5a24289898fd
sidecar.rs
//! XMP sidecars: `<stem>.xmp` (default) or `<file>.xmp` next to the original.
//!
//! **Write** ([`Session::save_sidecar`], `photo.saveMetadataToFile`, or automatically after every
//! change when [`XmpPrefs::auto_write`] is on): rating, colour label, title, caption, copyright,
//! creator, keywords, capture time and GPS in the standard namespaces, plus our complete develop
//! settings (`lc:settings`, JSON), the pick/reject flag (`lc:flag`) and location (`lc:location`).
//! A sidecar is written only beside an original that is an existing file at an absolute path;
//! otherwise the save fails.
//!
//! An existing sidecar is never replaced wholesale (issue #92): LightCraft's properties
//! ([`OWNED`], [`OWNED_IF_STATED`]) are merged into it ([`lightcraft_meta::merge_xmp`]) and
//! everything else — another editor's `crs:` develop settings, `xmpMM` history, unknown
//! namespaces — is kept byte for byte. A sidecar that can't be read as XMP is first copied to
//! `<name>.xmp.bak-<time>` (never overwritten), then replaced. Writes are atomic (temp file,
//! fsync, rename).
//!
//! **Shared names** (Stem naming): `IMG_0001.CR3` and `IMG_0001.JPG` would share `IMG_0001.xmp`.
//! When two catalogued files map to one stem sidecar, it belongs to one of them (a raw first, then
//! the first by file name — [`Session::sidecar_naming`]); the others use `<file>.xmp`
//! (`IMG_0001.JPG.xmp`) for writing and reading, so neither overwrites the other. An import ranks
//! the files it is adding against the library the same way ([`ImportStems`]): a JPEG or a TIFF
//! arriving beside its raw's `IMG_0001.xmp` reads its own `<file>.xmp`, never the raw's packet
//! (issue #1905).
//!
//! **Read** (on import, and `photo.readMetadataFromFile`): the sidecar wins for metadata, except
//! the capture time: a time embedded in the file wins, and the sidecar's (`exif:DateTimeOriginal`,
//! else `photoshop:DateCreated`, else `xmp:CreateDate`) fills in only when the file has none; develop
//! settings are restored from `lc:settings` when present, else mapped from interoperable `crs:`
//! fields ([`crate::crs`]). For raw/DNG files without a sidecar the file's embedded XMP is used.
//! Manual reads also accept the first complete standard XMP APP1 in a bounded JPEG header (see `read_jpeg_xmp`).
use std::collections::{HashMap, HashSet};
use std::io::{Read, Seek, SeekFrom};
use std::path::{Path, PathBuf};
use std::sync::Arc;
use lightcraft_catalog::{ColorLabel, Flag, MediaKind, Op, Photo, PhotoId, Source};
use lightcraft_develop::DevelopSettings;
use serde::{Deserialize, Serialize};
use serde_json::Value;
use crate::{EngineError, Result, Session};
/// Sidecar file naming.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "lowercase")]
pub enum SidecarNaming {
/// `IMG_0001.xmp` (the common convention; shared by `IMG_0001.CR3` and `IMG_0001.JPG`).
#[default]
Stem,
/// `IMG_0001.CR3.xmp` (unambiguous when raw+JPEG pairs live side by side).
Full,
}
/// Library preferences for XMP sidecars.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, Serialize, Deserialize)]
#[serde(default, rename_all = "camelCase")]
pub struct XmpPrefs {
/// Write the sidecar after every change to a photo's metadata or develop settings.
pub auto_write: bool,
pub naming: SidecarNaming,
}
/// The sidecar path for an original, by naming convention.
pub fn sidecar_path(original: &str, naming: SidecarNaming) -> PathBuf {
match naming {
SidecarNaming::Stem => Path::new(original).with_extension("xmp"),
SidecarNaming::Full => PathBuf::from(format!("{original}.xmp")),
}
}
/// An existing sidecar for `original`: the preferred naming first, then the other one (also
/// accepting `.XMP` and `.Xmp`, and a full name whose photo extension differs in case, such as
/// `IMG_1.cr3.xmp` beside `IMG_1.CR3`, which a case-sensitive file system would otherwise miss).
pub fn find_sidecar(original: &str, naming: SidecarNaming) -> Option<PathBuf> {
sidecar_candidates(original, naming).find(|p| p.is_file())
}
/// Where to look for a catalogued photo's sidecar: how the library names it, and whether the photo
/// may read the shared stem sidecar (`IMG_0001.xmp`) of its stem. A photo that shares its stem
/// with a file that owns that sidecar (see [`Session::sidecar_naming`]) reads its own `<file>.xmp`
/// and never the packet of the sibling that owns the stem one (issue #1523).
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub struct SidecarLookup {
pub naming: SidecarNaming,
/// False when another file with the same stem owns the stem sidecar.
pub owns_stem: bool,
}
impl SidecarLookup {
/// The names a sidecar for `original` can have, in the order they are looked for: those of
/// [`sidecar_candidates`], without the shared stem sidecar when the photo does not own it.
pub(crate) fn candidates(self, original: &str) -> impl Iterator<Item = PathBuf> {
let shared = (!self.owns_stem).then(|| xmp_cases(sidecar_path(original, SidecarNaming::Stem)));
sidecar_candidates(original, self.naming).filter(move |p| shared.as_ref().is_none_or(|shared| !shared.contains(p)))
}
/// [`find_sidecar`], except that a photo that does not own its stem sidecar never reads it.
pub fn find(self, original: &str) -> Option<PathBuf> {
self.candidates(original).find(|p| p.is_file())
}
}
/// Find the sidecar for a file an import is adding, by the ownership [`ImportStems`] worked out
/// (issue #1905): [`find_sidecar`] with the library's naming, except that a file which does not
/// own its stem sidecar reads its own `<file>.xmp` (the name it writes) and never the packet of
/// the sibling that owns the stem one.
