FSM workflow (for Node.js)


Machine and its definition.
Run npm install @opuscapita/fsm-workflow-core to get up and running.
``javascript`
import { MachineDefinition, Machine } from '@opuscapita/fsm-workflow-core';
Machine definition consists of:
- schema
- actions
- conditions guards & automatic
`javascript`
const machineDefinition = new MachineDefinition({
schema: {
name: "invoice approval",
initialState: "open",
finalStates: ["approved"],
states: [ // optional meta data for states
{
name: "open",
description: "Open",
// release guards specify if state can be released
// if not specified then release is always allowed
release: [
{
// if 'release[i].to' field is undefined then release guard is applied no matter what target state is
guards: [ // just like in transitions
{
expression: "object.enabled === true"
},
{
name: "someFunction",
params: [
{
name: "param1",
value: 100
}
]
}
]
},
{
to: 'approved', // if 'to' is a string then the folling guards are checked only when trying to transit from 'open' to 'approved'
guards: [...]
},
{
to: ['approved', 'awaitingConfirmation'], // the same as with string, but for a list of target states
guards: [...]
}
]
},
{ name: "approved", description: "Approved" }
],
transitions: [
{
from: "open",
to: "approved",
event: "approve",
// guards define weither this transition is available or not
guards: [
{
"name": "validate",
"params": [
{
"name": "param1",
"value": "value1"
},
{
"name": "param2",
"value": "value2"
}
]
},
// guards can also be simple expressions
{
"expression": "invoice.netAmount < 10000"
}
],
// actions are executed is sequence during transition; 'sendEvent' resolves only after all actions are finished
actions: [
{
"name": "archive",
"params": [
{
"name": "param1",
"value": "value1"
},
{
"name": "param2",
"value": "value2"
}
]
}
],
//automatic guards define if transition can be marked as automatic
automatic: [
{
"name": "lastlyUpdatedMoreThan24hAgo",
"params": [
{
"name": "param1",
"value": "value1"
},
{
"name": "param2",
"value": "value2"
}
],
"negate": true
}
]
}
]
},
// actions can be synchronous or asynchronous (return Promise)
actions: {
archive: function({ param1, param2 }) {}
},
// conditions can be synchronous or asynchronous (return Promise)
conditions: {
validate: function({ param1, param2 }) {},
lastlyUpdatedMoreThan24hAgo: function({ param1, param2 }) {}
},
objectConfiguration: {
stateFieldName: "status",
alias: "invoice",
example: {
"invoiceNo": "1111",
"customerId": "wefwefewfew",
"supplierId": "33333",
"netAmount": 1000,
"status": "reviewRequired"
},
schema: {
{
title: "Invoice",
type: "object",
properties: {
invoiceNo: {
type: "string"
},
customerId: {
type: "string"
},
supplierId: {
type: "string"
},
netAmount: {
type: "number"
},
status: {
type: "string"
}
},
required: ["invoiceNo"]
}
}
}
});
Defines machine transitions and initialization options. Could be presented as oriented graph, where each node represents state and directed edges are used to represent transition from one state to another.
#### Transitions
In schema you needs to define an array of available machine transitions. Typically a transition is triggered by an _event_ and happens between _from_ and _to_ states. Optionally each transition can have _actions_, _guards_ and/or _automatic_ (conditions).
#### Initial state
You can define the initial state by setting the _initialState_ property:
`javascript
var machineDefinition = new MachineDefinition({
schema: {
name: 'sprint'
initialState: 'start'
transitions: [
{ from: 'start', event: 'run', to: 'finish' }
]
}
});
const machine = new Machine({ machineDefinition });
machine.start({ object }).then(({ status: 'none' }) => {
console.log(machine.currentState({ object }));
// start
});
`
if initial state is not specified, then 'none' will be used (TBD)
#### Final states
You can define the final states (one or many) by setting the _finalStates_ property:
`javascript`
var machineDefinition = new MachineDefinition({
schema: {
initialState: 'start',
finalStates: ['finish'],
transitions: [
{from: 'start', event: 'run', to: 'finish'}
]
}
});
- Actions & conditions configuration and usage
- Ideas & thoughts
#### Action
Actions (action = function) are executed during transition (not while leaving/entering state). Action references specific function by name. Action implemented separately from schema. Each action accepts named arguments explicitly defined in transition and implicit arguments like _object_, _from_, _to_, etc. During transition machine executes each action in defined order. Each action gets _actionExecutionResults_ argument which serves as an accumulator from perviously called actions, where each property is an action name and value is value returned by action.
#### Guard (conditions)
Guards are used to protect transitions. Guard works as 'if' condition.
Technically guard is defined the same way like as action, it is a function.
The difference is that it should always return boolean value (true or false).
Condition(function) result could be inverted if its property _negate_ is set to true.
Guards could be also _sync_ and _async_ functions. In case you want to implement async guard, pay additional attention
to the value resolved by a guard - it should be only boolean value. In case your guard rejects some value
(error or smth else) - it will be taken as an error and _findAvailableTransitions_ will be rejected with error.
