librsvg 2.62.2

A library to render SVG images to Cairo surfaces. GNOME uses this to render SVG icons. Outside of GNOME, other desktop environments use it for similar purposes. Wikimedia uses it for Wikipedia's SVG diagrams.
Documentation
//! Various utilities for working with Cairo image surfaces.

use std::alloc;
use std::slice;

pub mod iterators;
pub mod shared_surface;
pub mod srgb;

// These two are for Cairo's platform-endian 0xaarrggbb pixels

#[cfg(target_endian = "little")]
use rgb::alt::BGRA8;
#[cfg(target_endian = "little")]
#[allow(clippy::upper_case_acronyms)]
pub type CairoARGB = BGRA8;

#[cfg(target_endian = "big")]
use rgb::alt::ARGB8;
#[cfg(target_endian = "big")]
#[allow(clippy::upper_case_acronyms)]
pub type CairoARGB = ARGB8;

use rgb::ColorComponentMap;

/// Analogous to `rgb::FromSlice`, to convert from `[T]` to `[CairoARGB]`
#[allow(clippy::upper_case_acronyms)]
pub trait AsCairoARGB {
    /// Reinterpret slice as `CairoARGB` pixels.
    fn as_cairo_argb(&self) -> &[CairoARGB];

    /// Reinterpret mutable slice as `CairoARGB` pixels.
    fn as_cairo_argb_mut(&mut self) -> &mut [CairoARGB];
}

// SAFETY: transmuting from u32 to CairoRGB is based on the following assumptions:
//  * there are no invalid bit representations for ARGB
//  * u32 and ARGB are the same size
//  * u32 is sufficiently aligned
impl AsCairoARGB for [u32] {
    fn as_cairo_argb(&self) -> &[CairoARGB] {
        const LAYOUT_U32: alloc::Layout = alloc::Layout::new::<u32>();
        const LAYOUT_ARGB: alloc::Layout = alloc::Layout::new::<CairoARGB>();
        let _: [(); LAYOUT_U32.size()] = [(); LAYOUT_ARGB.size()];
        let _: [(); 0] = [(); LAYOUT_U32.align() % LAYOUT_ARGB.align()];
        unsafe { slice::from_raw_parts(self.as_ptr() as *const _, self.len()) }
    }

    fn as_cairo_argb_mut(&mut self) -> &mut [CairoARGB] {
        unsafe { slice::from_raw_parts_mut(self.as_mut_ptr() as *mut _, self.len()) }
    }
}

/// Modes which specify how the values of out of bounds pixels are computed.
///
/// <https://www.w3.org/TR/filter-effects/#element-attrdef-fegaussianblur-edgemode>
#[derive(Debug, Clone, Copy, Eq, PartialEq)]
pub enum EdgeMode {
    /// The nearest inbounds pixel value is returned.
    Duplicate,
    /// The image is extended by taking the color values from the opposite of the image.
    ///
    /// Imagine the image being tiled infinitely, with the original image at the origin.
    Wrap,
    /// Zero RGBA values are returned.
    None,
}

/// Trait to convert pixels in various formats to our own Pixel layout.
pub trait ToPixel {
    fn to_pixel(&self) -> Pixel;
}

/// Trait to convert pixels in various formats to Cairo's endian-dependent 0xaarrggbb.
pub trait ToCairoARGB {
    fn to_cairo_argb(&self) -> CairoARGB;
}

impl ToPixel for CairoARGB {
    #[inline]
    fn to_pixel(&self) -> Pixel {
        Pixel {
            r: self.r,
            g: self.g,
            b: self.b,
            a: self.a,
        }
    }
}

impl ToPixel for image::Rgba<u8> {
    #[inline]
    fn to_pixel(&self) -> Pixel {
        Pixel {
            r: self.0[0],
            g: self.0[1],
            b: self.0[2],
            a: self.0[3],
        }
    }
}

impl ToCairoARGB for Pixel {
    #[inline]
    fn to_cairo_argb(&self) -> CairoARGB {
        CairoARGB {
            r: self.r,
            g: self.g,
            b: self.b,
            a: self.a,
        }
    }
}

/// Extension methods for `cairo::ImageSurfaceData`.
pub trait ImageSurfaceDataExt {
    /// Sets the pixel at the given coordinates. Assumes the `ARgb32` format.
    fn set_pixel(&mut self, stride: usize, pixel: Pixel, x: u32, y: u32);
}

/// A pixel consisting of R, G, B and A values.
pub type Pixel = rgb::RGBA8;

pub trait PixelOps {
    fn premultiply(self) -> Self;
    fn unpremultiply(self) -> Self;
    fn diff(&self, other: &Self) -> Self;
    fn to_luminance_mask(&self) -> Self;
    fn to_u32(&self) -> u32;
    fn from_u32(x: u32) -> Self;
}

impl PixelOps for Pixel {
    /// Returns an unpremultiplied value of this pixel.
    ///
    /// For a fully transparent pixel, a transparent black pixel will be returned.
    #[inline]
    fn unpremultiply(self) -> Self {
        if self.a == 0 {
            Self {
                r: 0,
                g: 0,
                b: 0,
                a: 0,
            }
        } else {
            let alpha = f32::from(self.a) / 255.0;
            self.map_colors(|x| ((f32::from(x) / alpha) + 0.5) as u8)
        }
    }

    /// Returns a premultiplied value of this pixel.
    #[inline]
    fn premultiply(self) -> Self {
        let a = self.a as u32;
        self.map_colors(|x| (((x as u32) * a + 127) / 255) as u8)
    }

    #[inline]
    fn diff(&self, other: &Pixel) -> Pixel {
        self.iter()
            .zip(other.iter())
            .map(|(l, r)| (l as i32 - r as i32).unsigned_abs() as u8)
            .collect()
    }

