Turbine is a JavaScript workflow engine. It vastly simplifies the development, deployment, and testing of complex web applications through the use of declarative workflows that express your app's program logic in a form that is simple to read and to understand.
Turbine is the ideal solution for apps (or parts of apps) with multi-step processes involving many possible branches, sub-flows, or permutations. Examples include:
- Signup forms
- Login forms
- Interactive tours
- Shopping carts
- Checkout flows
- Asset creation (i.e. upload photo -> add filter -> add caption -> tag friends -> share)
- etc.
The programming of these types of apps usually involves a tangled nightmare of conditionals, switches, callbacks, promises, and other strands of spaghetti code.
This tightly coupled code makes it almost impossible to A/B/n test different flows or variations -- any attempt to do so usually makes the problem even worse. It is also very difficult to follow the program logic to trace all the possible flows through the code.
The end result is code which is dense, brittle, untestable, obtuse, and just plain gross.
Turbine was written to solve these problems by taking the program logic of complex apps and abstracting it out into a simple document (the workflow) which is easy to read and understand.
Turbine is available via a number of popular package managers:
npm install turbine.js
jam install turbine.js
bower install turbine.js
Or you can download the latest tag from https://github.com/wmbenedetto/turbine.js/tags
In order to use Turbine, it's important to first define some key concepts. Once we have a common vocabulary and general conceptual understanding established, we can then begin talking about how you can use Turbine to power your app.
The workflow is the jet fuel that powers Turbine. It's an expressive, declarative syntax for defining the program flow of your application. It allows you to define all the logical branching of your app in a single document, in a format that is both human- and machine-readable.
Turbine workflows are declarative -- they are only concerned with what your app does, not how it does it. Although workflows are written in JavaScript, they should not contain any functional logic. They should be serializable to JSON -- and deserializable from JSON -- without any ill effects.
A workflow is essentially a series of questions (queries) and answers (responses). It's almost like a conversation between Turbine and your app.
Turbine: Is the user signed up?
Your app: Nope.
Turbine: Okay. Ask him to sign up. I'll wait.
Your app displays a signup form. The user fills it in and clicks Submit.
Your app: Alrighty, he signed up.
Turbine: Great. Is he over 18?
Your app: No, he's only 13.
Turbine: Damn. Ask him for his parent's email, then let me know.
Your app asks for the parent's email. The user submits it.
Your app: I got the parent's email.
Turbine: Is is valid?
Your app: Yep, looks good.
Turbine: Great! Let him in. We're done!
Now let's look at the same "conversation" expressed as a workflow:
var workflow = {
// Turbine: Is the user signed up?
isUserSignedUp : {
// Your app: Nope.
no : {
// Turbine: Okay. Ask him to sign up. I'll wait.
publish : {
message : 'Signup.stepOne.show'
},
waitFor : 'Signup.stepOne.submitted',
// Your app is listening for a Signup.stepOne.show message. It knows to handle
// that by displaying a signup form. The user fills it in and clicks Submit.
// This publishes a Signup.stepOne.submitted message. Which is equivalent to:
//
// Your app: Alrighty, he signed up.
//
// Turbine moves to its next "question"
then : 'isOver18'
},
yes : {
then : 'stop.'
},
},
// Turbine: Great. Is he over 18?
isOver18 : {
// Your app: No, he's only 13.
no : {
// Turbine: Damn. Ask him for his parent's email, then let me know.
publish : {
message : 'Signup.parentEmail.show'
},
waitFor : 'Signup.parentEmail.submitted',
// Your app is listening for a Signup.parentEmail.show message. It knows to handle
// that by displaying a form that asks the user for his parent's email address.
// The user enters the email and clicks Submit. This publishes a Signup.parentEmail.submitted message.
// Which is equivalent to:
//
// Your app: I got the parent's email.
//
// Turbine moves to its next "question"
then : 'isParentEmailValid'
},
yes : {
then : 'stop.'
}
},
// Turbine: Is is valid?
isParentEmailValid : {
// Your app: Yep, looks good.
yes : {
// Turbine: Great! Let him in. We're done!
publish : {
message : 'Signup.form.complete'
},
then : 'stop.'
},
no : {
publish : {
message : 'Signup.error.show.INVALID_EMAIL_ADDRESS'
},
then : 'isOver18'
}
}
};Queries are the questions that Turbine asks your app. Therefore, a query is a string typically written as a question, such as:
- isUserLoggedIn
- canOpenGoldDoor
- whichErrorCode
- howManyStars
Generally, queries should be phrased so that the affirmative answer (if there is one) is the answer you want/expect.
For example, say you want to confirm that a plugin is the latest version. There are two ways you could write the same query: isPluginOutdated or isPluginUpToDate.
Since you want and expect the plugin to be up-to-date, the preferred phrasing would be isPluginUpToDate.
Phrasing your queries this way means that the "happy path" through your app is a series of yesses. Is the plugin up to date? Yes! Is the game loaded? Yes! Is the user logged in? Yes! And so on.
When you instantiate an instance of Turbine, you can (but don't have to) define functions from your app that Turbine can use to execute the query.
