arazzo: 1.0.1
info:
  title: Azure Synapse Analytics Open Spark Session and Run Statement
  summary: Create an interactive Spark session, wait until idle, then run a statement.
  description: >-
    Interactive Spark sessions on a Synapse Spark pool let you run ad-hoc code
    statements through the Livy API. This workflow creates a session, polls it
    until it becomes idle and ready, and then submits a code statement to it.
    Every step spells out its request inline so the flow can be read and executed
    without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: sparkJobApi
  url: ../openapi/azure-synapse-analytics-spark-job-openapi.yml
  type: openapi
workflows:
- workflowId: open-spark-session-and-run-statement
  summary: Start a Spark session, wait for it to be idle, and run a code statement.
  description: >-
    Creates a Spark session, loops on reading the session until its livy state
    is idle, then submits a code statement against the ready session.
  inputs:
    type: object
    required:
    - sparkPoolName
    - sessionOptions
    - statement
    properties:
      sparkPoolName:
        type: string
        description: The name of the Spark pool to host the session.
      sessionOptions:
        type: object
        description: The SparkSessionOptions body. Must include name at minimum.
      statement:
        type: object
        description: >-
          The SparkStatementOptions body, including code and kind (e.g. spark,
          pyspark, sql).
  steps:
  - stepId: createSession
    description: >-
      Create a new interactive Spark session on the pool.
    operationId: SparkSession_CreateSparkSession
    parameters:
    - name: sparkPoolName
      in: path
      value: $inputs.sparkPoolName
    - name: detailed
      in: query
      value: true
    requestBody:
      contentType: application/json
      payload: $inputs.sessionOptions
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      sessionId: $response.body#/id
  - stepId: waitForIdle
    description: >-
      Poll the session until its livy currentState is idle. Loop back while it is
      still starting; proceed once it is ready.
    operationId: SparkSession_GetSparkSession
    parameters:
    - name: sparkPoolName
      in: path
      value: $inputs.sparkPoolName
    - name: sessionId
      in: path
      value: $steps.createSession.outputs.sessionId
    - name: detailed
      in: query
      value: true
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      currentState: $response.body#/livyInfo/currentState
    onSuccess:
    - name: notReady
      type: goto
      stepId: waitForIdle
      criteria:
      - context: $response.body
        condition: $.livyInfo.currentState != "idle"
        type: jsonpath
    - name: ready
      type: goto
      stepId: runStatement
      criteria:
      - context: $response.body
        condition: $.livyInfo.currentState == "idle"
        type: jsonpath
  - stepId: runStatement
    description: >-
      Submit a code statement to the ready Spark session.
    operationId: SparkSession_CreateSparkStatement
    parameters:
    - name: sparkPoolName
      in: path
      value: $inputs.sparkPoolName
    - name: sessionId
      in: path
      value: $steps.createSession.outputs.sessionId
    requestBody:
      contentType: application/json
      payload: $inputs.statement
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      statementId: $response.body#/id
      statementState: $response.body#/state
  outputs:
    sessionId: $steps.createSession.outputs.sessionId
    statementId: $steps.runStatement.outputs.statementId