Mithril 1.1.0-rc.1

Change log


v1.0.2

News

Bug fixes


v1.0.1

News

Bug fixes


Migrating from v0.2.x

v1.x is largely API-compatible with v0.2.x, but there are some breaking changes.

If you are migrating, consider using the mithril-codemods tool to help automate the most straightforward migrations.


m.prop removed

In v1.x, m.prop() is now a more powerful stream micro-library, but it's no longer part of core. You can read about how to use the optional Streams module in the documentation.

v0.2.x

var m = require("mithril")

var num = m.prop(1)

v1.x

var m = require("mithril")
var prop = require("mithril/stream")

var num = prop(1)
var doubled = num.map(function(n) {return n * 2})

m.component removed

In v0.2.x components could be created using either m(component) or m.component(component). v1.x only supports m(component).

v0.2.x

// These are equivalent
m.component(component)
m(component)

v1.x

m(component)

config function

In v0.2.x mithril provided a single lifecycle method, config. v1.x provides much more fine-grained control over the lifecycle of a vnode.

v0.2.x

m("div", {
    config : function(element, isInitialized) {
        // runs on each redraw
        // isInitialized is a boolean representing if the node has been added to the DOM
    }
})

v1.x

More documentation on these new methods is available in lifecycle-methods.md.

m("div", {
    // Called before the DOM node is created
    oninit : function(vnode) { /*...*/ },
    // Called after the DOM node is created
    oncreate : function(vnode) { /*...*/ },
    // Called before the node is updated, return false to cancel
    onbeforeupdate : function(vnode, old) { /*...*/ },
    // Called after the node is updated
    onupdate : function(vnode) { /*...*/ },
    // Called before the node is removed, return a Promise that resolves when
    // ready for the node to be removed from the DOM
    onbeforeremove : function(vnode) { /*...*/ },
    // Called before the node is removed, but after onbeforeremove calls done()
    onremove : function(vnode) { /*...*/ }
})

If available the DOM-Element of the vnode can be accessed at vnode.dom.


Changes in redraw behaviour

Mithril's rendering engine still operates on the basis of semi-automated global redraws, but some APIs and behaviours differ:

No more redraw locks

In v0.2.x, Mithril allowed 'redraw locks' which temporarily prevented blocked draw logic: by default, m.request would lock the draw loop on execution and unlock when all pending requests had resolved - the same behaviour could be invoked manually using m.startComputation() and m.endComputation(). The latter APIs and the associated behaviour has been removed in v1.x. Redraw locking can lead to buggy UIs: the concerns of one part of the application should not be allowed to prevent other parts of the view from updating to reflect change.

Cancelling redraw from event handlers

m.mount() and m.route() still automatically redraw after a DOM event handler runs. Cancelling these redraws from within your event handlers is now done by setting the redraw property on the passed-in event object to false.

v0.2.x

m("div", {
    onclick : function(e) {
        m.redraw.strategy("none")
    }
})

v1.x

m("div", {
    onclick : function(e) {
        e.redraw = false
    }
})

Synchronous redraw removed

In v0.2.x it was possible to force mithril to redraw immediately by passing a truthy value to m.redraw(). This behavior complicated usage of m.redraw() and caused some hard-to-reason about issues and has been removed.

v0.2.x

m.redraw(true) // redraws immediately & synchronously

v1.x

m.redraw() // schedules a redraw on the next requestAnimationFrame tick

m.startComputation/m.endComputation removed

They are considered anti-patterns and have a number of problematic edge cases, so they no longer exist in v1.x.


Component controller function

In v1.x there is no more controller property in components, use oninit instead.

v0.2.x

m.mount(document.body, {
    controller : function() {
        var ctrl = this

        ctrl.fooga = 1
    },

    view : function(ctrl) {
        return m("p", ctrl.fooga)
    }
})

v1.x

m.mount(document.body, {
    oninit : function(vnode) {
        vnode.state.fooga = 1
    },

    view : function(vnode) {
        return m("p", vnode.state.fooga)
    }
})

// OR

m.mount(document.body, {
    oninit : function(vnode) {
        var state = this  // this is bound to vnode.state by default

        state.fooga = 1
    },

    view : function(vnode) {
        var state = this // this is bound to vnode.state by default

        return m("p", state.fooga)
    }
})