For example, if your app only lets magicians open gold doors, you might define a query function like:
var initObj = {
queries : {
canOpenGoldDoor : user.isMagician.bind(user)
}
};When Turbine gets to the canOpenGoldDoor query in your workflow, it will execute user.isMagician(), which will return true or false. This is the "response."
Responses are the answers your app returns for queries.
Responses are often simple booleans: true or false get converted to "yes" or "no" by Turbine. However, the response really can be any arbitrary string or number. Responses for the query examples above might look like:
- isUserLoggedIn
- yes
- no
- canOpenGoldDoor
- yes
- no
- whichErrorCode
- INVALID_EMAIL
- INVALID_CREDIT_CARD
- MUST_CHECK_TERMS_BOX
- howManyStars
- 1
- 2
- 3
- 4
- 5
Responses can be defined in a couple of ways.
The most obvious way to get a response is as a value returned by the query function. When the canOpenGoldDoor query executes the user.isMagician() function, whatever that function returns gets set as the response to canOpenGoldDoor: true, which gets converted to "yes".
The second way a response can be defined is via Turbine's setResponse() method. This isn't really the preferred way of doing things, since it requires more tightly coupling your app with Turbine and your workflow. However, it's an option you can use if you need it.
For example, say you have some form validation logic. You could do something like:
if (user.email.indexOf('@') < 0){
turbine.setResponse('whichErrorCode','INVALID_EMAIL');
}When Turbine reaches the whichErrorCode query, it will first check if a query function has been defined. It hasn't been, so Turbine then checks to see which response was set via setResponse().
When Turbine is instantiated, you can define default responses in the init object passed to the constructor. Since Turbine defaults all responses to false, you only need to define defaults for non-false values.
var initObj = {
responses : {
isLoggedIn : true,
howManyStars : 3
}
};When Turbine executes a query and 1.) no query function has been defined, and 2.) no reponse has been set via setResponse(), then it will use the response from the init object (or false if none is defined).
Workflows don't always move inexorably forward in a straight line. Sometimes they need to backtrack, repeat, start over, etc. When this happens, you may need to reset some of the responses you previously set.
Just like query functions and default responses, resets are defined in the init object passed to the Turbine constructor. A reset can be either a function reference, or a simple value.
For example, consider a login form that limits a user to 3 login attempts before locking the login form. By default, the user can attempt to login, so the default response for canAttemptLogin is true.
In addition, a reset function is defined for the canAttemptLogin query. This will be executed whenever Turbine moves backwards through the workflow past canAttemptLogin.
var initObj = {
responses : {
canAttemptLogin : true
},
resets : {
canAttemptLogin : user.hasLoginAttemptsRemaining.bind(user)
}
};The login workflow is very simple. We check if the user can attempt a login. If he can, we publish a message telling the app to show the form, and we wait until the form is submitted. Then we check if the login is valid.
If the login is not valid, then we publish a message telling the app there has been an error, then we wait for a message telling us the user wants to retry. When we get that message, we rewind the workflow and replay the canAttemptLogin query.
Here's that workflow:
var workflow = {
canAttemptLogin : {
yes : {
publish : {
message : 'LoginForm.show'
},
waitFor : 'LoginForm.submit',
then : 'isLoginValid'
},
no : {
// publish message to disable login form
}
},
isLoginValid : {
yes : {
// publish message to let user into site
},
no : {
publish : {
message : 'LoginForm.error'
},
waitFor : 'LoginForm.retry',
then : 'canAttemptLogin'
}
}
}The key here is the rewind: whenever we go backwards in the workflow, Turbine checks to see if a reset function (or value) has been defined for each query we have already passed.
In this example, we defined a reset function for canAttemptLogin in our init object. Therefore, each time we rewind from isLoginValid back to canAttemptLogin, Turbine calls the user.hasLoginAttemptsRemaining() function.
The first time through, this will return true. Second time, true. Third time, true.
But when we rewind the fourth time, it will return false. So when the canAttemptLogin query is executed for the fourth time, it will follow the "no" response instead, locking the form against further attempts.
Turbine is an event-driven workflow engine. In the Turbine world, events are called messages. Turbine both publishes messages and waits for messages.
When Turbine publishes a message, the expectation is that your app is listening for that message. When your app gets the message, it goes off and does whatever it needs to do.
When your app is finished doing its thing, it publishes its own message saying it's done.
If Turbine is waiting for that message, it will pick up where it left off, executing the next query in the workflow.
By using publish and waitFor together like this, Turbine is basically telling your app, "Hey, go do some stuff, and let me know when you're done. Then I'll keep going."
Of course, there's no requirement that you wait for a return message after you publish. You can just publish and move on through the workflow. Likewise, you can wait for a message without having published one previously.
var workflow = {
isAppStarted : {
// You can wait for a message without having published one previously
yes : {
waitFor : 'App.stepOne.complete',
then : 'isAfterMidnight'
},
no : {
// do stuff
}
},
isAfterMidnight : {
// You can publish a message without waiting for a return message
yes : {
publish : {
message : 'App.theme.update.DARK_BACKGROUND'
},
then : 'isStepOneComplete'
},
no : {
// do stuff
}
},
isStepOneComplete : {
// You can publish a message then wait for a return message
yes : {
publish : {
message : 'App.stepTwo.show'
},
waitFor : 'App.stepTwo.complete'
then : 'stop.'
