$ gnpm install devtools-timeline-model
Parse raw trace data into the Chrome DevTools' structured profiling data models
If you use something like big-rig or automated-chrome-profiling you may end up with raw trace data. It's pretty raw. This module will parse that stuff into something a bit more consumable, and should help you with higher level analysis.
$ npm install --save devtools-timeline-model
var filename = 'demo/mdn-fling.json'
var events = require('fs').readFileSync(filename, 'utf8')
var DevtoolsTimelineModel = require('devtools-timeline-model');
// events can be either a string of the trace data or the JSON.parse'd equivalent
var model = new DevtoolsTimelineModel(events)
// tracing model
model.tracingModel()
// timeline model, all events
model.timelineModel()
// interaction model, incl scroll, click, animations
model.interactionModel()
// frame model, incl frame durations
model.frameModel()
// filmstrip model, incl screenshots
model.filmStripModel()
// topdown tree
model.topDown()
// bottom up tree
model.bottomUp()
// bottom up tree, grouped by URL
model.bottomUpGroupBy('URL') // accepts: None Category Subdomain Domain URL EventName
// see example.js for API examples.
These objects are huge. You'll want to explore them in a UI like devtool.
npm i
brew install entr
gls index.js lib/*.js | entr node example.js
Requiring the DevTools frontend looks rather straightforward at first. (global.WebInspector = {}
, then start require()
ing the files, in dependency order). However, there are two problems that crop up:
utilities.js
adds a number of methods to native object prototypes, such as Array, Object, and typed arrays.devtools-timeline-model
addresses that by sandboxing the WebInspector into it's own context. Here's how it works:
// First, sandboxed contexts don't have any globals from node, so we whitelist a few we'll provide for it.
var glob = { require: require, global: global, console: console, process, process, __dirname: __dirname }
// We read in our script to run, and create a vm.Script object
var script = new vm.Script(fs.readFileSync(__dirname + "/lib/timeline-model.js", 'utf8'))
// We create a new V8 context with our globals
var ctx = vm.createContext(glob)
// We evaluate the `vm.Script` in the new context
var output = script.runInContext(ctx)
// establish our sandboxed globals
this.window = this.self = this.global = this
// We locally eval, as the node module scope isn't appropriate for the browser-centric DevTools frontend
function requireval(path){
var filesrc = fs.readFileSync(__dirname + '/node_modules/' + path, 'utf8');
eval(filesrc + '\n\n//# sourceURL=' + path);
}
// polyfills, then the real chrome devtools frontend
requireval('../lib/api-stubs.js')
requireval('chrome-devtools-frontend/front_end/common/Object.js')
requireval('chrome-devtools-frontend/front_end/common/SegmentedRange.js')
requireval('chrome-devtools-frontend/front_end/platform/utilities.js')
requireval('chrome-devtools-frontend/front_end/sdk/Target.js')
// ...
// After that's all done, we pull the local `instance` variable out, to use as our proxy object
this.sandbox = ctx.instance;
Debugging is harder, as most tools aren't used to this setup. While devtool
doesn't work well, you can have it run lib/devtools-timeline-model.js
directly, which is fairly succesful. The classic node-inspector
does work pretty well with the sandboxed script, though the workflow is a little worse than devtool
's.
Apache © Paul Irish
Copyright 2013 - present © cnpmjs.org | Home |