pub(crate) fn find_sidecar_for_import(stems: &ImportStems, naming: SidecarNaming, original: &str) -> Option<PathBuf> {
SidecarLookup { naming, owns_stem: stems.owns(original) }.find(original)
}
/// The names `original`'s sidecar can have, in the order they are looked for: under `naming`,
/// then under the other naming (the full name also with the photo extension in the other case),
/// each as `.xmp`, `.XMP` and `.Xmp`.
pub(crate) fn sidecar_candidates(original: &str, naming: SidecarNaming) -> impl Iterator<Item = PathBuf> {
let other = if naming == SidecarNaming::Stem { SidecarNaming::Full } else { SidecarNaming::Stem };
let ext = Path::new(original).extension().map(|e| e.to_string_lossy().to_string()).unwrap_or_default();
let base = Path::new(original).with_extension("");
let recased = |e: String| (e != ext).then(|| format!("{}.{e}", base.to_string_lossy()));
let full_variants: Vec<String> = [recased(ext.to_lowercase()), recased(ext.to_uppercase())].into_iter().flatten().collect();
[naming, other]
.into_iter()
.flat_map(move |n| {
let mut v = vec![sidecar_path(original, n)];
if n == SidecarNaming::Full {
v.extend(full_variants.iter().map(|o| sidecar_path(o, n)));
}
v
})
.flat_map(xmp_cases)
}
/// The three ways one sidecar name can be written: `.xmp`, `.XMP` and `.Xmp`.
fn xmp_cases(p: PathBuf) -> [PathBuf; 3] {
[p.clone(), p.with_extension("XMP"), p.with_extension("Xmp")]
}
/// What a sidecar (or embedded XMP packet) says about a photo. `None` = not stated.
#[derive(Clone, Debug, Default, PartialEq)]
pub struct SidecarData {
pub rating: Option<u8>,
pub flag: Option<Flag>,
pub label: Option<Option<ColorLabel>>,
/// The label's text as written (`xmp:Label`), for [`SidecarData::resolve_label`].
pub label_text: Option<String>,
pub title: Option<String>,
pub caption: Option<String>,
pub copyright: Option<String>,
/// `xmpRights:Marked`: `Some(true)` copyrighted, `Some(false)` public domain.
pub copyright_marked: Option<bool>,
pub usage_terms: Option<String>,
pub copyright_url: Option<String>,
pub creator: Option<String>,
pub location: Option<String>,
pub city: Option<String>,
pub state: Option<String>,
pub country: Option<String>,
pub alt_text: Option<String>,
pub extended_description: Option<String>,
pub keywords: Option<Vec<String>>,
pub regions: Option<Vec<lightcraft_meta::Region>>,
/// The size the regions were drawn on, when they don't say how the photo was turned
/// ([`lightcraft_meta::XmpData::regions_drawn_on`]), and how the photo's stored pixels turn upright: the
/// sidecar's `tiff:Orientation`, or the file's own Exif orientation where the caller knows it. With both,
/// [`merge_into`] puts boxes drawn on the photo as stored where the upright photo has them (#2674).
pub regions_drawn_on: Option<(f64, f64)>,
pub orientation: Option<lightcraft_meta::Orientation>,
/// Capture time (ISO 8601) from `exif:DateTimeOriginal`, `photoshop:DateCreated` or
/// `xmp:CreateDate` (first found). Used only when the file itself has no capture time.
pub captured: Option<String>,
pub develop: Option<DevelopPatch>,
/// Actionable diagnostics for foreign develop settings that were not fully carried over.
/// These are surfaced in the photo import report and never persisted in the catalog.
pub warnings: Vec<crate::import::ImportWarning>,
}
#[derive(Clone, Debug, PartialEq)]
pub enum DevelopPatch {
/// Our own complete settings (`lc:settings`).
Full(Box<DevelopSettings>),
/// Mapped `crs:` fields (partial JSON, merged like a preset).
Partial(Value),
}
/// Parse a sidecar / XMP packet. `target` selects how a `crs:` white balance is read
/// ([`crate::crs::Target`]: absolute Kelvin for raws with a measured illuminant, a shift from the
/// as-shot white for everything developed relative to it).
pub fn parse_sidecar(xmp: &str, target: crate::crs::Target) -> std::result::Result<SidecarData, String> {
let d = lightcraft_meta::parse_xmp(xmp).map_err(|e| e.to_string())?;
let m = &d.metadata;
let lc = |k: &str| d.properties.get(&format!("lc:{k}")).and_then(|v| v.first()).cloned();
let mut out = SidecarData {
title: m.title.clone(),
caption: m.caption.clone(),
copyright: m.copyright.clone(),
copyright_marked: m.copyright_marked,
usage_terms: m.usage_terms.clone(),
copyright_url: m.copyright_url.clone(),
creator: m.artist.clone(),
location: lc("location").or_else(|| m.sublocation.clone()),
city: m.city.clone(),
state: m.state.clone(),
country: m.country.clone(),
alt_text: m.alt_text.clone(),
extended_description: m.extended_description.clone(),
keywords: (!m.keywords.is_empty() || !m.hierarchical_keywords.is_empty())
.then(|| lightcraft_catalog::keywords::from_file(&m.keywords, &m.hierarchical_keywords)),
// A sidecar that has `mwg-rs:Regions` at all (even an empty list) was written by an app that
// knows about regions, so it's authoritative: its list, empty or not, replaces the catalog's.