    /// Returns a 'mask' pixel with only the alpha channel
    ///
    /// Assuming, the pixel is linear RGB (not sRGB)
    /// y = luminance
    /// Y = 0.2126 R + 0.7152 G + 0.0722 B
    /// 1.0 opacity = 255
    ///
    /// When Y = 1.0, pixel for mask should be 0xFFFFFFFF
    /// (you get 1.0 luminance from 255 from R, G and B)
    ///
    /// r_mult = 0xFFFFFFFF / (255.0 * 255.0) * .2126 = 14042.45  ~= 14042
    /// g_mult = 0xFFFFFFFF / (255.0 * 255.0) * .7152 = 47239.69  ~= 47240
    /// b_mult = 0xFFFFFFFF / (255.0 * 255.0) * .0722 =  4768.88  ~= 4769
    ///
    /// This allows for the following expected behaviour:
    ///    (we only care about the most significant byte)
    /// if pixel = 0x00FFFFFF, pixel' = 0xFF......
    /// if pixel = 0x00020202, pixel' = 0x02......
    ///
    /// if pixel = 0x00000000, pixel' = 0x00......
    #[inline]
    fn to_luminance_mask(&self) -> Self {
        let r = u32::from(self.r);
        let g = u32::from(self.g);
        let b = u32::from(self.b);

        Self {
            r: 0,
            g: 0,
            b: 0,
            a: (((r * 14042 + g * 47240 + b * 4769) * 255) >> 24) as u8,
        }
    }

    /// Returns the pixel value as a `u32`, in the same format as `cairo::Format::ARgb32`.
    #[inline]
    fn to_u32(&self) -> u32 {
        (u32::from(self.a) << 24)
            | (u32::from(self.r) << 16)
            | (u32::from(self.g) << 8)
            | u32::from(self.b)
    }

    /// Converts a `u32` in the same format as `cairo::Format::ARgb32` into a `Pixel`.
    #[inline]
    fn from_u32(x: u32) -> Self {
        Self {
            r: ((x >> 16) & 0xFF) as u8,
            g: ((x >> 8) & 0xFF) as u8,
            b: (x & 0xFF) as u8,
            a: ((x >> 24) & 0xFF) as u8,
        }
    }
}

impl<'a> ImageSurfaceDataExt for cairo::ImageSurfaceData<'a> {
    #[inline]
    fn set_pixel(&mut self, stride: usize, pixel: Pixel, x: u32, y: u32) {
        let this: &mut [u8] = &mut *self;
        // SAFETY: this code assumes that cairo image surface data is correctly
        // aligned for u32. This assumption is justified by the Cairo docs,
        // which say this:
        //
        // https://cairographics.org/manual/cairo-Image-Surfaces.html#cairo-image-surface-create-for-data
        //
        // > This pointer must be suitably aligned for any kind of variable,
        // > (for example, a pointer returned by malloc).
        #[allow(clippy::cast_ptr_alignment)]
        let this: &mut [u32] =
            unsafe { slice::from_raw_parts_mut(this.as_mut_ptr() as *mut u32, this.len() / 4) };
        this.set_pixel(stride, pixel, x, y);
    }
}
impl ImageSurfaceDataExt for [u8] {
    #[inline]
    fn set_pixel(&mut self, stride: usize, pixel: Pixel, x: u32, y: u32) {
        let this = &mut self[y as usize * stride + x as usize * 4..];
        this[..4].copy_from_slice(&pixel.to_u32().to_ne_bytes());
    }
}
impl ImageSurfaceDataExt for [u32] {
    #[inline]
    fn set_pixel(&mut self, stride: usize, pixel: Pixel, x: u32, y: u32) {
        self[(y as usize * stride + x as usize * 4) / 4] = pixel.to_u32();
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use proptest::prelude::*;

    #[test]
    fn pixel_diff() {
        let a = Pixel::new(0x10, 0x20, 0xf0, 0x40);
        assert_eq!(a, a.diff(&Pixel::default()));
        let b = Pixel::new(0x50, 0xff, 0x20, 0x10);
        assert_eq!(a.diff(&b), Pixel::new(0x40, 0xdf, 0xd0, 0x30));
    }

    // Floating-point reference implementation
    fn premultiply_float(pixel: Pixel) -> Pixel {
        let alpha = f64::from(pixel.a) / 255.0;
        pixel.map_colors(|x| ((f64::from(x) * alpha) + 0.5) as u8)
    }

    prop_compose! {
        fn arbitrary_pixel()(a: u8, r: u8, g: u8, b: u8) -> Pixel {
            Pixel { r, g, b, a }
        }
    }

    proptest! {
        #[test]
        fn pixel_premultiply(pixel in arbitrary_pixel()) {
            prop_assert_eq!(pixel.premultiply(), premultiply_float(pixel));
        }

        #[test]
        fn pixel_unpremultiply(pixel in arbitrary_pixel()) {
            let roundtrip = pixel.premultiply().unpremultiply();
            if pixel.a == 0 {
                prop_assert_eq!(roundtrip, Pixel::default());
            } else {
                // roundtrip can't be perfect, the accepted error depends on alpha
                let tolerance = 0xff / pixel.a;
                let diff = roundtrip.diff(&pixel);
                prop_assert!(diff.r <= tolerance, "red component value differs by more than {}: {:?}", tolerance, roundtrip);
                prop_assert!(diff.g <= tolerance, "green component value differs by more than {}: {:?}", tolerance, roundtrip);
                prop_assert!(diff.b <= tolerance, "blue component value differs by more than {}: {:?}", tolerance, roundtrip);

                prop_assert_eq!(pixel.a, roundtrip.a);
            }
       }
    }
}