Megosztás a következőn keresztül:


Create and get relationships using the REST API

In this tutorial, you learn how to use Purview REST APIs to:

  1. Create assets and lineage relationships between the data assets.
  2. Query lineage relationships/paths.

APIs referenced in this article:

Prerequisites

The use of these APIs requires Data Curator and Data Reader roles. Check this tutorial for details on how to obtain the access token.

Concepts

Asset

In Purview, we have two types of base assets: DataSet and Process.

  • DataSet Assets that contain data such as Azure SQL Table, Oracle Table etc. should be inherited from DataSet.
  • Process Assets that process data, such as a data pipeline, query, function etc. should be inherited from Process.

Refer to assets and types to understand type definitions of DataSets and Processes.

Relationships & Lineage

In Microsoft Purview, we define three types of relationships for lineage:

  • dataset_process_inputs: connects DataSet to Process, which means the DataSet is the input of the Process
  • process_dataset_outputs: connects Process to DataSet, which means the Process produces the DataSet
  • direct_lineage_dataset_dataset: connects DataSet1 to DataSet2, which means DataSet1 is the upstream of DataSet2, although we wouldn’t know exactly which Process is between them

Example 1: Here’s an example of lineage with 2 DataSets, one Process and two lineage relationships:

Dataset ---> dataset_process_inputs ---> Process ---> process_dataset_outputs ---> DataSet

Screenshot showing DataSet to Process to DataSet lineage.

Example 2: Another example of lineage with 2 DataSets and one lineage relationship:

Dataset ---> direct_lineage_dataset_dataset ---> DataSet

Screenshot showing DataSet to DataSet lineage.

Create assets and lineage relationships between the data assets

In the following sections, let’s take hive_table as an example type for DataSet, and hive_query_process as an example type for Process. We'll create assets with those two types and create lineages between them. You can use any other type, which inherits from DataSet or Process to create lineages.

Example 1

Create assets via API

If the assets you want to create lineage for aren't created in Microsoft Purview yet, you can call the following API to create them.

API: Bulk Create Assets

Task: Create two hive_tables as datasets - table1 and table2, each with 2 hive_columns - column1 and column2.

POST {endpoint}/datamap/api/atlas/v2/entity/bulk

With the body:

{
  "entities": [
    {
      "typeName": "hive_table",
      "attributes": {
        "qualifiedName": "test_lineage.table1",
        "name": "table1"
      },
      "relationshipAttributes": {
        "columns": [
          {
            "guid": "-11",
            "typeName": "hive_column"
          },{
            "guid": "-12",
            "typeName": "hive_column"
          }
        ]
      },
      "guid": "-1"
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table1#column1",
        "name": "column1",
        "type": "int"
      },
      "guid": "-11",
      "relationshipAttributes": {
        "table": {
          "guid": "-1",
          "typeName": "hive_table"
        }
      }
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table1#column2",
        "name": "column2",
        "type": "int"
      },
      "guid": "-12",
      "relationshipAttributes": {
        "table": {
          "guid": "-1",
          "typeName": "hive_table"
        }
      }
    },
    {
      "typeName": "hive_table",
      "attributes": {
        "qualifiedName": "test_lineage.table2",
        "name": "table2"
      },
      "relationshipAttributes": {
        "columns": [
          {
            "guid": "-21",
            "typeName": "hive_column"
          },{
            "guid": "-22",
            "typeName": "hive_column"
          }
        ]
      },
      "guid": "-2"
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table2#column1",
        "name": "column1",
        "type": "int"
      },
      "guid": "-21",
      "relationshipAttributes": {
        "table": {
          "guid": "-2",
          "typeName": "hive_table"
        }
      }
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table2#column2",
        "name": "column2",
        "type": "int"
      },
      "guid": "-22",
      "relationshipAttributes": {
        "table": {
          "guid": "-2",
          "typeName": "hive_table"
        }
      }
    }    
  ]
}

Task: Create the process asset 'hive_view_query'

POST {endpoint}/datamap/api/atlas/v2/entity/bulk

With the body:

{
  "entities": [
    {
      "typeName": "hive_view_query",
      "attributes": {
        "qualifiedName": "test_lineage.HiveQuery1",
        "name": "HiveQuery1",
        "columnMapping": "[{\"DatasetMapping\":{\"Source\":\"test_lineage.table1\",\"Sink\":\"test_lineage.table2\"},\"ColumnMapping\":[{\"Source\":\"column1\",\"Sink\":\"column1\"},{\"Source\":\"column2\",\"Sink\":\"column2\"}]}]"
      },
      "guid": "-1"
    }
  ]
}

The above API calls result in creation of two hive_tables(Datasets) and one hive_view_query(Process).