// One without it (most writers, LightCraft's own included, which keeps another app's
// `mwg-rs:Regions` byte for byte but never writes one) says nothing about regions, and the
// photo's are kept.
regions: d.values.contains_key("mwg-rs:Regions").then(|| m.regions.clone()),
regions_drawn_on: d.regions_drawn_on(),
// (only what the regions need: a packet that says nothing else still says nothing)
orientation: d.regions_drawn_on().and(m.orientation),
captured: m.capture_time.map(|d| d.to_iso()),
..Default::default()
};
if let Some(r) = m.rating {
if r < 0 {
out.flag = Some(Flag::Reject);
out.rating = Some(0);
} else {
out.rating = Some(r.clamp(0, 5) as u8);
}
}
if let Some(f) = lc("flag").and_then(|f| Flag::parse(&f)) {
out.flag = Some(f);
}
if let Some(l) = &m.label {
out.label = Some(ColorLabel::parse(l.trim()));
out.label_text = Some(l.trim().to_string());
}
let full =
d.lc_settings.as_deref().and_then(|j| serde_json::from_str::<Value>(j).ok()).and_then(|v| DevelopSettings::from_untrusted_json(&v).ok());
out.develop = match full {
Some(s) => Some(DevelopPatch::Full(Box::new(s))),
None if crate::crs::has_adjustments(&d.properties) => {
let (partial, unmapped) = crate::crs::to_partial_report(&d.properties, Some(&d.values), target, crate::crs_masks::DEFAULT_ASPECT);
// Lightroom's default profiles are named by nearly every sidecar; our own starting look stands
// in for them, so only another profile is worth a warning
let profile = crate::crs::camera_profile_name(&d.properties)
.filter(|name| !matches!(name.trim(), "Adobe Standard" | "Adobe Color"))
.map(str::to_string);
if let Some(name) = profile {
out.warnings.push(crate::import::ImportWarning {
path: String::new(),
code: crate::import::ImportWarningCode::CameraProfile,
profile: Some(name),
mode: None,
fields: Vec::new(),
});
}
let upright = crate::crs::has_saved_upright_transform(&d.properties);
if upright {
let mode = d
.properties
.get("crs:PerspectiveUpright")
.and_then(|values| values.first())
.map(String::as_str)
.map(|mode| match mode {
"0" => crate::import::ImportUprightMode::Off,
"1" => crate::import::ImportUprightMode::Auto,
"2" => crate::import::ImportUprightMode::Level,
"3" => crate::import::ImportUprightMode::Vertical,
"4" => crate::import::ImportUprightMode::Full,
"5" => crate::import::ImportUprightMode::Guided,
_ => crate::import::ImportUprightMode::Unknown,
})
.unwrap_or(crate::import::ImportUprightMode::Unknown);
out.warnings.push(crate::import::ImportWarning {
path: String::new(),
code: crate::import::ImportWarningCode::UprightGeometry,
profile: None,
mode: Some(mode),
fields: Vec::new(),
});
}
let other: Vec<&str> = unmapped
.iter()
.map(String::as_str)
.filter(|name| {
*name != "CameraProfile"
&& !name.starts_with("UprightTransform_")
&& !matches!(*name, "UprightTransformCount" | "UprightCenterNormX" | "UprightCenterNormY" | "UprightVersion")
})
.collect();
if !other.is_empty() {
out.warnings.push(crate::import::ImportWarning {
path: String::new(),
code: crate::import::ImportWarningCode::UnmappedXmp,
profile: None,
mode: None,
fields: other.into_iter().map(str::to_string).collect(),
});
}
Some(DevelopPatch::Partial(partial))
}
None => None,
};
Ok(out)
}
/// Apply sidecar data to a photo record (the sidecar wins, except that a capture time embedded in
/// the file is kept). Returns true if develop changed.
pub fn merge_into(p: &mut Photo, sc: &SidecarData, now: &str) -> bool {
if p.captured.is_none() {
p.captured = sc.captured.clone();
}
if let Some(r) = sc.rating {
p.rating = r;
}
if let Some(f) = sc.flag {
p.flag = f;
}
if let Some(l) = sc.label {
p.label = l;
}
let m = &mut p.meta;
for (src, dst) in [
(&sc.title, &mut m.title),
(&sc.caption, &mut m.caption),
(&sc.copyright, &mut m.copyright),
(&sc.usage_terms, &mut m.usage_terms),
(&sc.copyright_url, &mut m.copyright_url),
(&sc.creator, &mut m.creator),
(&sc.location, &mut m.location),
(&sc.city, &mut m.city),
(&sc.state, &mut m.state),
(&sc.country, &mut m.country),
(&sc.alt_text, &mut m.alt_text),
(&sc.extended_description, &mut m.extended_description),
] {
if let Some(v) = src {
*dst = v.clone();
}
}
if let Some(marked) = sc.copyright_marked {
m.copyright_status = lightcraft_catalog::CopyrightStatus::from_marked(Some(marked));
}
if let Some(k) = &sc.keywords {
m.keywords = k.clone();
}
if let Some(r) = &sc.regions {
m.regions = r.clone();
// (the photo's size is the upright one)
if let (Some(drawn_on), Some(o)) = (sc.regions_drawn_on, sc.orientation) {
let stored = if o.swaps_axes() { (p.height, p.width) } else { (p.width, p.height) };
lightcraft_meta::upright_regions(&mut m.regions, drawn_on, o, stored);
}
}
let develop = match &sc.develop {
Some(DevelopPatch::Full(s)) => (**s).clone(),
Some(DevelopPatch::Partial(v)) => {
let mut v = v.clone();
if p.width > 0 && p.height > 0 {
crate::crs_masks::refit_radials(&mut v, crate::crs_masks::DEFAULT_ASPECT, p.width as f64 / p.height as f64);
}
lightcraft_develop::apply_partial(&p.develop, &v, 1.0)
}
None => return false,
};
if develop == *p.develop {
return false;
}
p.develop = Arc::new(develop);
p.edited = Some(now.to_string());
true
}
impl SidecarData {
/// Read the label through the catalog's label names: other editors write the label's name
/// (a custom set's "To Do" as well as "Red"); unknown names clear the label.
pub fn resolve_label(mut self, cat: &lightcraft_catalog::Catalog) -> Self {
if let Some(t) = &self.label_text {
self.label = Some(if t.is_empty() { None } else { cat.label_from_name(t) });
}
self
}
/// Take the orientation of the photo file at `original` (LightCraft shows the photo turned by it, and the packet
/// may not repeat it) when the regions need it: they were drawn on a stated size ([`SidecarData::regions_drawn_on`]).
