mirror of
https://github.com/storytold/pdfcraft.git
synced 2026-11-06 18:52:36 +00:00
M2: area-average shrunk scans and smooth enlarged ones
Scanned pages looked thin and broken at screen size. An image that does not ask for /Interpolate (most scans) was pre-shrunk with nearest-neighbour since about one sample in ten and dropped most strokes of its text, and the nearest sampler then read the shrunk raster again at a fractional offset. - Shrinking (vendored hayro patch 21): non-interpolated images outside Type 3 glyphs are area-averaged (area_shrink: each device pixel the coverage-weighted mean of the samples it covers, 16.16 fixed point whose weights sum to exactly one, premultiplied when masked) to their snapped device rectangle and drawn 1:1. /Interpolate governs smoothing when an image is enlarged (ISO 32000-2 8.9.5.3); averaging adds nothing between samples and drops none. PicScale's Box/Area filters were not usable: shrinking, they average a window four target pixels wide. - Enlarging (patch 28): an image of page-scan size (512+ samples on its shorter side) is drawn bilinearly when enlarged, so a zoomed scan has no staircase edges; icons, QR codes and pixel art keep their hard pixels. - A snapped image whose shrink failed keeps the full transform instead of being drawn at its source size (latent in the interpolated path too). Measured on a 13-page 300 dpi scan at 96, 120 and 192 dpi: every rendered sample is within 1 of the exact area average (it was up to 255 off, 17-23% of samples by more than 1). Bilinear zoom costs nothing measurable (four 1024 px tiles at 400%: 50 ms, nearest 48 ms; bicubic was 267 ms). Review fixes: - a tile snaps a shrunk image's half-pixel edges where the whole page does - a scan shrunk one way is smoothed the way it is stretched - a smaller soft mask doesn't let transparent colour through the area shrink Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
kronbiiandClaude Opus 5.5 committed
1 parent
a4d148bf48
commit
ff227b770c
6 files changed
+571
-66
No files matched your search
+2
-1
@@ -23,7 +23,7 @@
|
||||
|
||||
**Stage: alpha** · next: beta, ~25% (ready for real work 50% → 75%) and ~400–700 h away
|
||||
|
||||
> **Last reviewed:** 2026-10-29 | **Last updated:** 2026-10-29 | **Change:** minor (M7 #2517: a narrowed then widened paragraph keeps the typed spaces the wrap consumed, after save and reopen; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-29 | **Last updated:** 2026-10-29 | **Change:** minor (M2: scans keep their strokes at screen size and are smoothed when zoomed; M7 #2517: a narrowed then widened paragraph keeps the typed spaces the wrap consumed, after save and reopen; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-27 | **Last updated:** 2026-10-27 | **Change:** minor (M10: OCR model downloads use verified ONNX pairs with legacy discovery fallback; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-25 | **Last updated:** 2026-10-25 | **Change:** minor (M14 localization current state: Arabic logical catalog, explicit RTL label direction and mirrored stock chrome; native/browser RTL workflows unqualified. Native release Hans face via LXGW WenKai Lite; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-23 · **Last updated:** 2026-10-23 · **Change:** minor (M7, #2636 partial: pasting added text boxes and images in the engine and through `content_copy`; estimates unchanged)
|
||||
@@ -257,6 +257,7 @@ Ranked; detail and the full milestone table in [docs/roadmap.md](docs/roadmap.md
|
||||
- **2026-10-26 (M2, PR #4172):** Reconciled current main, preserving its Japanese serif/weight selection and both branches' progress records. Installed CJK viewer and preview fallback remains display-only; estimates unchanged.
|
||||
|
||||
- **2026-10-29 (M7, #2517):** Widening a paragraph rejoins a wrap with the number of spaces that wrap consumed, including a typed run and none for a Chinese or Japanese word split. The count survives an incremental save and a full rewrite, and a width-only `text_edit` uses it. Estimates unchanged.
|
||||
- **2026-10-29 (M2):** Scanned pages look like the paper again. An image that doesn't ask for /Interpolate was shrunk by nearest-neighbour (since #635), which kept one sample in N per axis, so the text of a 300 dpi scan at 100% lost most of its strokes; it is now area-averaged to whole device pixels (on a 13-page 300 dpi scan every sample at 96–192 dpi is within 1 of the exact area average; it was up to 255 off). Enlarged, an image of page-scan size is smoothed bilinearly, so a zoomed scan has no staircases; icons, QR codes and pixel art keep hard pixels. Estimates unchanged.
|
||||
- **2026-10-26 (M11, JBIG2 symbols):** Lossless JBIG2 now stores each distinct mark of a black-and-white image once, as a symbol, and places its exact copies, when that is smaller than the generic region: on six pages of rendered 300 dpi text 28% of the generic region's size and 17% of Group 4's, and the same bits back through two independent decoders (hayro-jbig2 and Poppler). Scans, whose letters are seldom identical, gain little until refinement coding; a dictionary shared between pages (`/JBIG2Globals`) is next. Estimates unchanged.
|
||||
- **2026-10-25 (M5, comment replies):** One reply can be deleted from the Comments panel without its thread. On the selected card each reply has its own "…" menu (Edit, Copy text, Delete); Delete takes that reply with its pop-up and the replies to it, as `comment_delete` already did headlessly, and the comment and its other replies stay. The card stays selected, and the panel's editor stays open, on what they were on though the delete moved it in `/Annots`. Undo and Redo close the editor, since either can move what it was open on. A locked reply can't be deleted. Estimates unchanged.
|
||||
- **2026-10-24 (M10, #4158 partial):** Word, HTML, RTF and PowerPoint export keep a block that continues another on its line (an italic case name, a link, a superscript) as a run in its own style inside that paragraph, with the source's word space, instead of a paragraph of its own. Style changes that wrap onto the next line still split paragraphs (the reading blocks keep style boundaries). Estimates unchanged.
|
||||
|
||||
@@ -6436,8 +6436,8 @@ fn draw_minified_image(
|
||||
/// making scans, screenshots and rasterized text look soft.
|
||||
#[test]
|
||||
fn non_interpolated_images_are_not_smoothed_when_minified() {
|
||||
// Top band K W R G, bottom band B Y M C. Nearest 8×4 → 4×2 sampling picks each block's
|
||||
// pixel, so every device pixel is one palette colour; Catmull-Rom blends across blocks.
|
||||
// Top band K W R G, bottom band B Y M C. Shrinking 8×4 → 4×2 averages each 2×2 block,
|
||||
// so every device pixel is one palette colour; Catmull-Rom blends across blocks.
|
||||
let palette: [[u8; 3]; 8] = [[0, 0, 0], [255, 255, 255], [255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 255, 0], [255, 0, 255], [0, 255, 255]];
|
||||
let rgb: Vec<u8> = block_source(palette).concat();
|
||||
let exact: Vec<[u8; 4]> = palette.iter().map(|c| [c[0], c[1], c[2], 255]).collect();
|
||||
@@ -6455,38 +6455,76 @@ fn non_interpolated_images_are_not_smoothed_when_minified() {
|
||||
assert!(pixels(&enabled).iter().any(|px| px[..3].iter().any(|&c| (32..=223).contains(&c))), "/Interpolate true made no mid-tone");
|
||||
}
|
||||
|
||||
/// A shrunk, interpolated image whose edges fall between device pixels is snapped to whole
|
||||
/// pixels (shrunk once to that size, drawn 1:1). Before, the shrunk raster was sampled a second
|
||||
/// time at the fractional offset, so every edge of a scan was a blended, soft pixel.
|
||||
/// A shrunk image whose edges fall between device pixels is snapped to whole pixels (shrunk
|
||||
/// once to that size, drawn 1:1), interpolated or not. Before, the shrunk raster was sampled a
|
||||
/// second time at the fractional offset, so every edge of a scan was a blended, soft pixel.
|
||||
#[test]
|
||||
fn interpolated_minified_image_edges_snap_to_whole_pixels() {
|
||||
let black = vec![0u8; 40 * 40];
|
||||
let content = "q 10 0 0 10 0.4 0.4 cm /Im0 Do Q";
|
||||
let image = hex(&black);
|
||||
let pdf = format!(
|
||||
"%PDF-1.4\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\
|
||||
fn minified_image_edges_snap_to_whole_pixels() {
|
||||
for interpolate in ["/Interpolate true ", ""] {
|
||||
let black = vec![0u8; 40 * 40];
|
||||
let content = "q 10 0 0 10 0.4 0.4 cm /Im0 Do Q";
|
||||
let image = hex(&black);
|
||||
let pdf = format!(
|
||||
"%PDF-1.4\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\
|
||||
2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n\
|
||||
3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 11 11] /Contents 4 0 R /Resources << /XObject << /Im0 5 0 R >> >> >> endobj\n\
|
||||
4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\
|
||||
5 0 obj << /Type /XObject /Subtype /Image /Width 40 /Height 40 /ColorSpace /DeviceGray /BitsPerComponent 8 /Interpolate true /Filter /ASCIIHexDecode /Length {} >> stream\n{image}>\nendstream endobj\n\
|
||||
5 0 obj << /Type /XObject /Subtype /Image /Width 40 /Height 40 /ColorSpace /DeviceGray /BitsPerComponent 8 {interpolate}/Filter /ASCIIHexDecode /Length {} >> stream\n{image}>\nendstream endobj\n\
|
||||
trailer << /Root 1 0 R >>\n%%EOF\n",
|
||||
content.len(),
|
||||
image.len() + 1
|
||||
);
|
||||
let mut renderer = PageRenderer::new(Arc::new(pdf.into_bytes()), RenderConfig::default());
|
||||
let page = renderer.render(RenderRequest { page: 0, kind: RequestKind::Pixels, scale: 1.0, ..Default::default() });
|
||||
assert!(page.error.is_none(), "{:?}", page.error);
|
||||
assert_eq!((page.width, page.height), (11, 11));
|
||||
let gray = |x: usize, y: usize| page.rgba[(y * 11 + x) * 4];
|
||||
// Device box 0.4..10.4 × 0.6..10.6 snaps to columns 0..10 and rows 1..11.
