写在前面
本来是想写个如何编写gulp插件的科普文的,突然探究欲又发作了,于是就有了这篇东西。。。翻了下源码看了下gulp.src()
的实现,不禁由衷感慨:肿么这么复杂。。。
进入正题
首先我们看下gulpfile
里面的内容是长什么样子的,很有express中间件的味道是不是~
我们知道.pipe()
是典型的流式操作的API。很自然的,我们会想到gulp.src()
这个API返回的应该是个Stream对象(也许经过层层封装)。本着一探究竟的目的,花了点时间把gulp的源码大致扫了下,终于找到了答案。
gulpfile.js
var gulp = require('gulp'),
preprocess = require('gulp-preprocess'); gulp.task('default', function() { gulp.src('src/index.html') .pipe(preprocess({USERNAME:'程序猿小卡'})) .pipe(gulp.dest('dest/')); });
提前剧透
此处有内容剧透,如有对剧透不适者,请自行跳过本段落。。。
gulp.src() 的确返回了定制化的Stream对象。可以在github上搜索
ordered-read-streams
这个项目。大致关系是:
ordered-read-streams --> glob-stream --> vinyl-fs --> gulp.src()
探究之路
首先,我们看下require('gulp')
返回了什么。从gulp的源码来看,返回了Gulp
对象,该对象上有src
、pipe
、dest
等方法。很好,找到了我们想要的src
方法。接着往下看
参考:https://github.com/gulpjs/gulp/blob/master/index.js#L62
gulp/index.js
var inst = new Gulp();
module.exports = inst;
从下面的代码可以看到,gulp.src
方法,实际上是vfs.src
。继续
参考:https://github.com/gulpjs/gulp/blob/master/index.js#L25
gulp/index.js
var vfs = require('vinyl-fs');
// 省略很多行代码
Gulp.prototype.src = vfs.src;
接下来我们看下vfs.src
这个方法。从vinyl-fs/index.js
可以看到,vfs.src
实际是vinyl-fs/lib/src/index.js
。
参考:https://github.com/wearefractal/vinyl-fs/blob/master/index.js
vinyl-fs/index.js
'use strict';
module.exports = {
src: require('./lib/src'),
dest: require('./lib/dest'), watch: require('glob-watcher') };
那么,我们看下vinyl-fs/lib/src/index.js
。可以看到,gulp.src()
返回的,实际是outputStream
这货,而outputStream
是gs.create(glob, options).pipe()
获得的,差不多接近真相了,还有几步而已。
参考:https://github.com/wearefractal/vinyl-fs/blob/master/lib/src/index.js#L37
vinyl-fs/lib/src/index.js
var defaults = require('lodash.defaults');
var through = require('through2'); var gs = require('glob-stream'); var File = require('vinyl'); // 省略非重要代码若干行 function src(glob, opt) { // 继续省略代码 var globStream = gs.create(glob, options); // when people write to use just pass it through var outputStream = globStream .pipe(through.obj(createFile)) .pipe(getStats(options)); if (options.read !== false) { outputStream = outputStream .pipe(getContents(options)); } // 就是这里了 return outputStream .pipe(through.obj()); }
我们再看看glob-stream/index.js
里的create
方法,最后的return aggregate.pipe(uniqueStream);
。好的,下一步就是真相了,我们去ordered-read-streams
这个项目一探究竟。
参考:https://github.com/wearefractal/glob-stream/blob/master/index.js#L89
glob-stream/index.js
var through2 = require('through2');
var Combine = require('ordered-read-streams'); var unique = require('unique-stream'); var glob = require('glob'); var minimatch = require('minimatch'); var glob2base = require('glob2base'); var path = require('path'); // 必须省略很多代码 // create 方法 create: function(globs, opt) { // 继续省略代码 // create all individual streams var streams = positives.map(function(glob){ return gs.createStream(glob, negatives, opt); }); // then just pipe them to a single unique stream and return it var aggregate = new Combine(streams); var uniqueStream = unique('path'); // TODO: set up streaming queue so items come in order return aggregate.pipe(uniqueStream);
真相来了,我们看下ordered-read-streams
的代码,可能刚开始看不是很懂,没关系,知道它实现了自己的Stream
就可以了(nodejs是有暴露相应的API让开发者对Stream进行定制的),具体可参考:http://www.nodejs.org/api/stream.html#stream_api_for_stream_implementors
代码来自:https://github.com/armed/ordered-read-streams/blob/master/index.js
ordered-read-streams/index.js
function OrderedStreams(streams, options) {
if (!(this instanceof(OrderedStreams))) { return new OrderedStreams(streams, options); } streams = streams || []; options = options || {}; if (!Array.isArray(streams)) { streams = [streams]; } options.objectMode = true; Readable.call(this, options); // stream data buffer this._buffs = []; if (streams.length === 0) { this.push(null); // no streams, close return; } streams.forEach(function (s, i) { if (!s.readable) { throw new Error('All input streams must be readable'); } s.on('error', function (e) { this.emit('error', e); }.bind(this)); var buff = []; this._buffs.push(buff); s.on('data', buff.unshift.bind(buff)); s.on('end', flushStreamAtIndex.bind(this, i)); }, this); }
参考:https://github.com/armed/ordered-read-streams/blob/master/index.js
写在后面
兜兜转转一大圈,终于找到了gulp.src()
的源头,大致流程如下,算是蛮深的层级。代码细节神马的,有兴趣的同学可以深究一下。
ordered-read-streams --> glob-stream --> vinyl-fs --> gulp.src()