/// Only then is the file read ([`file_orientation`]).
pub fn with_file_orientation(self, original: &str) -> Self {
if self.regions_drawn_on.is_none() {
return self;
}
SidecarData { orientation: file_orientation(original).or(self.orientation), ..self }
}
}
/// The orientation the photo file at `path` states, matching what the import stores on [`crate::media::ProbeInfo`]:
/// for a raw, the decoder's (`lightcraft_raw::probe_info`: CIFF for a CRW, the raw IFD for a MEF, …); otherwise Exif,
/// else XMP, as [`lightcraft_meta::extract`] reads it. A file shorter than [`ORIENTATION_SPAN`] is read whole. Most
/// non-raws state it in the Exif of their first [`ORIENTATION_SPAN`]; when that part doesn't, the Exif may come later
/// ([`late_orientation`]). A WebP's or PNG's chunks are always walked, since the last Exif chunk is the one that counts
/// ([`webp_orientation`], [`png_orientation`]).
fn file_orientation(path: &str) -> Option<lightcraft_meta::Orientation> {
let mut file = std::fs::File::open(path).ok()?;
let mut head = Vec::new();
(&mut file).take(ORIENTATION_SPAN).read_to_end(&mut head).ok()?;
// Import takes a raw's orientation from the decoder, not from Exif IFD0 / XMP alone. Re-read and sync must do the
// same, or face boxes turned at import are put back and Synchronize Folder reports changes every pass (#2857).
if lightcraft_raw::probe(&head).is_some() {
let bytes = if (head.len() as u64) < ORIENTATION_SPAN {
head
} else {
file.rewind().ok()?;
let mut all = Vec::new();
file.read_to_end(&mut all).ok()?;
all
};
return match lightcraft_raw::probe_info(&bytes) {
Ok(raw) => Some(raw.orientation),
// unsupported / preview-only: same fallback the import uses (`ProbeInfo::orientation` from meta)
Err(_) => lightcraft_meta::extract(&bytes).orientation,
};
}
if (head.len() as u64) < ORIENTATION_SPAN {
return lightcraft_meta::extract(&head).orientation;
}
if head.get(..4) == Some(b"RIFF") && head.get(8..12) == Some(b"WEBP") {
return webp_orientation(&mut file);
}
if head.starts_with(b"\x89PNG\r\n\x1a\n") {
return png_orientation(&mut file);
}
if let Some(o) = exif_orientation(&head) {
return Some(o);
}
late_orientation(&mut file, &head).or_else(|| lightcraft_meta::extract(&head).orientation)
}
/// The orientation the Exif in `bytes` (a file, or the start of one) states, leaving out a separate XMP packet (the
/// one the Exif holds in tag 700 counts, as in [`lightcraft_meta::read_exif`]).
fn exif_orientation(bytes: &[u8]) -> Option<lightcraft_meta::Orientation> {
lightcraft_meta::embedded(bytes).exif.as_deref().map(lightcraft_meta::read_exif).and_then(|m| m.orientation)
}
/// The orientation of a JPEG, TIFF or PSD whose `head` has no Exif stating it, found by the file's structure without
/// reading its image data (a JPEG's Exif can follow other long segments, many TIFF writers put the first IFD after the
/// strips, and a PSD's image resources can outgrow `head`). As the import does, a JPEG takes its first Exif segment,
/// else its first XMP segment.
fn late_orientation(file: &mut (impl Read + Seek), head: &[u8]) -> Option<lightcraft_meta::Orientation> {
let len = file.seek(SeekFrom::End(0)).ok()?;
file.rewind().ok()?;
match head.get(..4)? {
[0xff, 0xd8, ..] => {
let (mut exif_seen, mut xmp) = (false, None);
let exif = read_jpeg_app1(file, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS, |segment| {
let embedded = lightcraft_meta::embedded(segment);
if let Some(exif) = embedded.exif
&& !exif_seen
{
// the first Exif segment ends the walk if it states one; later ones don't count
exif_seen = true;
return lightcraft_meta::read_exif(&exif).orientation;
}
if xmp.is_none() {
xmp = embedded.xmp.map(|x| lightcraft_meta::parse_xmp(&x).ok().and_then(|x| x.metadata.orientation));
}
None
});
exif.or(xmp.flatten())
}
b"II*\0" | b"MM\0*" => tiff_orientation(file, 0, len),
b"8BPS" => psd_orientation(file),
_ => None,
}
}
/// A WebP's orientation as the import takes it: from its last `EXIF` chunk, else its last `XMP ` chunk, in whichever
/// order they come. The chunk headers are walked from the start of `file` (at most [`WEBP_CHUNK_STEPS`], ending at a
/// chunk that runs past the end of the file); then only the first IFD of the EXIF chunk is read, whatever its size
/// ([`tiff_orientation`]), and the XMP chunk if it is at most [`ORIENTATION_SPAN`].
