Implement PhantomJS's render() about go-webkit2

[sqs]

@dansouza, are you interested in submitting a patch for this? I will code review it and help out if so. A few other folks had expressed an interest in this, so it'd be a great feature.

[bryansum]

I was halfway through an implementation myself. WebKit2 returns a Cairo surface. I can see a few different implementations for this: either we consume a cgo Cairo lib (e.g., go-cairo) and return a Go Cairo type; we return some kind of Go image type (image.RGBA?); or we do something similar to PhantomJS and take in a file path and render directly as a PNG or similar. Thoughts?

[bryansum]

FWIW I like the RGBA solution, as Go already allows for easy image conversion using std libraries.

[sqs]

Awesome.

I would personally prefer one of the first 2 options (a Go Cairo type, or a Go image type), or a 3rd option of returning a typedef'd pointer that another lib could cast to get a C Cairo type (e.g., type RawCairoSurface *C.cairo_surface_t) if they really wanted (but that doesn't require go-webkit2 to pull in go-cairo). But whichever one you find simplest is good with me.

[dansouza]

@sqs, @bryansum alright guys, go ahead and implement this, since I can't promise a date to implement this due to my current work load. I agree that returning at a Go image type would be preferred. BTW, how does PhantomJS generate PDFs?

[bryansum]

Slight update: Good and bad news. Good news is that Saucy snapshotting works! Bad news is that I spent an inordinate amount of time debugging the Raring version, which fails reliably. Should hopefully have a PR today.

@dansouza It looks like the current webkit2 snapshot implementation returns bitmap data. I imagine PhantomJS is hooking more deeply into the rendering layer, allowing for vector output.

[sqs]

Moving discussion to @bryansum's pull request #5