Python SDK for Lineage

In this guide, we will use the Python SDK to create and fetch Lineage information. For simplicity, we are going to create lineage between Tables. However, this would work with ANY entity. You can find the Lineage Entity defined here, as well as the Entity defining the payload to add a new lineage: AddLineage.

Note that in OpenMetadata, the Lineage information is just a possible relationship between Entities. Other types of relationships for example could be:

Contains (a Database contains Schemas, which at the same time contain Tables),
or Ownership of any asset.

The point being, any Entity existent in OpenMetadata can be related to any other via Lineage.

In the following sections we will:

Create a Database Service, a Database, a Schema and two Tables,
Add Lineage between both Tables,
Get the Lineage information back.

A prerequisite for this section is to have previously gone through the following docs.

Creating the Entities

To prepare the necessary ingredients, execute the following steps. All functions that we are going to use related to Lineage can be found here.

1. Preparing the Client

from metadata.sdk import configure

configure(
    host="http://localhost:8585/api",
    jwt_token="<token>",
)

2. Creating the Database Service

We are mocking a MySQL instance. Note how we need to pass the right configuration class MysqlConnection, as a parameter for the generic DatabaseConnection type.

from metadata.sdk import DatabaseServices
from metadata.generated.schema.api.services.createDatabaseService import (
    CreateDatabaseServiceRequest,
)
from metadata.generated.schema.entity.services.connections.database.common.basicAuth import (
    BasicAuth,
)
from metadata.generated.schema.entity.services.connections.database.mysqlConnection import (
    MysqlConnection,
)
from metadata.generated.schema.entity.services.databaseService import (
    DatabaseConnection,
    DatabaseServiceType,
)

db_service = CreateDatabaseServiceRequest(
    name="test-service-db-lineage",
    serviceType=DatabaseServiceType.Mysql,
    connection=DatabaseConnection(
        config=MysqlConnection(
            username="username",
            authType=BasicAuth(password="password"),
            hostPort="http://localhost:1234",
        )
    ),
)

db_service_entity = DatabaseServices.create(db_service)

3. Creating the Database

Any Entity that is created and linked to another Entity, has to hold the fullyQualifiedName to the Entity it relates to. In this case, a Database is bound to a specific service.

from metadata.sdk import Databases
from metadata.generated.schema.api.data.createDatabase import CreateDatabaseRequest

create_db = CreateDatabaseRequest(
    name="test-db",
    service=db_service_entity.fullyQualifiedName,
)

create_db_entity = Databases.create(create_db)

4. Creating the Schema

The same happens with the Schemas. They are related to a Database.

from metadata.sdk import DatabaseSchemas
from metadata.generated.schema.api.data.createDatabaseSchema import (
    CreateDatabaseSchemaRequest,
)

create_schema = CreateDatabaseSchemaRequest(
    name="test-schema", database=create_db_entity.fullyQualifiedName
)

create_schema_entity = DatabaseSchemas.create(create_schema)

5. Creating the Tables

And finally, Tables are contained in a specific Schema, so we use the fullyQualifiedName here as well. We are doing a simple example with a single column.

from metadata.sdk import Tables
from metadata.generated.schema.api.data.createTable import CreateTableRequest
from metadata.generated.schema.entity.data.table import Column, DataType

table_a = CreateTableRequest(
    name="tableA",
    databaseSchema=create_schema_entity.fullyQualifiedName,
    columns=[Column(name="id", dataType=DataType.BIGINT)],
)

table_b = CreateTableRequest(
    name="tableB",
    databaseSchema=create_schema_entity.fullyQualifiedName,
    columns=[Column(name="id", dataType=DataType.BIGINT)],
)

table_a_entity = Tables.create(table_a)
table_b_entity = Tables.create(table_b)

6. Adding Lineage

With everything prepared, we can now create the lineage between both entities.

from metadata.sdk.api import Lineage

created_lineage = Lineage.add_lineage(
    from_entity_id=str(table_a_entity.id),
    from_entity_type="table",
    to_entity_id=str(table_b_entity.id),
    to_entity_type="table",
    description="test lineage",
)

The Python client will already return us a JSON object with the Lineage information about the fromEntity node we added:

{
  "entity": {
    "id": "e7bee99b-5c5e-43ec-805c-8beba04804f5",
    "type": "table",
    "name": "tableA",
    "fullyQualifiedName": "test-service-db-lineage.test-db.test-schema.tableA",
    "deleted": false,
    "href": "http://localhost:8585/api/v1/tables/e7bee99b-5c5e-43ec-805c-8beba04804f5"
  },
  "nodes": [
    {
      "id": "800caa0f-a149-48d2-a0ce-6ca84501767e",
      "type": "table",
      "name": "tableB",
      "fullyQualifiedName": "test-service-db-lineage.test-db.test-schema.tableB",
      "deleted": false,
      "href": "http://localhost:8585/api/v1/tables/800caa0f-a149-48d2-a0ce-6ca84501767e"
    }
  ],
  "upstreamEdges": [],
  "downstreamEdges": [
    {
      "fromEntity": "e7bee99b-5c5e-43ec-805c-8beba04804f5",
      "toEntity": "800caa0f-a149-48d2-a0ce-6ca84501767e"
    }
  ]
}