fn webp_orientation(file: &mut (impl Read + Seek)) -> Option<lightcraft_meta::Orientation> {
let len = file.seek(SeekFrom::End(0)).ok()?;
let (mut exif, mut xmp) = (None, None);
let mut at = 12u64;
for _ in 0..WEBP_CHUNK_STEPS {
let mut header = [0; 8];
if file.seek(SeekFrom::Start(at)).is_err() || file.read_exact(&mut header).is_err() {
break;
}
let [a, b, c, d, s0, s1, s2, s3] = header;
let size = u64::from(u32::from_le_bytes([s0, s1, s2, s3]));
let Some(end) = at.checked_add(8 + size).filter(|&end| end <= len) else { break };
match &[a, b, c, d] {
b"EXIF" => exif = Some((at + 8, end)),
b"XMP " => xmp = Some((at + 8, size)),
_ => {}
}
// chunks are padded to an even size
at = end + (size & 1);
}
if let Some(o) = exif.and_then(|(start, end)| exif_chunk_orientation(file, start, end)) {
return Some(o);
}
let (start, size) = xmp?;
xmp_orientation(file, start, size)
}
/// How many chunks [`webp_orientation`] passes over (an animated WebP has one per frame).
const WEBP_CHUNK_STEPS: usize = 4096;
/// The orientation the Exif chunk from `start` to `end` in `file` states: its TIFF stream comes after an `Exif\0\0` if
/// the chunk starts with one (as `strip_exif_header` takes it).
fn exif_chunk_orientation(file: &mut (impl Read + Seek), start: u64, end: u64) -> Option<lightcraft_meta::Orientation> {
let mut prefix = [0; 5];
file.seek(SeekFrom::Start(start)).ok()?;
let prefixed = start + 6 <= end && file.read_exact(&mut prefix).is_ok() && &prefix == b"Exif\0";
tiff_orientation(file, if prefixed { start + 6 } else { start }, end)
}
/// A PNG's orientation as the import takes it: from its last `eXIf` chunk, else its last uncompressed XMP `iTXt` chunk,
/// wherever they come (a writer may put them after the image data). The chunk headers are walked from the start of
/// `file` to `IEND` (at most [`png_chunk_steps`], ending at a chunk that runs past the end of the file); then only the
/// first IFD of the `eXIf` chunk is read, whatever its size ([`tiff_orientation`]), and the `iTXt` chunk if it is at
/// most [`ORIENTATION_SPAN`].
fn png_orientation(file: &mut (impl Read + Seek)) -> Option<lightcraft_meta::Orientation> {
let len = file.seek(SeekFrom::End(0)).ok()?;
let (mut exif, mut xmp) = (None, None);
let mut at = 8u64;
for _ in 0..png_chunk_steps(len) {
let (mut header, mut key) = ([0; 8], [0; 19]);
if at + 12 > len || file.seek(SeekFrom::Start(at)).is_err() || file.read_exact(&mut header).is_err() {
break;
}
let [s0, s1, s2, s3, a, b, c, d] = header;
let size = u64::from(u32::from_be_bytes([s0, s1, s2, s3]));
let Some(end) = at.checked_add(8 + size).filter(|&end| end <= len) else { break };
match &[a, b, c, d] {
b"eXIf" => exif = Some((at + 8, end)),
// the keyword, then compression flag 0
b"iTXt" if size >= 20 && file.read_exact(&mut key).is_ok() && &key == b"XML:com.adobe.xmp\0\0" => xmp = Some((at + 8, size)),
b"IEND" => break,
_ => {}
}
// past the chunk's CRC
at = end + 4;
}
if let Some(o) = exif.and_then(|(start, end)| exif_chunk_orientation(file, start, end)) {
return Some(o);
}
let (start, size) = xmp?;
let chunk = read_within_span(file, start, size)?;
// keyword\0 compression-flag compression-method language\0 translated\0 text
let text = chunk.get(20..)?.splitn(3, |&c| c == 0).nth(2)?;
lightcraft_meta::parse_xmp(&String::from_utf8_lossy(text)).ok()?.metadata.orientation
}
/// How many chunks [`png_orientation`] passes over in a file of `len` bytes: 4096 and one per KiB of the file, at most
/// 2^20. Image data is often split into many `IDAT` chunks (libpng writes 8 KiB ones by default), so the limit grows
/// with the file; each chunk costs two system calls, so a short file of empty chunks can't hold a worker for seconds.
fn png_chunk_steps(len: u64) -> u64 {
(4096 + len / 1024).min(1 << 20)
}
/// A PSD's or PSB's orientation as the import takes it, from its image resources: the file's start up to their end is
/// read if it is at most [`PSD_RESOURCES_SPAN`]. They come before the layers and the image data, after the 26-byte
/// header and the colour-mode data; each of those two sections starts with its length.
fn psd_orientation(file: &mut (impl Read + Seek)) -> Option<lightcraft_meta::Orientation> {
let mut length = [0; 4];
file.seek(SeekFrom::Start(26)).ok()?;
file.read_exact(&mut length).ok()?;
let resources = 30 + u64::from(u32::from_be_bytes(length));
file.seek(SeekFrom::Start(resources)).ok()?;
file.read_exact(&mut length).ok()?;
let end = resources + 4 + u64::from(u32::from_be_bytes(length));
if end > PSD_RESOURCES_SPAN {
return None;
}
file.rewind().ok()?;
let mut bytes = Vec::new();
(&mut *file).take(end).read_to_end(&mut bytes).ok()?;
lightcraft_meta::extract(&bytes).orientation
}
/// How far into a PSD or PSB [`psd_orientation`] reads.
const PSD_RESOURCES_SPAN: u64 = 16 * 1024 * 1024;
/// The orientation the TIFF stream from `start` to `end` in `file` states in its first IFD (a TIFF file, or the Exif
/// block of a WebP's EXIF chunk; its offsets count from `start`, and nothing past `end` is read). Only the IFD's 12-byte
/// entries are read (at most 65535 of them, under 1 MiB): its orientation tag, else (as [`lightcraft_meta::read_exif`]
/// does) the XMP packet its tag 700 holds, read if it is at most [`ORIENTATION_SPAN`].