Create lineage relationships between Datasets and the Process

API: Create Relationship

Task: Create lineage from table1 -> HiveQuery1 (that is, Dataset -> Process)

POST {endpoint}/datamap/api/atlas/v2/relationship

With the body:

{
        "typeName": "dataset_process_inputs",
        "guid": "-1",
        "end1": {
            "typeName": "hive_table",
            "uniqueAttributes": {
                "qualifiedName": "test_lineage.table1"
            }
        },
        "end2": {
            "typeName": "Process",
            "uniqueAttributes": {
                "qualifiedName": "test_lineage.HiveQuery1"
            }
        }
}

Task: Create lineage from HiveQuery1 -> table2 (that is, Process -> Dataset)

POST {endpoint}/datamap/api/atlas/v2/relationship

With the body:

{
        "typeName": "process_dataset_outputs",
        "guid": "-2",
        "end1": {
            "typeName": "Process",
            "uniqueAttributes": {
                "qualifiedName": "test_lineage.HiveQuery1"
            }
        },
        "end2": {
            "typeName": "hive_table",
            "uniqueAttributes": {
                "qualifiedName": "test_lineage.table2"
            }
        }
}

View lineage

Once the assets and lineage relationships are created, you can check the lineage graph in Microsoft Purview:

DataSet-Process-DataSet lineage.

Example 2

Create a hive table, table3, with two columns

API: Bulk Create Assets

Task: Create table3, with 2 hive_columns, column1 and column2

POST {endpoint}/datamap/api/atlas/v2/entity/bulk

With the body:

{
"entities": [
    {
      "typeName": "hive_table",
      "attributes": {
        "qualifiedName": "test_lineage.table3",
        "name": "table3"
      },
      "relationshipAttributes": {
        "columns": [
          {
            "guid": "-31",
            "typeName": "hive_column"
          },{
            "guid": "-32",
            "typeName": "hive_column"
          }
        ]
      },
      "guid": "-3"
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table3#column1",
        "name": "column1",
        "type": "int"
      },
      "guid": "-31",
      "relationshipAttributes": {
        "table": {
          "guid": "-3",
          "typeName": "hive_table"
        }
      }
    },
    {
      "typeName": "hive_column",
      "attributes": {
        "qualifiedName": "test_lineage.table3#column2",
        "name": "column2",
        "type": "int"
      },
      "guid": "-32",
      "relationshipAttributes": {
        "table": {
          "guid": "-3",
          "typeName": "hive_table"
        }
      }
    }
   ]
}

Create direct lineage between table 2 and table 3, with column mapping

API: Create Relationship

Task: Create lineage from table2 -> table3 (that is, Dataset -> Dataset) with column mapping

POST {endpoint}/datamap/api/atlas/v2/relationship

With the body:

{
    "typeName": "direct_lineage_dataset_dataset",
    "guid": "-1",
    "end1": {
        "typeName": "hive_table",
        "uniqueAttributes": {
            "qualifiedName": "test_lineage.table2"
        }
    },
    "end2": {
        "typeName": " hive_table ",
        "uniqueAttributes": {
            "qualifiedName": "test_lineage.table3"
        }
    },
    "attributes": {
      "columnMapping": "[{\"Source\":\"column1\",\"Sink\":\"column1\"},{\"Source\":\"column2\",\"Sink\":\"column2\"}]"
    }
}

View lineage

Now the lineage graph (together from Example 1 & Example 2 above) becomes:

DataSet to DataSet lineage.

Notice that table2 is directly linked to table3, without a HiveQuery between them.

Query lineage relationships/paths

API: Get Lineage by GUID

Task: Get lineage of table2 via REST API

GET {{endpoint}}/api/atlas/v2/lineage/{{guid_of_table2}}?direction=BOTH

Below you can JSON response payload:

{
    "baseEntityGuid": "2a12b3ff-5816-4222-833a-035bf82e06e0",
    "lineageDirection": "BOTH",
    "lineageDepth": 3,
    "lineageWidth": -1,
    "childrenCount": -1,
    "guidEntityMap": {
        "16b93b78-8683-4f88-9651-24c4a9d797b0": {
            "typeName": "hive_table",
            "attributes": {
                "temporary": false,
                "lastAccessTime": 0,
                "createTime": 0,
                "qualifiedName": "test_lineage.table3",
                "name": "table3",
                "retention": 0
            },
            "lastModifiedTS": "1",
            "guid": "16b93b78-8683-4f88-9651-24c4a9d797b0",
            "status": "ACTIVE",
            "displayText": "table3",
            "classificationNames": [],
            "meaningNames": [],
            "meanings": [],
            "isIncomplete": false,
            "labels": [],
            "isIndexed": true
        },
        "cb22ba23-47a2-4149-ade6-e3d9642fe592": {
            "typeName": "hive_table",
            "attributes": {
                "temporary": false,
                "lastAccessTime": 0,
                "createTime": 0,
                "qualifiedName": "test_lineage.table1",
                "name": "table1",
                "retention": 0
            },
            "lastModifiedTS": "1",
            "guid": "cb22ba23-47a2-4149-ade6-e3d9642fe592",
            "status": "ACTIVE",
            "displayText": "table1",
            "classificationNames": [],
            "meaningNames": [],
            "meanings": [],
            "isIncomplete": false,
            "labels": [],
            "isIndexed": true
        },
        "bbeacce6-5bde-46f7-8fe4-689cbb36ba51": {
            "typeName": "hive_view_query",
            "attributes": {
                "qualifiedName": "test_lineage.HiveQuery1",
                "name": "HiveQuery1",
                "columnMapping": "[{\"DatasetMapping\":{\"Source\":\"test_lineage.table1\",\"Sink\":\"test_lineage.table2\"},\"ColumnMapping\":[{\"Source\":\"column1\",\"Sink\":\"column1\"},{\"Source\":\"column2\",\"Sink\":\"column2\"}]}]"
            },
            "lastModifiedTS": "1",
            "guid": "bbeacce6-5bde-46f7-8fe4-689cbb36ba51",
            "status": "ACTIVE",
            "displayText": "HiveQuery1",
            "classificationNames": [],
            "meaningNames": [],
            "meanings": [],
            "isIncomplete": false,
            "labels": [],
            "isIndexed": true
        },
        "2a12b3ff-5816-4222-833a-035bf82e06e0": {
            "typeName": "hive_table",
            "attributes": {
                "temporary": false,
                "lastAccessTime": 0,
                "createTime": 0,
                "qualifiedName": "test_lineage.table2",
                "name": "table2",
                "retention": 0
            },
            "lastModifiedTS": "1",
            "guid": "2a12b3ff-5816-4222-833a-035bf82e06e0",
            "status": "ACTIVE",
            "displayText": "table2",
            "classificationNames": [],
            "meaningNames": [],
            "meanings": [],
            "isIncomplete": false,
            "labels": [],
            "isIndexed": true
        }
    },
    "includeParent": false,
    "relations": [
        {
            "fromEntityId": "2a12b3ff-5816-4222-833a-035bf82e06e0",
            "toEntityId": "16b93b78-8683-4f88-9651-24c4a9d797b0",
            "relationshipId": "23df8e3e-b066-40b2-be29-9fd90693c51b",
            "columnMapping": "[{\"Source\":\"column1\",\"Sink\":\"column1\"},{\"Source\":\"column2\",\"Sink\":\"column2\"}]"
        },
        {
            "fromEntityId": "bbeacce6-5bde-46f7-8fe4-689cbb36ba51",
            "toEntityId": "2a12b3ff-5816-4222-833a-035bf82e06e0",
            "relationshipId": "5fe8d378-39cd-4c6b-8ced-91b0152d3014"
        },
        {
            "fromEntityId": "cb22ba23-47a2-4149-ade6-e3d9642fe592",
            "toEntityId": "bbeacce6-5bde-46f7-8fe4-689cbb36ba51",
            "relationshipId": "73e084bf-98a3-45fb-a1e4-c56cc40661b8"
        }
    ],
    "parentRelations": [],
    "widthCounts": {
        "INPUT": {
            "cb22ba23-47a2-4149-ade6-e3d9642fe592": 0,
            "bbeacce6-5bde-46f7-8fe4-689cbb36ba51": 1,
            "2a12b3ff-5816-4222-833a-035bf82e06e0": 1
        },
        "OUTPUT": {
            "16b93b78-8683-4f88-9651-24c4a9d797b0": 0,
            "2a12b3ff-5816-4222-833a-035bf82e06e0": 1
        }
    }
}

Other Resources

Type Definitions

All Assets/Entities and Relationships are defined in the type system.

Call list all type definitions API to get current type definitions in your Microsoft Purview instance. If you’re not sure which typeName to use in the API calls described, you can check the type definitions API to find the appropriate Asset/Entity type.

Following is a sample response of this API:

TypeDef API response 1.

You can see that in the entityDefs in the above response, asset types (such as oracle_table, oracle_view etc.) are defined. A further look at the definition of the asset (in this example, oracle_view) shows that the asset is inherited DataSet type.

TypeDef API response 2.

Similarly, you can find that a process (for example, Oracle_function) is inherited from “Process” type as shown:

TypeDef API response 3.

Create New Custom Types

If you want to create custom assets or processes, you can use the following APIs.

API: Create Typedef API

Task: Create a Custom Process Type

POST {endpoint}/datamap/api/atlas/v2/types/typedefs

With the body:

{
  "enumDefs": [],
  "structDefs": [],
  "classificationDefs": [],
  "entityDefs": [
    {
      "name": "MyCustomServiceProcess",
      "superTypes": [
        "Process"
      ],
      "typeVersion": "1.0",
      "attributeDefs": []
    }
  ],
  "relationshipDefs": []
}

Task: Create a Custom DataSet Type

POST {endpoint}/datamap/api/atlas/v2/types/typedefs

With the body:

{
  "enumDefs": [],
  "structDefs": [],
  "classificationDefs": [],
  "entityDefs": [
    {
      "name": "MyCustomDataSet",
      "superTypes": [
        "DataSet"
      ],
      "typeVersion": "1.0",
      "attributeDefs": []
    }
  ],
  "relationshipDefs": []
}