If the node were to have other edges already, they would be showing up here. If we validate the Lineage from the UI, we will see:

7. Fetching Lineage

Finally, let’s fetch the lineage from the other node involved:

from metadata.sdk.api import Lineage

lineage = Lineage.get_lineage(
    "test-service-db-lineage.test-db.test-schema.tableB",
    upstream_depth=1,
    downstream_depth=1,
)

Which will give us the symmetric results from above

{
  "entity": {
    "id": "800caa0f-a149-48d2-a0ce-6ca84501767e",
    "type": "table",
    "name": "tableB",
    "fullyQualifiedName": "test-service-db-lineage.test-db.test-schema.tableB",
    "deleted": false,
    "href": "http://localhost:8585/api/v1/tables/800caa0f-a149-48d2-a0ce-6ca84501767e"
  },
  "nodes": [
    {
      "id": "e7bee99b-5c5e-43ec-805c-8beba04804f5",
      "type": "table",
      "name": "tableA",
      "fullyQualifiedName": "test-service-db-lineage.test-db.test-schema.tableA",
      "deleted": false,
      "href": "http://localhost:8585/api/v1/tables/e7bee99b-5c5e-43ec-805c-8beba04804f5"
    }
  ],
  "upstreamEdges": [
    {
      "fromEntity": "e7bee99b-5c5e-43ec-805c-8beba04804f5",
      "toEntity": "800caa0f-a149-48d2-a0ce-6ca84501767e"
    }
  ],
  "downstreamEdges": []
}

You can also get lineage by entity type and ID using Lineage.get_entity_lineage(entity_type="table", entity_id="entity-id", upstream_depth=1, downstream_depth=1).

Lineage Details

Note how when adding lineage information we give to the API an AddLineage Request. This is composed of an Entity Edge, whose definition you can find here. In a nutshell, an Entity Edge has:

The Entity Reference as the lineage origin,
The Entity Reference as the lineage destination,
Optionally, Lineage Details.

In the Lineage Details property we can pass further information specific about Table to Table lineage:

sqlQuery specifying the transformation,
An array of columnsLineage as an object with an array of source and destination columns, as well as their own specific transformation function,
Optionally, the Entity Reference of a Pipeline powering the transformation from Table A to Table B.

The simple Lineage.add_lineage(...) helper covers direct edges. For advanced payloads, construct an AddLineageRequest and submit it with Lineage.add_lineage_request(...). Let’s see how to do that and play with the possible combinations. First, import the required classes and create a new table:

from metadata.sdk import Tables
from metadata.generated.schema.api.lineage.addLineage import AddLineageRequest
from metadata.generated.schema.type.entityReference import EntityReference
from metadata.generated.schema.type.entityLineage import (
    ColumnLineage,
    EntitiesEdge,
    LineageDetails,
)

# Prepare a new table
table_c = CreateTableRequest(
    name="tableC",
    databaseSchema=create_schema_entity.ifullyQualifiedName,
    columns=[Column(name="id", dataType=DataType.BIGINT)],
)

table_c_entity = metadata.create_or_update(data=table_c)

Column Level Lineage

We can start by linking our columns together. For that we are going to create:

A ColumnLineage object, linking our Table A column ID -> Table C column ID. Note that this can be a list!
A LineageDetails object, passing the column lineage and the SQL query that powers the transformation.

from metadata.sdk.api import Lineage

column_lineage = ColumnLineage(
    fromColumns=["test-service-db-lineage.test-db.test-schema.tableA.id"],
    toColumn="test-service-db-lineage.test-db.test-schema.tableC.id"
)

lineage_details = LineageDetails(
    sqlQuery="SELECT * FROM AWESOME",
    columnsLineage=[column_lineage],
)

add_lineage_request = AddLineageRequest(
    edge=EntitiesEdge(
        fromEntity=EntityReference(id=table_a_entity.id, type="table"),
        toEntity=EntityReference(id=table_c_entity.id, type="table"),
        lineageDetails=lineage_details,
    ),
)

created_lineage = Lineage.add_lineage_request(add_lineage_request)

This information will now be reflected in the UI as well:

Adding a Pipeline Reference

We can as well pass the reference to the pipeline used to create the lineage (e.g., the ETL feeding the tables). To prepare this example, we need to start by creating the Pipeline Entity. Again, we’ll need first to prepare the Pipeline Service:

from metadata.generated.schema.type.entityReference import EntityReference
from metadata.generated.schema.api.data.createPipeline import CreatePipelineRequest
from metadata.generated.schema.api.services.createPipelineService import (
    CreatePipelineServiceRequest,
)
from metadata.generated.schema.entity.services.pipelineService import (
    PipelineConnection,
    PipelineService,
    PipelineServiceType,
)
from metadata.generated.schema.entity.services.connections.pipeline.backendConnection import (
    BackendConnection,
)

pipeline_service = CreatePipelineServiceRequest(
    name="test-service-pipeline",
    serviceType=PipelineServiceType.Airflow,
    connection=PipelineConnection(
        config=AirflowConnection(
            hostPort="http://localhost:8080",
            connection=BackendConnection(),
        ),
    ),
)

pipeline_service_entity = metadata.create_or_update(data=pipeline_service)

create_pipeline = CreatePipelineRequest(
    name="test",
    service=pipeline_service_entity.fullyQualifiedName,
)

pipeline_entity = metadata.create_or_update(data=create_pipeline)

With these ingredients ready, we can then follow the code above and add there a pipeline argument as an Entity Reference:

lineage_details = LineageDetails(
    sqlQuery="SELECT * FROM AWESOME",
    columnsLineage=[column_lineage],
    pipeline=EntityReference(id=pipeline_entity.id, type="pipeline"),
)

The UI currently supports showing the column lineage information. Data about the SQL queries and the Pipeline Entities will be surfaced soon. Thanks!

Automated SQL Lineage

To have lineage generated from a SQL query, call add_lineage_by_query. metadata.sdk.api.Lineage doesn’t expose it, but the client returned by configure(...) does, through its ometa property. This is the same call the CLI makes.

from metadata.sdk import configure
from metadata.generated.schema.entity.services.databaseService import DatabaseService

ometa = configure(
    host="http://localhost:8585/api",
    jwt_token="<token>",
).ometa

database_service = ometa.get_by_name(entity=DatabaseService, fqn="local_mysql")
if database_service is None:
    raise ValueError("Database service local_mysql not found")

ometa.add_lineage_by_query(
    database_service=database_service,
    sql="INSERT INTO target_table (id) SELECT id FROM source_table",
    database_name="my_db",  # Optional: database for table names without one
    schema_name="my_schema",  # Optional: schema for table names without one
    timeout=300,  # Parsing timeout in seconds
    check_patch=True,
)

This example creates lineage from source_table to target_table within the local_mysql database service. add_lineage_by_query takes these parameters:

database_service: The database service that contains the tables in the query. The SQL dialect is taken from its service type.
sql: The query to parse.
database_name and schema_name: (Optional) The database and schema to use for table names that the query doesn’t qualify.
timeout: Timeout for the lineage parsing process, in seconds.
check_patch: Defaults to False. The CLI sets it to True.

add_lineage_by_query returns None and doesn’t raise on failure. Parsing and API errors are only logged, so enable logging to see them. Check that get_by_name found the service before calling it: when database_service is None, the method does nothing.

Automated SQL Lineage via CLI

To create the automated sql lineage via CLI, you need to make sure that you have installed the openmetadata-ingestion package in your local environment using command pip install openmetadata-ingestion. Once that is done you will have to prepare a yaml file as follows.

serviceName: local_mysql
query: insert into target_table(id) select id from source_table
# filePath: test.sql
# parseTimeout: 360 # timeout in seconds
workflowConfig:
  # loggerLevel: DEBUG  # DEBUG, INFO, WARN or ERROR
  openMetadataServerConfig:
    hostPort: <OpenMetadata host and port>
    authProvider: <OpenMetadata auth provider>

serviceName: Name of the database service which contains the table involved in query.
query: You can specify the raw sql query within the yaml file itself.
filePath: In case the query is too big then you can also save query in a file and pass the path to the file in this field.
parseTimeout: Timeout for the lineage parsing process.
workflowConfig: The main property here is the openMetadataServerConfig, where you can define the host and security provider of your OpenMetadata installation.

Once the yaml file is prepare you can run the command

metadata lineage -c path/to/your_config_yaml.yaml

API Reference

Data Assets

Collaboration

AI SDK

SDK

Discovery

Lineage

Data Contracts

Teams And Users

Data Governance

Data Quality

Alerts

Incidents

Main Concepts

Python SDK for Lineage

Python SDK for Lineage

Creating the Entities

1. Preparing the Client

2. Creating the Database Service

3. Creating the Database

4. Creating the Schema

5. Creating the Tables

6. Adding Lineage

7. Fetching Lineage

Lineage Details

Column Level Lineage

Adding a Pipeline Reference

Automated SQL Lineage

Automated SQL Lineage via CLI