DPI-based natural sizing for images (#3571)

4 files changed, 177 insertions, 25 deletions

diff --git a/crates/typst/Cargo.toml b/crates/typst/Cargo.toml
index 77809ed1..889d41e7 100644
--- a/crates/typst/Cargo.toml
+++ b/crates/typst/Cargo.toml
@@ -46,6 +46,7 @@ once_cell = { workspace = true }
 palette = { workspace = true }
 qcms = { workspace = true }
 phf = { workspace = true }
+png = { workspace = true }
 portable-atomic = { workspace = true }
 rayon = { workspace = true }
 regex = { workspace = true }
@@ -72,5 +73,8 @@ wasmi = { workspace = true }
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 stacker = { workspace = true }
 
+[dev-dependencies]
+typst-dev-assets = { workspace = true }
+
 [lints]
 workspace = true
diff --git a/crates/typst/src/visualize/image/mod.rs b/crates/typst/src/visualize/image/mod.rs
index e7c91cd6..199564c6 100644
--- a/crates/typst/src/visualize/image/mod.rs
+++ b/crates/typst/src/visualize/image/mod.rs
@@ -27,7 +27,7 @@ use crate::loading::Readable;
 use crate::model::Figurable;
 use crate::syntax::{Span, Spanned};
 use crate::text::{families, Lang, LocalName, Region};
-use crate::util::{option_eq, LazyHash, Numeric};
+use crate::util::{option_eq, LazyHash};
 use crate::visualize::Path;
 use crate::World;
 
@@ -198,20 +198,30 @@ impl LayoutSingle for Packed<ImageElem> {
         let region_ratio = region.x / region.y;
 
         // Find out whether the image is wider or taller than the target size.
-        let pxw = image.width() as f64;
-        let pxh = image.height() as f64;
+        let pxw = image.width();
+        let pxh = image.height();
         let px_ratio = pxw / pxh;
         let wide = px_ratio > region_ratio;
 
         // The space into which the image will be placed according to its fit.
         let target = if expand.x && expand.y {
+            // If both width and height are forced, take them.
             region
-        } else if expand.x || (!expand.y && wide && region.x.is_finite()) {
+        } else if expand.x {
+            // If just width is forced, take it.
             Size::new(region.x, region.y.min(region.x.safe_div(px_ratio)))
-        } else if region.y.is_finite() {
+        } else if expand.y {
+            // If just height is forced, take it.
             Size::new(region.x.min(region.y * px_ratio), region.y)
         } else {
-            Size::new(Abs::pt(pxw), Abs::pt(pxh))
+            // If neither is forced, take the natural image size at the image's
+            // DPI bounded by the available space.
+            let dpi = image.dpi().unwrap_or(Image::DEFAULT_DPI);
+            let natural = Axes::new(pxw, pxh).map(|v| Abs::inches(v / dpi));
+            Size::new(
+                natural.x.min(region.x).min(region.y * px_ratio),
+                natural.y.min(region.y).min(region.x.safe_div(px_ratio)),
+            )
         };
 
         // Compute the actual size of the fitted image.
@@ -219,7 +229,7 @@ impl LayoutSingle for Packed<ImageElem> {
         let fitted = match fit {
             ImageFit::Cover | ImageFit::Contain => {
                 if wide == (fit == ImageFit::Contain) {
-                    Size::new(target.x, target.x / px_ratio)
+                    Size::new(target.x, target.x.safe_div(px_ratio))
                 } else {
                     Size::new(target.y * px_ratio, target.y)
                 }
@@ -320,6 +330,10 @@ pub enum ImageKind {
 }
 
 impl Image {
+    /// When scaling an image to it's natural size, we default to this DPI
+    /// if the image doesn't contain DPI metadata.
+    pub const DEFAULT_DPI: f64 = 72.0;
+
     /// Create an image from a buffer and a format.
     #[comemo::memoize]
     #[typst_macros::time(name = "load image")]
@@ -394,6 +408,14 @@ impl Image {
         }
     }
 
+    /// The image's pixel density in pixels per inch, if known.
+    pub fn dpi(&self) -> Option<f64> {
+        match &self.0.kind {
+            ImageKind::Raster(raster) => raster.dpi(),
+            ImageKind::Svg(_) => None,
+        }
+    }
+
     /// A text describing the image.
     pub fn alt(&self) -> Option<&str> {
         self.0.alt.as_deref()
diff --git a/crates/typst/src/visualize/image/raster.rs b/crates/typst/src/visualize/image/raster.rs
index 3c9afe43..995e3483 100644
--- a/crates/typst/src/visualize/image/raster.rs
+++ b/crates/typst/src/visualize/image/raster.rs
@@ -1,3 +1,4 @@
+use std::cmp::Ordering;
 use std::hash::{Hash, Hasher};
 use std::io;
 use std::sync::Arc;
@@ -22,6 +23,7 @@ struct Repr {
     format: RasterFormat,
     dynamic: image::DynamicImage,
     icc: Option<Vec<u8>>,
+    dpi: Option<f64>,
 }
 
 impl RasterImage {
@@ -46,11 +48,19 @@ impl RasterImage {
         }
         .map_err(format_image_error)?;
 