},
no : {
// do stuff
}
}
};.
Each instance of Turbine is initialized by calling the Turbine constructor and passing it a single initObj object literal.
The documentation below has more detail about what each property means, but here is the basic structure. Note that only workflow is required; the rest are optional.
// Create initialization object
var initObj = {
// REQUIRED
workflow : {},
// OPTIONAL
name : '',
logLevel : '',
queries : {},
responses : {},
resets : {},
shortcuts : {},
variables : {},
mixins : {},
always : {},
init : function(){},
log : function(){},
publish : function(){},
listen : function(){},
remove : function(){},
report : function(){}
};
// Instantiate new Turbine instance using initObj
var turbine = new Turbine(initObj);Now let's look at what each of these properties means.
[OBJECT] Defines the control flow of your application
The workflow is the jet fuel that powers Turbine. It tells your app what to do, and where to go next after doing it.
Since workflows are a whole topic unto themselves, see the Elements of a workflow section for more details.
[STRING] Gives your Turbine instance a unique name for logging purposes.
This can be useful when you have multiple Turbine instances running simultaneously (or sequentially) and you want to disambiguate the log messages from each instance.
For example, if initObj.name is set to "CartExample", then the console logs will look like:
[CartExample.start()] Starting Turbine
[CartExample.publish()] Publishing message: Turbine|workflow|started
[CartExample.publish()] Publishing message: Turbine|workflow|waiting
If you were to spin up another Turbine instance with an initObj.name of "SignupExample", you'd be able to tell the difference from CartExample:
[CartExample.publish()] Publishing message: Turbine|workflow|waiting
[SingupExample.start()] Starting Turbine
[SingupExample.publish()] Publishing message: Turbine|workflow|started
If no name property is set, then the default value will be "Turbine", i.e. [Turbine.start()] Starting Turbine
[STRING] Determines the verbosity of the logs being output to the console.
Logging is only available in the non-minified version of Turbine.js. In the minified version, all logging functionality is stripped out to reduce file size.
Valid values for initObj.logLevel, in order of increasing verbosity, are:
OFFERRORWARNINFODEBUGTRACE
The default value is ERROR.
[OBJECT] Functions used to resolve queries and return responses.
The initObj.queries object is a collection of key:value pairs. Each key is the name of a query that appears in your workflow; each corresponding value is a reference to a function that will return the result of the query (a.k.a. the response).
For example:
var initObj = {
queries : {
isFoo : someGlobalFunction,
isLoggedIn : app.isLoggedIn.bind(app),
isCartEmpty : cart.isCartEmpty.bind(cart),
getsSpecialOffer : cart.getsSpecialOffer.bind(cart),
whichItemMissing : cart.getMissingItem.bind(cart)
}
}As Turbine steps through each query in your workflow, it looks for a corresponding function defined in initObj.queries. If found, the query function is executed. The return value tells Turbine which response to process.
Defining query functions here promotes decoupling of your workflow from the rest of your app. The functions don't need to know anything about the workflow or how all the pieces of the app fit together -- they just need to be able to return a valid response, and Turbine (along with your workflow) does the rest.
It's important to note that the functions in initObj.queries are just references -- they are not actually called here. You can see this most clearly with the isFoo : someGlobalFunction example. Notice that someGlobalFunction does not have parentheses after it. That's what makes it a reference instead of an executed function.
So what's this bind stuff at the end of the other functions? In a nutshell, bind tells the function what the scope of this should be when it's used inside the function.
bind is actually a feature of the Function prototype in ECMAScript 5 (JavaScript 1.8.5), and is supported by all modern browsers. In other words, everything but IE8.
Never fear though ... Turbine includes an implementation of bind, so you can use it even in browsers that don't technically support it.
To ensure that your function is always being called with this scoped correctly, you should always use bind when defining your query functions.
[OBJECT] Default responses to workflow queries.
The responses object is a collection of key:value pairs. Each key is the name of a query that appears in your workflow; each corresponding value is the default response for that query.
For example:
var initObj = {
responses : {
isCartEmpty : true,
whichItemMissing : 'playstation'
}
}Responses are false by default, so initObj.responses is your chance to define a non-false default response for a query.
When Turbine is instantiated, it imports these default responses. If no query function is defined in initObj.queries, and the response isn't explicitly set in your app via the setResponse() method, then the value from initObj.responses is used.
[OBJECT] Functions or values used to reset query responses when rewinding a workflow
The resets object is a collection of key:value pairs. Each key is the name of a query that appears in your workflow; each corresponding value is either a function or a value to use when rewinding the workflow.
For example:
var initObj = {
resets : {
isCartEmpty : cart.isCartEmpty.bind(cart),
isCheckoutStarted : true
}
}Sometimes Turbine executes a query response that requires it to go backwards in the workflow, to an earlier query. For example, a user may get several steps through a checkout process, then decides to backtrack a few steps to remove a product from the cart.
When this happens, it may be necessary to reset some of the responses that are being rewound. For example, if a user is on Step 4 of the checkout, and wants to go back to Step 2, you may need to reset the value of a isStepThreeComplete query to false.
