Workflows help you automate business processes in a standardized manner to bridge the gap between expressing and modelling business logic.

Workflow definitions are written based on Serverless Workflow specification. For the Crosswork Workflow Manager version 1.0, only a subset of the specification is supported. This chapter describes all the supported features and gives practicable examples for each.

A new workflow can be defined in either JSON or YAML formatting. The structure of the workflow definition is described in the specification.

The supported high-level components are as follows:

• id

• name

• description

• version

• start

• retries

• errors

• functions

• states

• metadata

States define the building blocks of workflow execution logic. Different types of states provide control flow logic to the Execution Engine and also allow you to define which activities to execute.

As per serverless workflow specification, operation states define sets of actions to be executed in sequence or parallel. Crosswork Workflow Manager only supports execution of actions in sequence.

An action can define invocation of 3 different types of services:

• Execution of function definition.

• Execution of another workflow definition as a child workflow (not supported in current implementation).

• Referencing events that may be "produced" or "consumed" (not supported in current implementation).

Note	Only execution of function definition is supported in current implementation.

Switch states enable you to define decision points to route the workflow to a given path based on certain conditions. Workflow Serverless specification supports Data-based conditions and Event-based conditions. CWM only supports Data-based conditions.

Sleep state pauses workflow execution for a given duration.  

    {
        "name": "Sleep3Minutes",
        "type": "sleep",
        "duration": "PT3M",
        "transition": {
            "nextState": "NextState"
        }
    }

Inject state is used to inject static data into the State Data.

    {
        "id": "example",
        "version": "1.0",
        "specVersion": "0.9",
        "start": "HelloWorld",
        "states": [
            {
                "name": "HelloWorld",
                "type": "inject",
                "data": {
                    "name": "Cisco",
                    "message": "Hello World"
                },
                "stateDataFilter":{
                    "output": "${ .message + \" from \" + .name + \"!\"  }"
                },
                "end": {
                    "terminate": "true"
                }
            }
        ]
    }

ForEach state allows you to define a set of actions to execute for each element in an array or list defined in State Data. For example, for Each device in device array, check the devices is in sync. Whilst the serverless workflow specification defines support for Parallel and Sequential execution of actions, current implementation only supports sequential execution of actions for each element in array.

    {
        "id": "example",
        "version": "1.0",
        "specVersion": "0.9",
        "start": "InjectData",
        "functions": [
            {
                "name": "HelloWorld",
                "operation": "HelloWorld"
            }
        ],
        "states": [
            {
                "name": "InjectData",
                "type": "inject",
                "data": {
                    "people": [
                        {
                            "Firstname": "Peter",
                            "Surname": "Parker"
                        },
                        {
                            "Firstname": "Thor",
                            "Surname": "Odinson"
                        },
                        {
                            "Firstname": "Bruce",
                            "Surname": "Banner"
                        }
                    ]
                },
                "transition":{
                    "nextStat": "SayHelloToEveryone"
                }
            },
            {
                "name": "SayHelloToEveryone",
                "type": "foreach",
                "inputCollection": "${ .people }",
                "iterationParam": "person",
                "outputCollection": "${  .messages }",
                "actions": [
                    {
                        "name": "SayHello",
                        "functionRef":{
                            "refName": "HelloWorld",
                            "arguments": {
                                "name": "${ .person.Firstname + \" \" + .person.Surname }"
                            }
                        }
                    }
                ],
                "end": {
                    "terminate": "true"
                }
            }
        ]
    }
 

Parallel state allows you to define a collection of branches that are executed in parallel. Each branch in a state can define its own set of actions. Once the execution has completed, the parallel branches are joined into current path based on the completionType attribute.

The completionType attribute can define 2 values:

• allOf: All branches must complete execution before state can transition/end. This is the default value.

• atLeast: State can transition/end if the number of branches specified in atLeast has completed execution. If completionType attribute is "atLeast", numCompleted must also be set.

State data plays an important role during the lifecycle of the workflow. A state can filter data, inject data, and add data. Jq plays an important role in data filtering, creation and manipulation. For more information on how Data can be handled, see the Serverless Workflow specification.

When creating workflows, the following rules will apply when it comes to data management within CMW:

• Initial data passed into workflow execution is passed into State data as input.

• Data output from the last executed state is workflow output.

• If no State Input Filter is specified, all the data is passed into the state.

• If no State Output Filter is specified, all the data is passed into the next state. 

• Workflow expressions in jq allow you to filter and manipulate data.

• Actions also allow for filtering data and also, if return data from action should be merged back into state data.

• Filters must return JSON object, if a jq workflow expression results in a string literal, this will result in an error. 

• When working with jq, it is highly recommended to use https://jqplay.org/ to test the jq expressions. Alternatively, you can download jq locally and use it for testing.