mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer-rs.git
synced 2024-11-25 11:01:10 +00:00
examples: Make use of glib::closure! for the overlay-composition example
This commit is contained in:
Sebastian Dröge
Sebastian Dröge
parent
eecce101da
commit
1ebfaee644
1 changed files with 114 additions and 115 deletions
|
|
@ -116,7 +116,6 @@ fn create_pipeline() -> Result<gst::Pipeline, Error> {
|
|||
info: None,
|
||||
}));
|
||||
|
||||
let drawer_clone = drawer.clone();
|
||||
// Connect to the overlaycomposition element's "draw" signal, which is emitted for
|
||||
// each videoframe piped through the element. The signal handler needs to
|
||||
// return a gst_video::VideoOverlayComposition to be drawn on the frame
|
||||
|
|
@ -128,126 +127,124 @@ fn create_pipeline() -> Result<gst::Pipeline, Error> {
|
|||
//
|
||||
// In this case, the signal passes the gst::Element and a gst::Sample with
|
||||
// the current buffer
|
||||
overlay.connect("draw", false, move |args| {
|
||||
use std::f64::consts::PI;
|
||||
overlay.connect_closure(
|
||||
"draw",
|
||||
false,
|
||||
glib::closure!(@strong drawer => move |_overlay: &gst::Element,
|
||||
sample: &gst::Sample| {
|
||||
use std::f64::consts::PI;
|
||||
|
||||
let drawer = &drawer_clone;
|
||||
let drawer = drawer.lock().unwrap();
|
||||
let drawer = drawer.lock().unwrap();
|
||||
|
||||
// Get the signal's arguments
|
||||
let _overlay = args[0].get::<gst::Element>().unwrap();
|
||||
let sample = args[1].get::<gst::Sample>().unwrap();
|
||||
let buffer = sample.buffer().unwrap();
|
||||
let timestamp = buffer.pts().unwrap();
|
||||
let buffer = sample.buffer().unwrap();
|
||||
let timestamp = buffer.pts().unwrap();
|
||||
|
||||
let info = drawer.info.as_ref().unwrap();
|
||||
let layout = drawer.layout.borrow();
|
||||
let info = drawer.info.as_ref().unwrap();
|
||||
let layout = drawer.layout.borrow();
|
||||
|
||||
let angle = 2.0 * PI * (timestamp % (10 * gst::ClockTime::SECOND)).nseconds() as f64
|
||||
/ (10.0 * gst::ClockTime::SECOND.nseconds() as f64);
|
||||
let angle = 2.0 * PI * (timestamp % (10 * gst::ClockTime::SECOND)).nseconds() as f64
|
||||
/ (10.0 * gst::ClockTime::SECOND.nseconds() as f64);
|
||||
|
||||
/* Create a Cairo image surface to draw into and the context around it. */
|
||||
let surface = cairo::ImageSurface::create(
|
||||
cairo::Format::ARgb32,
|
||||
info.width() as i32,
|
||||
info.height() as i32,
|
||||
)
|
||||
.unwrap();
|
||||
let cr = cairo::Context::new(&surface).expect("Failed to create cairo context");
|
||||
/* Create a Cairo image surface to draw into and the context around it. */
|
||||
let surface = cairo::ImageSurface::create(
|
||||
cairo::Format::ARgb32,
|
||||
info.width() as i32,
|
||||
info.height() as i32,
|
||||
)
|
||||
.unwrap();
|
||||
let cr = cairo::Context::new(&surface).expect("Failed to create cairo context");
|
||||
|
||||
cr.save().expect("Failed to save state");
|
||||
cr.set_operator(cairo::Operator::Clear);
|
||||
cr.paint().expect("Failed to clear background");
|
||||
cr.restore().expect("Failed to restore state");
|
||||
|
||||
// The image we draw (the text) will be static, but we will change the
|
||||
// transformation on the drawing context, which rotates and shifts everything
|
||||
// that we draw afterwards. Like this, we have no complicated calulations
|
||||
// in the actual drawing below.
|
||||
// Calling multiple transformation methods after each other will apply the
|
||||
// new transformation on top. If you repeat the cr.rotate(angle) line below
|
||||
// this a second time, everything in the canvas will rotate twice as fast.
|
||||
cr.translate(
|
||||
f64::from(info.width()) / 2.0,
|
||||
f64::from(info.height()) / 2.0,
|
||||
);
|
||||
cr.rotate(angle);
|
||||
|
||||
// This loop will render 10 times the string "GStreamer" in a circle
|
||||
for i in 0..10 {
|
||||
// Cairo, like most rendering frameworks, is using a stack for transformations
|
||||
// with this, we push our current transformation onto this stack - allowing us
|
||||
// to make temporary changes / render something / and then returning to the
|
||||
// previous transformations.
