To achieve this, the framework provides:

• Way to set a condition that needs to be satisfied for a transformation to be applied.
• Action to be taken on the entity once the condition is met.

The existing transformation frameworks allowed this to happen.

Reason for New Transformation Framework

While the existing framework provided the basic benefits of the transformation framework, it did not have support for some commonly used Atlas types. Which meant that users of this framework would have to meticulously define transformations for every type they are working with. This can be tedious and potentially error-prone. The new framework addresses this problem by providing built-in transformations for some commonly used types. It can also be extended to accommodate new types.

Approach

The approach used by the new transformation framework creates a transformation by:

• Specifying a condition.
• Specifying action(s) to be taken if the condition is met.

Conditions

Following are built-in conditions.

Actions

Built-in Transforms

Add Classification

During import, hivedb entity whose _qualifiedName is stocks@cl1 will get the classification clSrcImported.

{
    "conditions": {
        "hive_db.qualifiedName": "stocks@cl1"
    },
    "action": {
        "__entity": "ADD_CLASSIFICATION: clSrcImported"
    }
}

Every imported entity will get the classification by simply changing the condition. The __entity is a special condition which matches entity.

{
    "conditions": {
        "__entity": ""
    },
    "action": {
        "__entity": "ADD_CLASSIFICATION: clSrcImported"
    }
}

To add the classification to only the top-level entity (an entity that is used as a starting point for an export), use:

{
    "conditions": {
        "__entity": "topLevel:"
    },
    "action": {
        "__entity": "ADD_CLASSIFICATION: clSrcImported"
    }
}

Replace Prefix

This action works on string values. The first parameter is the prefix that is searched for a match, once matched, it is replaced with the provided replacement string.

The sample below searches for /aa/bb/, once found replaces it with /xx/yy/.

{
    "conditions": {
        "hdfs_path.clusterName": "EQUALS: CL1"
    },
    "action": {
        "hdfs_path.path": "REPLACE_PREFIX: = :/aa/bb/=/xx/yy/"
    }
}

To Lower

Entity whose hdfs_path.clusterName is CL1 will get its path attribute converted to lower case.

{
    "conditions": {
        "hdfs_path.clusterName": "EQUALS: CL1"
    },
    "action": {
        "hdfs_path.path": "TO_LOWER:"
    }
}

Clear

Entity whose hdfspath.clusterName has value set, will get its _replicatedTo attribute value cleared.

{
    "conditions": {
        "hdfs_path.clusterName": "HAS_VALUE:"
    },
    "action": {
        "hdfs_path.replicatedTo": "CLEAR:"
    }
}

Additional Examples

Please look at these tests for examples using Java classes.