Compose provides many modifiers for common behaviors right out of the box, but you can also create your own custom modifiers.
Modifiers have multiple parts:
- A modifier factory
- This is an extension function on
Modifier
, which provides an idiomatic API for your modifier and allows modifiers to be easily chained together. The modifier factory produces the modifier elements used by Compose to modify your UI.
- This is an extension function on
- A modifier element
- This is where you can implement the behavior of your modifier.
There are multiple ways to implement a custom modifier depending on the
functionality needed. Often, the easiest way to implement a custom modifier is
just to implement a custom modifier factory which combines other already defined
modifier factories together. If you need more custom behavior, implement the
modifier element using the Modifier.Node
APIs, which are lower level but
provide more flexibility.
Chain existing modifiers together
It is often possible to create custom modifiers just by using existing
modifiers. For example, Modifier.clip()
is implemented using the
graphicsLayer
modifier. This strategy uses existing modifier elements, and you
provide your own custom modifier factory.
Before implementing your own custom modifier, see if you can use the same strategy.
fun Modifier.clip(shape: Shape) = graphicsLayer(shape = shape, clip = true)
Or, if you find you are repeating the same group of modifiers often, you can wrap them into your own modifier:
fun Modifier.myBackground(color: Color) = padding(16.dp) .clip(RoundedCornerShape(8.dp)) .background(color)
Create a custom modifier using a composable modifier factory
You can also create a custom modifier using a composable function to pass values to an existing modifier. This is known as a composable modifier factory.
Using a composable modifier factory to create a modifier also allows for using
higher level compose APIs, such as animate*AsState
and other Compose
state backed animation APIs. For example, the following snippet shows a
modifier that animates an alpha change when enabled/disabled:
@Composable fun Modifier.fade(enable: Boolean): Modifier { val alpha by animateFloatAsState(if (enable) 0.5f else 1.0f) return this then Modifier.graphicsLayer { this.alpha = alpha } }
If your custom modifier is a convenience method to provide default values from a
CompositionLocal
, the easiest way to implement this is to use a composable
modifier factory:
@Composable fun Modifier.fadedBackground(): Modifier { val color = LocalContentColor.current return this then Modifier.background(color.copy(alpha = 0.5f)) }
This approach has some caveats detailed below.
CompositionLocal
values are resolved at the call site of the modifier factory
When creating a custom modifier using a composable modifier factory, composition locals take the value from the composition tree where they are created, not used. This can lead to unexpected results. For example, take the composition local modifier example from above, implemented slightly differently using a composable function:
@Composable fun Modifier.myBackground(): Modifier { val color = LocalContentColor.current return this then Modifier.background(color.copy(alpha = 0.5f)) } @Composable fun MyScreen() { CompositionLocalProvider(LocalContentColor provides Color.Green) { // Background modifier created with green background val backgroundModifier = Modifier.myBackground() // LocalContentColor updated to red CompositionLocalProvider(LocalContentColor provides Color.Red) { // Box will have green background, not red as expected. Box(modifier = backgroundModifier) } } }
If this is not how you expect your modifier to work, use a custom
Modifier.Node
instead, as composition locals will be
correctly resolved at the usage site and can be safely hoisted.
Composable function modifiers are never skipped
Composable factory modifiers are never skipped because composable functions that have return values cannot be skipped. This means your modifier function will be called on every recomposition, which may be expensive if it recomposes frequently.
Composable function modifiers must be called within a composable function
Like all composable functions, a composable factory modifier must be called from within composition. This limits where a modifier can be hoisted to, as it can never be hoisted out of composition. In comparison, non-composable modifier factories can be hoisted out of composable functions to allow easier reuse and improve performance:
val extractedModifier = Modifier.background(Color.Red) // Hoisted to save allocations @Composable fun Modifier.composableModifier(): Modifier { val color = LocalContentColor.current.copy(alpha = 0.5f) return this then Modifier.background(color) } @Composable fun MyComposable() { val composedModifier = Modifier.composableModifier() // Cannot be extracted any higher }
Implement custom modifier behavior using Modifier.Node
Modifier.Node
is a lower level API for creating modifiers in Compose. It
is the same API that Compose implements its own modifiers in and is the most
performant way to create custom modifiers.
Implement a custom modifier using Modifier.Node
There are three parts to implementing a custom modifier using Modifier.Node:
- A
Modifier.Node
implementation that holds the logic and state of your modifier. - A
ModifierNodeElement
that creates and updates modifier node instances. - An optional modifier factory as detailed above.