-        if let Some(rotation) = exif_rotation(&data) {
+        let exif = exif::Reader::new()
+            .read_from_container(&mut std::io::Cursor::new(&data))
+            .ok();
+
+        // Apply rotation from EXIF metadata.
+        if let Some(rotation) = exif.as_ref().and_then(exif_rotation) {
             apply_rotation(&mut dynamic, rotation);
         }
 
-        Ok(Self(Arc::new(Repr { data, format, dynamic, icc })))
+        // Extract pixel density.
+        let dpi = determine_dpi(&data, exif.as_ref());
+
+        Ok(Self(Arc::new(Repr { data, format, dynamic, icc, dpi })))
     }
 
     /// The raw image data.
@@ -73,6 +83,11 @@ impl RasterImage {
         self.dynamic().height()
     }
 
+    /// The image's pixel density in pixels per inch, if known.
+    pub fn dpi(&self) -> Option<f64> {
+        self.0.dpi
+    }
+
     /// Access the underlying dynamic image.
     pub fn dynamic(&self) -> &image::DynamicImage {
         &self.0.dynamic
@@ -133,13 +148,11 @@ impl TryFrom<image::ImageFormat> for RasterFormat {
     }
 }
 
-/// Get rotation from EXIF metadata.
-fn exif_rotation(data: &[u8]) -> Option<u32> {
-    let reader = exif::Reader::new();
-    let mut cursor = std::io::Cursor::new(data);
-    let exif = reader.read_from_container(&mut cursor).ok()?;
-    let orient = exif.get_field(exif::Tag::Orientation, exif::In::PRIMARY)?;
-    orient.value.get_uint(0)
+/// Try to get the rotation from the EXIF metadata.
+fn exif_rotation(exif: &exif::Exif) -> Option<u32> {
+    exif.get_field(exif::Tag::Orientation, exif::In::PRIMARY)?
+        .value
+        .get_uint(0)
 }
 
 /// Apply an EXIF rotation to a dynamic image.
@@ -163,6 +176,87 @@ fn apply_rotation(image: &mut DynamicImage, rotation: u32) {
     }
 }
 
