(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 transformation framework, it did not have support for some of the 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 of the 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 condition is met.

Conditions

Following are built-in conditions.

Condition Types	Description
ENTITY_ALL	Any/every entity
ENTITY_TOP_LEVEL	Entity that is the top-level entity. This is also the entity present specified in AtlasExportRequest.
EQUALS	Entity attribute equals to the one specified in the condition.
EQUALS_IGNORE_CASE	Entity attribute equals to the one specified in the condition ignoring case.
STARTS_WITH	Entity attribute starts with.
STARTS_WITH_IGNORE_CASE	Entity attribute starts with ignoring case.
HAS_VALUE	Entity attribute has value.

Actions

Action Type	Description
ADD_CLASSIFICATION	Add classifiction
REPLACE_PREFIX	Replace value starting with another value.
TO_LOWER	Convert value of an attribute to lower case.
SET	Set the value of an attribute
CLEAR	Clear value of an attribute

Built-in Transforms

Add Classification

During import, hive_db 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 special condition which matches entity.

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

To add classification to only the top-level entity (entity that is used as 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 hdfs_path.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.