ModifierNodeElement
classes are stateless and new instances are allocated each
recomposition, whereas Modifier.Node
classes can be stateful and will survive
across multiple recompositions, and can even be reused.
The following section describes each part and shows an example of building a custom modifier to draw a circle.
Modifier.Node
The Modifier.Node
implementation (in this example, CircleNode
) implements
the
functionality of your custom modifier.
// Modifier.Node private class CircleNode(var color: Color) : DrawModifierNode, Modifier.Node() { override fun ContentDrawScope.draw() { drawCircle(color) } }
In this example, it draws the circle with the color passed in to the modifier function.
A node implements Modifier.Node
as well as zero or more node types. There are
different node types based on the functionality your modifier requires. The
example above needs to be able to draw, so it implements DrawModifierNode
, which
allows it to override the draw method.
The available types are as follows:
Node |
Usage |
Sample Link |
A |
||
A |
||
Implementing this interface allows your |
||
A |
||
A |
||
A |
||
A |
||
A |
||
|
||
A This can be useful to compose multiple node implementations into one. |
||
Allows |
Nodes are automatically invalidated when update is called on their corresponding
element. Because our example is a DrawModifierNode
, any time update is called on
the element, the node triggers a redraw and its color correctly updates. It is
possible to opt out of auto-invalidation as detailed below.
ModifierNodeElement
A ModifierNodeElement
is an immutable class that holds the data to create or
update your custom modifier:
// ModifierNodeElement private data class CircleElement(val color: Color) : ModifierNodeElement<CircleNode>() { override fun create() = CircleNode(color) override fun update(node: CircleNode) { node.color = color } }
ModifierNodeElement
implementations need to override the following methods:
create
: This is the function that instantiates your modifier node. This gets called to create the node when your modifier is first applied. Usually, this amounts to constructing the node and configuring it with the parameters that were passed in to the modifier factory.update
: This function is called whenever this modifier is provided in the same spot this node already exists, but a property has changed. This is determined by theequals
method of the class. The modifier node that was previously created is sent as a parameter to theupdate
call. At this point, you should update the nodes' properties to correspond with the updated parameters. The ability for nodes to be reused this way is key to the performance gains thatModifier.Node
brings; therefore, you must update the existing node rather than creating a new one in theupdate
method. In our circle example, the color of the node is updated.
Additionally, ModifierNodeElement
implementations also need to implement equals
and hashCode
. update
will only get called if an equals comparison with the
previous element returns false.
The example above uses a data class to achieve this. These methods are used to
check if a node needs updating or not. If your element has properties that do
not contribute to whether a node needs to be updated, or you want to avoid data
classes for binary compatibility reasons, then you can manually implement equals
and hashCode
e.g. the padding modifier element.
Modifier factory
This is the public API surface of your modifier. Most implementations simply create the modifier element and add it to the modifier chain:
// Modifier factory fun Modifier.circle(color: Color) = this then CircleElement(color)
Complete example
These three parts come together to create the custom modifier to draw a circle
using the Modifier.Node
APIs:
// Modifier factory fun Modifier.circle(color: Color) = this then CircleElement(color) // ModifierNodeElement private data class CircleElement(val color: Color) : ModifierNodeElement<CircleNode>() { override fun create() = CircleNode(color) override fun update(node: CircleNode) { node.color = color } } // Modifier.Node private class CircleNode(var color: Color) : DrawModifierNode, Modifier.Node() { override fun ContentDrawScope.draw() { drawCircle(color) } }
Common situations using Modifier.Node
When creating custom modifiers with Modifier.Node
, here are some common situations you might
encounter.
Zero parameters
If your modifier has no parameters, then it never needs to update and, furthermore, does not need to be a data class. Here is a sample implementation of a modifier that applies a fixed amount of padding to a composable:
fun Modifier.fixedPadding() = this then FixedPaddingElement data object FixedPaddingElement : ModifierNodeElement<FixedPaddingNode>() { override fun create() = FixedPaddingNode() override fun update(node: FixedPaddingNode) {} } class FixedPaddingNode : LayoutModifierNode, Modifier.Node() { private val PADDING = 16.dp override fun MeasureScope.measure( measurable: Measurable, constraints: Constraints ): MeasureResult { val paddingPx = PADDING.roundToPx() val horizontal = paddingPx * 2 val vertical = paddingPx * 2 val placeable = measurable.measure(constraints.offset(-horizontal, -vertical)) val width = constraints.constrainWidth(placeable.width + horizontal) val height = constraints.constrainHeight(placeable.height + vertical) return layout(width, height) { placeable.place(paddingPx, paddingPx) } } }
Referencing composition locals
Modifier.Node
modifiers do not automatically observe changes to Compose state
objects, like CompositionLocal
. The advantage Modifier.Node
modifiers have over
modifiers that are just created with a composable factory is that they can read
the value of the composition local from where the modifier is used in your UI
tree, not where the modifier is allocated, using currentValueOf
.