fn tiff_orientation(file: &mut (impl Read + Seek), start: u64, end: u64) -> Option<lightcraft_meta::Orientation> {
// where the `len` bytes at `offset` in the stream are in the file, when all of them are in the stream
let within = |offset: u32, len: u64| {
let at = start.checked_add(u64::from(offset))?;
(at.checked_add(len)? <= end).then_some(at)
};
file.seek(SeekFrom::Start(within(0, 8)?)).ok()?;
let mut header = [0; 8];
file.read_exact(&mut header).ok()?;
let [b0, b1, b2, b3, i0, i1, i2, i3] = header;
let le = match &[b0, b1, b2, b3] {
b"II*\0" => true,
b"MM\0*" => false,
_ => return None,
};
let u16_of = |b: [u8; 2]| if le { u16::from_le_bytes(b) } else { u16::from_be_bytes(b) };
let u32_of = |b: [u8; 4]| if le { u32::from_le_bytes(b) } else { u32::from_be_bytes(b) };
let ifd = u32_of([i0, i1, i2, i3]);
file.seek(SeekFrom::Start(within(ifd, 2)?)).ok()?;
let mut count = [0; 2];
file.read_exact(&mut count).ok()?;
let count = u16_of(count);
within(ifd, 2 + 12 * u64::from(count))?;
let mut entries = vec![0; usize::from(count) * 12];
file.read_exact(&mut entries).ok()?;
let mut xmp = None;
for &[t0, t1, y0, y1, c0, c1, c2, c3, v0, v1, v2, v3] in entries.as_chunks::<12>().0 {
let (tag, kind, count) = (u16_of([t0, t1]), u16_of([y0, y1]), u32_of([c0, c1, c2, c3]));
match (tag, kind) {
// exactly one SHORT (with more, the field also holds a second value, or an offset to them all)
(274, 3) if count == 1 => {
let v = u16_of([v0, v1]);
if (1..=8).contains(&v) {
return Some(lightcraft_meta::Orientation::from_exif(v));
}
}
// BYTE or UNDEFINED, at an offset (a packet longer than the field's four bytes)
(700, 1 | 7) if count > 4 => xmp = Some((u32_of([v0, v1, v2, v3]), u64::from(count))),
_ => {}
}
}
let (offset, len) = xmp?;
xmp_orientation(file, within(offset, len)?, len)
}
/// The orientation the XMP packet of `len` bytes at `at` in `file` states, read if it is at most [`ORIENTATION_SPAN`].
fn xmp_orientation(file: &mut (impl Read + Seek), at: u64, len: u64) -> Option<lightcraft_meta::Orientation> {
let packet = read_within_span(file, at, len)?;
lightcraft_meta::parse_xmp(&String::from_utf8_lossy(&packet)).ok()?.metadata.orientation
}
/// The `len` bytes at `at` in `file`, read if they are at most [`ORIENTATION_SPAN`].
fn read_within_span(file: &mut (impl Read + Seek), at: u64, len: u64) -> Option<Vec<u8>> {
if len > ORIENTATION_SPAN {
return None;
}
file.seek(SeekFrom::Start(at)).ok()?;
let mut bytes = Vec::new();
(&mut *file).take(len).read_to_end(&mut bytes).ok()?;
Some(bytes)
}
/// How much of a photo file [`file_orientation`] reads first.
const ORIENTATION_SPAN: u64 = 1 << 20;
/// The sidecar packet for a photo; the colour label is written by its name in `cat` (custom
/// label names included, as other editors do).
pub fn sidecar_packet(p: &Photo, cat: &lightcraft_catalog::Catalog) -> String {
sidecar_packet_with(p, cat, false)
}
/// [`sidecar_packet`], without any location (GPS, sublocation, city, state, country) when
/// `remove_location` is set (an export with "Remove location info").
pub fn sidecar_packet_with(p: &Photo, cat: &lightcraft_catalog::Catalog, remove_location: bool) -> String {
let nz = |s: &str| (!s.trim().is_empty()).then(|| s.to_string());
// Names in `dc:subject`, paths in `lr:hierarchicalSubject` (`keywords::to_file`).
let keywords = lightcraft_catalog::keywords::to_file(&p.meta.keywords);
let meta = lightcraft_meta::Metadata {
software: Some("LightCraft".into()),
title: nz(&p.meta.title),
caption: nz(&p.meta.caption),
alt_text: nz(&p.meta.alt_text),
extended_description: nz(&p.meta.extended_description),
sublocation: nz(&p.meta.location),
city: nz(&p.meta.city),
state: nz(&p.meta.state),
country: nz(&p.meta.country),
copyright: nz(&p.meta.copyright),
copyright_marked: p.meta.copyright_status.marked(),
usage_terms: nz(&p.meta.usage_terms),
copyright_url: nz(&p.meta.copyright_url),
artist: nz(&p.meta.creator),
keywords: keywords.flat,
hierarchical_keywords: keywords.hierarchical,
rating: Some(p.rating.min(5) as i8),
label: p.label.map(|l| cat.label_name(l)),
capture_time: p.captured.as_deref().and_then(lightcraft_meta::DateTime::parse_iso),
gps: p.meta.gps.map(|(latitude, longitude)| lightcraft_meta::Gps { latitude, longitude, altitude: None }),
..Default::default()
};
let mut meta = meta;
if remove_location {
(meta.sublocation, meta.city, meta.state, meta.country, meta.gps) = (None, None, None, None, None);
}
let settings = serde_json::to_string(&*p.develop).unwrap_or_default();
let flag = match p.flag {
Flag::None => "none",
Flag::Pick => "pick",
Flag::Reject => "reject",
};
let mut lc = vec![("settings", settings.as_str()), ("flag", flag)];
if !remove_location && !p.meta.location.trim().is_empty() {
lc.push(("location", p.meta.location.as_str()));
}
lightcraft_meta::write_xmp_lc(&meta, &lc)
}
/// The properties LightCraft writes and owns in a sidecar: replaced on every save (removed when
/// LightCraft has no value, so clearing a field clears it in the file too). All of them are read
/// back on import and by Read Metadata from File.
