Set browser width/height

[clanstyles]

Set browser width/height

[kenshaw]

You need to set the width/height using a command line option. Unfortunately, this is dependent on the operating system (and I'm not sure if this is going to change for the headless version, either).

Basically, you pass a runner option to chromedp.New():

ctxt, cancel := context.WithCancel(context.Background());
defer cancel()
c, err := chromedp.New(ctxt, chromedp.WithRunnerOptions(runner.WindowSize(width, height)))

[kenshaw]

BTW, in the future I do plan to be able to prove some way to resize the window dynamically, but the issue with that is that currently Chrome (as far as I can tell) does not support that via the Remote Debugging Protocol, and I would need to write cross-platform support to resize the window via either system calls or via some other mechanism, and I just don't have the bandwidth at the moment to do the research needed. Also, as this is geared mostly towards running remote chrome instances, it wouldn't work when you're running chrome that way.

[clanstyles]

There has to be some way, how else would you emulate mobile?

I found

metrics := emulation.SetDeviceMetricsOverride(1024, 4068, 0, false, false)

// set the device metrics
if err := c.Run(ctxt, metrics); err != nil {
	log.Fatal(err)
}

Which didn't set the actual device though. I've also noticed if you try to issue a "Click" on a button that's not within the window's view, it can't click it and dies.

[kenshaw]

The rendering and events of Chrome are completely dependent on the items being in view, which, yes, I know, is a giant pain. You need to scroll to the element first (ie, via javascript) before applying an action against something. I am in the process of overhauling the existing actions to take this into account. With headless chrome, this may not be necessary. I am definitely not an expert with the Chrome API yet, but I also had tried the emulation domain to override settings, and it did not do what I had hoped.

[clanstyles]

How would you go about scrolling to the element?

[clanstyles]

I was able to get further with:

	vp := emulation.ForceViewport(1920, 1080, 1.0)
	metrics := emulation.SetDeviceMetricsOverride(1920, 1080, 1, false, false)

	if err = c.Run(ctxt, vp); err != nil {
		log.Fatal(err)
	}

	if err = c.Run(ctxt, metrics); err != nil {
		log.Fatal(err)
	}

It seems like it's able to interact with buttons that would be out of the window, but it's now within the view port. The downfall is that the page is gray. I was able to get the page to render by pressing "F12 (dev tools)".

[clanstyles]

https://groups.google.com/a/chromium.org/forum/#!topic/headless-dev/qcvOjVUZ-WQ

This seems to talk about it. From what I can tell, the buffer thinks it's that size, but the window might not actually be the size (that's why we see gray). So in theory, this is working, it's just not viewable.

[kenshaw]

I have a big fix rollup that I'll be pushing either late today or early tomorrow, that should fix the issues for elements that are off screen.

Basically, a "Click" action actually becomes a composition of the following:

1. Get dimensions (DOM.getBoxModel) of content
2. window.scrollTo() to the element's location via Runtime.evaluate
3. DOM.focus()
4. Input.dispatchMouseEvent with button down / up

From my simple testing so far, this seems to work 100% of the time -- I'm not going to try to claim it's 100% perfect, but I haven't found any failure scenarios yet.