|
||||
cr.save().expect("Failed to save state");
|
||||
|
||||
let angle = (360. * f64::from(i)) / 10.0;
|
||||
let red = (1.0 + f64::cos((angle - 60.0) * PI / 180.0)) / 2.0;
|
||||
cr.set_source_rgb(red, 0.0, 1.0 - red);
|
||||
cr.rotate(angle * PI / 180.0);
|
||||
|
||||
// Update the text layout. This function is only updating pango's internal state.
|
||||
// So e.g. that after a 90 degree rotation it knows that what was previously going
|
||||
// to end up as a 200x100 rectangle would now be 100x200.
|
||||
pangocairo::functions::update_layout(&cr, &**layout);
|
||||
let (width, _height) = layout.size();
|
||||
// Using width and height of the text, we can properly possition it within
|
||||
// our canvas.
|
||||
cr.move_to(
|
||||
-(f64::from(width) / f64::from(pango::SCALE)) / 2.0,
|
||||
-(f64::from(info.height())) / 2.0,
|
||||
);
|
||||
// After telling the layout object where to draw itself, we actually tell
|
||||
// it to draw itself into our cairo context.
|
||||
pangocairo::functions::show_layout(&cr, &**layout);
|
||||
|
||||
// Here we go one step up in our stack of transformations, removing any
|
||||
// changes we did to them since the last call to cr.save();
|
||||
cr.set_operator(cairo::Operator::Clear);
|
||||
cr.paint().expect("Failed to clear background");
|
||||
cr.restore().expect("Failed to restore state");
|
||||
}
|
||||
|
||||
/* Drop the Cairo context to release the additional reference to the data and
|
||||
* then take ownership of the data. This only works if we have the one and only
|
||||
* reference to the image surface */
|
||||
drop(cr);
|
||||
let stride = surface.stride();
|
||||
let data = surface.take_data().unwrap();
|
||||
// The image we draw (the text) will be static, but we will change the
|
||||
// transformation on the drawing context, which rotates and shifts everything
|
||||
// that we draw afterwards. Like this, we have no complicated calulations
|
||||
// in the actual drawing below.
|
||||
// Calling multiple transformation methods after each other will apply the
|
||||
// new transformation on top. If you repeat the cr.rotate(angle) line below
|
||||
// this a second time, everything in the canvas will rotate twice as fast.
|
||||
cr.translate(
|
||||
f64::from(info.width()) / 2.0,
|
||||
f64::from(info.height()) / 2.0,
|
||||
);
|
||||
cr.rotate(angle);
|
||||
|
||||
/* Create an RGBA buffer, and add a video meta that the videooverlaycomposition expects */
|
||||
let mut buffer = gst::Buffer::from_mut_slice(data);
|
||||
// This loop will render 10 times the string "GStreamer" in a circle
|
||||
for i in 0..10 {
|
||||
// Cairo, like most rendering frameworks, is using a stack for transformations
|
||||
// with this, we push our current transformation onto this stack - allowing us
|
||||
// to make temporary changes / render something / and then returning to the
|
||||
// previous transformations.
|
||||
cr.save().expect("Failed to save state");
|
||||
|
||||
gst_video::VideoMeta::add_full(
|
||||
buffer.get_mut().unwrap(),
|
||||
gst_video::VideoFrameFlags::empty(),
|
||||
gst_video::VideoFormat::Bgra,
|
||||
info.width(),
|
||||
info.height(),
|
||||
&[0],
|
||||
&[stride],
|
||||
)
|
||||
.unwrap();
|
||||
let angle = (360. * f64::from(i)) / 10.0;
|
||||
let red = (1.0 + f64::cos((angle - 60.0) * PI / 180.0)) / 2.0;
|
||||
cr.set_source_rgb(red, 0.0, 1.0 - red);
|
||||
cr.rotate(angle * PI / 180.0);
|
||||
|
||||
/* Turn the buffer into a VideoOverlayRectangle, then place
|
||||
* that into a VideoOverlayComposition and return it.