pub const OWNED: &[&str] = &[
"xmp:Rating",
"xmp:Label",
"dc:title",
"dc:description",
"dc:rights",
"dc:creator",
"dc:subject",
"lr:hierarchicalSubject",
"Iptc4xmpCore:Location",
"Iptc4xmpCore:AltTextAccessibility",
"Iptc4xmpCore:ExtDescrAccessibility",
"photoshop:City",
"photoshop:State",
"photoshop:Country",
"xmpRights:Marked",
"xmpRights:UsageTerms",
"xmpRights:WebStatement",
"lc:*",
];
/// Properties replaced only when LightCraft has a value (capture time; GPS, which LightCraft
/// doesn't read from sidecars): otherwise the sidecar's own value stays. Anything else LightCraft
/// writes (`xmp:CreatorTool`) goes only into new sidecars.
pub const OWNED_IF_STATED: &[&str] = &["exif:DateTimeOriginal", "photoshop:DateCreated", "exif:GPSLatitude", "exif:GPSLongitude"];
/// Write `data` to `path` atomically: a temp file next to it, synced, then renamed over it.
fn write_atomic(path: &Path, data: &[u8]) -> std::io::Result<()> {
lightcraft_catalog::safe_file::write_atomic(path, data)
}
/// Copy `path` to a new `<name>.bak-<stamp>` (`-1`, `-2`… if taken; never overwriting).
fn backup(path: &Path, stamp: &str) -> std::io::Result<PathBuf> {
use std::io::Write;
// only a plain file is copied aside: a link under the sidecar's name would have its target's
// contents copied into a new, readable file
if !std::fs::symlink_metadata(path)?.is_file() {
return Err(std::io::Error::other(format!("{} is not a regular file", path.display())));
}
let bytes = crate::read_file::read_regular(path, SIDECAR_READ_MAX)?;
let stamp: String = stamp.chars().filter(char::is_ascii_alphanumeric).collect();
let name = path.file_name().map(|n| n.to_string_lossy().to_string()).unwrap_or_default();
for k in 0..1000 {
let b = path.with_file_name(if k == 0 { format!("{name}.bak-{stamp}") } else { format!("{name}.bak-{stamp}-{k}") });
match std::fs::OpenOptions::new().write(true).create_new(true).open(&b) {
Ok(mut f) => {
f.write_all(&bytes)?;
f.sync_all()?;
return Ok(b);
}
Err(e) if e.kind() == std::io::ErrorKind::AlreadyExists => continue,
Err(e) => return Err(e),
}
}
Err(std::io::Error::other("no free backup name"))
}
/// What [`Session::save_sidecar`] did.
#[derive(Clone, Debug, PartialEq)]
pub struct SidecarSaved {
pub path: PathBuf,
/// An existing sidecar was merged into (its other contents kept).
pub merged: bool,
/// An existing sidecar that couldn't be read as XMP was copied here before being replaced.
pub backup: Option<PathBuf>,
}
/// The new contents for the sidecar at `path`: `fresh` merged into the existing file, if any.
/// Returns (contents, merged, backup).
fn sidecar_contents(path: &Path, fresh: String, stamp: &str) -> std::io::Result<(String, bool, Option<PathBuf>)> {
let bytes = match crate::read_file::read_regular(path, SIDECAR_READ_MAX) {
Ok(b) => b,
Err(e) if e.kind() == std::io::ErrorKind::NotFound => return Ok((fresh, false, None)),
// can't see what is there: never overwrite it blindly
Err(e) => return Err(e),
};
if bytes.iter().all(u8::is_ascii_whitespace) {
return Ok((fresh, false, None));
}
let rules = lightcraft_meta::MergeRules { owned: OWNED, owned_if_present: OWNED_IF_STATED };
let why = match std::str::from_utf8(&bytes) {
Ok(old) => match lightcraft_meta::merge_xmp(old, &fresh, rules) {
Ok(merged) => return Ok((merged, true, None)),
// a readable sidecar whose merge fails (e.g. the merged packet exceeds the XMP limits), or one
// that is valid but over the parser's complexity limits: fail the save and keep the file
// as it is, rather than replace metadata that could be read
// (text with no `rdf:RDF` element parses as empty metadata but holds nothing to keep: it is replaced
// after a backup, as any sidecar that is not XMP)
Err(why)
if why == lightcraft_meta::XmpError::TooComplex
|| (lightcraft_meta::parse_xmp(old.strip_prefix('\u{feff}').unwrap_or(old)).is_ok()
&& !matches!(&why, lightcraft_meta::XmpError::Xml(m) if m.contains("no rdf:RDF element"))) =>
{
return Err(std::io::Error::other(format!("the sidecar is readable but could not be merged ({why}); it was left unchanged")));
}
Err(why) => why.to_string(),
},
Err(why) => why.to_string(),
};
let b = backup(path, stamp)?;
log::warn!("{}: not readable as XMP ({why}); kept a copy as {} before replacing it", path.display(), b.display());
Ok((fresh, false, Some(b)))
}
/// Key of the stem sidecar a file maps to (case-insensitive).
fn stem_key(original: &str) -> String {
sidecar_path(original, SidecarNaming::Stem).to_string_lossy().to_lowercase()
}
/// Where a plain path is cut to make its [`stem_key`]: at the dot of its extension, which is
/// what `with_extension` replaces. A plain path is ASCII and ends in its file's name with an
/// extension, which is what a library's paths are; `None` for any other shape.