|
||||
for y in 1..11 {
|
||||
for x in 0..10 {
|
||||
assert_eq!(gray(x, y), 0, "pixel ({x}, {y}) inside the snapped image");
|
||||
content.len(),
|
||||
image.len() + 1
|
||||
);
|
||||
let mut renderer = PageRenderer::new(Arc::new(pdf.into_bytes()), RenderConfig::default());
|
||||
let page = renderer.render(RenderRequest { page: 0, kind: RequestKind::Pixels, scale: 1.0, ..Default::default() });
|
||||
assert!(page.error.is_none(), "{:?}", page.error);
|
||||
assert_eq!((page.width, page.height), (11, 11));
|
||||
let gray = |x: usize, y: usize| page.rgba[(y * 11 + x) * 4];
|
||||
// Device box 0.4..10.4 × 0.6..10.6 snaps to columns 0..10 and rows 1..11.
|
||||
for y in 1..11 {
|
||||
for x in 0..10 {
|
||||
assert_eq!(gray(x, y), 0, "{interpolate:?}: pixel ({x}, {y}) inside the snapped image");
|
||||
}
|
||||
assert_eq!(gray(10, y), 255, "{interpolate:?}: column 10 is outside the snapped image (row {y})");
|
||||
}
|
||||
for x in 0..11 {
|
||||
assert_eq!(gray(x, 0), 255, "{interpolate:?}: row 0 is outside the snapped image (column {x})");
|
||||
}
|
||||
assert_eq!(gray(10, y), 255, "column 10 is outside the snapped image (row {y})");
|
||||
}
|
||||
for x in 0..11 {
|
||||
assert_eq!(gray(x, 0), 255, "row 0 is outside the snapped image (column {x})");
|
||||
}
|
||||
|
||||
/// A tile draws exactly the pixels of the whole page under it, also where a shrunk image's
|
||||
/// edge falls on a half pixel. The edges were snapped by rounding halves away from zero, so
|
||||
/// the tile at x = 5, which moves the left edge from 2.5 to -2.5, snapped it one pixel left of
|
||||
/// where the whole page did and shifted every column of the image.
|
||||
#[test]
|
||||
fn a_tile_snaps_half_pixel_image_edges_as_the_whole_page_does() {
|
||||
for interpolate in ["/Interpolate true ", ""] {
|
||||
let ramp: Vec<u8> = (0..40 * 40).map(|i| (i % 40 * 6) as u8).collect();
|
||||
let content = "q 10 0 0 10 2.5 2.5 cm /Im0 Do Q";
|
||||
let image = hex(&ramp);
|
||||
let pdf = format!(
|
||||
"%PDF-1.4\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\
|
||||
2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n\
|
||||
3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 20 20] /Contents 4 0 R /Resources << /XObject << /Im0 5 0 R >> >> >> endobj\n\
|
||||
4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\
|
||||
5 0 obj << /Type /XObject /Subtype /Image /Width 40 /Height 40 /ColorSpace /DeviceGray /BitsPerComponent 8 {interpolate}/Filter /ASCIIHexDecode /Length {} >> stream\n{image}>\nendstream endobj\n\
|
||||
trailer << /Root 1 0 R >>\n%%EOF\n",
|
||||
content.len(),
|
||||
image.len() + 1
|
||||
);
|
||||
let mut renderer = PageRenderer::new(Arc::new(pdf.into_bytes()), RenderConfig::default());
|
||||
let req = RenderRequest { page: 0, kind: RequestKind::Pixels, scale: 1.0, ..Default::default() };
|
||||
let full = renderer.render(req);
|
||||
let tile = Tile { x: 5, y: 5, w: 10, h: 10 };
|
||||
let part = renderer.render(RenderRequest { tile: Some(tile), ..req });
|
||||
assert!(full.error.is_none() && part.error.is_none(), "{:?} {:?}", full.error, part.error);
|
||||
assert_eq!((full.width, part.width, part.height), (20, 10, 10));
|
||||
for y in 0..10usize {
|
||||
for x in 0..10usize {
|
||||
let whole = full.rgba[((y + 5) * 20 + x + 5) * 4];
|
||||
assert_eq!(part.rgba[(y * 10 + x) * 4], whole, "{interpolate:?}: pixel ({x}, {y}) of the tile");
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -6570,28 +6608,132 @@ fn non_interpolated_gray_images_are_not_smoothed_when_minified() {
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn non_integer_minification_samples_the_same_columns() {
|
||||
// Ten columns drawn six pixels wide: centre sampling reads columns 0, 2, 4, 5, 7 and 9.
|
||||
fn non_integer_minification_averages_the_columns_each_pixel_covers() {
|
||||
// Ten columns drawn six pixels wide: each pixel covers 10/6 columns. Pixel 0 is 60% of
|
||||
// column 0 and 40% of column 1 (0.6 × 10 + 0.4 × 20 = 14), pixel 1 is 20% of column 1,
|
||||
// all of column 2's 60% and 20% of column 3 (30), and so on. Nearest-neighbour read whole
|
||||
// columns 0, 2, 4, 5, 7 and 9 instead, dropping the four others.
|
||||
let ramp = [10u8, 20, 30, 40, 50, 60, 70, 80, 90, 100];
|
||||
let page = draw_minified_image("DeviceGray", (10, 1), None, &ramp, None, (6, 1));
|
||||
let grays: Vec<u8> = page.rgba.as_chunks::<4>().0.iter().map(|px| px[0]).collect();
|
||||
assert_eq!(grays, [10, 30, 50, 60, 80, 100]);
|
||||
assert_eq!(grays, [14, 30, 46, 64, 80, 96]);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn non_integer_minification_samples_colour_and_mask_from_the_same_columns() {
|
||||
// Ten gray columns with a soft mask, drawn six pixels wide: columns 0, 2, 4, 5, 7 and 9 are
|
||||
// sampled, so the mask must pick the same columns as the colours.
|
||||
fn non_integer_minification_averages_colour_and_mask_together() {
|
||||
// Ten gray columns with a soft mask, drawn six pixels wide. Colours are averaged
|
||||
// premultiplied, over the same columns as the mask, and composited over the white page.
|
||||
// Pixel 0: 60% of column 0 (gray 10, opaque) and 40% of column 1 (transparent) is
|
||||
// premultiplied gray 6 at alpha 153, so 6 + 255 × (1 − 153/255) = 108. Pixels 1 and 4
|
||||
// cover only transparent columns and show the page.
|
||||
let ramp = [10u8, 20, 30, 40, 50, 60, 70, 80, 90, 100];
|
||||
let mask = [255u8, 0, 0, 0, 255, 255, 0, 0, 0, 255];
|
||||
let page = draw_minified_image("DeviceGray", (10, 1), None, &ramp, Some(&mask), (6, 1));
|
||||
// Opaque columns keep their gray; transparent columns show the white page.
|
||||
assert_eq!(
|
||||
page.rgba.as_chunks::<4>().0.to_vec(),
|
||||
vec![[10, 10, 10, 255], [255, 255, 255, 255], [50, 50, 50, 255], [60, 60, 60, 255], [255, 255, 255, 255], [100, 100, 100, 255]]
|
||||
vec![[108, 108, 108, 255], [255, 255, 255, 255], [132, 132, 132, 255], [138, 138, 138, 255], [255, 255, 255, 255], [162, 162, 162, 255]]
|
||||
);
|
||||
}
|
||||
|
||||
/// A scan shown at screen size: a stroke one sample wide in an image drawn at a third of its
|
||||
/// resolution. Nearest-neighbour read columns 1, 4 and 7 and lost the stroke in column 3, which
|
||||
/// is how the text of 300 dpi scans broke up at 100% zoom. Area averaging keeps it as the
|
||||
/// third of pixel 1 it covers.
|
||||
#[test]
|
||||
fn non_interpolated_minification_keeps_thin_strokes() {
|
||||
let mut stroke = [255u8; 9];
|
||||
stroke[3] = 0;
|
||||
let page = draw_minified_image("DeviceGray", (9, 1), None, &stroke, None, (3, 1));
|
||||
let grays: Vec<u8> = page.rgba.as_chunks::<4>().0.iter().map(|px| px[0]).collect();
|
||||
assert_eq!(grays, [255, 170, 255]);
|
||||
}
|
||||
|
||||
/// A zoomed scan: an image of page-scan size drawn three times its resolution without
|
||||
/// `/Interpolate`. Nearest-neighbour turned each sample into a 3 × 3 block, so the edges of
|
||||
/// scanned text were staircases; it is smoothed (bilinearly) instead, as browsers draw it.
|
||||
#[test]
|
||||
fn large_non_interpolated_images_are_smoothed_when_enlarged() {
|
||||
let side = 512usize;
|
||||
let half: Vec<u8> = (0..side * side).map(|i| if i % side < side / 2 { 0 } else { 255 }).collect();
|
||||
let page = draw_minified_image("DeviceGray", (512, 512), None, &half, None, (1536, 1536));
|
||||
let row = 768 * 1536;
|
||||
let edge: Vec<u8> = (760..776).map(|x| page.rgba[(row + x) * 4]).collect();
|
||||
assert!(edge.iter().any(|v| (32..=223).contains(v)), "a hard edge, not a smoothed one: {edge:?}");
|
||||
assert_eq!((page.rgba[(row + 700) * 4], page.rgba[(row + 836) * 4]), (0, 255), "away from the edge");
|
||||
}
|
||||
|
||||
/// Small images (icons, QR codes, pixel art) keep their hard pixels when enlarged.
|
||||
#[test]
|
||||
fn small_non_interpolated_images_keep_hard_pixels_when_enlarged() {
|
||||
let checker: Vec<u8> = (0..64).map(|i| if (i / 8 + i % 8) % 2 == 0 { 0 } else { 255 }).collect();
|
||||
let page = draw_minified_image("DeviceGray", (8, 8), None, &checker, None, (80, 80));
|
||||
assert!(page.rgba.as_chunks::<4>().0.iter().all(|px| px[0] == 0 || px[0] == 255), "a small image was smoothed");
|
||||
}
|
||||
|
||||
/// A scan shrunk across and stretched down (a fax shown at screen size, zoomed in): the shrunk
|
||||
/// side is still averaged, and the stretched side is smoothed as a scan enlarged both ways
|
||||
/// is. Shrinking one side kept the nearest sampler for the other, so its edges were staircases.
|
||||
#[test]
|
||||
fn a_scan_shrunk_one_way_is_smoothed_the_way_it_is_stretched() {
|
||||
// Top half black; bottom half columns alternating black and white, which averages to gray.
|
||||
let side = 512usize;
|
||||
let source: Vec<u8> = (0..side * side).map(|i| if i / side < side / 2 || i % 2 == 0 { 0 } else { 255 }).collect();
|
||||
let page = draw_minified_image("DeviceGray", (512, 512), None, &source, None, (256, 1536));
|
||||
let gray = |x: usize, y: usize| page.rgba[(y * 256 + x) * 4];
|
||||
let edge: Vec<u8> = (760..776).map(|y| gray(128, y)).collect();
|
||||
assert!(edge.iter().any(|v| (16..=112).contains(v)), "a hard edge down the stretched side: {edge:?}");
|
||||
assert_eq!(gray(128, 700), 0, "above the edge");
|
||||
assert!((0..256).all(|x| gray(x, 836).abs_diff(128) <= 1), "the shrunk side is averaged at 1:1");
|
||||
}
|
||||
|
||||
/// An image shrunk across and stretched down (a 204 × 98 dpi fax shown at screen size): the
|
||||
/// shrunk side is still averaged. The area shrink was asked to enlarge the other side too,
|
||||
/// refused, and the image fell back to its source resolution and the nearest sampler, which
|
||||
/// read single columns.
|
||||
#[test]
|
||||
fn an_image_shrunk_one_way_and_stretched_the_other_is_still_averaged() {
|
||||
let ramp: Vec<u8> = [10u8, 20, 30, 40, 50, 60, 70, 80].repeat(2);
|
||||
let page = draw_minified_image("DeviceGray", (8, 2), None, &ramp, None, (4, 4));
|
||||
let grays: Vec<u8> = page.rgba.as_chunks::<4>().0.iter().map(|px| px[0]).collect();
|
||||
assert_eq!(grays, [15, 35, 55, 75].repeat(4), "each pixel is the mean of the two columns it covers");
|
||||
}
|
||||
|
||||
/// Averaging straight colours would tint an opaque sample with the colour of a transparent
|
||||
/// neighbour: white beside transparent black, shrunk to one pixel, came out gray over the white
|
||||
/// page. Premultiplied, the transparent sample adds nothing but its transparency.
|
||||
#[test]
|
||||
fn non_interpolated_minification_does_not_bleed_transparent_colour() {
|
||||
let page = draw_minified_image("DeviceGray", (2, 1), None, &[255, 0], Some(&[255, 0]), (1, 1));
|
||||
assert_eq!(page.rgba.as_chunks::<4>().0.to_vec(), vec![[255, 255, 255, 255]]);
|
||||
}
|
||||
|
||||
/// The same with a soft mask smaller than its image: red, red, blue, blue under a mask that is
|
||||
/// opaque, then transparent, shrunk to one pixel, is half-covered red over the white page.
|
||||
/// Colour and mask were averaged apart and multiplied, so the transparent blue came through
|
||||
/// as purple (191, 127, 191).
|
||||
#[test]
|
||||
fn a_smaller_soft_mask_does_not_bleed_transparent_colour_when_minified() {
|
||||
let content = "q 1 0 0 1 0 0 cm /Im0 Do Q";
|
||||
let (image, mask) = ("FF0000FF00000000FF0000FF", "FF00");
|
||||
let pdf = format!(
|
||||
"%PDF-1.4\n1 0 obj << /Type /Catalog /Pages 2 0 R >> endobj\n\
|
||||
2 0 obj << /Type /Pages /Kids [3 0 R] /Count 1 >> endobj\n\
|
||||
3 0 obj << /Type /Page /Parent 2 0 R /MediaBox [0 0 1 1] /Contents 4 0 R /Resources << /XObject << /Im0 5 0 R >> >> >> endobj\n\
|
||||
4 0 obj << /Length {} >> stream\n{content}\nendstream endobj\n\
|
||||
5 0 obj << /Type /XObject /Subtype /Image /Width 4 /Height 1 /ColorSpace /DeviceRGB /BitsPerComponent 8 /SMask 6 0 R /Filter /ASCIIHexDecode /Length {} >> stream\n{image}>\nendstream endobj\n\
|
||||
6 0 obj << /Type /XObject /Subtype /Image /Width 2 /Height 1 /ColorSpace /DeviceGray /BitsPerComponent 8 /Filter /ASCIIHexDecode /Length {} >> stream\n{mask}>\nendstream endobj\n\
|
||||
trailer << /Root 1 0 R >>\n%%EOF\n",
|
||||
content.len(),
|
||||
image.len() + 1,
|
||||
mask.len() + 1
|
||||
);
|
||||
let mut renderer = PageRenderer::new(Arc::new(pdf.into_bytes()), RenderConfig::default());
|
||||
let page = renderer.render(RenderRequest { page: 0, kind: RequestKind::Pixels, scale: 1.0, ..Default::default() });
|
||||
assert!(page.error.is_none(), "{:?}", page.error);
|
||||
let [r, g, b, _] = page.rgba.as_chunks::<4>().0.first().copied().unwrap_or_default();
|
||||
assert!(r >= 254 && g.abs_diff(128) <= 1 && b == g, "half-covered red over white, got ({r}, {g}, {b})");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn non_interpolated_alpha_images_keep_their_colours_and_mask_when_minified() {
|
||||
let palette: [[u8; 3]; 8] = [[0, 0, 0], [255, 255, 255], [255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 255, 0], [255, 0, 255], [0, 255, 255]];
|
||||
@@ -7716,7 +7858,7 @@ fn stitched_render_partitions_match_a_whole_render() {
|
||||
/// and square caps, a sharp miter, a rotated and a stretched matrix, a hairline, dashes, a
|
||||
/// curve, text, and a page shown rotated.
|
||||
///
|
||||
/// Images are skipped the same way (vendored patch (10)): an image on a tile, one crossing
|
||||
/// Images are skipped the same way (vendored patch (11)): an image on a tile, one crossing
|
||||
/// tile lines, and one fully off the page.
|
||||
#[test]
|
||||
fn skipping_paths_off_the_canvas_changes_no_pixel() {
|
||||
|
||||
+2
-1
@@ -22,7 +22,7 @@
|
||||
> **Last reviewed:** 2026-10-26 · **Last updated:** 2026-10-26 · **Change:** patch (M2, PR #4172: document installed-font helpers and regression contracts; behavior and estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-26 · **Last updated:** 2026-10-26 · **Change:** patch (M2, PR #4172: print-preview font policy matches output, with isolated caches and GBK switching regressions; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-26 · **Last updated:** 2026-10-26 · **Change:** patch (M2, PR #4172: reconcile current main and retain Japanese serif/weight selection alongside display-only CJK fallback; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-29 | **Last updated:** 2026-10-29 | **Change:** minor (M7 #2517: a narrowed then widened paragraph keeps the typed spaces the wrap consumed, after save and reopen; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-29 | **Last updated:** 2026-10-29 | **Change:** minor (M2: scans keep their strokes at screen size and are smoothed when zoomed; M7 #2517: a narrowed then widened paragraph keeps the typed spaces the wrap consumed, after save and reopen; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-27 | **Last updated:** 2026-10-27 | **Change:** minor (M10: OCR model downloads use verified ONNX pairs with legacy discovery fallback; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-25 | **Last updated:** 2026-10-25 | **Change:** minor (M5: one reply in the Comments panel is edited, copied or deleted from its own "…" menu; estimates unchanged)
|
||||
> **Last reviewed:** 2026-10-26 · **Last updated:** 2026-10-26 · **Change:** minor (M11: lossless JBIG2 stores each distinct mark once, as a symbol; estimates unchanged)
|
||||
@@ -321,6 +321,7 @@ Estimates unchanged.
|
||||
| 2026-10-23 | minor | M11: black-and-white images (1-bit, and image masks) are stored losslessly as CCITT Group 4 when that is smaller: PDF Optimizer ▸ Images ▸ Monochrome Images (on by default, so Reduce File Size too), `doc_optimize` `mono`; `pdfcraft_filters::encode_g4`. JBIG2 and monochrome downsampling not yet. Estimates unchanged. |
|
||||
| 2026-10-23 | minor | M6: calculation checks: a formula reading a field the form doesn't have, and a script building the names it reads, are flagged by `form_calculation_order` and on the Calculation order page (a red mark, a card naming the broken formulas, a note for built names); `form_calculation_order` gives each formula and `form_fields` a custom Calculate script. Estimates unchanged. |
|
||||
| 2026-10-29 | minor | M7 #2517: widening a paragraph rejoins a wrap with the spaces that wrap consumed, including a typed run and none for a Chinese or Japanese word split, after an incremental save and a full rewrite. Estimates unchanged. |
|
||||
| 2026-10-29 | minor | M2: non-interpolated images drawn below their resolution are area-averaged to whole device pixels instead of nearest-neighbour (scanned text kept one sample in N per axis); page-scan-size images are smoothed when enlarged. Estimates unchanged. |
|
||||
| 2026-10-27 | minor | M10: OCR stages SHA-256-verified ONNX models after the upstream RTEN URL brownout; discovery prefers a complete ONNX pair per directory, falls back to a complete RTEN pair, and never mixes formats or directories; language/web limits and estimates unchanged. |
|
||||
| 2026-10-24 | minor | M3, #3970 follow-up: the button release given back after a window move or resize goes before the pointer leaving, so hover doesn't stick. Estimates unchanged. |
|
||||
| 2026-10-23 | minor | M8, #3205 follow-up: phone-pattern redaction includes the +39 country code on labelled Italian geographic numbers with a slash; dates, protocol identifiers and whole-run guards remain. Estimates unchanged. |
|
||||
|
||||
@@ -734,8 +734,8 @@ milestone = "M2"
|
||||
status = "partial"
|
||||
commands = []
|
||||
tools = ["page_render"]
|
||||
evidence = ["crates/render/src/raster/tests.rs::renders_a_blue_rectangle", "crates/automation/tests/it/automation.rs::open_inspect_render_and_find", "crates/render/src/raster/tests.rs::user_unit_scales_the_page_and_everything_on_it"]
|
||||
notes = "hayro bootstrap renderer; own DisplayList devices in M2.6. Pages with /UserUnit are shown at their size in points, as Acrobat reads the value"
|
||||
evidence = ["crates/render/src/raster/tests.rs::renders_a_blue_rectangle", "crates/automation/tests/it/automation.rs::open_inspect_render_and_find", "crates/render/src/raster/tests.rs::user_unit_scales_the_page_and_everything_on_it", "crates/render/src/raster/tests.rs::non_interpolated_minification_keeps_thin_strokes", "crates/render/src/raster/tests.rs::minified_image_edges_snap_to_whole_pixels", "crates/render/src/raster/tests.rs::large_non_interpolated_images_are_smoothed_when_enlarged", "crates/render/src/raster/tests.rs::small_non_interpolated_images_keep_hard_pixels_when_enlarged"]
|
||||
notes = "hayro bootstrap renderer; own DisplayList devices in M2.6. Pages with /UserUnit are shown at their size in points, as Acrobat reads the value. An image drawn below its resolution is shrunk once to whole device pixels: by area averaging unless it asks for /Interpolate (then Catmull-Rom), so the text of a 300 dpi scan at 100% keeps its strokes (it was nearest-neighbour, which dropped them). Enlarged, an image of page-scan size (512 samples or more on its shorter side) is smoothed bilinearly whatever its /Interpolate, so a zoomed scan has no staircases; smaller images keep their hard pixels. Not measured against Acrobat"
|
||||
|
||||
[[feature]]
|
||||
id = "view.render-shadings-patterns"
|
||||
|
||||
Vendored
+1
-1
@@ -40,7 +40,7 @@ Other resource categories retain their existing parent-chain lookup.
|
||||
| Crate | Version | Licence | Patches | Test |
|
||||
|---|---|---|---|---|
|
||||
| hayro-interpret | 0.7.0 | Apache-2.0 OR MIT | (1) select `/AP /N` appearance states via `/AS` (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to `/Encoding` glyph names, then printable ASCII; (3) `Type3Glyph::advance_width`; (4) `InterpreterSettings::ocg_overrides` for viewer layer toggles; (5) soft masks whose group lacks `/CS` were dropped (Chrome gradient text); (6) paint nesting cap `MAX_PAINT_NESTING` for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in `/W` and `/W2` clamped to `MAX_CID` (a `0 4294967295 w` range hung); (8) `InterpreterSettings::hide_comments` skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) `image_size_ok`: images and masks over 2^28 pixels are not decoded (a fuzzed `/W 4294967295` stencil mask allocated 4 GiB); (10) `MAX_NESTED_PAINTS`: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (`Charset::glyph_id` returns `Option`); (12) annotation appearances whose `/BBox` or `/Matrix` collapses to a line or a point are skipped (mapping them onto `/Rect` divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its `/QuadPoints`, `/C` and `/CA` with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font `/Widths` (`read_widths`): entries only for codes 0–255 (one entry per code below `/FirstChar` was allocated: `4294967295` asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a `/LastChar` below `/FirstChar` gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto `/Rect` corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from `/Rect` was drawn shifted); (16) a clip set by `W`/`W*` takes effect after whichever path-painting operator ends the path (`S`, `f`, `B`, … as well as `n`; ISO 32000-2 §8.5.4): it was applied only on `n`, so `re W* S` clipped nothing and the clip stayed pending for a later, unrelated `n`; (17) `InterpreterSettings::cancelled`: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) `MAX_PAGE_CONTENT`: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (`decode_content`; once it runs out, `InterpreterWarning::ContentTruncated` is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most `MAX_SAMPLE_ENTRIES` (2^21) table points and `MAX_SAMPLE_VALUES` (2^23) samples, read no further than `/Size` asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (`ShadingPattern::encode_within`, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (`TriangleSampler`), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most `MAX_MESH_PATCHES` (2^16) patches and `MAX_MESH_TRIANGLES` (2^20) triangles, a lattice needs two to 2^19 vertices per row (`/VerticesPerRow 0` pushed empty rows for ever; one huge row read the whole stream), and a page samples at most `MAX_PAGE_MESH_TRIANGLES` (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws `.nLine truncated
|
||||
| hayro | 0.7.1 | Apache-2.0 OR MIT | (1) `RenderSettings::{x_offset, y_offset}` to render one tile of a large page; (2) images over `MAX_IMAGE_PIXELS` (2^28) are skipped (a fuzzed `/W 4294967295` hung in resampling); (3) `tiling_cell_scale` keeps a tiling pattern's cell pixmap within 3000 px a side whatever its `/XStep`/`/YStep` (a fuzzed step far beyond `/BBox` allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed `/LW 9223372036854775807` allocated 10 GB in stroke expansion); (5) `usable_dash_pattern`: a stroke whose dash pattern could cut it into more than `MAX_DASHES_PER_STROKE` (250,000) pieces, counting kurbo's restart of the pattern at every subpath and `ClosePath`, is drawn solid (a fuzzed `/D [[11] 0]` on a line from x = 92234775807 allocated over 5 GB, `[0.000001] 0 d` on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid (`[-1 -1] 0 d` looped for ever in kurbo); (7) `render_size` and `render_into`: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); `render` is `render_into` on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) `may_paint_canvas`: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); `set_skip_offscreen_paths` turns this off for the regression test; (10) `/Interpolate` governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour, planned without the Catmull-Rom filter limits (`image_resampling_size` takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) `may_paint_rect`: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (`InterpreterSettings::preview`: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under `PREVIEW_MIN_IMAGE_AREA` device pixels, while final renders always draw them; (13) reading colours (#897): `RenderSettings::color_replacement` takes a `ColorReplacement` (text colour, background, `only_black`, `line_art`) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; `render_pdf` gains a `color_replacement: Option<ColorReplacement>` argument (pass `None` for true colours); the upstream-only `[dev-dependencies.log]` entry is removed from `Cargo.toml` (log is a normal dependency since the logging patch), so `cargo test -p hayro` runs this crate's unit tests; (14) a shrunk, interpolated image whose axes lie along the device's (upright, mirrored or a quarter turn, as on a page under `/Rotate`) is snapped to whole device pixels: shrunk once to its rounded device rectangle and drawn 1:1 from the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (`snap_minified_image`; scans looked soft next to Acrobat, and until #5530 a turned one still did, up to 200 grey levels off the same page drawn upright). An image counts as axis-aligned when the terms the snap drops move its far edges by at most a hundredth of a device pixel (`SNAP_SKEW`), not by a tolerance on the coefficients, which let a slight skew of a very wide image through; and the snapped 1:1 transform is used only when the shrink produced the snapped size, so a failed resize keeps the full transform instead of drawing the source raster at its source size; (15) a soft mask of another size than its image is filtered when shrunk, whatever its `/Interpolate` (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroke thinner than the shrink factor; (16) under a soft mask or a blend mode, axial and radial shadings are sampled instead of drawn as native gradients (#Line truncated
|
||||
| hayro | 0.7.1 | Apache-2.0 OR MIT | (1) `RenderSettings::{x_offset, y_offset}` to render one tile of a large page; (2) images over `MAX_IMAGE_PIXELS` (2^28) are skipped (a fuzzed `/W 4294967295` hung in resampling); (3) `tiling_cell_scale` keeps a tiling pattern's cell pixmap within 3000 px a side whatever its `/XStep`/`/YStep` (a fuzzed step far beyond `/BBox` allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed `/LW 9223372036854775807` allocated 10 GB in stroke expansion); (5) `usable_dash_pattern`: a stroke whose dash pattern could cut it into more than `MAX_DASHES_PER_STROKE` (250,000) pieces, counting kurbo's restart of the pattern at every subpath and `ClosePath`, is drawn solid (a fuzzed `/D [[11] 0]` on a line from x = 92234775807 allocated over 5 GB, `[0.000001] 0 d` on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid (`[-1 -1] 0 d` looped for ever in kurbo); (7) `render_size` and `render_into`: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); `render` is `render_into` on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) `may_paint_canvas`: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); `set_skip_offscreen_paths` turns this off for the regression test; (10) `/Interpolate` governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour (since (21), only inside Type 3 glyphs), planned without the Catmull-Rom filter limits (`image_resampling_size` takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) `may_paint_rect`: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (`InterpreterSettings::preview`: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under `PREVIEW_MIN_IMAGE_AREA` device pixels, while final renders always draw them; (13) reading colours (#897): `RenderSettings::color_replacement` takes a `ColorReplacement` (text colour, background, `only_black`, `line_art`) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; `render_pdf` gains a `color_replacement: Option<ColorReplacement>` argument (pass `None` for true colours); the upstream-only `[dev-dependencies.log]` entry is removed from `Cargo.toml` (log is a normal dependency since the logging patch), so `cargo test -p hayro` runs this crate's unit tests; (14) a shrunk image that is interpolated (or, since (21), area-averaged) and whose axes lie along the device's (upright, mirrored or a quarter turn, as on a page under `/Rotate`) is snapped to whole device pixels: shrunk once to its rounded device rectangle (halves round up, so a tile snaps an edge where the whole page does) and drawn 1:1 from the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (`snap_minified_image`; scans looked soft next to Acrobat, and until #5530 a turned one still did, up to 200 grey levels off the same page drawn upright). An image counts as axis-aligned when the terms the snap drops move its far edges by at most a hundredth of a device pixel (`SNAP_SKEW`), not by a tolerance on the coefficients, which let a slight skew of a very wide image through; and the snapped 1:1 transform is used only when the shrink produced the snapped size, so a failed resize keeps the full transform instead of drawing the source raster at its source size; (15) a soft mask of another size than its image is filtered when shrunk, whatever its `/Interpolate` (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroLine truncated
|
||||
| hayro-jbig2 | 0.3.0 | Apache-2.0 OR MIT | (1) at most 2^28 pixels per bitmap and page bitmap, and 2^25 cells per halftone grid and grayscale image (each side could be 65535: 4.3 billion pixels decoded one by one, and a 4-byte value per cell); (2) at most 2^26 symbol pixels per symbol dictionary (a fuzzed dictionary decoded for minutes); (3) a symbol dictionary's height classes are capped at one per new symbol and its export runs at `2 × symbols + 1`, each plus `EXTRA_EMPTY_STEPS` (1024): an empty height class or a zero-length export run doesn't advance, and past the end of its data the arithmetic decoder can decode them for ever (one changed byte hung the renderer); (4) a segment header's long-form referred-to segment count is capped at `u16::MAX` and a larger count is refused before its retention or referred-to bytes are read: the 29-bit count went straight to `Vec::with_capacity`, a reservation of about 2 GiB whose failure aborts the process (from storytold/craft-libs#176); (5) what a page's segments hold for later ones (intermediate regions, pattern and symbol dictionaries with their headers and retained contexts, and Huffman tables) is at most 64 MiB, two square bitmaps of 16,384 a side (`MAX_RETAINED_WORDS`), and a dictionary's new and exported symbols are counted against it as they are made: each was bounded on its own, but a few dozen bytes an intermediate region held 33 MB each, and 8 MiB symbols exported again and again killed the process, so a few kilobytes could exhaust memory; (6) a segment may refer to at most 2^20 symbols, checked before any are collected: references may repeat, and 65,535 of them to one dictionary, in a 78 KB stream, asked for 17 GB; (7) a pattern dictionary keeps at most 16 MiB of shifted copies, which only speed up drawing, and none that wouldn't fit what the page has left (`MAX_SHIFTED_WORDS`); (8) a custom Huffman table has at most 2^16 lines, and a stream with its globals at most 2^20 segments and 2^22 references, checked as they are read: both grew before anything counted them, and PDFs of a kilobyte or so reached 5 GB and more behind nested Flate; (9) a page past the bounds of (1) is refused when the stream is opened: a caller sizes its output from the header before the page bitmap is made, and an Indexed image in a 743-byte PDF asked for 137 GB, which aborts; (10) a stream may ask for at most 2^30 units of decoding work, four of the largest pages, about 6 s to 17 s (`MAX_WORK`): decoding a bitmap costs a pixel each and 4 a row, copying one a word each and 1 a row, and each integer an item reads costs 38, what one takes at worst, so one with no pixels pays too, before it is decoded: each bitmap was bounded, but nothing bounded how many there were, and a 795-byte PDF took 265 s; (11) `Image::decode_within` spends from what a run of images may still do, starting with its page's pixels, and `charge_reading` charges reading a stream and its globals before anything is parsed, so an operation over a document can bound what `MAX_WORK` bounds for one image, and images that fail or decode nothing add up too: a file of thousands of images made to reach that limit took as many times as long; (12) adding to a location, a reference offset, a symbol's corner or a Huffman range line near the end of an `i32`, and negating a decoded integer of 2^31, overflowed: a panic wherever overflow is checked. The refinement context, a text region's corners, the integer decoder and the Huffman table wrap, as a release build did, and `Bitmap::combine` saturates, which comes to the same, as a location that overflows is past the end of any page. What is drawn doesn't change | (1) `raster::tests::jbig2_regions_of_billions_of_pixels_are_refused`, `raster::tests::jbig2_halftone_grids_past_their_bound_are_refused`; (2) the nightly `cargo xtask fuzz` finding, which now finishes in about a second; (3) `raster::tests::jbig2_symbol_dictionaries_that_never_advance_terminate`; (4) `segment::tests::test_referred_to_count_is_capped` (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`); (5) `raster::tests::jbig2_regions_held_past_a_pages_bound_are_refused`, `jbig2_symbols_without_pixels_count_towards_a_pages_bound`, `jbig2_retained_contexts_count_towards_a_pages_bound`, `jbig2_huffman_tables_count_towards_a_pages_bound`, `jbig2_new_symbols_count_towards_a_pages_bound`, `jbig2_new_symbols_are_held_to_what_a_page_has_left`, `jbig2_pattern_dictionaries_count_towards_a_pages_bound`, and in the crate (`cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml`) `decode::symbol::tests::exported_symbols_are_counted_before_they_are_copied`, `tests::the_room_left_shrinks_as_segments_are_stored`; (6) `raster::tests::jbig2_segments_referring_to_too_many_symbols_are_refused`; (7) `raster::tests::jbig2_pattern_dictionaries_too_large_to_shift_still_draw`, `jbig2_shifted_patterns_never_refuse_an_image`; (8) `raster::tests::jbig2_huffman_tables_of_too_many_lines_are_refused`, `jbig2_streams_of_too_many_segments_are_reLine truncated
|
||||
| lopdf | 0.45.0 | MIT | (1) PNG predictor rows longer than the data are refused before their buffers are allocated (a fuzzed `/Columns 4294967295` allocated two 4 GiB rows), and empty frames return without row buffers; (2) predictor row widths use saturating arithmetic or reject overflow; (3) sub-byte TIFF Predictor 2 scratch is bounded by the samples present in the input instead of the declared `/Colors`, aligned with [lopdf#589](https://github.com/J-F-Liu/lopdf/pull/589); (4) the object lookup keeps its cycle history through compressed objects, requires an ordinary in-file container, and caps the chain of indirect lookups at 16 (a chain of new objects, each a `/Length` compressed in a container whose own `/Length` names the next, or a stream whose `/Length` names the next stream, overflowed the stack: a dev build's 2 MiB thread within 800 hops); (5) object-stream members are parsed within their own extent (up to the next distinct offset), an entry whose offset repeats an earlier entry's is skipped (so a repeated object number can resolve to its earlier entry), and offsets are processed in index order; (6) `find_latest_trailer`: a rebuild takes a `/Type /XRef` stream dictionary as the trailer when the file has no `trailer` keyword (the latest in the file wins, as a later `trailer` would), so a concatenated or truncated file keeps its /Info and /ID instead of failing to load; a candidate whose /Root is the catalog beats one whose /Root is another dictionary, and an in-file /Root must match its generation (#2365) | (1) `inspect::tests::png_predictor_rows_longer_than_the_data_are_refused_before_allocating`, `inspect::tests::empty_png_predictor_frames_do_not_allocate_rows`; (2) `inspect::tests::tiff_subbyte_predictor_row_width_overflow_is_refused`; (3) `inspect::tests::tiff_subbyte_predictor_scratch_is_bounded_by_available_samples`, `inspect::tests::tiff_subbyte_predictors_keep_components_rows_and_padding`; (4) `inspect::tests::compressed_object_self_container_does_not_recurse`, `inspect::tests::compressed_object_mutual_containers_do_not_recurse`, `inspect::tests::object_stream_length_cycle_does_not_recurse`, `inspect::tests::compressed_object_container_chain_does_not_recurse`, `inspect::tests::compressed_length_chain_does_not_recurse`, `inspect::tests::plain_stream_length_chain_does_not_recurse`, `inspect::tests::compressed_objects_and_indirect_stream_length_still_load`, `inspect::tests::compressed_length_in_a_container_with_an_indirect_length_still_loads`; (5) `inspect::tests::object_stream_member_extents_bound_owned_strings`, `inspect::tests::object_stream_member_extents_preserve_index_order_and_whitespace`; (6) `inspect::tests::rebuilt_file_keeps_the_info_of_its_cross_reference_stream`, `inspect::tests::rebuilt_file_prefers_a_trailer_whose_root_is_the_catalog` |
|
||||
| hayro-syntax | 0.7.2 | Apache-2.0 OR MIT | (1) Page-tree cycle guard in `resolve_pages` (visited set + depth cap): a `/Kids` loop among compressed objects overflowed the stack and aborted the process; (2) `ccitt_size_ok`: CCITT images wider than 2^20 columns or over 2^28 pixels are not decoded (a fuzzed `/Columns 4294967295` allocated 4 GiB); (3) Flate/LZW predictor row widths are saturating `usize`, a zero width is refused and a row longer than the data is not allocated (a fuzzed `/Columns 9223372036854775807` wrapped to a 2^61-byte allocation and aborted); (4) the inline-image end-of-data check (`EI`) doesn't nest inside its own look-ahead and examines at most 256 candidates and steps (50 images with "EI" in their data ran for over ten minutes); an EI followed by a delimiter (EI%..., EI[...) ends the image like one followed by white space; only a regular character after EI skips the candidate; (5) a stream whose `/Length` is wrong is recovered only when `stream` is the next token — scanning ahead to a later object's `stream` made an indirect appearance-state dictionary (Quartz checkboxes: `/AP /N` → `<< /Yes … /Off … >>`) render as a non-form, so the widget was skipped; (6) JPX rescaling uses an exact integer sample range for 1–32-bit components (32 shifted past u32, and floating-point maxima at 25–31 bits wrapped to zero), and refuses zero-sized rows; (7) `decode_limit`: Flate (through flate2 and the fallback inflater), LZW, RunLength and ASCII85 never write past `MAX_DECODED_STREAM` (256 MiB), or, for the last filter of an image of known size, past twice its pixel bytes (a predictor tag byte per byte), at most 1 GiB; `Stream::decoded_within` for budgeted callers; a page's `/Contents` array shares one `MAX_DECODED_STREAM` (a 2 MB page whose content stream or 8 × 8 image inflated to 2 GB took the renderer past 3 GB in under a second; naming the stream twenty times in `/Contents`, 5 GB); (8) high-precision numbers (jsPDF MediaBox coordinates with over 19 digits) parse without overflowing the u64 mantissa: fractional digits past it are dropped, integer digits past it scale the value (a real, kept finite); (9) large dictionary indexes are built on first use and shared across repeated reads within each PDF (#307); the validated-byte cache is bounded, reader contexts remain separate, repairs and inline/content dictionaries bypass sharing, and small dictionaries stay eager; (10) decoded object-stream slots publish once under contention, cold loading rejects recursive dependencies, and xref repair serializes without holding recursive read guards; this permits one parser per native render pool with private worker caches; (11) `sync.rs`: a poisoned lock is recovered instead of unwrapped (a worker panic no longer cascades into every reader of the same PDF), and xref repair takes a blocking write lock instead of asserting a non-blocking one; (12) `SegmentList` (`util.rs`) returns `None` for an out-of-range index instead of panicking; (13) a page's `/UserUnit` scales `render_dimensions` and `initial_transform`, so the page and everything on it are sized in points, as Acrobat shows them; it is read as Acrobat reads it (the page's own entry, not inherited; a number from 1 up, at most 75,000; anything else counts as 1), and a page without it keeps exactly the transform it had; (14) revision 2–4 passwords are tried as given and then in PDFDocEncoding (`legacy_passwords` in `crypto/mod.rs`, which resolves the upstream `TODO: Convert to PDFDocEncoding`): passwords arrive as UTF-8 text, so a password with characters outside ASCII (`Çok güzel`, `ılık€`) written in PDFDocEncoding, as ISO 32000-2 §7.6.4.3.2 asks and qpdf and Acrobat write it, was always wrong; (15) page content is read a piece at a time (`page::ContentReader`, `Page::content_reader`): plain streams and Flate streams without a predictor are decoded as they are read, with a budget of 64 times their compressed size and never less than `MAX_DECODED_STREAM` (a stream that expands past it is cut and reported); other streams are decoded whole, as before, and share `MAX_DECODED_STREAM`. The content tokenizer reads a window at a time (`UntypedIter::new_window`, `TypedIter::needs_more`, `TypedIter::committed`): an instruction that runs past the window is not decided until more data follows, which `Reader` records by flagging reads and `at_end` past the end of the data, including through the clones the parser makes. The Flate decoder keeps its own 32 KiB history, so it pauses at any output size. `ContentReader::new` reuses raw/decrypted stream bytes for budget accounting and piece creation, and `decoded_raw_within` applies the existing Whole-stream filters and limits to those bytes without decrypting twice; (16) revision 5–6 (AES-256) passwords are tried in their SASLprep form (RFC 4013, then UTF-8, at most 127 bytes, as ISO 32000-2 §7.6.4.3.3 Algorithm 2.A asks and pdfcraft-crypt writes them) and then, if that differs, as given (`aes256_passwLine truncated
|
||||
|
||||
Vendored
+386
-25
@@ -193,6 +193,27 @@ fn image_buffer(len: usize) -> Option<Vec<u8>> {
|
||||
Some(out)
|
||||
}
|
||||
|
||||
/// PdfCraft patch: `mask` sampled onto the grid of a `width` × `height` image, each image sample
|
||||
/// taking the mask sample under its centre. `None` for an empty side or a buffer over the limits.
|
||||
fn mask_on_image_grid(mask: &LumaData, width: u32, height: u32) -> Option<Vec<u8>> {
|
||||
if width == 0 || height == 0 || mask.width == 0 || mask.height == 0 {
|
||||
return None;
|
||||
}
|
||||
let mut out = image_buffer(source_byte_len(width, height, 1, MAX_IMAGE_PIXELS)?)?;
|
||||
let mut columns = Vec::new();
|
||||
columns.try_reserve_exact(width as usize).ok()?;
|
||||
columns.extend((0..width).map(|x| nearest_source_index(mask.width, width, x)));
|
||||
let mask_width = mask.width as usize;
|
||||
for y in 0..height {
|
||||
let start = nearest_source_index(mask.height, height, y).checked_mul(mask_width)?;
|
||||
let row = mask.data.get(start..start.checked_add(mask_width)?)?;
|
||||
for &x in &columns {
|
||||
out.push(*row.get(x)?);
|
||||
}
|
||||
}
|
||||
Some(out)
|
||||
}
|
||||
|
||||
/// PdfCraft patch (#624): the source index that PicScale's nearest-neighbour plan copies for
|
||||
/// destination `index` of `dst` samples taken from `src`. It is the same fixed-point centre
|
||||
/// sampling as `ResampleNearestPlan`, so these are the samples a nearest resize would copy.
|
||||
@@ -236,6 +257,177 @@ fn nearest_premultiplied_rgba(
|
||||
Some(out)
|
||||
}
|
||||
|
||||
/// PdfCraft patch: the shorter side, in samples, from which an image that did not ask for
|
||||
/// `/Interpolate` is still smoothed when enlarged. A page scanned at 72 dpi is 612 samples wide;
|
||||
/// icons, QR codes and pixel art are usually far smaller and keep their hard pixels.
|
||||
const SMOOTHED_IMAGE_SIDE: u32 = 512;
|
||||
|
||||
/// PdfCraft patch: fixed-point unit of [`area_shrink`]'s weights.
|
||||
const AREA_ONE: u64 = 1 << 16;
|
||||
|
||||
/// PdfCraft patch: the source samples target sample `t` covers when `src` samples shrink to `dst`
|
||||
/// (`dst <= src`), with their weights in [`AREA_ONE`] units. Target `t` spans `[t·src, (t+1)·src)`
|
||||
/// and source `s` spans `[s·dst, (s+1)·dst)`, both in units of 1/(src·dst) of the axis, so every
|
||||
/// overlap is exact. The weights are differences of one rounded running sum, so they add up to
|
||||
/// exactly [`AREA_ONE`] and a flat area keeps its exact value.
|
||||
fn area_taps(src: u32, dst: u32, t: u32) -> impl Iterator<Item = (usize, u64)> {
|
||||
let (src, dst, t) = (u64::from(src), u64::from(dst), u64::from(t));
|
||||
// `t < dst <= src <= 2^28` (MAX_IMAGE_PIXELS), so every product below fits in a u64.
|
||||
let (lo, hi) = (t * src, (t + 1) * src);
|
||||
let fixed = move |x: u64| (x * AREA_ONE + src / 2) / src;
|
||||
(lo / dst..hi.div_ceil(dst)).map(move |s| {
|
||||
let from = (s * dst).max(lo) - lo;
|
||||
let to = ((s + 1) * dst).min(hi) - lo;
|
||||
(s as usize, fixed(to) - fixed(from))
|
||||
})
|
||||
}
|
||||
|
||||
/// PdfCraft patch: [`area_taps`] of every target sample of one axis, worked out once: for each
|
||||
/// target, the first source sample it covers and the weights of the samples from there on (the
|
||||
/// samples a target covers are consecutive). There are at most `src + dst` taps in all.
|
||||
struct AreaAxis {
|
||||
first: Vec<usize>,
|
||||
/// Where each target's weights start in `weights`, and where the last one's end.
|
||||
offsets: Vec<usize>,
|
||||
weights: Vec<u32>,
|
||||
}
|
||||
|
||||
impl AreaAxis {
|
||||
fn new(src: u32, dst: u32) -> Option<Self> {
|
||||
let n = dst as usize;
|
||||
let (mut first, mut offsets, mut weights) = (Vec::new(), Vec::new(), Vec::new());
|
||||
first.try_reserve_exact(n).ok()?;
|
||||
offsets.try_reserve_exact(n.checked_add(1)?).ok()?;
|
||||
weights.try_reserve_exact((src as usize).checked_add(n.checked_mul(2)?)?).ok()?;
|
||||
offsets.push(0);
|
||||
for t in 0..dst {
|
||||
let mut start = None;
|
||||
for (s, w) in area_taps(src, dst, t) {
|
||||
start.get_or_insert(s);
|
||||
// A weight is at most AREA_ONE, so it fits in a u32.
|
||||
weights.push(w as u32);
|
||||
}
|
||||
first.push(start?);
|
||||
offsets.push(weights.len());
|
||||
}
|
||||
Some(Self { first, offsets, weights })
|
||||
}
|
||||
|
||||
/// Target `t`'s first source sample and its weights.
|
||||
fn get(&self, t: usize) -> Option<(usize, &[u32])> {
|
||||
let (lo, hi) = (*self.offsets.get(t)?, *self.offsets.get(t.checked_add(1)?)?);
|
||||
Some((*self.first.get(t)?, self.weights.get(lo..hi)?))
|
||||
}
|
||||
}
|
||||
|
||||
/// PdfCraft patch: shrinks `N`-channel samples from `source` to `target` (no side larger than its
|
||||
/// source side) by area averaging: each target sample is the mean of the source it covers,
|
||||
/// weighted by how much of it it covers. Nothing is invented between samples, which is what
|
||||
/// `/Interpolate` governs (ISO 32000-2 §8.9.5.3, for images drawn above their resolution), and
|
||||
/// nothing is dropped: nearest-neighbour keeps one sample in N per axis, so a scan shown at screen
|
||||
/// size lost the strokes of its text. Premultiplied input stays premultiplied. Work memory is one
|
||||
/// source row of `u32` accumulators, the horizontal taps (at most `sw + 2 × tw`) and a vertically
|
||||
/// shrunk copy no larger than the source; a side over `MAX_RESAMPLING_SOURCE_SIDE` is refused, so
|
||||
/// none of it exceeds a few times one source row. `None` if nothing shrinks, the sizes don't fit
|
||||
/// or a buffer cannot be allocated.
|
||||
fn area_shrink<const N: usize>(data: &[u8], source: (u32, u32), target: (u32, u32)) -> Option<Vec<u8>> {
|
||||
let ((sw, sh), (tw, th)) = (source, target);
|
||||
// A side over MAX_RESAMPLING_SOURCE_SIDE would make the accumulator row and the tap table
|
||||
// several times larger than the samples (a 2^27 × 2 image of zeros is a tiny file).
|
||||
if tw == 0 || th == 0 || tw > sw || th > sh || sw > MAX_RESAMPLING_SOURCE_SIDE || sh > MAX_RESAMPLING_SOURCE_SIDE {
|
||||
return None;
|
||||
}
|
||||
let (sw_us, tw_us) = (sw as usize, tw as usize);
|
||||
if data.len() != source_byte_len(sw, sh, N, MAX_IMAGE_PIXELS)? {
|
||||
return None;
|
||||
}
|
||||
// Weights sum to AREA_ONE (2^16) and samples are at most 255, so every sum fits in a u32.
|
||||
let round = |acc: u32| ((acc + (AREA_ONE as u32) / 2) >> 16).min(255) as u8;
|
||||
// Vertical pass first: groups of whole source rows averaged into each target row. Rows are
|
||||
// contiguous, so this pass is cheap, and it leaves the horizontal pass, which gathers samples
|
||||
// per target, a third as many rows when a scan shrinks threefold.
|
||||
let tall = if th == sh {
|
||||
None
|
||||
} else {
|
||||
let row_len = sw_us * N;
|
||||
let mut rows = image_buffer(source_byte_len(sw, th, N, MAX_IMAGE_PIXELS)?)?;
|
||||
let mut acc: Vec<u32> = Vec::new();
|
||||
acc.try_reserve_exact(row_len).ok()?;
|
||||
acc.resize(row_len, 0);
|
||||
for t in 0..th {
|
||||
acc.fill(0);
|
||||
for (s, w) in area_taps(sh, th, t) {
|
||||
let row = data.get(s * row_len..(s + 1) * row_len)?;
|
||||
// A weight is at most AREA_ONE, so it fits in a u32.
|
||||
let w = w as u32;
|
||||
for (a, v) in acc.iter_mut().zip(row) {
|
||||
*a += w * u32::from(*v);
|
||||
}
|
||||
}
|
||||
rows.extend(acc.iter().map(|a| round(*a)));
|
||||
}
|
||||
Some(rows)
|
||||
};
|
||||
if tw == sw {
|
||||
return tall;
|
||||
}
|
||||
let rows = tall.as_deref().unwrap_or(data);
|
||||
// Horizontal pass: each of those rows shrunk to `tw` samples. The taps depend only on the
|
||||
// widths, so they are worked out once, not again for each row.
|
||||
let taps = AreaAxis::new(sw, tw)?;
|
||||
let mut out = image_buffer(image_byte_len(tw, th, N, MAX_IMAGE_PIXELS)?)?;
|
||||
for row in rows.chunks_exact(sw_us * N) {
|
||||
for t in 0..tw_us {
|
||||
let (first, weights) = taps.get(t)?;
|
||||
let samples = row.get(first * N..(first + weights.len()) * N)?;
|
||||
let mut acc = [0_u32; N];
|
||||
for (w, px) in weights.iter().zip(samples.chunks_exact(N)) {
|
||||
for (a, v) in acc.iter_mut().zip(px) {
|
||||
*a += w * u32::from(*v);
|
||||
}
|
||||
}
|
||||
out.extend(acc.map(round));
|
||||
}
|
||||
}
|
||||
Some(out)
|
||||
}
|
||||
|
||||
/// PdfCraft patch: [`area_shrink`] for an image's pixel format. If it cannot run, the image keeps
|
||||
/// its source resolution, as when [`Renderer::resize_image_data`] fails.
|
||||
fn area_resize(
|
||||
data: Vec<u8>,
|
||||
source: (u32, u32),
|
||||
target: (u32, u32),
|
||||
format: ImagePixelFormat,
|
||||
) -> (Vec<u8>, u32, u32) {
|
||||
let shrunk = match format {
|
||||
ImagePixelFormat::Luma => area_shrink::<1>(&data, source, target),
|
||||
ImagePixelFormat::Rgb => area_shrink::<3>(&data, source, target),
|
||||
ImagePixelFormat::Rgba => area_shrink::<4>(&data, source, target),
|
||||
};
|
||||
match shrunk {
|
||||
Some(out) => (out, target.0, target.1),
|
||||
None => {
|
||||
log::warn!(
|
||||
"Image area shrink failed or exceeded its buffer limit; falling back to source data"
|
||||
);
|
||||
(data, source.0, source.1)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// PdfCraft patch: premultiplies straight RGBA samples in place.
|
||||
fn premultiply_rgba(rgba: &mut [u8]) {
|
||||
let (chunks, _) = rgba.as_chunks_mut::<4>();
|
||||
for chunk in chunks {
|
||||
let [r, g, b, a] = *chunk;
|
||||
*chunk = AlphaColor::from_rgba8(r, g, b, a)
|
||||
.premultiply()
|
||||
.to_rgba8()
|
||||
.to_u8_array();
|
||||
}
|
||||
}
|
||||
|
||||
/// PdfCraft patch: most pieces one dashed stroke may be cut into. Stroke expansion keeps every
|
||||
/// dash, so a pattern that is tiny next to its path asks for billions: a fuzzed `/D [[11] 0]` on
|
||||
/// a line from x = 92234775807 allocated over 5 GB, and `[0.000001] 0 d` on a 20 pt line over
|
||||
@@ -423,9 +615,12 @@ fn snap_minified_image(
|
||||
planned: (u32, u32),
|
||||
) -> Option<(u32, u32, Affine)> {
|
||||
let [a, b, c, d, e, f] = transform.as_coeffs();
|
||||
// Halves round up, not away from zero: a tile's origin moves edges by whole pixels, and only
|
||||
// then do they snap where the whole page snaps them (-2.5 must go to -2 as 2.5 goes to 3).
|
||||
let snap = |v: f64| (v + 0.5).floor();
|
||||
// The device span of a source edge from `origin` across `side` samples of `scale`, rounded.
|
||||
let edge = |origin: f64, scale: f64, side: u32| {
|
||||
let (p, q) = (origin.round(), (origin + scale * f64::from(side)).round());
|
||||
let (p, q) = (snap(origin), snap(origin + scale * f64::from(side)));
|
||||
let (lo, hi) = (p.min(q), p.max(q));
|
||||
(hi - lo >= 1.0 && lo.abs() < 1e6 && hi.abs() < 1e6).then_some((lo, hi))
|
||||
};
|
||||
@@ -935,17 +1130,51 @@ impl Renderer {
|
||||
return;
|
||||
}
|
||||
let interpolate = image_data.interpolate();
|
||||
if let Some(a) = &alpha_data {
|
||||
let requested_resize = x_scale < 1.0 || y_scale < 1.0;
|
||||
// PdfCraft patch: an image that did not ask for interpolation and is drawn below its
|
||||
// resolution is shrunk by area averaging (`area_shrink`), not nearest-neighbour, which
|
||||
// broke up the text of scans shown at screen size. Type 3 glyph images keep the nearest
|
||||
// shrink of #624: they are drawn with a frame of their own and never snapped.
|
||||
let area = !interpolate
|
||||
&& requested_resize
|
||||
&& !self.in_type3_glyph
|
||||
// Wider sources keep #624's bounded nearest-neighbour shrink (`area_shrink` refuses
|
||||
// them): skipping them, as a refused shrink at source size would, loses the image.
|
||||
&& source_width.max(source_height) <= MAX_RESAMPLING_SOURCE_SIDE;
|
||||
let mut alpha_data = alpha_data;
|
||||
if let Some(a) = alpha_data.take() {
|
||||
if source_byte_len(a.width, a.height, 1, MAX_IMAGE_PIXELS) != Some(a.data.len()) {
|
||||
log::warn!("Image has invalid or over-limit alpha dimensions/data and was skipped");
|
||||
return;
|
||||
}
|
||||
if a.width != source_width || a.height != source_height || a.interpolate != interpolate
|
||||
{
|
||||
if let Some(alpha) = alpha_data {
|
||||
self.draw_image_with_alpha_mask(image_data, alpha);
|
||||
// PdfCraft patch: a soft mask no larger than its area-shrunk image is sampled onto
|
||||
// the image's grid, so colour and mask are averaged together, premultiplied. Drawn
|
||||
// as a mask layer over the image instead, the averaged colour (transparent samples
|
||||
// included) was multiplied by the averaged mask, and the colour of transparent
|
||||
// samples showed at the mask's edges. A larger mask carries detail the image does
|
||||
// not (#1649) and keeps the layer.
|
||||
let on_grid = (area && a.width <= source_width && a.height <= source_height)
|
||||
.then(|| mask_on_image_grid(&a, source_width, source_height))
|
||||
.flatten();
|
||||
match on_grid {
|
||||
Some(data) => {
|
||||
alpha_data = Some(LumaData {
|
||||
data,
|
||||
width: source_width,
|
||||
height: source_height,
|
||||
interpolate,
|
||||
..a
|
||||
});
|
||||
}
|
||||
None => {
|
||||
self.draw_image_with_alpha_mask(image_data, a);
|
||||
return;
|
||||
}
|
||||
}
|
||||
return;
|
||||
} else {
|
||||
alpha_data = Some(a);
|
||||
}
|
||||
}
|
||||
let mut quality = if interpolate {
|
||||
@@ -955,7 +1184,6 @@ impl Renderer {
|
||||
};
|
||||
let has_alpha = alpha_data.is_some();
|
||||
let mut may_have_transparency = has_alpha || self.force_images_may_have_transparency();
|
||||
let requested_resize = x_scale < 1.0 || y_scale < 1.0;
|
||||
let (new_width, new_height) = match image_resampling_size(
|
||||
source_width,
|
||||
source_height,
|
||||
@@ -985,12 +1213,28 @@ impl Renderer {
|
||||
if requested_resize && self.in_type3_glyph && interpolate {
|
||||
quality = ImageQuality::High;
|
||||
}
|
||||
// PdfCraft patch: an interpolated image that is shrunk, with its axes along the device's
|
||||
// (upright, mirrored or a quarter turn), lands on whole device pixels. It is shrunk once to
|
||||
// exactly its snapped device rectangle and drawn 1:1 from the snapped corner. Otherwise the
|
||||
// shrunk raster (rounded up to a whole size) is sampled a second time at a fractional scale
|
||||
// and offset, which blurs scans and photos.
|
||||
let snapped = if interpolate
|
||||
// PdfCraft patch: an image the size of a page scan is smoothed when drawn above its
|
||||
// resolution, even without `/Interpolate`: bilinearly, as images that ask for it are.
|
||||
// Scans rarely ask, and drawn nearest-neighbour each sample of a zoomed scan became a
|
||||
// block, so the edges of its text were staircases. Bicubic looked the same on a zoomed
|
||||
// 300 dpi scan and drew its tiles five times slower. Smaller images (icons, QR codes,
|
||||
// pixel art) keep their hard pixels. A scan shrunk on one side and stretched on the other
|
||||
// is smoothed too: the area shrink leaves the shrunk side at 1:1, where bilinear adds
|
||||
// nothing, and the stretched side would otherwise be drawn nearest-neighbour.
|
||||
if !interpolate
|
||||
&& !self.in_type3_glyph
|
||||
&& (x_scale > 1.0 || y_scale > 1.0)
|
||||
&& source_width.min(source_height) >= SMOOTHED_IMAGE_SIDE
|
||||
{
|
||||
quality = ImageQuality::Medium;
|
||||
}
|
||||
// PdfCraft patch: an image that is shrunk (filtered or area-averaged), with its axes along
|
||||
// the device's (upright, mirrored or a quarter turn), lands on whole device pixels. It is
|
||||
// shrunk once to exactly its snapped device rectangle and drawn 1:1 from the snapped
|
||||
// corner. Otherwise the shrunk raster (rounded up to a whole size) is sampled a second time
|
||||
// at a fractional scale and offset, which blurs scans and photos (or, by the nearest
|
||||
// sampler of a non-interpolated image, drops samples again).
|
||||
let snapped = if (interpolate || area)
|
||||
&& requested_resize
|
||||
&& !self.in_type3_glyph
|
||||
&& (new_width, new_height) != (source_width, source_height)
|
||||
@@ -1000,13 +1244,22 @@ impl Renderer {
|
||||
None
|
||||
};
|
||||
let (new_width, new_height) = snapped.map_or((new_width, new_height), |s| (s.0, s.1));
|
||||
// PdfCraft patch: an area shrink never enlarges. A side drawn at or above its resolution
|
||||
// (an image shrunk across and stretched down, as a 204 × 98 dpi fax is, or a snapped side
|
||||
// rounded past its source) keeps its samples and is enlarged by the sampler; the snap
|
||||
// then no longer matches, so the full transform places the image.
|
||||
let (new_width, new_height) = if area {
|
||||
(new_width.min(source_width), new_height.min(source_height))
|
||||
} else {
|
||||
(new_width, new_height)
|
||||
};
|
||||
let needs_resize = (new_width, new_height) != (source_width, source_height);
|
||||
|
||||
// PdfCraft patch (#624): a minified image with a soft mask that is not interpolated is
|
||||
// nearest-sampled straight into premultiplied target RGBA. That skips the full-resolution
|
||||
// RGBA copy below and the premultiply pass after it.
|
||||
// PdfCraft patch (#624): a minified Type 3 glyph image with a soft mask that is not
|
||||
// interpolated is nearest-sampled straight into premultiplied target RGBA. That skips the
|
||||
// full-resolution RGBA copy below and the premultiply pass after it.
|
||||
let direct = match &alpha_data {
|
||||
Some(alpha) if needs_resize && !interpolate => nearest_premultiplied_rgba(
|
||||
Some(alpha) if needs_resize && !interpolate && !area => nearest_premultiplied_rgba(
|
||||
&image_data,
|
||||
&alpha.data,
|
||||
(source_width, source_height),
|
||||
@@ -1014,7 +1267,7 @@ impl Renderer {
|
||||
),
|
||||
_ => None,
|
||||
};
|
||||
let premultiplied = direct.is_some();
|
||||
let mut premultiplied = direct.is_some();
|
||||
// Preserve the single-channel/RGB fast paths. With alpha, expand directly to RGBA;
|
||||
// every source and alpha length was validated before zipping or allocating.
|
||||
let (data, format) = match (direct, image_data, alpha_data) {
|
||||
@@ -1057,6 +1310,15 @@ impl Renderer {
|
||||
};
|
||||
let (data, mut img_width, mut img_height) = if premultiplied {
|
||||
(data, new_width, new_height)
|
||||
} else if needs_resize && area {
|
||||
let mut data = data;
|
||||
if matches!(format, ImagePixelFormat::Rgba) {
|
||||
// Averaging straight colours would bleed the colour of transparent samples into
|
||||
// their neighbours; premultiplied colours count each by its coverage.
|
||||
premultiply_rgba(&mut data);
|
||||
premultiplied = true;
|
||||
}
|
||||
area_resize(data, (source_width, source_height), (new_width, new_height), format)
|
||||
} else if needs_resize {
|
||||
self.resize_image_data(
|
||||
data,
|
||||
@@ -1103,14 +1365,7 @@ impl Renderer {
|
||||
rgba
|
||||
};
|
||||
if has_alpha && !premultiplied {
|
||||
let (chunks, _) = rgba_data.as_chunks_mut::<4>();
|
||||
for chunk in chunks {
|
||||
let [r, g, b, a] = *chunk;
|
||||
*chunk = AlphaColor::from_rgba8(r, g, b, a)
|
||||
.premultiply()
|
||||
.to_rgba8()
|
||||
.to_u8_array();
|
||||
}
|
||||
premultiply_rgba(&mut rgba_data);
|
||||
}
|
||||
// Type 3 glyphs get a transparent two-pixel frame for interpolation. Check the padded
|
||||
// RGBA allocation and u16 dimensions too, before narrowing or growing either side.
|
||||
@@ -2635,3 +2890,109 @@ mod snap_tests {
|
||||
assert!(snap_minified_image(&turned_noisy, source, (20_000, 200)).is_some());
|
||||
}
|
||||
}
|
||||
|
||||
/// PdfCraft patch: the area-averaging shrink of non-interpolated images.
|
||||
#[cfg(test)]
|
||||
mod area_shrink_tests {
|
||||
use super::{
|
||||
AREA_ONE, AreaAxis, LumaData, MAX_IMAGE_PIXELS, area_shrink, area_taps, mask_on_image_grid,
|
||||
};
|
||||
|
||||
fn mask(data: &[u8], width: u32, height: u32) -> LumaData {
|
||||
LumaData { data: data.to_vec(), width, height, interpolate: false, scale_factors: (1.0, 1.0) }
|
||||
}
|
||||
|
||||
/// Each image sample takes the mask sample under its centre, at whole and fractional ratios.
|
||||
#[test]
|
||||
fn a_mask_is_sampled_onto_the_image_grid() {
|
||||
assert_eq!(mask_on_image_grid(&mask(&[10, 20], 2, 1), 4, 2), Some(vec![10, 10, 20, 20, 10, 10, 20, 20]));
|
||||
// Centres of 7 samples over 3: 0.5/7·3 = 0.21 → 0, 1.5/7·3 = 0.64 → 0, 2.5/7·3 = 1.07 → 1, …
|
||||
assert_eq!(mask_on_image_grid(&mask(&[1, 2, 3], 3, 1), 7, 1), Some(vec![1, 1, 2, 2, 2, 3, 3]));
|
||||
assert_eq!(mask_on_image_grid(&mask(&[1, 2, 3, 4], 1, 4), 1, 2), Some(vec![2, 4]));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn malformed_masks_are_not_sampled() {
|
||||
assert_eq!(mask_on_image_grid(&mask(&[], 0, 1), 4, 1), None, "an empty mask");
|
||||
assert_eq!(mask_on_image_grid(&mask(&[1, 2], 2, 1), 0, 1), None, "an empty image");
|
||||
assert_eq!(mask_on_image_grid(&mask(&[1], 2, 1), 4, 1), None, "mask data shorter than its size");
|
||||
}
|
||||
|
||||
/// Every target sample's weights add up to exactly one, and together the targets take every
|
||||
/// source sample exactly once (each `dst / src` of it, in fixed point within rounding).
|
||||
#[test]
|
||||
fn taps_are_exact_partitions_of_unity() {
|
||||
for src in 1..=40_u32 {
|
||||
for dst in 1..=src {
|
||||
let mut taken = vec![0_u64; src as usize];
|
||||
for t in 0..dst {
|
||||
let taps: Vec<_> = area_taps(src, dst, t).collect();
|
||||
assert_eq!(taps.iter().map(|(_, w)| w).sum::<u64>(), AREA_ONE, "{src} → {dst}, target {t}");
|
||||
assert!(taps.iter().all(|(_, w)| *w > 0), "{src} → {dst}, target {t}: an empty tap");
|
||||
for (s, w) in taps {
|
||||
taken[s] += w;
|
||||
}
|
||||
}
|
||||
// Source s is dst/src of one target in all; scaled by src/dst, that is one.
|
||||
for (s, w) in taken.iter().enumerate() {
|
||||
let share = (*w * u64::from(src)) as f64 / (AREA_ONE * u64::from(dst)) as f64;
|
||||
assert!((share - 1.0).abs() < 1e-3, "{src} → {dst}: source {s} weighs {share}");
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// The table the horizontal pass reads holds exactly the taps of every target, in order.
|
||||
#[test]
|
||||
fn the_tap_table_holds_every_targets_taps() {
|
||||
for src in 1..=40_u32 {
|
||||
for dst in 1..=src {
|
||||
let axis = AreaAxis::new(src, dst).expect("table");
|
||||
for t in 0..dst {
|
||||
let (first, weights) = axis.get(t as usize).expect("target");
|
||||
let expected: Vec<(usize, u64)> = area_taps(src, dst, t).collect();
|
||||
let got: Vec<(usize, u64)> = weights.iter().enumerate().map(|(k, w)| (first + k, u64::from(*w))).collect();
|
||||
assert_eq!(got, expected, "{src} → {dst}, target {t}");
|
||||
}
|
||||
assert_eq!(axis.get(dst as usize), None, "{src} → {dst}: no target past the last");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// The largest side an image may have still computes in u64 without overflowing.
|
||||
#[test]
|
||||
fn taps_at_the_pixel_limit_do_not_overflow() {
|
||||
let side = MAX_IMAGE_PIXELS as u32;
|
||||
for (dst, t) in [(1, 0), (3, 2), (side - 1, side - 2), (side, side - 1)] {
|
||||
assert_eq!(area_taps(side, dst, t).map(|(_, w)| w).sum::<u64>(), AREA_ONE, "{side} → {dst}, target {t}");
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn flat_images_keep_their_exact_value() {
|
||||
for (source, target) in [((7, 5), (3, 2)), ((10, 3), (6, 3)), ((4, 9), (4, 4)), ((13, 11), (1, 1))] {
|
||||
let data = vec![77_u8; (source.0 * source.1 * 3) as usize];
|
||||
let out = area_shrink::<3>(&data, source, target).expect("shrinks");
|
||||
assert_eq!(out.len(), (target.0 * target.1 * 3) as usize);
|
||||
assert!(out.iter().all(|v| *v == 77), "{source:?} → {target:?}: {out:?}");
|
||||
}
|
||||
}
|
||||
|
||||
/// A side over the resampling limit is refused before any work buffer is allocated.
|
||||
#[test]
|
||||
fn sides_over_the_resampling_limit_are_refused() {
|
||||
let wide = (super::MAX_RESAMPLING_SOURCE_SIDE + 1) as usize;
|
||||
assert_eq!(area_shrink::<1>(&vec![0_u8; wide], (wide as u32, 1), (10, 1)), None, "too wide");
|
||||
assert_eq!(area_shrink::<1>(&vec![0_u8; wide], (1, wide as u32), (1, 10)), None, "too tall");
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn malformed_shrinks_are_refused() {
|
||||
let data = vec![0_u8; 4 * 4];
|
||||
assert_eq!(area_shrink::<1>(&data, (4, 4), (4, 4)), None, "nothing shrinks");
|
||||
assert_eq!(area_shrink::<1>(&data, (4, 4), (5, 2)), None, "a side grows");
|
||||
assert_eq!(area_shrink::<1>(&data, (4, 4), (0, 2)), None, "an empty side");
|
||||
assert_eq!(area_shrink::<1>(&data[1..], (4, 4), (2, 2)), None, "data shorter than the source");
|
||||
assert_eq!(area_shrink::<3>(&data, (4, 4), (2, 2)), None, "data of another channel count");
|
||||
}
|
||||
}
|
||||
Reference in new issue
Block a user