To do this you could either set initObj.resets.isStepThreeComplete = false, or your could set it to a function to be called to determine the reset value, i.e. initObj.resets.isStepThreeComplete = app.isStepThreeComplete.bind(app).
If no reset for a query is defined in initObj.resets, then the response is not reset during a rewind.
[OBJECT] Aliases for workflow queries
Shortcuts are a way for you to reference a query by an alias instead of using it directly. This creates greater flexibility in your workflow by decoupling intention from expression. The shortcut name can be any arbitrary string, and you can define as many shortcuts as you want.
For example, say you sometimes want your workflow to go back to the beginning of the checkout process based on some query response. You can define a checkout shortcut like this:
var initObj = {
shortcuts : {
checkout : 'isCheckoutStarted'
}
}To use the shortcut in your workflow, you would reference it with an @ symbol, like @checkout:
workflow : {
isCheckoutStarted : {
yes : {
then : 'isCheckoutCancelled'
},
no : {
// do stuff
}
},
isCheckoutCancelled : {
yes : {
then : '@checkout'
},
no : {
// do stuff
}
}
}By using the shortcut, your workflow doesn't need to know which query is the starting query -- it just needs to know to go back to the query defined by @checkout.
If, in the future, you add additional queries to the beginning of your checkout flow, you only need to change the definition of the checkout shortcut in the config.
You can use any arbitrary string for a shortcut name, but there's one that has special meaning: @start. If you define a start shortcut, Turbine will use that as the first query to execute when Turbine is started.
[OBJECT] Keys representing scalar values (string, boolean, numeric, null)
As you might expect, variables in Turbine work just like those in any programming language: the variable is replaced with the value defined in the config.
var initObj = {
variables : {
cartTimeout : 36000
}
}To use the variable in your workflow, you would reference it with an $ symbol, like $cartTimeout:
workflow : {
isCheckoutStarted : {
yes : {
timeout : {
after : '$cartTimeout',
publish : 'Cart.timeout.expired',
then : 'stop.'
}
},
no : {
// do stuff
}
}
}The one caveat is that variables can only be used for string, boolean, numeric, or null values. If you want a variable-like way to represent object literals, use a mixin instead.
[OBJECT] Keys representing object literals
A mixin is basically a variable representing an object literal. Mixins are replaced recursively, so you can use mixins within mixins. You can also use variables and shortcuts within mixins.
var initObj = {
mixins : {
invalidLogin : {
publish : {
message : 'Cart.login.failed'
}
}
}
}To use the mixin in your workflow, you would reference it with a + symbol, like +invalidLogin:
workflow : {
whichError : {
USERNAME_NOT_FOUND : '+invalidLogin',
PASSWORD_INCORRECT : '+invalidLogin',
CAPTCHA_INCORRECT : '+invalidLogin'
}
}When Turbine imports your workflow, it replaces the mixins like this:
workflow : {
whichError : {
USERNAME_NOT_FOUND : {
publish : {
message : 'Cart.login.failed'
}
},
PASSWORD_INCORRECT : {
publish : {
message : 'Cart.login.failed'
}
},
CAPTCHA_INCORRECT : {
publish : {
message : 'Cart.login.failed'
}
}
}
}For a more complex implementation of mixins, see the example app in the /examples directory.
[OBJECT]
The always object is a way to define things that should be added to every query that is executed. This saves you from needing to duplicate the same code over and over.
var initObj = {
always : {
timeout : {},
waitFor : [],
using : {}
}
}The timeout property allows you to define a global timeout for the entire workflow.
For example, you may want to ask the user if they're still there when there has been no activity for a few minutes. Or you may want to raise an error if you app has become unresponsive for some reason. The format of the timeout property is the same as when timeout is defined in a response (see docs below).
timeout : {
after : 300000,
publish : {
message : "Cart.issue.detected.GLOBAL_TIMEOUT"
},
then : "stop."
},The waitFor property defines messages for which to listen, as well as an optional then that tells the workflow where to go when a message is received. Whenever your app is waiting for messages, these global waitFor messages will be listened for as well.
The format of the waitFor property is the same as when waitFor is defined in a response (see docs below).
The using property is an object literal that will be merged with the using property whenever a message is published from Turbine.
using : {
timestamp : new Date().getTime()
}workflow : {
isAppStarted : {
yes : {
publish : {
message : 'Cart.app.started',
using : {
storeName : 'My Store'
}
}
},
no : {
// do stuff
}
}
}When this workflow runs and the yes response to isAppStarted is executed, Turbine will publish the Cart.app.started message. The message payload will be an object containing storeName, as well as timestamp from the global using object.
[FUNCTION] Initialization function called at the end of the Turbine constructor
The init function is an optional function that can be defined to be called once Turbine's constructor is complete. It is passed one argument: the Turbine instance that was just instantiated. This might be useful if you want your app to wait for Turbine to be fully instantiated before doing something.
[FUNCTION] Custom logging function
By default, Turbine outputs all its logs via the standard console methods: log, warn, and error. If you would rather send the logs to some other function, you can define it here and Turbine will use that instead.
