m.deferred
- Usage
- Retrieving a value via the getter-setter API
- Integrating to the Mithril redrawing system
- Differences from Promises/A+
- The exception monitor
- Signature
This is a low-level method in Mithril. It's a modified version of the Thenable API.
A deferred is an asynchrony monad. It exposes a promise
property which can bind callbacks to build a computation tree.
The deferred object can then apply a value by calling either resolve
or reject
, which then dispatches the value to be processed to the computation tree.
Each computation function takes a value as a parameter and is expected to return another value, which in turns is forwarded along to the next computation function (or functions) in the tree.
The deferred object returned by m.deferred
has two methods: resolve
and reject
, and one property called promise
. The methods can be called to dispatch a value to the promise tree. The promise
property is the root of the promise tree. It has a method then
which takes a successCallback
and a errorCallback
callbacks. Calling the then
method attaches the computations represented by successCallback
and errorCallback
to the promise, which will be called when either resolve
or reject
is called. The then
method returns a child promise, which, itself, can have more child promises, recursively.
The promise
object is actually a function - specifically, it's an m.prop
getter-setter, which gets populated with the value returned by successCallback
if the promise is resolved successfully.
Note that Mithril promises are not automatically integrated to its automatic redrawing system. If you wish to use third party asynchronous libraries (for example, jQuery.ajax
), you should also consider using m.startComputation
/ m.endComputation
if you want views to redraw after requests complete.
Usage
//standalone usage
var greetAsync = function() {
var deferred = m.deferred();
setTimeout(function() {
deferred.resolve("hello");
}, 1000);
return deferred.promise;
};
greetAsync()
.then(function(value) {return value + " world"})
.then(function(value) {console.log(value)}); //logs "hello world" after 1 second
Retrieving a value via the getter-setter API
The promise object is actually a getter-setter function that gets populated when the promise is fulfilled.
//asynchronous service
var greetAsync = function() {
var deferred = m.deferred();
setTimeout(function() {
deferred.resolve("hello");
}, 1000);
return deferred.promise;
};
//asynchronous consumer
var greeting = greetAsync()
var processed = greeting.then(function(value) {return value + " world"})
console.log(greeting()) // undefined - because `deferred.resolve` has not been called yet
setTimeout(function() {
//now `deferred.resolve` has been called
console.log(greeting()) // "hello"
console.log(processed()) // "hello world"
}, 2000)
Integrating to the Mithril redrawing system
By default, promises are not integrated to the Mithril auto-redrawing system. When dealing with asynchronous functions, you must call [m.startComputation
/ m.endComputation
] if you want the asynchronous payload to affect the view.
//asynchronous service
var greetAsync = function() {
//tell Mithril to wait for this service to complete before redrawing
m.startComputation();
var deferred = m.deferred();
setTimeout(function() {
deferred.resolve("hello");
//the service is done, tell Mithril that it may redraw
m.endComputation();
}, 1000);
return deferred.promise;
};
Some cases may not require a redraw upon completion of the asynchronous callbacks. In such cases, simply omit the m.startComputation/m.endComputation calls.
Some asynchronous operations might need to affect redrawing both before and after their completion. In those cases, you can call m.redraw
instead of using m.startComputation/m.endComputation.
//asynchronous service
var greetAsync = function() {
//don't wait for this service; redraw right away
var deferred = m.deferred();
setTimeout(function() {
deferred.resolve("hello");
//redraw again
m.redraw()
}, 1000);
return deferred.promise;
};
Differences from Promises/A+
For the most part, Mithril promises behave as you'd expect a Promise/A+ promise to behave, but have one difference: Mithril promises attempt to execute synchronously if possible.
Synchronous execution
Mithril promises attempt to execute synchronously if possible. To illustrate the difference between Mithril and A+ promises, consider the code below:
var deferred = m.deferred()
deferred.promise.then(function() {
console.log(1)
})
deferred.resolve("value")
console.log(2)
In the example above, A+ promises are required to log 2
before logging 1
, whereas Mithril logs 1
before 2
. Typically resolve
/reject
are called asynchronously after the then
method is called, so normally this difference does not matter.
