vue-loader 源码解析系列之 selector

笔者系 vue-loader 贡献者之一(#16)

前言

vue-loader 源码解析系列之一,阅读该文章之前,请大家首先参考大纲 vue-loader 源码解析系列之 整体分析

selector 做了什么

1
2
3
4
5
6
7
8
9
10
11
12
13
const path = require('path')
const parse = require('./parser')
const loaderUtils = require('loader-utils')

module.exports = function (content) {
// 略
const query = loaderUtils.getOptions(this) || {}
// 略
const parts = parse(content, filename, this.sourceMap, sourceRoot, query.bustCache)
let part = parts[query.type]
// 略
this.callback(null, part.content, part.map)
}

大家可以看到,selector的代码非常简单,
通过 parser 将 .vue 解析成对象 parts, 里面有分别 style, script, template。可以根据不同的 query, 返回对应的部分。
很明显那么这个 parser 完成了分析分解 .vue 的工作,那么让我们继续深入 parser

parser 做了什么

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
const compiler = require('vue-template-compiler')
const cache = require('lru-cache')(100)

module.exports = (content, filename, needMap, sourceRoot, bustCache) => {
const cacheKey = hash(filename + content)
// 略
let output = cache.get(cacheKey)
if (output) return output
output = compiler.parseComponent(content, { pad: 'line' })
if (needMap) {
// 略去了生成 sourceMap 的代码
}
cache.set(cacheKey, output)
return output
}

同样的,为了方便读者理解主要流程,笔者去掉了部分代码。

从上面代码可以看到,.vue 解析的工作其实是交给了 compiler.parseComponent 去完成,那么我们需要继续深入 compiler。
注意,这里 vue-template-compiler 并不是 vue-loader 的一部分了,从 vue-template-compiler 的 npm 主页可以了解到, vue-template-compiler 是 vue 本体的一部分
并不是一个单独的 package。通过查看文档可知,compiler.parseComponent 的逻辑在 vue/src/sfc/parser.js 里。

源码如下

parseComponent 做了什么

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
/**
* Parse a single-file component (*.vue) file into an SFC Descriptor Object.
*/
export function parseComponent (
content: string,
options?: Object = {}
): SFCDescriptor {
const sfc: SFCDescriptor = {
template: null,
script: null,
styles: [],
customBlocks: []
}
let depth = 0
let currentBlock: ?(SFCBlock | SFCCustomBlock) = null

function start (
tag: string,
attrs: Array<Attribute>,
unary: boolean,
start: number,
end: number
) {
// 略
}

function checkAttrs (block: SFCBlock, attrs: Array<Attribute>) {
// 略
}

function end (tag: string, start: number, end: number) {
// 略
}

function padContent (block: SFCBlock | SFCCustomBlock, pad: true | "line" | "space") {
// 略
}

parseHTML(content, {
start,
end
})

return sfc
}

parseComponent 里面有以下变量

  • 处理对象 sfc

    把 .vue 里的 css, javaScript, html 抽离出来之后,存放到找个对象里面

  • 变量 depth

    当前正在处理的节点的深度,比方说,对于 <template><div><p>foo</p></div></template来说,处理到 foo 时,当前深度就是 3, 处理到 </div> 时,当前深度就是 2 。

  • currentBlock

    当前正在处理的节点,以及该节点的 attr 和 content 等信息。

  • 函数 start

    遇到 openTag 节点时,对 openTag 的相关处理。逻辑不是很复杂,读者可以直接看源码。有一点值得注意的是,style 是用 array 形式存储的

  • 函数 end

    遇到 closeTag 节点时,对 closeTag 的相关处理。

  • 函数 checkAttrs

    对当前节点的 attrs 的相关处理

  • 函数 parseHTML

    这是和一个外部的函数,传入了 content (其实也就是 .vue 的内容)以及由 start和 end 两个函数组成的对象。看来,这个 parseHTML 之才是分解分析 .vue 的关键

    跟之前一样,我们要继续深入 parseHTML 函数来分析,它到底对 .vue 做了些什么,源码如下

parseHTML 做了什么

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

export function parseHTML (html, options) {
const stack = []
const expectHTML = options.expectHTML
const isUnaryTag = options.isUnaryTag || no
const canBeLeftOpenTag = options.canBeLeftOpenTag || no
let index = 0
let last, lastTag
while (html) {
last = html
if (!lastTag || !isPlainTextElement(lastTag)) {
// 这里分离了template
} else {
// 这里分离了style/script
}

// 略

// 前进n个字符
function advance (n) {
// 略
}

// 解析 openTag 比如 <template>
function parseStartTag () {
// 略
}

// 处理 openTag
function handleStartTag (match) {
// 略
if (options.start) {
options.start(tagName, attrs, unary, match.start, match.end)
}
}

// 处理 closeTag
function parseEndTag (tagName, start, end) {
// 略
if (options.start) {
options.start(tagName, [], false, start, end)
}
if (options.end) {
options.end(tagName, start, end)
}
}
}

深入到这一步,我想再提醒一下读者,selector的目的是将 .vue 中的 template, javaScript, css 分离出来。带着这个目的意识,我们再来审视这个 parseHTML。

parseHTML 整个函数的组成是:

  • 一个 while 循环

    在 while 循环中,存在两个大的分支,一个用来分析 template,一个是用来分析 script 和 style。

  • 函数 advance

    向前跳过文本

  • 函数 parseStartTag

    判断当前的 node 是不是 openTag

  • 函数 handleStartTag

    处理 openTag, 这里就用到了之前提到的 start() 函数

  • 函数 parseEndTag

    判断当前的 node 是不是 closeTag,同时这里也用到了 end() 函数

通过以上各个函数的组合,在while循环中就将 sfc 分割成了三个不同的部分,读者可以对比我的注释和源码自行解读源码逻辑。

顺便在这里吐个槽,很明显这里的 parseHTML 是函数名是有问题的,parseHTML 应该叫做 parseSFC 比较合适。