Component arguments

Arguments to a component in v1.x must be an object, simple values like String/Number/Boolean will be treated as text children. Arguments are accessed within the component by reading them from the vnode.attrs object.

v0.2.x

var component = {
    controller : function(options) {
        // options.fooga === 1
    },

    view : function(ctrl, options) {
        // options.fooga == 1
    }
}

m("div", m.component(component, { fooga : 1 }))

v1.x

var component = {
    oninit : function(vnode) {
        // vnode.attrs.fooga === 1
    },

    view : function(vnode) {
        // vnode.attrs.fooga == 1
    }
}

m("div", m(component, { fooga : 1 }))

view() parameters

In v0.2.x view functions are passed a reference to the controller instance and (optionally) any options passed to the component. In v1.x they are passed only the vnode, exactly like the controller function.

v0.2.x

m.mount(document.body, {
    controller : function() {},

    view : function(ctrl, options) {
        // ...
    }
})

v1.x

m.mount(document.body, {
    oninit : function(vnode) {
        // ...
    },

    view : function(vnode) {
        // Use vnode.state instead of ctrl
        // Use vnode.attrs instead of options
    }
})

Passing components to m()

In v0.2.x you could pass components as the second argument of m() w/o any wrapping required. To help with consistency in v1.x they must always be wrapped with a m() invocation.

v0.2.x

m("div", component)

v1.x

m("div", m(component))

Passing vnodes to m.mount() and m.route()

In v0.2.x, m.mount(element, component) tolerated vnodes as second arguments instead of components (even though it wasn't documented). Likewise, m.route(element, defaultRoute, routes) accepted vnodes as values in the routes object.

In v1.x, components are required instead in both cases.

v0.2.x

m.mount(element, m('i', 'hello'))
m.mount(element, m(Component, attrs))

m.route(element, '/', {
    '/': m('b', 'bye')
})

v1.x

m.mount(element, {view: function () {return m('i', 'hello')}})
m.mount(element, {view: function () {return m(Component, attrs)}})

m.route(element, '/', {
    '/': {view: function () {return m('b', 'bye')}}
})

m.route.mode

In v0.2.x the routing mode could be set by assigning a string of "pathname", "hash", or "search" to m.route.mode. In v.1.x it is replaced by m.route.prefix(prefix) where prefix can be #, ?, or an empty string (for "pathname" mode). The new API also supports hashbang (#!), which is the default, and it supports non-root pathnames and arbitrary mode variations such as querybang (?!)

v0.2.x

m.route.mode = "pathname"
m.route.mode = "search"

v1.x

m.route.prefix("")
m.route.prefix("?")

m.route() and anchor tags

Handling clicks on anchor tags via the mithril router is similar to v0.2.x but uses a new lifecycle method and API.

v0.2.x

// When clicked this link will load the "/path" route instead of navigating
m("a", {
    href   : "/path",
    config : m.route
})

v1.x

// When clicked this link will load the "/path" route instead of navigating
m("a", {
    href     : "/path",
    oncreate : m.route.link
})

Reading/writing the current route

In v0.2.x all interaction w/ the current route happened via m.route(). In v1.x this has been broken out into two functions.

v0.2.x

// Getting the current route
m.route()

// Setting a new route
m.route("/other/route")

v1.x

// Getting the current route
m.route.get()

// Setting a new route
m.route.set("/other/route")

Accessing route params

In v0.2.x reading route params was entirely handled through m.route.param(). This API is still available in v1.x, and additionally any route params are passed as properties in the attrs object on the vnode.

v0.2.x

m.route(document.body, "/booga", {
    "/:attr" : {
        controller : function() {
            m.route.param("attr") // "booga"
        },
        view : function() {
            m.route.param("attr") // "booga"
        }
    }
})

v1.x

m.route(document.body, "/booga", {
    "/:attr" : {
        oninit : function(vnode) {
            vnode.attrs.attr // "booga"
            m.route.param("attr") // "booga"
        },
        view : function(vnode) {
            vnode.attrs.attr // "booga"
            m.route.param("attr") // "booga"
        }
    }
})

Building/Parsing query strings

v0.2.x used methods hanging off of m.route, m.route.buildQueryString() and m.route.parseQueryString(). In v1.x these have been broken out and attached to the root m.