Note that log messages are only output by the non-minified Turbine.js. Logging is stripped out of the minified version to reduce file size.
[FUNCTION] Function to use when publishing events
By default, Turbine will use jQuery.trigger() to publish events. If you would rather use some other event publishing method, such as Backbone.Events.trigger(), you can define that method here.
Turbine will pass your publish method two arguments:
message[String] Event to publishpayload[Object] Optional data object
Your events library may not be expecting those arguments, or in that order, so you may have to wrap your library's function in your own function that translates those arguments into something your library understands.
For example, maybe your fictional PubSub library requires a single object literal defining event and data instead of two arguments for message and payload. Then you might wrap it like this:
var initObj = {
publish : function(message,payload){
yourPubSub.trigger({
event : message,
data : payload
});
}
}[FUNCTION] Function to use when listening for events
By default, Turbine will use jQuery.on() to listen for events. If you would rather use some other event listener, such as Backbone.Events.on(), you can define that method here.
Turbine will pass your listen method two arguments:
message[String] Event to listen forhandler[Function] Function to call when event is triggered
When the handler is triggered, it will be passed two arguments:
message[String] Event that triggered the handlerpayload[Object] Optional data object
Like publish, you may need to wrap your own event listener method in a custom function to translate these arguments into something your listener understands.
[FUNCTION] Function to use when removing event listeners
By default, Turbine will use jQuery.off() to remove listeners. If you would rather use some other event library, such as Backbone.Events.off(), you can define that method here.
Turbine will pass your remove method one argument:
message[String] Event for which listeners should be removed
Like publish and listen, you may need to wrap your own method in a custom function to translate these arguments into something your listener understands.
[FUNCTION] Function to use when reporting errors or events
By default, Turbine will report any internal errors to the browser console via console.error().
If you'd rather have issues reported through some event logging or analytics system, you can define your own custom report function here.
Turbine will pass your report method one argument:
obj[Object] Data object
If an internal Turbine issue is being reported, this object will contain two properties:
handleShort string identifying the issue (WORKFLOW_ISSUE_REPORTED is the default)descriptionHuman-readable description of the issue
The report function isn't just for errors though -- it can be used in the workflow to report any arbitrary event or activity. In this case, the data object passed to report is entirely defined in your workflow.
For example, say you have a isUserBanned query in your workflow. When a banned user tries to access your app, you want to report that activity to a security monitor. You might have this in your workflow:
var workflow = {
isUserBanned : {
yes : {
report : {
errorType : 'FATAL'
handle : 'BANNED_USER_LOGIN',
description : 'A banned user tried to log into the site',
username : app.getUserName()
timestamp : new Date().getTime()
},
then : 'stop.'
},
no : {
then : 'isUserLoggedIn'
}
}
};Your report function would be passed whatever is defined in the workflow. You can then use that data to report the issue however your system requires.
.
A workflow is an object literal defined in the init object passed to the Turbine constructor. It is the only mandatory property of the init object.
The workflow is essentially a dialog between Turbine and your app. Turbine executes a query and receives a response. That response tells Turbine what to do next, and which query to execute next.
isUserLoggedIn : {
yes : {
// do something
},
no : {
// do something different
}
}In this example:
isUserLoggedInis the queryyesandnoare the responses- The object literal values of the
yesandnoresponses are the response bodies
When Turbine starts your workflow, it begins with the first query in the workflow (or the @start shortcut, if defined).
To get the response to the query, Turbine checks a few things:
- Has a query function been set in
initObj.queries? If so, Turbine executes the function and processes its response. - If there's no query function, has a response been set using
Turbine.setResponse()? If so, Turbine uses that value. - If no response has been set, has any default response been set in
initObj.responses? If so, Turbine uses that value. - If none of the above exist, then Turbine returns false and processes the "no" response.
Responses can be boolean, strings, or numbers. If a response is boolean, true is converted to "yes" and false is converted to "no". In addition, null and undefined responses are also converted to "no".
A query's responses work similiarly to a JavaScript switch/case statement. If the value of the query's response matches any of the responses in the workflow, that response is processed.
In addition, there is the special default response. If default is defined, and the query's response doesn't match any of the responses defined in the workflow, Turbine will use the default response.
whichError : {
INVALID_EMAIL : {
// display invalid email error
},
INVALID_USERNAME : {
// display invalid username error
},
default : {
// display generic error
}
}In this example, if the whichError query doesn't return either INVALID_EMAIL or INVALID_USERNAME, then the default response will be processed.
Once a query has been executed and a response has been received, we need to know what to do next. This is expressed in the response body.
We'll get into details about all the things that can go into response bodies in a minute. For now, let's just focus on the most important property: then.
The then property tells Turbine which query to execute next. When you see it in action, it's pretty self-explanatory:
isUserLoggedIn : {
yes : {
then : 'isUserOver18'
},
no : {
then : 'doesAccountExist'
}
},
isUserOver18 : {
yes : {
// let the user in
},
no : {
// don't let the user in
}
},
doesAccountExist : {
yes : {
// make the user log in
},
no : {
// ask the user to create an account
}
}Turbine's expressive workflow syntax makes it simple to see how the program will flow. For example:
- Is the user logged in? Yes. Is the user over 18? Yes. Then let him in.
