Applier

Kotlin |Java

interface Applier<N>

androidx.compose.runtime.Applier

Known Direct Subclasses

AbstractApplier

An abstract Applier implementation.

Known Indirect Subclasses

VectorApplier

An Applier is responsible for applying the tree-based operations that get emitted during a composition. Every Composer has an Applier which it uses to emit a ComposeNode.

A custom Applier implementation will be needed in order to utilize Compose to build and maintain a tree of a novel type.

import androidx.compose.runtime.Composition
import androidx.compose.runtime.ComposeNode
import androidx.compose.runtime.mutableStateOf
import androidx.compose.runtime.remember

// Provided we have a tree with a node base type like the following
abstract class Node {
    val children = mutableListOf<Node>()
}

// We would implement an Applier class like the following, which would teach compose how to
// manage a tree of Nodes.
class NodeApplier(root: Node) : AbstractApplier<Node>(root) {
    override fun insertTopDown(index: Int, instance: Node) {
        current.children.add(index, instance)
    }

    override fun insertBottomUp(index: Int, instance: Node) {
        // Ignored as the tree is built top-down.
    }

    override fun remove(index: Int, count: Int) {
        current.children.remove(index, count)
    }

    override fun move(from: Int, to: Int, count: Int) {
        current.children.move(from, to, count)
    }

    override fun onClear() {
        root.children.clear()
    }
}

// A function like the following could be created to create a composition provided a root Node.
fun Node.setContent(
    parent: CompositionContext,
    content: @Composable () -> Unit
): Composition {
    return Composition(NodeApplier(this), parent).apply {
        setContent(content)
    }
}

// assuming we have Node sub-classes like "TextNode" and "GroupNode"
class TextNode : Node() {
    var text: String = ""
    var onClick: () -> Unit = {}
}
class GroupNode : Node()

// Composable equivalents could be created
@Composable fun Text(text: String, onClick: () -> Unit = {}) {
    ComposeNode<TextNode, NodeApplier>(::TextNode) {
        set(text) { this.text = it }
        set(onClick) { this.onClick = it }
    }
}

@Composable fun Group(content: @Composable () -> Unit) {
    ComposeNode<GroupNode, NodeApplier>(::GroupNode, {}, content)
}

// and then a sample tree could be composed:
fun runApp(root: GroupNode, parent: CompositionContext) {
    root.setContent(parent) {
        var count by remember { mutableStateOf(0) }
        Group {
            Text("Count: $count")
            Text("Increment") { count++ }
        }
    }
}

Summary

Public methods
abstract Unit	`clear()` Move to the root and remove all nodes from the root, preparing both this Applier and its root to be used as the target of a new composition in the future.
abstract Unit	`down(node: N)` Indicates that the applier is getting traversed "down" the tree.
abstract Unit	`insertBottomUp(index: Int, instance: N)` Indicates that instance should be inserted as a child of current at index.
abstract Unit	`insertTopDown(index: Int, instance: N)` Indicates that instance should be inserted as a child to current at index.
abstract Unit	`move(from: Int, to: Int, count: Int)` Indicates that count children of current should be moved from index from to index to.
open Unit	`onBeginChanges()` Called when the Composer is about to begin applying changes using this applier.
open Unit	`onEndChanges()` Called when the Composer is finished applying changes using this applier.
abstract Unit	`remove(index: Int, count: Int)` Indicates that the children of current from index to index + count should be removed.
abstract Unit	`up()` Indicates that the applier is getting traversed "up" the tree.

Properties
abstract N	`current` The node that operations will be applied on at any given time.

Public methods

clear

abstract fun clear(): Unit

Move to the root and remove all nodes from the root, preparing both this Applier and its root to be used as the target of a new composition in the future.

down

abstract fun down(node: N): Unit

Indicates that the applier is getting traversed "down" the tree. When this gets called, node is expected to be a child of current, and after this operation, node is expected to be the new current.

insertBottomUp

abstract fun insertBottomUp(
    index: Int, 
    instance: N
): Unit

Indicates that instance should be inserted as a child of current at index. An applier should insert the node into the tree either in insertTopDown or insertBottomUp, not both. See the description of insertTopDown to which describes when to implement insertTopDown and when to use insertBottomUp.

insertTopDown

abstract fun insertTopDown(
    index: Int, 
    instance: N
): Unit

Indicates that instance should be inserted as a child to current at index. An applier should insert the node into the tree either in insertTopDown or insertBottomUp, not both.

The insertTopDown method is called before the children of instance have been created and inserted into it. insertBottomUp is called after all children have been created and inserted.

Some trees are faster to build top-down, in which case the insertTopDown method should be used to insert the instance. Other tress are faster to build bottom-up in which case insertBottomUp should be used.

To give example of building a tree top-down vs. bottom-up consider the following tree,

where the node B is being inserted into the tree at R. Top-down building of the tree first inserts B into R, then inserts A into B followed by inserting C into B`. For example,

    1           2           3
    R           R           R
    |           |           |
    B           B           B
               /           / \
              A           A   C

A bottom-up building of the tree starts with inserting A and C into B then inserts B tree into R.

    1           2           3
    B           B           R
    |          / \          |
    A         A   C         B
                           / \
                          A   C

To see how building top-down vs. bottom-up can differ significantly in performance consider a tree where whenever a child is added to the tree all parent nodes, up