However, modifier node instances do not automatically observe state changes. To automatically react to a composition local changing, you can read its current value inside a scope:
DrawModifierNode
:ContentDrawScope
LayoutModifierNode
:MeasureScope
&IntrinsicMeasureScope
SemanticsModifierNode
:SemanticsPropertyReceiver
This example observes the value of LocalContentColor
to draw a background based
on its color. As ContentDrawScope
does observe snapshot changes, this
automatically redraws when the value of LocalContentColor
changes:
class BackgroundColorConsumerNode : Modifier.Node(), DrawModifierNode, CompositionLocalConsumerModifierNode { override fun ContentDrawScope.draw() { val currentColor = currentValueOf(LocalContentColor) drawRect(color = currentColor) drawContent() } }
To react to state changes outside of a scope and automatically update your
modifier, use an ObserverModifierNode
.
For example, Modifier.scrollable
uses this technique to
observe changes in LocalDensity
. A simplified example is shown below:
class ScrollableNode : Modifier.Node(), ObserverModifierNode, CompositionLocalConsumerModifierNode { // Place holder fling behavior, we'll initialize it when the density is available. val defaultFlingBehavior = DefaultFlingBehavior(splineBasedDecay(UnityDensity)) override fun onAttach() { updateDefaultFlingBehavior() observeReads { currentValueOf(LocalDensity) } // monitor change in Density } override fun onObservedReadsChanged() { // if density changes, update the default fling behavior. updateDefaultFlingBehavior() } private fun updateDefaultFlingBehavior() { val density = currentValueOf(LocalDensity) defaultFlingBehavior.flingDecay = splineBasedDecay(density) } }
Animating modifier
Modifier.Node
implementations have access to a coroutineScope
. This allows for
use of the Compose Animatable APIs. For example, this snippet modifies the
CircleNode
from above to fade in and out repeatedly:
class CircleNode(var color: Color) : Modifier.Node(), DrawModifierNode { private val alpha = Animatable(1f) override fun ContentDrawScope.draw() { drawCircle(color = color, alpha = alpha.value) drawContent() } override fun onAttach() { coroutineScope.launch { alpha.animateTo( 0f, infiniteRepeatable(tween(1000), RepeatMode.Reverse) ) { } } } }
Sharing state between modifiers using delegation
Modifier.Node
modifiers can delegate to other nodes. There are many use cases
for this, such as extracting common implementations across different modifiers,
but it can also be used to share common state across modifiers.
For example, a basic implementation of a clickable modifier node that shares interaction data:
class ClickableNode : DelegatingNode() { val interactionData = InteractionData() val focusableNode = delegate( FocusableNode(interactionData) ) val indicationNode = delegate( IndicationNode(interactionData) ) }
Opting out of node auto-invalidation
Modifier.Node
nodes invalidate automatically when their corresponding
ModifierNodeElement
calls update. Sometimes, in a more complex modifier, you may
want to opt out of this behavior to have more fine-grained control over when
your modifier invalidates phases.
This can be particularly useful if your custom modifier modifies both layout and
draw. Opting out of auto-invalidation allows you to just invalidate draw when
only draw-related properties, like color
, change, and not invalidate layout.
This can improve the performance of your modifier.
A hypothetical example of this is shown below with a modifier that has a color
,
size
, and onClick
lambda as properties. This modifier only invalidates what is
required, and skips any invalidation that isn't:
class SampleInvalidatingNode( var color: Color, var size: IntSize, var onClick: () -> Unit ) : DelegatingNode(), LayoutModifierNode, DrawModifierNode { override val shouldAutoInvalidate: Boolean get() = false private val clickableNode = delegate( ClickablePointerInputNode(onClick) ) fun update(color: Color, size: IntSize, onClick: () -> Unit) { if (this.color != color) { this.color = color // Only invalidate draw when color changes invalidateDraw() } if (this.size != size) { this.size = size // Only invalidate layout when size changes invalidateMeasurement() } // If only onClick changes, we don't need to invalidate anything clickableNode.update(onClick) } override fun ContentDrawScope.draw() { drawRect(color) } override fun MeasureScope.measure( measurable: Measurable, constraints: Constraints ): MeasureResult { val size = constraints.constrain(size) val placeable = measurable.measure(constraints) return layout(size.width, size.height) { placeable.place(0, 0) } } }