- Is the user logged in? No. Does an account exist? No. Then ask the user to create one.
- Is the user logged in? No. Does an account exist? Yes. Then ask the user to log in.
- And so on ...
Because then tells your workflow where to go next, it is required for every response body.
(There's a slight caveat to that rule when using the repeat or delay property -- more on that later.)
If you leave then out, your app will basically freeze -- Turbine will get to the response that has no then in the response body, and it won't know where to go from there. Instead, it will throw an exception.
Of course, there will be times where your workflow really has no place else to go. In this case, you can set the value of then to either stop. or kill!. Using these special values allows your workflow to clearly indicate that it intends to stop.
Setting then to stop. tells Turbine to stop. There are no ill effects -- you can restart Turbine later using start(), and it will start over from the beginning of the workflow.
Setting then to kill! not only tells Turbine to stop, but it also prevents it from being started again. If you call start() after using kill!, Turbine will simply report an error.
Note that both stop. and kill! include punctuation -- that's required in order for Turbine to recognize them as special values.
Continuing to flesh out the example above, we can add stop. to isUserOver18.no:
isUserLoggedIn : {
yes : {
then : 'isUserOver18'
},
no : {
then : 'doesAccountExist'
}
},
isUserOver18 : {
yes : {
// let the user in
},
no : {
// don't let the user in
then : 'stop.'
}
},
doesAccountExist : {
yes : {
// make the user log in
},
no : {
// ask the user to create an account
}
}Turbine is an event-driven workflow engine, so it communicates with your app by publishing messages from the workflow using the publish function defined in your initObj, or jQuery.trigger() by default.
To tell Turbine to publish a message, you define a publish object in the response body. This object has two properties:
message[String or Array] The message to publish, or an array of messages to publishusing[Object] Optional data object to accompany published message
Let's publish some messages in our example:
isUserLoggedIn : {
yes : {
then : 'isUserOver18'
},
no : {
then : 'doesAccountExist'
}
},
isUserOver18 : {
yes : {
// let the user in
publish : {
message : 'App.view.show',
using : {
view : 'loginAccepted',
content : 'ENTRY_ACCEPTED'
}
},
then : 'stop.'
},
no : {
// don't let the user in
publish : {
message : [
'User.age.rejected',
'App.view.show'
],
using : {
view : 'loginRejected',
content : 'NOT_OLD_ENOUGH'
}
}
then : 'stop.'
}
},
doesAccountExist : {
yes : {
// make the user log in
publish : {
message : 'App.view.show',
using : {
view : 'logIn'
}
}
},
no : {
// ask the user to create an account
publish : {
message : 'App.view.show',
using : {
view : 'createAccount'
}
}
}
}Of course, publishing a message doesn't do much if there's nothing listening for it in your app. The expectation is that when your workflow publishes App.view.show, your app is listening for that message. Your app can then use the values from the using object to determine which view to show, and which content to use when showing it.
In the isUserOver18 query above, both responses have then : 'stop.' in their response bodies. In those cases, after the workflow publishes its message, it's done. There's nothing left to do after the appropriate view is shown.
But what about in the doesAccountExist query? We want the user to either log in or create an account, so we published an App.view.show message from your workflow, your app was listening for it, it showed the view ... now what?
We need a way for the app to tell Turbine that it is done doing whatever it needed to do. For that, we use waitFor.
The waitFor property accepts either a message or array of messages for which Turbine should listen. Once Turbine receives a message it's waiting for, it continues where it left off, going wherever the then property tells it to go.
Let's add some waitFor and then properties to the doesAccountExist response bodies. We'll also need to add two new queries: isLoginValid and isAccountValid:
isUserLoggedIn : {
yes : {
then : 'isUserOver18'
},
no : {
then : 'doesAccountExist'
}
},
isUserOver18 : {
yes : {
// let the user in
publish : {
message : 'App.view.show',
using : {
view : 'loginAccepted',
content : 'ENTRY_ACCEPTED'
}
},
then : 'stop.'
},
no : {
// don't let the user in
publish : {
message : [
'User.age.rejected',
'App.view.show'
],
using : {
view : 'loginRejected',
content : 'NOT_OLD_ENOUGH'
}
}
then : 'stop.'
}
},
doesAccountExist : {
yes : {
// make the user log in
publish : {
message : 'App.view.show',
using : {
view : 'logIn'
}
},
waitFor : 'App.login.submitted',
then : 'isLoginValid'
},
no : {
// ask the user to create an account
publish : {
message : 'App.view.show',
using : {
view : 'createAccount'
}
},
waitFor : 'App.account.created',
then : 'isAccountValid'
}
},
isLoginValid : {
yes : {
then : 'isUserOver18'
},
no : {
// show login error
}
},
isAccountValid : {
yes : {
then : 'isUserOver18'
},
no : {
// show login error
}
}Let's assume our user has an account, so we showed him a login form. When that form is submitted, your app publishes a App.login.submitted message. Since Turbine is waiting for that message, it follows then then property to isLoginValid.