+/// Try to determine the DPI (dots per inch) of the image.
+fn determine_dpi(data: &[u8], exif: Option<&exif::Exif>) -> Option<f64> {
+    // Try to extract the DPI from the EXIF metadata. If that doesn't yield
+    // anything, fall back to specialized procedures for extracting JPEG or PNG
+    // DPI metadata. GIF does not have any.
+    exif.and_then(exif_dpi)
+        .or_else(|| jpeg_dpi(data))
+        .or_else(|| png_dpi(data))
+}
+
+/// Try to get the DPI from the EXIF metadata.
+fn exif_dpi(exif: &exif::Exif) -> Option<f64> {
+    let axis = |tag| {
+        let dpi = exif.get_field(tag, exif::In::PRIMARY)?;
+        let exif::Value::Rational(rational) = &dpi.value else { return None };
+        Some(rational.first()?.to_f64())
+    };
+
+    [axis(exif::Tag::XResolution), axis(exif::Tag::YResolution)]
+        .into_iter()
+        .flatten()
+        .max_by(|a, b| a.partial_cmp(b).unwrap_or(Ordering::Equal))
+}
+
+/// Tries to extract the DPI from raw JPEG data (by inspecting the JFIF APP0
+/// section).
+fn jpeg_dpi(data: &[u8]) -> Option<f64> {
+    let validate_at = |index: usize, expect: &[u8]| -> Option<()> {
+        data.get(index..)?.starts_with(expect).then_some(())
+    };
+    let u16_at = |index: usize| -> Option<u16> {
+        data.get(index..index + 2)?.try_into().ok().map(u16::from_be_bytes)
+    };
+
+    validate_at(0, b"\xFF\xD8\xFF\xE0\0")?;
+    validate_at(6, b"JFIF\0")?;
+    validate_at(11, b"\x01")?;
+
+    let len = u16_at(4)?;
+    if len < 16 {
+        return None;
+    }
+
+    let units = *data.get(13)?;
+    let x = u16_at(14)?;
+    let y = u16_at(16)?;
+    let dpu = x.max(y) as f64;
+
+    Some(match units {
+        1 => dpu,        // already inches
+        2 => dpu * 2.54, // cm -> inches
+        _ => return None,
+    })
+}
+
+/// Tries to extract the DPI from raw PNG data.
+fn png_dpi(mut data: &[u8]) -> Option<f64> {
+    let mut decoder = png::StreamingDecoder::new();
+    let dims = loop {
+        let (consumed, event) = decoder.update(data, &mut Vec::new()).ok()?;
+        match event {
+            png::Decoded::PixelDimensions(dims) => break dims,
+            // Bail as soon as there is anything data-like.
+            png::Decoded::ChunkBegin(_, png::chunk::IDAT)
+            | png::Decoded::ImageData
+            | png::Decoded::ImageEnd => return None,
+            _ => {}
+        }
+        data = data.get(consumed..)?;
+        if consumed == 0 {
+            return None;
+        }
+    };
+
+    let dpu = dims.xppu.max(dims.yppu) as f64;
+    match dims.unit {
+        png::Unit::Meter => Some(dpu * 0.0254), // meter -> inches
+        png::Unit::Unspecified => None,
+    }
+}
+
 /// Format the user-facing raster graphic decoding error message.
 fn format_image_error(error: image::ImageError) -> EcoString {
     match error {
@@ -170,3 +264,24 @@ fn format_image_error(error: image::ImageError) -> EcoString {
         err => eco_format!("failed to decode image ({err})"),
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::{RasterFormat, RasterImage};
+    use crate::foundations::Bytes;
+
+    #[test]
+    fn test_image_dpi() {
+        #[track_caller]
+        fn test(path: &str, format: RasterFormat, dpi: f64) {
+            let data = typst_dev_assets::get(path).unwrap();
+            let bytes = Bytes::from_static(data);
+            let image = RasterImage::new(bytes, format).unwrap();
+            assert_eq!(image.dpi().map(f64::round), Some(dpi));
+        }
+
+        test("images/f2t.jpg", RasterFormat::Jpg, 220.0);
+        test("images/tiger.jpg", RasterFormat::Jpg, 72.0);
+        test("images/graph.png", RasterFormat::Png, 144.0);
+    }
+}
diff --git a/crates/typst/src/visualize/image/svg.rs b/crates/typst/src/visualize/image/svg.rs
index fe496807..9685e454 100644
--- a/crates/typst/src/visualize/image/svg.rs
+++ b/crates/typst/src/visualize/image/svg.rs
@@ -11,6 +11,7 @@ use crate::diag::{format_xml_like_error, StrResult};
 use crate::foundations::Bytes;
 use crate::layout::Axes;
 use crate::text::{FontVariant, FontWeight};
+use crate::visualize::Image;
 use crate::World;
 
 /// A decoded SVG.
@@ -29,8 +30,7 @@ impl SvgImage {
     /// Decode an SVG image without fonts.
     #[comemo::memoize]
     pub fn new(data: Bytes) -> StrResult<SvgImage> {
-        let opts = usvg::Options::default();
-        let tree = usvg::Tree::from_data(&data, &opts).map_err(format_usvg_error)?;
+        let tree = usvg::Tree::from_data(&data, &options()).map_err(format_usvg_error)?;
         Ok(Self(Arc::new(Repr {
             data,
             size: tree_size(&tree),
@@ -47,13 +47,8 @@ impl SvgImage {
         world: Tracked<dyn World + '_>,
         families: &[String],
     ) -> StrResult<SvgImage> {
-        // Disable usvg's default to "Times New Roman". Instead, we default to
-        // the empty family and later, when we traverse the SVG, we check for
-        // empty and non-existing family names and replace them with the true
-        // fallback family. This way, we can memoize SVG decoding with and without
-        // fonts if the SVG does not contain text.
-        let opts = usvg::Options { font_family: String::new(), ..Default::default() };
-        let mut tree = usvg::Tree::from_data(&data, &opts).map_err(format_usvg_error)?;
+        let mut tree =
+            usvg::Tree::from_data(&data, &options()).map_err(format_usvg_error)?;
         let mut font_hash = 0;
         if tree.has_text_nodes() {
             let (fontdb, hash) = load_svg_fonts(world, &mut tree, families);
@@ -126,6 +121,22 @@ impl Hash for Repr {
     }
 }
 
+/// The conversion options.
+fn options() -> usvg::Options {
+    // Disable usvg's default to "Times New Roman". Instead, we default to
+    // the empty family and later, when we traverse the SVG, we check for
+    // empty and non-existing family names and replace them with the true
+    // fallback family. This way, we can memoize SVG decoding with and without
+    // fonts if the SVG does not contain text.
+    usvg::Options {
+        font_family: String::new(),
+        // We override the DPI here so that we get the correct the size when
+        // scaling the image to its natural size.
+        dpi: Image::DEFAULT_DPI as f32,
+        ..Default::default()
+    }
+}
+
 /// Discover and load the fonts referenced by an SVG.
 fn load_svg_fonts(
     world: Tracked<dyn World + '_>,