Example 4: A more complicated workflow with follow

Workflow definition

{  "workflow_id": "example4",  "config": {},  "tasks": [    {      "id": "1",      "execution_configurations" : [        {          "function_def_id" : "firstFunc",          "priority" : 1        }      ],      "successors": [        "2",        "3"      ],      "properties": {        "position": "start"      }    },    {      "id": "2",      "execution_configurations" : [        {          "function_def_id" : "multFunc",          "priority" : 1        }      ],      "successors": [        "4"      ],      "properties": {        "scatter": "predecessor_outputs['1']['tile']"      }    },    {      "id": "3",      "execution_configurations" : [        {          "function_def_id" : "lambda_func2",          "priority" : 1        }      ],      "successors": [        "4"      ]    },    {      "id": "4",      "execution_configurations" : [        {          "function_def_id" : "multFunc",          "priority" : 1        }      ],      "successors": [        "5"      ],      "properties": {        "follow": "2"      }    },    {      "id": "5",      "execution_configurations" : [        {          "function_def_id" : "lambda_func2",          "priority" : 1        }      ],      "successors": []    }  ]}

If the function firstFunc generates the same output as the examle above, and task 2 is expanded in the same manner:

Then task 4 will be expanded to tasks 4_A, 4_B, 4_C, each of which becomes a successor of one of the tasks expanded from task 2:

DAG representation

When task 1 completes, task 2 can be expanded. In addition, task 4 can also immediately be expanded (without waiting for all task 2 instances to complete) since its multiplicity depends on the multiplicity of task 2, which has already been decided.