Note: similar to Spring State Machine Guards
#### Automatic (conditions)
Transition could be marked as automatic using corresponding property. It could be:
- true (boolean value) - e.g. this transition is always automatic
- array of conditions(functions, each return true or false), condition(function) result could be inverted if its property _negate_ is set to true
Check for whether object in current state has (at least one) automatic transition could be done via task manager (inside the application). Basing on evaluated results task manager will be able to take a decision to send event without user interaction.
Machine does not have own state, all the transitions are performed over object which state is changed by machine. Object is used by Machine as a mutable parameter passed to guards and actions.
`javascript
var machineDefinition = new MachineDefinition({
schema: {
initialState: 'start'
finalStates: ['finish'],
transitions: [
{ from: 'start', event: 'run', to: 'finish' }
]
}
});
const object = { status: 'none' };
const machine = new Machine(machineDefinition);
machine.start({ object }).then(({ object }) => {
console.log(machine.currentState({ object }));
// start
return machine.sendEvent({ object, event: 'start' })
}).then(({ object }) => {
console.log(machine.currentState({ object }));
// finish
});
`
Machine configuration
``
var machine = new Machine({ history, ... });
history is a DAO that provides a possibility to create and read object workflow history records. You can find its API (and DB specific implementation) here.
Machine writes history records for all object transitions within the workflow.
It happens when you start workflow
``
machine.start({ object, user, description })`
or you send an event`
machine.sendEvent({ object, event, user, description })
In both cases, new history records are created.
Here
- object (required) - business object of the following structure
- user (required) - user identifier who initiated an event
- description (optional) - custom text that describes transition/object
All this info together is stored in workflow history.
Note: In case of start method call history fields from and event are filled with string value NULL (4 upper cases letters: N, U, L, L)
Machine provides possibility to get(search) history records via getHistory method:
Getting/searching specific workflow history. You can either search by:
- specific object history
- initiated by specific user
- additionally, you can restrict query using finishedBy to get history within a specific period of time.
`javascript`
machine.getHistory(searchParameters, paging, sorting)`
where
- searchParameters
javascript`
{
object,
user, // example: 'john.miller'
finishedOn: {
gte, // example: Date("2018-03-05T21:00:00.000Z")
gt, // (syntax like 'today', 'yesterday', 'week ago' could be introduced later, if required)
lt,
lte
}
}
`
- paging:
javascript`
{
max, // example: '25', default value: 100
offset // example: '50', default value: 0
}
`
- sorting:
javascript`
{
by, // example: 'user', default value: 'finishedOn',
// possible values: ['event', 'from', 'to', 'user', 'description', 'finishedOn']
order // example: 'asc', default value: 'desc'
}
`
as a result you get a promise that results into array of objects of the following structure:
javascript`
{
event,
from,
to,
object: {
businessObjectId,
businessObjectType
},
user,
description,
finishedOn
}
Note: while writing workflow object history or searching history records by
object machine uses configured convertObjectToReference callback to convert
real business object into reference object that has the following structure
{businessObjType, businessObjId}
In case one needs to find out why particular transitions are available or unavailable there's a lower-level function MachineDefinition.inspectTransitions. According to provided input params it returns all analyzed transitions with their guards and results of their evaluation. This helps to determine why particular transition was (un)available in particular case.
In case one needs to inspect release guards defined for states, there's MachineDefinition.inspectReleaseConditions function which is similar to MachineDefinition.inspectTransitions.
For details about these functions see MachineDefinition.js source.
`javascript
var machineDefinition = new MachineDefinition({ schema, conditions, actions })
// register workflow
var machine = new Machine(machineDefinition, context);
// start/initialize machine/workflow
machine.start({ object })
// returns a list of available transitions: {event, from, to, request..}, e.g. event
// request is used to pass parameters to guards for some dynamic calculation, e.g. when event availability depends
// on current user information as roles and etc.
machine.availableTransitions({ object })
// returns a list of available automatic transitions: {event, from, to, ..}, e.g. event
// if machine schema is adequate then there should be not more than 1 such transition
machine.availableAutomaticTransitions({})
// send 'event' and pass addition 'request' data that is posted by user/app
// returns promise, in case of successful transition then function will be called
// with one parameter that is an JSON with the following structure:
// - object - object in new state (the same reference that is passed as parameter)
machine.sendEvent({object, event, request})
machine.currentState({ object }) // gets current state
machine.is({ object, state}) // is object in state
machine.isInFinalState({ object }) // returns true iff object is in one of final states
machine.can({ object, event }) // whether event is available
machine.cannot({ object, event }) // whether event is not available
machine.canBeReleased({ object, to, request }) // weither object can be released from its current state; 'to' is an optional target state
// hooks (tbd)
machine.onStartTransition() // returns promise
machine.onFinishTransition() // returns promise
``
| | Alexey Sergeev |
| :---: | :---: |
| | Daniel Zhitomirsky |
| | Egor Stambakio |
| | Andrei Mourzenkov |
OpusCapita FSM Workflow is licensed under the Apache License, Version 2.0. See LICENSE for the full license text.