|
||||
*
|
||||
* A VideoOverlayComposition can take a Vec of such rectangles
|
||||
* spaced around the video frame, but we're just outputting 1
|
||||
* here */
|
||||
let rect = gst_video::VideoOverlayRectangle::new_raw(
|
||||
&buffer,
|
||||
0,
|
||||
0,
|
||||
info.width(),
|
||||
info.height(),
|
||||
gst_video::VideoOverlayFormatFlags::PREMULTIPLIED_ALPHA,
|
||||
);
|
||||
// Update the text layout. This function is only updating pango's internal state.
|
||||
// So e.g. that after a 90 degree rotation it knows that what was previously going
|
||||
// to end up as a 200x100 rectangle would now be 100x200.
|
||||
pangocairo::functions::update_layout(&cr, &**layout);
|
||||
let (width, _height) = layout.size();
|
||||
// Using width and height of the text, we can properly possition it within
|
||||
// our canvas.
|
||||
cr.move_to(
|
||||
-(f64::from(width) / f64::from(pango::SCALE)) / 2.0,
|
||||
-(f64::from(info.height())) / 2.0,
|
||||
);
|
||||
// After telling the layout object where to draw itself, we actually tell
|
||||
// it to draw itself into our cairo context.
|
||||
pangocairo::functions::show_layout(&cr, &**layout);
|
||||
|
||||
// Here we go one step up in our stack of transformations, removing any
|
||||
// changes we did to them since the last call to cr.save();
|
||||
cr.restore().expect("Failed to restore state");
|
||||
}
|
||||
|
||||
/* Drop the Cairo context to release the additional reference to the data and
|
||||
* then take ownership of the data. This only works if we have the one and only
|
||||
* reference to the image surface */
|
||||
drop(cr);
|
||||
let stride = surface.stride();
|
||||
let data = surface.take_data().unwrap();
|
||||
|
||||
/* Create an RGBA buffer, and add a video meta that the videooverlaycomposition expects */
|
||||
let mut buffer = gst::Buffer::from_mut_slice(data);
|
||||
|
||||
gst_video::VideoMeta::add_full(
|
||||
buffer.get_mut().unwrap(),
|
||||
gst_video::VideoFrameFlags::empty(),
|
||||
gst_video::VideoFormat::Bgra,
|
||||
info.width(),
|
||||
info.height(),
|
||||
&[0],
|
||||
&[stride],
|
||||
)
|
||||
.unwrap();
|
||||
|
||||
/* Turn the buffer into a VideoOverlayRectangle, then place
|
||||
* that into a VideoOverlayComposition and return it.
|
||||
*
|
||||
* A VideoOverlayComposition can take a Vec of such rectangles
|
||||
* spaced around the video frame, but we're just outputting 1
|
||||
* here */
|
||||
let rect = gst_video::VideoOverlayRectangle::new_raw(
|
||||
&buffer,
|
||||
0,
|
||||
0,
|
||||
info.width(),
|
||||
info.height(),
|
||||
gst_video::VideoOverlayFormatFlags::PREMULTIPLIED_ALPHA,
|
||||
);
|
||||
|
||||
Some(
|
||||
gst_video::VideoOverlayComposition::new(Some(&rect))
|
||||
.unwrap()
|
||||
.to_value(),
|
||||
)
|
||||
});
|
||||
}),
|
||||
);
|
||||
|
||||
// Add a signal handler to the overlay's "caps-changed" signal. This could e.g.
|
||||
// be called when the sink that we render to does not support resizing the image
|
||||
|
|
@ -256,15 +253,17 @@ fn create_pipeline() -> Result<gst::Pipeline, Error> {
|
|||
// resize our canvas's size.
|
||||
// Another possibility for when this might happen is, when our video is a network
|
||||
// stream that dynamically changes resolution when enough bandwith is available.
|
||||
overlay.connect("caps-changed", false, move |args| {
|
||||
let _overlay = args[0].get::<gst::Element>().unwrap();
|
||||
let caps = args[1].get::<gst::Caps>().unwrap();
|
||||
|
||||
let mut drawer = drawer.lock().unwrap();
|
||||
drawer.info = Some(gst_video::VideoInfo::from_caps(&caps).unwrap());
|
||||
|
||||
None
|
||||
});
|
||||
overlay.connect_closure(
|
||||
"caps-changed",
|
||||
false,
|
||||
glib::closure!(move |_overlay: &gst::Element,
|
||||
caps: &gst::Caps,
|
||||
_width: u32,
|
||||
_height: u32| {
|
||||
let mut drawer = drawer.lock().unwrap();
|
||||
drawer.info = Some(gst_video::VideoInfo::from_caps(caps).unwrap());
|
||||
}),
|
||||
);
|
||||
|
||||
Ok(pipeline)
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in a new issue