There are a couple of reasons why Mithril runs callbacks synchronously. Conforming to the spec requires either a setImmediate
polyfill (which is a significantly large library), or setTimeout
(which is required to take at least 4 milliseconds per call, according to its specs). Neither of these trade-offs are acceptable, given Mithril's focus on nimbleness and performance.
Unchecked Error Handling
By default, Mithril does not swallow errors if these errors are subclasses of the Error class. Manually throwing an instance of the Error class itself (or any other objects or primitives) does trigger the rejection callback path as per the Promises/A+ spec.
This deviation from the spec is there to make it easier for developers to find common logical errors such as typos that lead to null reference exceptions. By default, the spec requires that all thrown errors trigger rejection, which result in silent failures if the developer forgets to explicitly handle the failure case.
For example, there is simply never a case where a developer would want to programmatically handle the error of accessing the property of a nullable entity without first checking for its existence. The only reasonable course of action to prevent the potential null reference exceptions in this case is to add the existence check in the source code. It is expected that such an error would bubble up to the console and display a developer-friendly error message and line number there.
m.request({method: "GET", url: "/things"})
.then(function(items) {
item.foreach(doSomething) //programmer error: typo will throw runtime error to the console
})
The other side of the coin is still supported: if a developer needs to signal an exceptional condition within a promise callback, they can manually throw a new Error
(for example, if a validation rule failed, and there should be an error message displayed to the user).
var error = m.prop()
m.request({method: "GET", url: "/user/:id", data: {id: 1}})
.then(function(user) {
if (user.isAdmin) throw new Error("Sorry, you don't have permissions")
})
.then(null, error) //handle the application error: bind to a getter-setter for diplaying it on the template
Note that the default promise exception handling semantics can be modified. See the next section.
The exception monitor
Any time an exception is thrown inside a promise callback, Mithril calls m.deferred.onerror(e)
.
By default, this event handler rethrows the exception to the console if an error is a subclass of Error (but not an instance of Error itself). Otherwise it follows the Promises/A+ specifications. It does this because people expect unexpected errors like null reference exceptions to be thrown to the console for debugging purposes, and these errors are always subclasses of Error.
On the other hand, javascript developers rarely ever throw errors that are subclasses of Error, and for the purposes of application error handling, the underlying prototypal chain of the error class is typically not relevant.
The onerror
function can be safely replaced if the default error monitoring semantics are not desired.
//swallow all errors
m.deferred.onerror = function() {}
//only log errors
m.deferred.onerror = function(e) {console.error(e)}
Signature
Deferred deferred() {void onerror(Error e)}
where:
Deferred :: Object { Promise promise, void resolve(any value), void reject(any value) }
Promise :: GetterSetter { Promise then(any successCallback(any value), any errorCallback(any value)) }
GetterSetter :: any getterSetter([any value])
GetterSetter { Promise then([any successCallback(any value) [, any errorCallback(any value)]]) } promise
A promise has a method called
then
which takes two computation callbacks as parameters.The
then
method returns another promise whose computations (if any) receive their inputs from the parent promise's computation.A promise is also a getter-setter (see
m.prop
). After a call to eitherresolve
orreject
, it holds the result of the parent's computation (or theresolve
/reject
value, if the promise has no parent promises)Promise then([any successCallback(any value) [, any errorCallback(any value)]])
This method accepts two callbacks which process a value passed to the
resolve
andreject
methods, respectively, and pass the processed value to the returned promiseany successCallback(any value) (optional)
The
successCallback
is called ifresolve
is called in the rootdeferred
.The default value (if this parameter is falsy) is the identity function
function(value) {return value}
If this function returns undefined, then it passes the
value
argument to the next step in the thennable queue, if anyany errorCallback(any value) (optional)
The
errorCallback
is called ifreject
is called in the rootdeferred
.The default value (if this parameter is falsy) is the identity function
function(value) {return value}
If this function returns undefined, then it passes the
value
argument to the next step in the thennable queue, if anyreturns Promise promise
void resolve(any value)
This method passes a value to the
successCallback
of the deferred object's child promisevoid reject(any value)
This method passes a value to the
errorCallback
of the deferred object's child promise-
m.deferred.onerror
void onerror(Error e)
This method gets called every time an exception is thrown inside a promise callback. By default, it rethrows to the console if an error is a subclass of Error (but not an instance of Error itself). Otherwise it follows the Promises/A+ specifications.