Get task context in a UDF

Use the TaskContext PySpark API to get context information while running a Batch Unity Catalog Python UDF or PySpark UDF.

For example, context information such as the user's identity and cluster tags can verify a user's identity to access external services.

Requirements

• TaskContext is supported on Databricks Runtime versions 16.3 and above.

• TaskContext is supported on the following UDF types:

  • Batch Unity Catalog Python UDFs
  • Scalar Python UDFs

Use TaskContext to get context information

Select a tab to see TaskContext examples for PySpark UDFs or Batch Unity Catalog Python UDFs .

PySpark UDF

The following PySpark UDF example prints the user's context:

@udf
def log_context():
  import json
  from pyspark.taskcontext import TaskContext
  tc = TaskContext.get()

  # Returns current user executing the UDF
  session_user = tc.getLocalProperty("user")

  # Returns cluster tags
  tags = dict(item.values() for item in json.loads(tc.getLocalProperty("spark.databricks.clusterUsageTags.clusterAllTags  ") or "[]"))

  # Returns current version details
  current_version = {
    "dbr_version": tc.getLocalProperty("spark.databricks.clusterUsageTags.sparkVersion"),
    "dbsql_version": tc.getLocalProperty("spark.databricks.clusterUsageTags.dbsqlVersion")
  }

  return {
    "user": session_user,
    "job_group_id": job_group_id,
    "tags": tags,
    "current_version": current_version
  }

Batch Unity Catalog Python UDF

The following Batch Unity Catalog Python UDF example gets the user's identity to call an AWS Lambda function using a service credential:

%sql
CREATE OR REPLACE FUNCTION main.test.call_lambda_func(data STRING, debug BOOLEAN) RETURNS STRING LANGUAGE PYTHON
PARAMETER STYLE PANDAS
HANDLER 'batchhandler'
CREDENTIALS (
  `batch-udf-service-creds-example-cred` DEFAULT
)
AS $$
import boto3
import json
import pandas as pd
import base64
from pyspark.taskcontext import TaskContext


def batchhandler(it):
  # Automatically picks up DEFAULT credential:
  session = boto3.Session()

  client = session.client("lambda", region_name="us-west-2")

  # Can propagate TaskContext information to lambda context:
  user_ctx = {"custom": {"user": TaskContext.get().getLocalProperty("user")}}

  for vals, is_debug in it:
    payload = json.dumps({"values": vals.to_list(), "is_debug": bool(is_debug[0])})

    res = client.invoke(
      FunctionName="HashValuesFunction",
      InvocationType="RequestResponse",
      ClientContext=base64.b64encode(json.dumps(user_ctx).encode("utf-8")).decode(
        "utf-8"
      ),
      Payload=payload,
    )

    response_payload = json.loads(res["Payload"].read().decode("utf-8"))
    if "errorMessage" in response_payload:
      raise Exception(str(response_payload))

    yield pd.Series(response_payload["values"])
$$;

Call the UDF after it is registered:

SELECT main.test.call_lambda_func(data, false)
FROM VALUES
('abc'),
('def')
AS t(data)

TaskContext properties

The TaskContext.getLocalProperty() method has the following property keys: