mirror of
https://github.com/storytold/lightcraft.git
synced 2026-10-24 10:09:01 -06:00
LR-IMP-PREVIEW-READ: a CR3 or a TIFF raw's preview is read from the plan's ranges
The header-only path added in #755 answers for a Fuji RAF, whose preview pointer sits in the first bytes. A Canon CR3 keeps its preview in a JPEG track, and a TIFF raw (CR2, ARW, DNG, RW2, 3FR …) can keep its preview tag in an IFD the prefix cuts short: both fell back to `std::fs::read` of the whole file — tens of megabytes for a JPEG of a few, on the format of the largest bucket in gaps.md (Canon, 33 % of photographers). `preview_window` now asks `lightcraft_raw::sparse::preview_plan` — the module #961 added for the import path, which had no production caller — which ranges `embedded_preview` reads, and reads exactly those: `planned_preview` fills a buffer of the file's length, reads the ranges the plan names and asks again until it says `Ready`. The buffer is the file's length, which the memory gate counted already, and its untouched pages stay zero, so what this path costs is the bytes read — on macOS a freed buffer of that size may sit in the allocator's cache rather than go back to the OS (measured on a 38k-photo import, #832). The loop gives up past 60 % of the file, the bound step 2's partial read uses, and a plan that does not fit, a file past 1 GiB or a window that does not decode all end in the whole-file fallback. The quick path hands the JPEG to the decoder alone, so two container facts the whole-file path applies have to be visible in the bytes it is handed, or the file declines and the whole file keeps the preview: - a preview colour space of Adobe RGB (`decode_raw_preview` applies it; a JPEG-only decode has the JPEG's own space, else sRGB). A container that says sRGB does not decline — that is what the decode assumes anyway, and declining it took the fast path away from a real Sony ARW; - the file's own EXIF orientation, when the preview JPEG carries none of its own: the plan's header is a bounded 64 KiB (the header route grows its prefix to the preview's declared extent instead), and an orientation tag it cannot see would show the preview unrotated. Measured over the whole CC0 corpus (111 raws): 52 files take this route, reading a median 7.8 % of their bytes (0.9-27.7 %) — RW2 3.3 %, Leica RWL 4.1 %, ARW 7.5 %, CR3 8.4 %, DNG 9.7 %, CR2 14.3 % — while the 21 RAFs stay on the header route (median 4.5 %). The other 38 read whole: NEF, NRW, ORF, PEF and SRW because the planner does not describe them (an SRW plans to the whole file, which the 60 % budget then declines), plus the ARWs, CR2s, CR3s and DNGs that declare Adobe RGB or whose plan wants more than that budget — the bytes they read are the ones the plan would have read anyway. Across all 111 files the plan handed on the same preview as the whole file, with the skipped bytes zeroed and again 0xFF. `LIGHTCRAFT_PROFILE=1` prints [profile] preview …/arw-sony-a7rm4a-compressed.arw: read 688128 of 61832192 bytes (plan, window 483441) Tests: a synthetic TIFF/DNG whose preview only its tags locate (identical pixels to the whole-file path, the CFA strip unread), the same structure without the tag that makes it a raw, a file whose orientation only the whole file shows, and a file that states Adobe RGB against one that states sRGB; the opt-in real-file test reports which path answered. Co-Authored-By: DeepSeek V3.2 <noreply@deepseek.com>
This commit is contained in:
goujicheng
co-authored by
DeepSeek V3.2
DeepSeek V3.2
parent
76b5377868
commit
3c7f450137
@@ -314,7 +314,7 @@ LightCraft is young and moving fast: **alpha**. **Where we honestly stand** (det
|
||||
- **camera colour calibration:** Sony, Nikon, Panasonic, Fujifilm and Canon CR3 raws have guarded estimates from their camera JPEGs, with built-in ILCE-7CR, ILCE-7M4, X-H2S and X-T4 profiles; measured calibration is missing, and other raws or rejected fits retain a neutral matrix;
|
||||
- **compressed Olympus raws and unsupported CR3 variants:** these use embedded JPEG previews when present. Fujifilm lossless/lossy compressed RAF now decodes sensor data; [verification and existing-library reload instructions](docs/raf-compression.md);
|
||||
- **AI masks and denoise:** subject and sky selection are classical heuristics;
|
||||
- **network libraries:** a preview reads the raw's header and its embedded JPEG only (56.5 MB → 5.9 MB, 10 s → 1 s on the test share), but the *developed* picture still needs the sensor data over the wire; see [docs/raw-preview-reads.md](docs/raw-preview-reads.md);
|
||||
- **network libraries:** a preview reads the raw's header — or, for a CR3 or a TIFF raw whose preview tag sits past it, exactly the ranges the sparse planner names — and its embedded JPEG only (56.5 MB → 5.9 MB, 10 s → 1 s on the test share), but the *developed* picture still needs the sensor data over the wire; see [docs/raw-preview-reads.md](docs/raw-preview-reads.md);
|
||||
- **on-screen HDR display, video and the Classic Print / Book / Map modules.**
|
||||
- **What's next:** see [the ranked gaps](docs/gaps.md) and [the roadmap](docs/roadmap.md).
|
||||
|
||||
|
||||
+267
-5
@@ -681,9 +681,10 @@ fn fit_box(max_edge: usize) -> lightcraft_codecs::DecodeOptions {
|
||||
/// Filesystem-backed embedded-preview hook (native).
|
||||
///
|
||||
/// The quick path reads the file's header and its embedded preview and nothing else (see
|
||||
/// [`preview_window`]); a format whose preview the header cannot point at falls back to the
|
||||
/// whole-file path, which is what every format used to cost. The memory gate still counts the whole
|
||||
/// file, which is what the decode of one can hold.
|
||||
/// [`preview_window`]); a container the header cannot point into is planned from its tags
|
||||
/// (`lightcraft_raw::sparse::preview_plan`, the same module the import path probes with), and only a
|
||||
/// file the planner cannot describe falls back to the whole-file path. The memory gate still counts
|
||||
/// the whole file, which is what the decode of one can hold.
|
||||
pub fn fs_preview_loader() -> PreviewLoader {
|
||||
Arc::new(|path: &str, max_edge: usize| {
|
||||
let len = std::fs::metadata(path).map(|m| usize::try_from(m.len()).unwrap_or(usize::MAX)).unwrap_or(0);
|
||||
@@ -706,6 +707,10 @@ pub struct PreviewRead {
|
||||
/// Whether the preview came from a bounded read (`false`: the whole file was read, which is what
|
||||
/// every preview cost before the header path existed).
|
||||
pub header_path: bool,
|
||||
/// Whether the sparse planner named the ranges, rather than a format's own header fields: the
|
||||
/// path a Canon CR3 and the TIFF raws whose preview tag sits past the prefix take (`false` on
|
||||
/// the header path, and on the whole-file fallback).
|
||||
pub planned: bool,
|
||||
}
|
||||
|
||||
impl PreviewRead {
|
||||
@@ -745,7 +750,13 @@ pub fn preview_window(path: &str) -> Result<Option<(Vec<u8>, Vec<u8>, PreviewRea
|
||||
{
|
||||
prefix = read_prefix_at(&mut file, path, extent.min(total))?;
|
||||
}
|
||||
let Some(range) = lightcraft_raw::bounded_preview(&prefix) else { return Ok(None) };
|
||||
let Some(range) = lightcraft_raw::bounded_preview(&prefix) else {
|
||||
// The header carries no pointer this path reads — a Canon CR3's boxes, a TIFF raw whose
|
||||
// preview tag sits past the prefix, the containers whose preview only a scan finds: ask the
|
||||
// sparse planner which ranges `embedded_preview` reads before handing the caller the
|
||||
// whole-file fallback.
|
||||
return planned_preview(&mut file, path, total);
|
||||
};
|
||||
// A preview at the file's end may have been padded past it: clamp rather than fail.
|
||||
let end = range.offset.saturating_add(range.window_len).min(total);
|
||||
let len = usize::try_from(end.saturating_sub(range.offset)).map_err(|_| format!("{path}: preview too large"))?;
|
||||
@@ -767,7 +778,7 @@ pub fn preview_window(path: &str) -> Result<Option<(Vec<u8>, Vec<u8>, PreviewRea
|
||||
if lightcraft_raw::embedded_preview_from_range(&window, range).is_none() {
|
||||
return Ok(None);
|
||||
}
|
||||
let read = PreviewRead { read: bytes_read, total, header_path: true };
|
||||
let read = PreviewRead { read: bytes_read, total, header_path: true, planned: false };
|
||||
// On a network share the bytes are the wait, so say what this preview actually cost. A window
|
||||
// taken from the prefix cost nothing beyond the prefix itself.
|
||||
if lightcraft_pipeline::profiling() {
|
||||
@@ -776,6 +787,107 @@ pub fn preview_window(path: &str) -> Result<Option<(Vec<u8>, Vec<u8>, PreviewRea
|
||||
Ok(Some((prefix, window, read)))
|
||||
}
|
||||
|
||||
/// The most times the sparse planner may name ranges before the preview gives up: the plan is
|
||||
/// monotone (it never asks for a range twice), so a real file needs a handful.
|
||||
const MAX_PREVIEW_PLAN_ROUNDS: usize = 8;
|
||||
/// Tenths of a file the planner may have read before the preview gives up and the whole-file path
|
||||
/// takes over. The import path bounds its own loop the same way; a "preview" that wants most of the
|
||||
/// file has stopped being one.
|
||||
const MAX_PREVIEW_PLAN_SHARE: u64 = 6;
|
||||
/// The largest file the planner path builds its buffer for. The raws it serves are far below this;
|
||||
/// a file past it is read whole as before rather than allocating its length up front.
|
||||
const MAX_PREVIEW_PLAN_BYTES: u64 = 1 << 30;
|
||||
/// The header the quick path is handed on the planner route. The header route grows its prefix to
|
||||
/// the preview's declared extent, so `quick_image_srgb` can read the file's own EXIF orientation out
|
||||
/// of it when the preview JPEG carries none; this route reads a bounded prefix instead, big enough
|
||||
/// for IFD0 (which is where that tag lives) and for a CR3's metadata boxes.
|
||||
const PLAN_PREFIX: u64 = 64 * 1024;
|
||||
|
||||
/// The preview of a container the header-only path cannot answer for, from the byte ranges
|
||||
/// [`lightcraft_raw::sparse::preview_plan`] says [`lightcraft_raw::embedded_preview`] reads.
|
||||
///
|
||||
/// The planner is asked about a buffer of the file's whole length in which only the ranges named so
|
||||
/// far hold bytes, and only those ranges are read: the sensor data in the bulk ranges is not. The
|
||||
/// buffer is the file's length, which the loader's memory gate already counts, and its untouched
|
||||
/// pages stay zero and are never written — though on macOS a freed buffer of that size may stay in
|
||||
/// the allocator's cache rather than going back to the OS (measured on a 38k-photo import, #832), so
|
||||
/// this path's cost is the bytes read, not a promise about RSS.
|
||||
///
|
||||
/// `None` — the caller's cue to read the file whole, exactly as when the header declines — when the
|
||||
/// planner says it cannot describe the file (NEF, ORF and PEF probe from samples; a structure it
|
||||
/// does not trust), when the ranges it names do not fit the buffer or want most of the file, or when
|
||||
/// the quick path could not reproduce what the whole-file path shows (the two container facts
|
||||
/// below). The planner's contract is the one its corpus test asserts over real files (what it loaded
|
||||
/// gives what the whole file gives), so a window that still fails to decode ends in the same
|
||||
/// whole-file fallback in [`fs_preview_loader`], never in a wrong picture.
|
||||
fn planned_preview(file: &mut std::fs::File, path: &str, total: u64) -> Result<Option<(Vec<u8>, Vec<u8>, PreviewRead)>, String> {
|
||||
use std::io::{Read, Seek, SeekFrom};
|
||||
|
||||
if total > MAX_PREVIEW_PLAN_BYTES {
|
||||
return Ok(None);
|
||||
}
|
||||
let Ok(len) = usize::try_from(total) else { return Ok(None) };
|
||||
let mut buf = vec![0u8; len];
|
||||
let header = read_prefix_at(file, path, PLAN_PREFIX.min(total))?;
|
||||
let Some(head) = buf.get_mut(..header.len()) else { return Ok(None) };
|
||||
head.copy_from_slice(&header);
|
||||
let mut loaded = lightcraft_raw::sparse::Ranges::new();
|
||||
loaded.insert(0, header.len() as u64);
|
||||
for _ in 0..MAX_PREVIEW_PLAN_ROUNDS {
|
||||
match lightcraft_raw::sparse::preview_plan(&buf, &loaded) {
|
||||
lightcraft_raw::sparse::HeaderPlan::Need(ranges) => {
|
||||
let wanted = loaded.total().saturating_add(ranges.total());
|
||||
if wanted.saturating_mul(10) > total.saturating_mul(MAX_PREVIEW_PLAN_SHARE) {
|
||||
return Ok(None);
|
||||
}
|
||||
for &(start, end) in ranges.spans() {
|
||||
let (Ok(a), Ok(b)) = (usize::try_from(start), usize::try_from(end)) else { return Ok(None) };
|
||||
let Some(dst) = buf.get_mut(a..b) else { return Ok(None) };
|
||||
file.seek(SeekFrom::Start(start)).map_err(|e| format!("{path}: {e}"))?;
|
||||
file.read_exact(dst).map_err(|e| format!("{path}: {e}"))?;
|
||||
loaded.insert(start, end);
|
||||
}
|
||||
}
|
||||
lightcraft_raw::sparse::HeaderPlan::Ready => {
|
||||
let Some(window) = lightcraft_raw::embedded_preview(&buf) else { return Ok(None) };
|
||||
// The quick path hands the JPEG to the decoder alone, so what the *container* says
|
||||
// about the preview has to be something the bytes it is handed can show. Two container
|
||||
// facts change the pixels and are not in the JPEG: the preview's colour space (which
|
||||
// the whole-file path applies through `decode_raw_preview`) and, when the JPEG carries
|
||||
// no orientation of its own, the file's EXIF orientation. Where either is present and
|
||||
// the header cannot answer for it, the whole-file path keeps the preview — exactly as
|
||||
// it already does for a container-declared colour space. A container that says sRGB
|
||||
// agrees with the JPEG-only decode (its own declaration, else sRGB), so only the other
|
||||
// space declines.
|
||||
if lightcraft_raw::embedded_preview_color_space(&buf) == Some(lightcraft_raw::PreviewColorSpace::AdobeRgb) {
|
||||
return Ok(None);
|
||||
}
|
||||
// `extract` copies a TIFF's bytes to hand on its Exif block, so ask the cheap side
|
||||
// first: a JPEG that carries its own orientation needs nothing from the container.
|
||||
if lightcraft_meta::extract(&window).orientation.is_none()
|
||||
&& lightcraft_meta::extract(&header).orientation.is_none()
|
||||
&& lightcraft_meta::extract(&buf).orientation.is_some()
|
||||
{
|
||||
return Ok(None);
|
||||
}
|
||||
let read = PreviewRead { read: loaded.total(), total, header_path: true, planned: true };
|
||||
// A plan-driven preview is the one that used to cost the whole file, so it is worth
|
||||
// saying what it cost now too.
|
||||
if lightcraft_pipeline::profiling() {
|
||||
eprintln!("[profile] preview {path}: read {} of {} bytes (plan, window {})", read.read, total, window.len());
|
||||
}
|
||||
// The planner's buffer is the file's length, so give its pages back the way a decode
|
||||
// does (see `fs_pair_loader`).
|
||||
drop(buf);
|
||||
crate::memory::release();
|
||||
return Ok(Some((header, window, read)));
|
||||
}
|
||||
lightcraft_raw::sparse::HeaderPlan::Whole(_) => return Ok(None),
|
||||
}
|
||||
}
|
||||
Ok(None)
|
||||
}
|
||||
|
||||
/// The file's first [`lightcraft_raw::PREVIEW_PREFIX`] bytes (or the whole file when it is shorter),
|
||||
/// which is what every format [`lightcraft_raw::bounded_preview`] covers keeps its pointer in.
|
||||
fn read_prefix(file: &mut std::fs::File, path: &str, total: u64) -> Result<Vec<u8>, String> {
|
||||
@@ -967,6 +1079,7 @@ mod tests {
|
||||
let (header, window, read) = preview_window(&path).expect("the file reads").expect("it has a preview");
|
||||
assert_eq!(header.len(), lightcraft_raw::PREVIEW_PREFIX, "the header prefix is bounded");
|
||||
assert!(read.header_path, "the header path is what answered");
|
||||
assert!(!read.planned, "a RAF pointer in the header needs no planner");
|
||||
assert_eq!(read.total, file.len() as u64);
|
||||
assert!(read.fraction() < 0.5, "the read is a fraction of the file: {read:?}");
|
||||
assert!(window.len() < file.len() / 2, "the raw tail must not be read: {} of {} bytes", window.len(), file.len());
|
||||
@@ -985,6 +1098,155 @@ mod tests {
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
/// What a fixture's IFD0 says about itself: the tag that makes `probe` call the file a raw, the
|
||||
/// file's own orientation, and the EXIF colour space the container declares for its preview.
|
||||
#[derive(Clone, Copy, Default)]
|
||||
struct Fixture {
|
||||
dng: bool,
|
||||
sony: bool,
|
||||
orientation: Option<u16>,
|
||||
/// EXIF `ColorSpace` (0xA001): 1 = sRGB, 0xFFFF with an "R03" InteropIndex = Adobe RGB.
|
||||
color_space: Option<u16>,
|
||||
}
|
||||
|
||||
/// A TIFF-family raw no header field points into: IFD0 is the JPEG-compressed preview strip (the
|
||||
/// shape a DNG's `PreviewImage`, a CR2's or an ARW's preview takes), IFD1 a CFA "sensor" strip of
|
||||
/// `tail` bytes. The writer puts each IFD's table after its own chunk, so a `strip` bigger than
|
||||
/// the plan's bounded header puts IFD0 (and any orientation tag in it) out of that header's reach.
|
||||
fn tiff_with_preview_strip(strip: &[u8], tail: usize, what: Fixture) -> Vec<u8> {
|
||||
use lightcraft_tiff::{IfdBuilder, ImageData, TiffWriter, Value, tags as t};
|
||||
let mut preview = IfdBuilder::new();
|
||||
if what.dng {
|
||||
preview.set(t::DNG_VERSION, Value::Byte(vec![1, 4, 0, 0]));
|
||||
}
|
||||
if what.sony {
|
||||
preview.set(t::MAKE, Value::Ascii("SONY".into()));
|
||||
}
|
||||
if let Some(o) = what.orientation {
|
||||
preview.set(t::ORIENTATION, Value::Short(vec![o]));
|
||||
}
|
||||
if let Some(space) = what.color_space {
|
||||
let mut exif = IfdBuilder::new();
|
||||
exif.set(0xA001, Value::Short(vec![space]));
|
||||
if space == 0xFFFF {
|
||||
let mut interop = IfdBuilder::new();
|
||||
interop.set(0x0001, Value::Undefined(b"R03".to_vec()));
|
||||
exif.set_child(t::INTEROP_IFD, interop);
|
||||
}
|
||||
preview.set_child(t::EXIF_IFD, exif);
|
||||
}
|
||||
preview.set(t::COMPRESSION, Value::Short(vec![7]));
|
||||
preview.set(t::PHOTOMETRIC, Value::Short(vec![t::photometric::RGB]));
|
||||
preview.set(t::IMAGE_WIDTH, Value::Long(vec![48]));
|
||||
preview.set(t::IMAGE_LENGTH, Value::Long(vec![32]));
|
||||
preview.set_image(ImageData::Strips { rows_per_strip: 32, strips: vec![strip.to_vec()] });
|
||||
let mut raw = IfdBuilder::new();
|
||||
raw.set(t::COMPRESSION, Value::Short(vec![1]));
|
||||
raw.set(t::PHOTOMETRIC, Value::Short(vec![t::photometric::CFA]));
|
||||
raw.set(t::IMAGE_WIDTH, Value::Long(vec![6000]));
|
||||
raw.set(t::IMAGE_LENGTH, Value::Long(vec![4000]));
|
||||
raw.set_image(ImageData::Strips { rows_per_strip: 4000, strips: vec![vec![0x33u8; tail]] });
|
||||
TiffWriter::default().write(&[preview, raw]).expect("the fixture writes")
|
||||
}
|
||||
|
||||
/// A raw whose preview only its tags locate — a Canon CR3's JPEG track, a TIFF raw whose preview
|
||||
/// tag sits past the prefix — is read in part too: `sparse::preview_plan` names the ranges
|
||||
/// `embedded_preview` reads, so the preview strip is read and the sensor strip is not. The pixels
|
||||
/// must equal the whole-file path's, exactly; a file the planner cannot describe must decline
|
||||
/// rather than guess, so the caller keeps reading it whole.
|
||||
#[test]
|
||||
fn preview_loading_plans_the_ranges_it_needs() {
|
||||
let (w, h) = (48u32, 32u32);
|
||||
let rgb: Vec<u8> = (0..w * h * 3).map(|i| (i * 29 % 241) as u8).collect();
|
||||
let img = EncodeImage::new(w, h, 3, Samples::U8(&rgb));
|
||||
let jpeg = encode_jpeg(&img, 90, ChromaSubsampling::S420, &EncodeMeta::default()).unwrap();
|
||||
let file = tiff_with_preview_strip(&jpeg, 400_000, Fixture { dng: true, ..Default::default() });
|
||||
let dir = crate::tests_xmp::temp_dir("preview-plan");
|
||||
let path = dir.join("preview.dng");
|
||||
std::fs::write(&path, &file).unwrap();
|
||||
let path = path.to_string_lossy().into_owned();
|
||||
let max_edge = 512;
|
||||
|
||||
let (header, window, read) = preview_window(&path).expect("the file reads").expect("the plan locates its preview");
|
||||
assert_eq!(header.len(), PLAN_PREFIX as usize, "the plan's header is bounded");
|
||||
assert!(read.planned, "the planner is what answered");
|
||||
assert!(read.header_path, "and it is still a bounded read");
|
||||
assert_eq!(read.total, file.len() as u64);
|
||||
assert!(read.fraction() < 0.5, "the read is a fraction of the file: {read:?} of {} bytes", file.len());
|
||||
assert!(window.len() < file.len() / 2, "the sensor strip must not be read: {} of {} bytes", window.len(), file.len());
|
||||
|
||||
// The plan path and the whole-file path agree byte for byte.
|
||||
let quick = quick_image_srgb(&header, &window, max_edge).expect("the window holds a decodable preview");
|
||||
let whole = embedded_preview_srgb(&file, max_edge).expect("the whole file decodes too");
|
||||
assert_eq!((quick.width, quick.height), (whole.width, whole.height), "same preview size");
|
||||
assert_eq!(quick.data, whole.data, "byte for byte");
|
||||
|
||||
// The same structure without the tag that makes it a raw: the planner declines, and the
|
||||
// caller's whole-file read is what answers (no guessing at what a file might hold).
|
||||
let not_a_raw = tiff_with_preview_strip(&jpeg, 400_000, Fixture::default());
|
||||
std::fs::write(&path, ¬_a_raw).unwrap();
|
||||
assert!(preview_window(&path).expect("the file reads").is_none(), "an unknown container falls back");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
/// A file whose preview the quick path could not show the way the whole-file path does is left to
|
||||
/// the whole file. Here the orientation tag sits in an IFD0 past the bounded header (its chunk is
|
||||
/// bigger than that header) and the preview JPEG carries none of its own, so a window read would
|
||||
/// come out unrotated where the whole-file path rotates it — the plan must decline instead of
|
||||
/// handing on a wrong picture.
|
||||
#[test]
|
||||
fn preview_loading_declines_an_orientation_only_the_whole_file_shows() {
|
||||
let (w, h) = (48u32, 32u32);
|
||||
let rgb: Vec<u8> = (0..w * h * 3).map(|i| (i * 17 % 239) as u8).collect();
|
||||
let img = EncodeImage::new(w, h, 3, Samples::U8(&rgb));
|
||||
let jpeg = encode_jpeg(&img, 90, ChromaSubsampling::S420, &EncodeMeta::default()).unwrap();
|
||||
// the strip (what the IFD's chunk covers) is padded past the plan's header, which is what puts
|
||||
// IFD0 out of its reach; `embedded_preview` trims back to the JPEG's end-of-image
|
||||
let mut strip = jpeg.clone();
|
||||
strip.extend(std::iter::repeat_n(0u8, PLAN_PREFIX as usize));
|
||||
let file = tiff_with_preview_strip(&strip, 400_000, Fixture { dng: true, orientation: Some(6), ..Default::default() });
|
||||
let dir = crate::tests_xmp::temp_dir("preview-plan-orientation");
|
||||
let path = dir.join("rotated.dng");
|
||||
std::fs::write(&path, &file).unwrap();
|
||||
let path = path.to_string_lossy().into_owned();
|
||||
|
||||
assert_eq!(lightcraft_meta::extract(&file).orientation, Some(Orientation::from_exif(6)), "the fixture states an orientation");
|
||||
assert!(lightcraft_meta::extract(&file[..PLAN_PREFIX as usize]).orientation.is_none(), "which the bounded header cannot see");
|
||||
assert!(lightcraft_meta::extract(&jpeg).orientation.is_none(), "and the preview JPEG carries none of its own");
|
||||
assert!(preview_window(&path).expect("the file reads").is_none(), "the whole-file path keeps it");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
/// A container can declare what its preview JPEG is in, and the whole-file path applies that
|
||||
/// declaration (`decode_raw_preview`). A JPEG-only decode has the JPEG's own space, else sRGB, so
|
||||
/// a container that declares Adobe RGB must keep the whole-file path — while one that declares
|
||||
/// sRGB (what most bodies write) must **not** take the fast path away: that half is a regression
|
||||
/// test, since a guard that declined every declared space did exactly that to a real Sony ARW.
|
||||
#[test]
|
||||
fn preview_loading_declines_only_a_container_colour_space_it_cannot_apply() {
|
||||
use lightcraft_raw::PreviewColorSpace;
|
||||
let (w, h) = (48u32, 32u32);
|
||||
let rgb: Vec<u8> = (0..w * h * 3).map(|i| (i * 23 % 233) as u8).collect();
|
||||
let img = EncodeImage::new(w, h, 3, Samples::U8(&rgb));
|
||||
let jpeg = encode_jpeg(&img, 90, ChromaSubsampling::S420, &EncodeMeta::default()).unwrap();
|
||||
let dir = crate::tests_xmp::temp_dir("preview-plan-space");
|
||||
let path = dir.join("space.arw");
|
||||
let path = path.to_string_lossy().into_owned();
|
||||
|
||||
let adobe = tiff_with_preview_strip(&jpeg, 400_000, Fixture { sony: true, color_space: Some(0xFFFF), ..Default::default() });
|
||||
assert_eq!(lightcraft_raw::embedded_preview_color_space(&adobe), Some(PreviewColorSpace::AdobeRgb), "the fixture declares Adobe RGB");
|
||||
std::fs::write(&path, &adobe).unwrap();
|
||||
assert!(preview_window(&path).expect("the file reads").is_none(), "the whole-file path applies the space");
|
||||
|
||||
let srgb = tiff_with_preview_strip(&jpeg, 400_000, Fixture { sony: true, color_space: Some(1), ..Default::default() });
|
||||
assert_eq!(lightcraft_raw::embedded_preview_color_space(&srgb), Some(PreviewColorSpace::Srgb), "the fixture declares sRGB");
|
||||
std::fs::write(&path, &srgb).unwrap();
|
||||
let (_, window, read) = preview_window(&path).expect("the file reads").expect("sRGB is what the JPEG-only decode assumes");
|
||||
assert!(read.planned, "the plan still answers");
|
||||
assert!(window.starts_with(&[0xff, 0xd8]), "and hands on the JPEG");
|
||||
let _ = std::fs::remove_dir_all(&dir);
|
||||
}
|
||||
|
||||
/// Issue #367: probes read headers, not pixels, and report what the decode-based probe did:
|
||||
/// oriented dimensions (EXIF orientation 6 swaps them), metadata, the error for a truncated
|
||||
/// file or one that isn't an image.
|
||||
|
||||
@@ -32,10 +32,16 @@ fn a_preview_reads_only_the_header_and_the_preview() {
|
||||
read.total,
|
||||
frac * 100.0,
|
||||
elapsed.as_secs_f64(),
|
||||
if read.header_path { " [header path]" } else { " [whole file]" }
|
||||
if read.planned {
|
||||
" [plan]"
|
||||
} else if read.header_path {
|
||||
" [header path]"
|
||||
} else {
|
||||
" [whole file]"
|
||||
}
|
||||
);
|
||||
|
||||
assert!(read.header_path, "the header path must answer for a raw whose pointer is in its header");
|
||||
assert!(read.header_path, "the header or the plan must answer for a raw whose preview is not the whole file");
|
||||
assert_eq!(read.total, file_len, "the report covers the whole file");
|
||||
assert!(frac < 0.5, "a preview must not cost most of the file: read {} of {} bytes", read.read, read.total);
|
||||
assert!(preview.starts_with(&[0xff, 0xd8]), "the preview bytes are a JPEG: {:?}", preview.get(..4));
|
||||
|
||||
File diff suppressed because one or more lines are too long
@@ -28,11 +28,12 @@ libraries. Measured on one real file (`DSCF9492.RAF`, 56.5 MB) over a GVFS/SMB s
|
||||
4. When the prefix already covers the range — the usual case, since it was grown to it — the window
|
||||
is a slice of the prefix. Reading it again would double a 5.6 MB preview.
|
||||
|
||||
Formats whose preview pointer is not in the header (CR3 scans boxes; a TIFF pointer tag can sit in
|
||||
an IFD the prefix cuts short; CRW/MRW/X3F are found by scanning) keep the whole-file path, so
|
||||
nothing regresses: `preview_window` returns `None` and the loader reads the file as before. A pointer
|
||||
that lies, a short read or a file that changes underneath us ends in the same fallback — the user's
|
||||
preview is never lost to a damaged header.
|
||||
Formats whose preview pointer is not in the header — a Canon CR3's boxes, a TIFF raw whose preview
|
||||
tag sits in an IFD the prefix cuts short — are planned instead (next section but one), and what the
|
||||
planner cannot describe keeps the whole-file path, so nothing regresses: `preview_window` returns
|
||||
`None` and the loader reads the file as before. A pointer that lies, a short read, a plan that does
|
||||
not fit or a file that changes underneath us ends in the same fallback — the user's preview is never
|
||||
lost to a damaged header.
|
||||
|
||||
`LIGHTCRAFT_PROFILE=1` prints what each preview cost:
|
||||
|
||||
@@ -40,6 +41,53 @@ preview is never lost to a damaged header.
|
||||
[profile] preview …/DSCF9492.RAF: read 5857257 of 59277632 bytes (header 5857257, window 5857109)
|
||||
```
|
||||
|
||||
## What the header cannot point at: the plan
|
||||
|
||||
A Canon CR3 keeps its preview in a JPEG track, and a TIFF raw (CR2, ARW, DNG, RW2, 3FR …) can keep
|
||||
its preview tag in an IFD the header route's prefix cuts short. Neither has a field step 2 above can
|
||||
read, so `preview_window` asks `lightcraft_raw::sparse::preview_plan` — the module the
|
||||
Lightroom-catalog import probes with — which ranges `embedded_preview` reads, and reads exactly
|
||||
those:
|
||||
|
||||
- the buffer is the file's length with only the planned ranges filled in; every other byte stays zero
|
||||
and its pages are never written. Its cost is the bytes read — on macOS a freed buffer of that size
|
||||
may stay in the allocator's cache instead of going back to the OS (measured on a 38k-photo import,
|
||||
[#832](https://github.com/storytold/lightcraft/issues/832)) — and the loader's memory gate counts
|
||||
the whole file either way;
|
||||
- the module's own guarantee is what this path needs, and its corpus test asserts it: with the
|
||||
skipped bytes `0` and again `0xFF`, what a partly loaded buffer yields is what the whole file
|
||||
yields. A second corpus pass now asks `preview_plan` for **every** file, including the ones whose
|
||||
header probe reads samples and whose header plan is `Whole`, which is where that plan was untested
|
||||
before;
|
||||
- the quick path must be able to show what the whole-file path would. Two container facts are not in
|
||||
the JPEG, so a file that states one where the bounded header cannot see it declines and the whole
|
||||
file keeps the preview: a preview colour space of **Adobe RGB** (the whole-file path applies it
|
||||
through `decode_raw_preview`; a JPEG-only decode has the JPEG's own space, else sRGB — a container
|
||||
that says sRGB does not decline, since that is what the decode assumes anyway), and the file's own
|
||||
EXIF orientation when the preview JPEG carries none of its own;
|
||||
- a format the planner cannot describe — NEF, ORF and PEF probe from samples, CRW/MRW/X3F are found
|
||||
by scanning — a plan that does not fit the buffer or wants more than 60 % of the file (the bound the
|
||||
import path's own partial read uses), or a window that does not decode all answer `None`, and the
|
||||
loader reads the file whole exactly as before: the plan can cost bytes but never the picture.
|
||||
|
||||
Measured over the CC0 corpus (111 raws, `cargo xtask corpus --download`): **52 files take the plan
|
||||
route**, reading a median **7.8 %** of their bytes (0.9 %–27.7 %) — by format RW2 3.3 %, Leica RWL
|
||||
4.1 %, ARW 7.5 %, CR3 8.4 %, DNG 9.7 %, CR2 14.3 % — while the 21 RAFs stay on the header route
|
||||
(median 4.5 %). The remaining 38 read whole: NEF, NRW, ORF, PEF and Samsung SRW because the planner
|
||||
does not describe them (the first four say so; an SRW plans to the whole file, and the 60 % budget
|
||||
then keeps the whole-file read), and the ARWs, CR2s, CR3s and DNGs that declare Adobe RGB or whose
|
||||
plan wants more than that budget. Across the whole corpus the plan handed on **the same preview as
|
||||
the whole file, every time** — and wrote it down: `cargo test -p lightcraft-raw --release --test
|
||||
sparse_corpus -- --nocapture` prints these numbers. `LIGHTCRAFT_PROFILE=1` says which route answered:
|
||||
|
||||
```
|
||||
[profile] preview …/arw-sony-a7rm4a-compressed.arw: read 688128 of 61832192 bytes (plan, window 483441)
|
||||
```
|
||||
|
||||
A Canon CR3's preview is a JPEG track instead (issue #832 measured 3 MB of one R5 Mark II file,
|
||||
~15 %), and Canon is the largest bucket in [gaps.md](gaps.md)'s ranking (33 % of photographers), so
|
||||
before this change every unedited CR3 thumbnail paid for the whole file.
|
||||
|
||||
## The grid thumbnail is the embedded JPEG too
|
||||
|
||||
A 512 px grid thumbnail of an unedited raw now keeps the camera's own JPEG instead of rendering the
|
||||
@@ -66,7 +114,14 @@ render queue stayed empty.
|
||||
clamped prefix, a padded extent, a late pointer, a lying pointer and a non-JPEG window are all
|
||||
covered.
|
||||
- `crates/engine/src/files.rs`: on a synthetic RAF with a 200 KB tail, the quick path and the
|
||||
whole-file path produce identical pixels, and the read stays under half the file.
|
||||
whole-file path produce identical pixels, and the read stays under half the file. On a synthetic
|
||||
TIFF/DNG whose preview only its tags locate — the plan path — the same identity holds and the CFA
|
||||
strip is not read; the same structure without the tag that makes it a raw declines instead of
|
||||
guessing, and the caller keeps reading it whole. So do a file whose own orientation or Adobe RGB
|
||||
colour space only the whole-file path can show, while one that states sRGB stays on the fast path.
|
||||
- `crates/raw/tests/sparse_corpus.rs`: `preview_plan` is asked for **every** corpus file, not only
|
||||
the ones whose header plan is `Ready`, and the preview it yields must equal the whole file's with
|
||||
the skipped bytes `0` and `0xFF`. The test prints what each format reads.
|
||||
- `crates/engine/tests/preview_read.rs` (opt-in, `#[ignore]`): the same measurement on a **real**
|
||||
file, typically on a share:
|
||||
|
||||
|
||||
Reference in New Issue
Block a user