介面定義
能夠對於文字、段落乃至任何元素的精準定位 並做出增刪改查，都是在開發一款富文字編輯器時一項最基本也是最重要的功能之一。讓我們先來看看Slate中對於如何在文件樹中定位元素是怎麼定義的[原始碼]：
/**
* The `Location` interface is a union of the ways to refer to a specific
* location in a Slate document: paths, points or ranges.
*
* Methods will often accept a `Location` instead of requiring only a `Path`,
* `Point` or `Range`. This eliminates the need for developers to manage
* converting between the different interfaces in their own code base.
*/
export type Location = Path | Point | Range
Location是一個包含了Path、Point及Range的聯合型別，代指了Slate中所有關於“定位”的概念，同時也方便了開發。例如在幾乎所有的Transforms方法中，都可以透過傳遞Location引數來決定Transforms方法需要應用到文件中的哪些位置上[連結]。
All transforms support a parameter
options. This includes options specific to the transform, and generalNodeOptionsto specify which Nodes in the document that the transform is applied to.interface NodeOptions { at?: Location match?: (node: Node, path: Location) => boolean mode?: 'highest' | 'lowest' voids?: boolean }
Path
Path是三個中最基本的概念，也是唯一一個不可擴充的型別。
/**
* `Path` arrays are a list of indexes that describe a node's exact position in
* a Slate node tree. Although they are usually relative to the root `Editor`
* object, they can be relative to any `Node` object.
*/
export type Path = number[]
Path型別就是一個陣列，用來表示Slate文件樹中自根節點root到指定node的絕對路徑。我們以下邊的示例來演示下各個node所代表的路徑：
const initialValue: Descendant[] = [
// path: [0]
{
type: 'paragraph',
children: [
{ text: 'This is editable ' }, // path: [0, 0]
{ text: 'rich text!', bold: true } // path: [0, 1]
]
},
// path: [1]
{
type: 'paragraph',
children: [
{ text: 'It\' so cool.' } // path: [1, 0]
]
}
]
雖然Path所代表的路徑通常是以頂層Editor作為root節點的，但也會有其他情況，比如由Node提供的get方法中傳入的Path引數則代表的是相對路徑[原始碼]：
/**
* Get the descendant node referred to by a specific path. If the path is an
* empty array, it refers to the root node itself.
*/
get(root: Node, path: Path): Node {
let node = root
for (let i = 0; i < path.length; i++) {
const p = path[i]
if (Text.isText(node) || !node.children[p]) {
throw new Error(
`Cannot find a descendant at path [${path}] in node: ${Scrubber.stringify(
root
)}`
)
}
node = node.children[p]
}
return node
}
Point
Point是在Path的基礎上封裝而來的概念：
/**
* `Point` objects refer to a specific location in a text node in a Slate
* document. Its path refers to the location of the node in the tree, and its
* offset refers to the distance into the node's string of text. Points can
* only refer to `Text` nodes.
*/
export interface BasePoint {
path: Path
offset: number
}
export type Point = ExtendedType<'Point', BasePoint>
用於定位單個字元在文件中的位置；先用Path定位到字元所屬的Node，再根據offset欄位資訊精確到字元是在該Node的text文字中的偏移量。
我們仍然以前面的示例做說明，如果想要將游標位置定位到第一句話"This is editable rich text!"的感嘆號之後，其Point值為：
const initialValue: Descendant[] = [
// path: [0]
{
type: 'paragraph',
children: [
{ text: 'This is editable ' }, // path: [0, 0]
{ text: 'rich text!', bold: true } // { path: [0, 1], offset: 10 }
]
},
// path: [1]
{
type: 'paragraph',
children: [
{ text: 'It\' so cool.' } // path: [1, 0]
]
}
]
Range
最後一個Range則是再在Point基礎上延伸封裝而來的概念：
/**
* `Range` objects are a set of points that refer to a specific span of a Slate
* document. They can define a span inside a single node or a can span across
* multiple nodes.
*/
export interface BaseRange {
anchor: Point
focus: Point
}
export type Range = ExtendedType<'Range', BaseRange>
它代表的是一段文字的集合；包含有兩個Point型別的欄位anchor和focus。看到這，應該能發現Slate中Range的概念其實與DOM中的Selection物件是一樣的，anchor和focus分別對應原生Selection中的錨點anchorNode和焦點focusNode；這正是Slate對於原生Selection做的抽象，使之在自身API中更方便地透過游標選區來獲取文件樹中的內容。
我們在上一篇文章中有提到過的Editor.selection是一個Selections型別，其本身就是一個Range型別，專門用來指定編輯區域中的游標位置[原始碼]：
export type BaseSelection = Range | null
export type Selection = ExtendedType<'Selection', BaseSelection>
Refs
當我們需要長期追蹤某些Node時，可以透過獲取對應Node的Path/Point/Range值並儲存下來以達到目的。但這種方式存在的問題是，在Slate文件樹經過insert、remove等操作後，原先的Path/Point/Range可能會產生變動或者直接作廢掉。
Refs的出現就是為了解決上述問題。
三者的ref的定義分別在slate/src/interfaces/下的path-ref.ts、point-ref.ts和range-ref.ts檔案中：
/**
* `PathRef` objects keep a specific path in a document synced over time as new
* operations are applied to the editor. You can access their `current` property
* at any time for the up-to-date path value.