v0.2.x

var qs = m.route.buildQueryString({ a : 1 });

var obj = m.route.parseQueryString("a=1");

v1.x

var qs = m.buildQueryString({ a : 1 });

var obj = m.parseQueryString("a=1");

Preventing unmounting

It is no longer possible to prevent unmounting via onunload's e.preventDefault(). Instead you should explicitly call m.route.set when the expected conditions are met.

v0.2.x

var Component = {
    controller: function() {
        this.onunload = function(e) {
            if (condition) e.preventDefault()
        }
    },
    view: function() {
        return m("a[href=/]", {config: m.route})
    }
}

v1.x

var Component = {
    view: function() {
        return m("a", {onclick: function() {if (!condition) m.route.set("/")}})
    }
}

Run code on component removal

Components no longer call this.onunload when they are being removed. They now use the standardized lifecycle hook onremove.

v0.2.x

var Component = {
    controller: function() {
        this.onunload = function(e) {
            // ...
        }
    },
    view: function() {
        // ...
    }
}

v1.x

var Component = {
    onremove : function() {
        // ...
    }
    view: function() {
        // ...
    }
}

m.request

Promises returned by m.request are no longer m.prop getter-setters. In addition, initialValue, unwrapSuccess and unwrapError are no longer supported options.

In addition, requests no longer have m.startComputation/m.endComputation semantics. Instead, redraws are always triggered when a request promise chain completes (unless background:true is set).

v0.2.x

var data = m.request({
    method: "GET",
    url: "https://api.github.com/",
    initialValue: [],
})

setTimeout(function() {
    console.log(data())
}, 1000)

v1.x

var data = []
m.request({
    method: "GET",
    url: "https://api.github.com/",
})
.then(function (responseBody) {
    data = responseBody
})

setTimeout(function() {
    console.log(data) // note: not a getter-setter
}, 1000)

Additionally, if the extract option is passed to m.request the return value of the provided function will be used directly to resolve the request promise, and the deserialize callback is ignored.


m.deferred removed

v0.2.x used its own custom asynchronous contract object, exposed as m.deferred, which was used as the basis for m.request. v1.x uses Promises instead, and implements a polyfill in non-supporting environments. In situations where you would have used m.deferred, you should use Promises instead.

v0.2.x

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

v1.x

var greetAsync = function() {
    return new Promise(function(resolve){
        setTimeout(function() {
            resolve("hello")
        }, 1000)
    })
}

greetAsync()
    .then(function(value) {return value + " world"})
    .then(function(value) {console.log(value)}) //logs "hello world" after 1 second

m.sync removed

Since v1.x uses standards-compliant Promises, m.sync is redundant. Use Promise.all instead.

v0.2.x

m.sync([
    m.request({ method: 'GET', url: 'https://api.github.com/users/lhorie' }),
    m.request({ method: 'GET', url: 'https://api.github.com/users/isiahmeadows' }),
])
.then(function (users) {
    console.log("Contributors:", users[0].name, "and", users[1].name)
})

v1.x

Promise.all([
    m.request({ method: 'GET', url: 'https://api.github.com/users/lhorie' }),
    m.request({ method: 'GET', url: 'https://api.github.com/users/isiahmeadows' }),
])
.then(function (users) {
    console.log("Contributors:", users[0].name, "and", users[1].name)
})

In v0.2.x, the xlink namespace was the only supported attribute namespace, and it was supported via special casing behavior. Now namespace parsing is fully supported, and namespaced attributes should explicitly declare their namespace.

v0.2.x

m("svg",
    // the `href` attribute is namespaced automatically
    m("image[href='image.gif']")
)

v1.x

m("svg",
    // User-specified namespace on the `href` attribute
    m("image[xlink:href='image.gif']")
)

Nested arrays in views

Arrays now represent fragments, which are structurally significant in v1.x virtual DOM. Whereas nested arrays in v0.2.x would be flattened into one continuous list of virtual nodes for the purposes of diffing, v1.x preserves the array structure - the children of any given array are not considered siblings of those of adjacent arrays.


vnode equality checks

If a vnode is strictly equal to the vnode occupying its place in the last draw, v1.x will skip that part of the tree without checking for mutations or triggering any lifecycle methods in the subtree. The component documentation contains more detail on this issue.


License: MIT. © Leo Horie.