VirtualDom
API to the core diffing algorithm. Can serve as a foundation for libraries that expose more helper functions for HTML or SVG.
Create
An immutable chunk of data representing a DOM node. This can be HTML or SVG.
Just put plain text in the DOM. It will escape the string so that it appears exactly as you specify.
text "Hello World!"
Create a DOM node with a tag name, a list of HTML properties that can
include styles and event listeners, a list of CSS properties like color
, and
a list of child nodes.
import Json.Encode as Json
hello : Node msg
hello =
node "div" [] [ text "Hello!" ]
greeting : Node msg
greeting =
node "div"
[ property "id" (Json.string "greeting") ]
[ text "Hello!" ]
Create a namespaced DOM node. For example, an SVG <path>
node could be
defined like this:
path : Array (Attribute msg) -> Array (Node msg) -> Node msg
path attrubutes children =
nodeNS "http://www.w3.org/2000/svg" "path" attributes children
Attributes
When using HTML and JS, there are two ways to specify parts of a DOM node.
-
Attributes — You can set things in HTML itself. So the
class
in<div class="greeting"></div>
is called an attribute. -
Properties — You can also set things in JS. So the
className
indiv.className = 'greeting'
is called a property.
So the class
attribute corresponds to the className
property. At first
glance, perhaps this distinction is defensible, but it gets much crazier.
There is not always a one-to-one mapping between attributes and properties!
Yes, that is a true fact. Sometimes an attribute exists, but there is no
corresponding property. Sometimes changing an attribute does not change the
underlying property. For example, as of this writing, the webkit-playsinline
attribute can be used in HTML, but there is no corresponding property!
Specify a style.
greeting : Node msg
greeting =
node "div"
[ style "backgroundColor" "red"
, style "height" "90px"
, style "width" "100%"
]
[ text "Hello!"
]
Create a property.
import Json.Encode as Encode
buttonLabel : Node msg
buttonLabel =
node "label" [ property "htmlFor" (Encode.string "button") ] [ text "Label" ]
Notice that you must give the property name, so we use htmlFor
as it
would be in JavaScript, not for
as it would appear in HTML.
Create an attribute. This uses JavaScript’s setAttribute
function
behind the scenes.
buttonLabel : Node msg
buttonLabel =
node "label" [ attribute "for" "button" ] [ text "Label" ]
Notice that you must give the attribute name, so we use for
as it would
be in HTML, not htmlFor
as it would appear in JS.
Would you believe that there is another way to do this?! This uses
JavaScript's setAttributeNS
function behind the scenes. It is doing pretty
much the same thing as attribute
but you are able to have namespaced
attributes. As an example, the elm/svg
package defines an attribute
like this:
xlinkHref : String -> Attribute msg
xlinkHref value =
attributeNS "http://www.w3.org/1999/xlink" "xlink:href" value
Events
Create custom event handlers.
You can define onClick
like this:
import Json.Decode as Decode
onClick : msg -> Attribute msg
onClick msg =
on "click" (Normal (Decode.succeed msg))
Note: These event handlers trigger in the bubble phase. You can learn more about what that means here. There is not support within Gren for doing tricks with the capture phase. We recommend doing that in JS through ports.
When using on
you can customize the event behavior
a bit. There are two ways to do this:
-
stopPropagation
means the event stops traveling through the DOM. So if propagation of a click is stopped, it will not trigger any other event listeners. -
preventDefault
means any built-in browser behavior related to the event is prevented. This can be handy with key presses or touch gestures.
Note 1: A passive event listener will be created if you use Normal
or MayStopPropagation
. In both cases preventDefault
cannot be used, so
we can enable optimizations for touch, scroll, and wheel events in some
browsers.
Note 2: Some actions, like uploading and downloading files, are only
allowed when the JavaScript event loop is running because of user input. This
is for security! So when an event occurs, we call update
and send any port
messages immediately, all within the same tick of the event loop. This makes
it possible to handle user-instigated events in ports.
Note 3: Normally the view
is shown in the next requestAnimationFrame
call. This allows us to save some work if messages are coming in very quickly.
But if stopPropagation
is used, we update the DOM immediately, within the
same tick of the event loop. This is useful for DOM nodes that hold their own
state, like <input type="text">
. If someone types very fast, the state in the
DOM can diverge from the state in your Model
while waiting on the next
requestAnimationFrame
call. So updating the DOM synchronously makes this
divergence impossible.
Routing Messages
This function is useful when nesting components with the Elm Architecture. It lets you transform the messages produced by a subtree.
Say you have a node named button
that produces ()
values when it is
clicked. To get your model updating properly, you will probably want to tag
this ()
value like this:
type Msg = Click | ...
update msg model =
case msg of
Click ->
...
view model =
map (\_ -> Click) button
So now all the events produced by button
will be transformed to be of type
Msg
so they can be handled by your update function!
Transform the messages produced by a Attribute
.
Keyed Nodes
Works just like node
, but you add a unique identifier to each child
node. You want this when you have a list of nodes that is changing: adding
nodes, removing nodes, etc. In these cases, the unique identifiers help make
the DOM modifications more efficient.
Create a keyed and namespaced DOM node. For example, an SVG <g>
node
could be defined like this:
g : Array (Attribute msg) -> Array ( String, Node msg ) -> Node msg
g =
keyedNodeNS "http://www.w3.org/2000/svg" "g"
Lazy Nodes
A performance optimization that delays the building of virtual DOM nodes.
Calling (view model)
will definitely build some virtual DOM, perhaps a lot of
it. Calling (lazy view model)
delays the call until later. During diffing, we
can check to see if model
is referentially equal to the previous value used,
and if so, we just stop. No need to build up the tree structure and diff it,
we know if the input to view
is the same, the output must be the same!
Same as lazy
but checks on two arguments.
Same as lazy
but checks on three arguments.
Same as lazy
but checks on four arguments.
Same as lazy
but checks on five arguments.
Same as lazy
but checks on six arguments.
Same as lazy
but checks on seven arguments.
Same as lazy
but checks on eight arguments.