*/
export interface PathRef {
current: Path | null
affinity: 'forward' | 'backward' | null
unref(): Path | null
}
/**
* `PointRef` objects keep a specific point in a document synced over time as new
* operations are applied to the editor. You can access their `current` property
* at any time for the up-to-date point value.
*/
export interface PointRef {
current: Point | null
affinity: TextDirection | null
unref(): Point | null
}
/**
* `RangeRef` objects keep a specific range in a document synced over time as new
* operations are applied to the editor. You can access their `current` property
* at any time for the up-to-date range value.
*/
export interface RangeRef {
current: Range | null
affinity: 'forward' | 'backward' | 'outward' | 'inward' | null
unref(): Range | null
}
都包含以下三個欄位：
- current：同React ref用法一樣，用current欄位儲存最新值
- affinity：當前定位所代表的節點 在文件樹變動時如果受到影響的話，所採取的調整策略
- unref：解除安裝方法；徹底刪除當前的ref確保能夠被引擎GC掉
另外我們先來看下各種Refs在Slate儲存的方式，跳到slate/src/utils/weak-maps.ts中[原始碼]：
export const PATH_REFS: WeakMap<Editor, Set<PathRef>> = new WeakMap()
export const POINT_REFS: WeakMap<Editor, Set<PointRef>> = new WeakMap()
export const RANGE_REFS: WeakMap<Editor, Set<RangeRef>> = new WeakMap()
可以看到Refs的儲存區在一個Set資料結構中的；而對於不同Editor下Set<xxxRef>的儲存則是放在雜湊表WeakMap中的，使其不會影響到GC（WeakMap同Map原理一樣，都是ES6之後新出的雜湊資料結構，與Map不同點在於其持有的引用算作弱引用）。
生成三種Ref以及獲取相應Refs的方法定義在EditorInterface介面上[原始碼]：
export interface EditorInterface {
// ...
pathRef: (
editor: Editor,
path: Path,
options?: EditorPathRefOptions
) => PathRef
pointRef: (
editor: Editor,
point: Point,
options?: EditorPointRefOptions
) => PointRef
rangeRef: (
editor: Editor,
range: Range,
options?: EditorRangeRefOptions
) => RangeRef
pathRefs: (editor: Editor) => Set<PathRef>
pointRefs: (editor: Editor) => Set<PointRef>
rangeRefs: (editor: Editor) => Set<RangeRef>
}
Path、Point及Range三者的實現邏輯都差不多，下面就以Path為例作介紹。
/**
* Create a mutable ref for a `Point` object, which will stay in sync as new
* operations are applied to the editor.
*/
pointRef(
editor: Editor,
point: Point,
options: EditorPointRefOptions = {}
): PointRef {
const { affinity = 'forward' } = options
const ref: PointRef = {
current: point,
affinity,
unref() {
const { current } = ref
const pointRefs = Editor.pointRefs(editor)
pointRefs.delete(ref)
ref.current = null
return current
},
}
const refs = Editor.pointRefs(editor)
refs.add(ref)
return ref
}
實現邏輯非常簡單，就是根據傳入的引數放入ref物件中，並新增解除安裝方法unref。然後透過pathRefs拿到對應的Set，講當前的ref物件新增進去。unref方法中實現的則是相反的操作：透過pathRefs拿到對應的Set後，將當前ref物件移除掉，然後再把ref.current的值置空。
/**
* Get the set of currently tracked path refs of the editor.
*/
pathRefs(editor: Editor): Set<PathRef> {
let refs = PATH_REFS.get(editor)
if (!refs) {
refs = new Set()
PATH_REFS.set(editor, refs)
}
return refs
}
程式碼非常簡短，類似懶載入的方式做Set的初始化，然後呼叫get方法獲取集合後返回。
Ref同步
前一篇文章我們提到過，用於修改內容的單個Transform方法會包含有多個Operation；Operation則是Slate中的原子化操作。而在Slate文件樹更新之後，解決Ref同步更新的方式就是：在執行了任意Operation之後，對所有的Ref根據執行的Operation型別做相應的調整。
看到create-editor.ts中的apply方法，該方法是所有Operation執行的入口[原始碼]：
apply: (op: Operation) => {
for (const ref of Editor.pathRefs(editor)) {
PathRef.transform(ref, op)
}
for (const ref of Editor.pointRefs(editor)) {
PointRef.transform(ref, op)
}
for (const ref of Editor.rangeRefs(editor)) {
RangeRef.transform(ref, op)
}
// ...
}
在apply方法的最開頭就是三組for of迴圈，對所有的Ref執行對應的Ref.transform方法並傳入當前執行的Operation。
同樣以Path為例，看下path-ref.ts中的PathRef.transform方法[原始碼]：
export const PathRef: PathRefInterface = {
/**
* Transform the path ref's current value by an operation.
*/
transform(ref: PathRef, op: Operation): void {
const { current, affinity } = ref
if (current == null) {
return
}
const path = Path.transform(current, op, { affinity })
ref.current = path
if (path == null) {
ref.unref()
}
}
}
將當前ref中的資料和Operation作為引數，傳遞給對應的Path.transform，返回更新後的path值並賦值給ref.current。如果path為空，則說明當前ref指代的位置已經失效，呼叫解除安裝方法unref。
至於Path.transform的細節在本篇就不展開了，我們留到後續的Transform篇再統一講解: )