Sometimes you might want your app to execute a different query depending on which waitFor message it receives. To do this, you can specify waitFor as an object or array of objects containing two properties:
message[String or Array] The message to wait for, or an array of messages to wait forthen[String] Query to execute when one of the messages is received
For example:
isAgeGateRequired : {
yes : {
publish : {
message : 'App.view.show',
using : {
view : 'ageGate',
content : 'Are you over 18?'
}
},
waitFor : [
{
message : ['App.button.clicked.YES','App.button.clicked.NO'],
then : 'isOldEnough'
}
{
message : 'App.button.clicked.HELP',
then : 'isHelpLoaded'
}
]
},
no : {
// do stuff
}
},
isOldEnough : {
yes : {
// let them in
},
no : {
// don't let them in
}
}
isHelpLoaded : {
yes : {
// show Help
},
no : {
// load Help
}
}If you specify a then to accompany a waitFor message, it will override any then that is specified outside waitFor.
isAgeGateRequired : {
yes : {
publish : {
message : 'App.view.show',
using : {
view : 'ageGate',
content : 'Are you over 18?'
}
},
waitFor : [
{
message : ['App.button.clicked.YES','App.button.clicked.NO'],
then : 'isOldEnough'
},
{
message : 'App.button.clicked.HELP',
then : 'isHelpLoaded'
}
],
then : 'isLoggedIn'
},
no : {
// do stuff
}
}In the example above, isLoggedIn will never be executed, because each waitFor message has its own then property.
However, if you don't specify a then to accompany a waitFor message, the then that is specified outside waitFor will be used.
isAgeGateRequired : {
yes : {
publish : {
message : 'App.view.show',
using : {
view : 'ageGate',
content : 'Are you over 18?'
}
},
waitFor : [
{
message : ['App.button.clicked.YES','App.button.clicked.NO']
}
{
message : 'App.button.clicked.HELP',
then : 'isHelpLoaded'
}
],
then : 'isOldEnough'
},
no : {
// do stuff
}
}In the example above, the App.button.clicked.YES and App.button.clicked.NO will execute isOldEnough next, whereas App.button.clicked.HELP will execute isHelpLoaded.
While this approach technically will work, it is not recommended -- if you're using multiple then options, then each waitFor message really should have its own then, for clarity's sake.
Sometimes you may want to repeat the same query over and over again, such as when you are polling a server for a particular response. To do this, you add a repeat object in the response body.
The repeat object is used in lieu of then -- by using repeat, you are implicitly saying "execute this query, then execute this query again".
The repeat object contains one required property:
limit[Number or null] The maximum times the query will be repeated. If null, the query will repeat infinitely.
In addition, the repeat object can contain anything that a response body can contain: publish, waitFor, then, etc. If the limit is reached, the repeat object is processed as a response body.
isUploadComplete : {
no : {
waitFor : 'App.upload.updated',
repeat : {
limit : 100,
publish : {
message : 'App.upload.failed',
using : {
reason : 'UPLOAD_CHECK_LIMIT_EXCEEDED'
}
},
then : 'stop.'
}
},
yes : {
// display Done message
}
}In the example above, Turbine waits for an App.upload.updated message. When it gets one, it repeats the isUploadComplete query. If the response is still no, then it again waits for App.upload.updated.
This continues until isUploadComplete is yes, or the query repeats 100 times. If the limit is reached, then Turbine executes the repeat object as a response body, publishing App.upload.failed and then stopping.
One of the drawbacks of an event-driven workflow engine is that if it's waiting for a message that never comes, it is basically stuck. To help avoid that situation, a timeout object is available.
The timeout object allows you to specify an alternate response body to process after a certain amount of time elapses. It contains one required property:
after[Number] The number of milliseconds after which the timeout will fire
In addition, the timeout object can contain anything that a response body can contain: publish, waitFor, then, etc. If the timeout is exceeded, the timeout object is processed as a response body.
isTransactionComplete : {
no : {
waitFor : 'App.transaction.completed',
timeout : {
after : 300000,
publish : {
message : 'App.transaction.failed',
using : {
reason : 'TIMEOUT_EXCEEDED'
}
},
then : 'stop.'
},
then : 'isTransactionComplete'
},
yes : {
// display Done message
}
}In the example above, Turbine is waiting for an App.transaction.complete message. If it doesn't receive it after 300000 milliseconds (5 mins), it publishes an App.transaction.failed message, then stops.
Notice that the no response body still has its own then value -- that is required so Turbine knows where to go if it does receive the App.transaction.complete message before the timeout.
Sometimes you may want to wait a little while before processing a response body. For example, say you're using Turbine to prototype a loading screen -- you might want to add a 3-second delay, then publish a message to move to the next screen. That's where the delay object is useful.
The delay object is used in lieu of then -- by using delay, you are implicitly saying "wait for a while, then process the delayed response".
The delay object contains one required property:
for[Number] The number of milliseconds to wait before processing the delayed response body
In addition, the delay object can contain anything that a response body can contain: publish, waitFor, then, etc. After the delay elapses, the delay object is processed as a response body.
isAppLoaded : {
no : {
delay : {
for : 3000,
publish : {
message : 'App.view.show',
using : {
view : 'appLoaded'
}
},
then : 'stop.'
}
},
yes : {
// display Loaded message
}
}In the example above, isAppLoaded gets a no response. Turbine waits for 3000 ms (3 seconds), publishes the App.view.show message, then stops.
Note that delay doesn't have to live alone in the response body. It can co-exist with other properties as well. For example:
isAppLoaded : {
no : {
publish : {
message : 'App.view.show',
using : {
view : 'appLoading'
}
},
delay : {
for : 3000,
publish : {
message : 'App.view.show',
using : {
view : 'appLoaded'
}
},
then : 'stop.'
}
},
yes : {
// display Loaded message
}
}In many ways, Turbine is just a big state machine. As such, it is a centralized mechanism for monitoring the state of your application. If your application is in a state that is worth reporting, you can define report in your response body.
The value of report can be anything you want: a string, an object, an array, etc.
By default, report just passes its value to console.error(). However, you can define your own custom report function in initObj, allowing you to send reports to Google Analytics, Omniture, or whatever your preferred analytics tool may be.
Consider the example we used for repeat earlier. We checked 100 times whether the upload was complete, and it never was. That's the sort of thing you might want to report.
isUploadComplete : {
no : {
waitFor : 'App.upload.updated',
repeat : {
limit : 100,
publish : {
message : 'App.upload.failed',
using : {
reason : 'UPLOAD_CHECK_LIMIT_EXCEEDED'
}
},
report : 'APP_UPLOAD_FAILURE',
then : 'stop.'
}
},
yes : {
// display Done message
}
}Now APP_UPLOAD_FAILURE will be sent to your reporting system so you can investigate why it failed.
The Turbine API is extremely simple -- most of the logic and complexity is implemented in the workflow itself. There are just a few methods available:
Once an instance of Turbine has been created, calling start() will actually start the workflow.
var turbine = new Turbine(initObj);
turbine.start();You can also chain start() directly to the new Turbine instance, if you want it to start right away:
var turbine = new Turbine(initObj).start();If you want to know whether Turbine has been started, you can use isStarted().
var turbine = new Turbine(initObj);
turbine.isStarted(); // returns false
turbine.start();
turbine.isStarted(); // returns trueTo stop Turbine, simple call stop().
var turbine = new Turbine(initObj);
turbine.isStarted(); // returns false
turbine.start();
turbine.isStarted(); // returns true
turbine.stop();
turbine.isStarted(); // returns falseThe getVar() method retrieves the value of the variable set in Turbine's init object.
varName[String] The name of the variable to retrieve
var initObj = {
variables : {
maxRetries : 100
}
};
var turbine = new Turbine(initObj);
turbine.getVar('maxRetries'); // returns 100The setResponse method sets the response to a query.
query[String] The query to set the response forresponse[String or Number or Boolean] The response to the query
var turbine = new Turbine(initObj);
turbine.setResponse('isLoggedIn',true);Inside the /examples directory, you'll find a sample app that uses most of the concepts described in these docs.
Just open /examples/index.html in a browser, select which type of workflow you want to load, then click Start Turbine.
The app has a simple list of products, along with a mock shopping cart. You can add and remove items in the cart, do a simulated login and signup, and execute a mock checkout.
Some things to try:
- Try adding an item without logging in. You'll be prompted to log in. Click the Log In button, then try adding the item again. You'll be able to add it now.
- If you try adding an item 3 times without logging in, you'll be forced to sign up.
- If you add a PlayStation and NBA 2K13, you'll get a discount.
- If, during signup, you select Male and Basketball, you'll get bonus SuperShopper points.
- You can add items to the cart immediately
- If you add a PlayStation and wait a few seconds, you'll be prompted to add a DualShock controller.
- If you add a DualShock, you'll be prompted to add a charging station.
- If you try to check out without logging in, you'll be prompted to log in first.
To see how this all fits together as workflows, check out /examples/js/init.js. There you'll find the initObj that sets up the workflows.
IMPORTANT NOTE: This example app is meant to show how to implement Turbine and its workflows. It is not a good example of how to actually write a web app. There's some kludgy code, there's HTML commingled in the JavaScript, etc. It's pretty gross.
Sure. As a matter of fact, the signup flow in the example app is a separate workflow from the shopping cart flow. Take a look.
First, instantiate an instance of Turbine and start it up. When you get to the part of the workflow where you want to kick off another workflow, publish a message that tells your app to start the other workflow, then wait for a message that lets you know when the other workflow is done.
For example, your first workflow might have:
isSignupRequired : {
yes : {
publish : {
message : 'App.signup.start'
},
waitFor : 'App.signup.complete',
then : 'isSignupValid'
},
no : {
// let them in
}
}
isSignupValid : {
yes : {
// let them in
},
no : {
// don't let them in
}
}Your app can listen for App.signup.start. When it receives that message, it creates a new Turbine instance for the signup workflow. When that workflow is complete, it publishes App.signup.complete. Since the first workflow is waiting for that message, it will execute the isSignupValid query. Voila! Nested workflows!
Please submit all bugs, questions, and suggestions via the Issues section so everyone can benefit from the answer.
If you need to contact me directly, email [email protected].
Copyright (c) 2012 Warren Benedetto [email protected]
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.