fn plain_cut(original: &str) -> Option<usize> {
let path = Path::new(original);
let (name, stem, ext) = (path.file_name()?.to_str()?, path.file_stem()?.to_str()?, path.extension()?.to_str()?);
// (the path ends in its file's name, as it stands: no `/.` or separator after it; and the
// name has a stem and an extension of its own, which `..jpg` and `a.` have not)
let plain = original.is_ascii() && original.ends_with(name) && !ext.is_empty() && !stem.is_empty() && stem != "." && stem != "..";
let cut = original.len().checked_sub(ext.len().saturating_add(1))?;
(plain && original.as_bytes().get(cut) == Some(&b'.')).then_some(cut)
}
/// [`stem_key`] written into `key`, which is reused: nothing is allocated for a plain path
/// ([`plain_cut`]). Any other shape goes the long way.
fn stem_key_into(key: &mut String, original: &str) {
key.clear();
match plain_cut(original) {
Some(cut) => {
key.push_str(original.get(..cut).unwrap_or_default());
key.make_ascii_lowercase();
key.push_str(".xmp");
}
None => key.push_str(&stem_key(original)),
}
}
fn file_path(p: &Photo) -> Option<&str> {
match &p.source {
Source::File { path } => Some(path),
Source::Demo { .. } => None,
}
}
/// The XMP packet that describes a photo on disk: its sidecar, else embedded raw/DNG or JPEG XMP.
pub fn read_packet(original: &str, kind: MediaKind, lookup: SidecarLookup) -> Option<(String, PathBuf)> {
if let Some(sc) = lookup.find(original) {
// A discovered sidecar keeps precedence even when reading it fails.
return crate::read_file::read_regular(&sc, SIDECAR_READ_MAX).ok().and_then(|b| String::from_utf8(b).ok()).map(|s| (s, sc));
}
if kind == MediaKind::Raw {
let bytes = crate::read_file::read_regular(original, crate::read_file::MAX_MEDIA_BYTES).ok()?;
let x = lightcraft_meta::embedded(&bytes).xmp?;
return Some((x, PathBuf::from(original)));
}
if kind == MediaKind::Image {
let mut file = std::fs::File::open(original).ok()?;
let x = read_jpeg_xmp(&mut file, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS)?;
return Some((x, PathBuf::from(original)));
}
None
}
/// Most bytes of a sidecar read whole (the opened file must be a regular file).
const SIDECAR_READ_MAX: u64 = 256 << 20;
const JPEG_XMP_HEADER_SPAN: u64 = 16 * 1024 * 1024;
const JPEG_XMP_HEADER_STEPS: usize = 4096;
/// Read from SOI at the current position, stopping at the first standard XMP packet or SOS/EOI.
/// The span includes SOI, segment headers and skipped payloads; each marker/fill step after SOI
/// consumes one iteration. Only a complete APP1 payload (at most 65533 bytes) reaches the shared
/// parser, in a single SOI+APP1 wrapper (at most 65539 bytes). This does not validate the whole JPEG.
fn read_jpeg_xmp(reader: &mut (impl Read + Seek), header_span: u64, marker_steps: usize) -> Option<String> {
read_jpeg_app1(reader, header_span, marker_steps, |segment| lightcraft_meta::jpeg_segments(segment).xmp)
}
/// Like [`read_jpeg_xmp`], stopping at the first APP1 segment (in its SOI+APP1 wrapper) that `pick` finds something in.
fn read_jpeg_app1<T>(reader: &mut (impl Read + Seek), header_span: u64, marker_steps: usize, mut pick: impl FnMut(&[u8]) -> Option<T>) -> Option<T> {
let mut header = reader.take(header_span);
let mut soi = [0; 2];
header.read_exact(&mut soi).ok()?;
if soi != [0xff, 0xd8] {
return None;
}
let mut filled = false;
for _ in 0..marker_steps {
if !filled {
let mut prefix = [0];
header.read_exact(&mut prefix).ok()?;
if prefix != [0xff] {
return None;
}
}
let mut code = [0];
header.read_exact(&mut code).ok()?;
let [marker] = code;
filled = marker == 0xff;
match marker {
0xff | 0xd8 | 0xd0..=0xd7 | 0x01 => continue,
0xd9 | 0xda => return None,
_ => {}
}
let mut length = [0; 2];
header.read_exact(&mut length).ok()?;
let payload_len = u16::from_be_bytes(length).checked_sub(2)?;
let remaining = header.limit().checked_sub(u64::from(payload_len))?;
if marker != 0xe1 {
// Seeking must spend the same span budget as reading, before the seek is attempted.
header.set_limit(remaining);
header.get_mut().seek(SeekFrom::Current(i64::from(payload_len))).ok()?;
continue;
}
let segment_len = 6usize.checked_add(usize::from(payload_len))?;
let [hi, lo] = length;
let mut segment = Vec::with_capacity(segment_len);
segment.extend_from_slice(&[0xff, 0xd8, 0xff, 0xe1, hi, lo]);
segment.resize(segment_len, 0);
header.read_exact(segment.get_mut(6..)?).ok()?;
if let Some(found) = pick(&segment) {
return Some(found);
}
}
None
}
/// Photo ids touched by an op (for auto-write).
pub(crate) fn op_photos(op: &Op, out: &mut Vec<PhotoId>) {
match op {
Op::SetRating { id, .. } | Op::SetFlag { id, .. } | Op::SetLabel { id, .. } | Op::SetDevelop { id, .. } | Op::SetMeta { id, .. }
if !out.contains(id) =>
{
out.push(*id);
}
Op::Batch { ops } => ops.iter().for_each(|o| op_photos(o, out)),
_ => {}
}
}
/// Where a file stands in the order a shared stem sidecar is owned by: a raw first, then the file
/// name in lower case, then the catalog id (a file an import is adding has no id yet, and ranks
/// last among files of its kind and name: see [`ImportStems`]).
type Rank = (bool, String, u64);
/// The rank of the file at `path`, of kind `kind`, catalogued as `id`.
fn rank_of(path: &str, kind: MediaKind, id: u64) -> Rank {
(kind != MediaKind::Raw, path.to_lowercase(), id)
}
/// Which photo owns each stem sidecar that several catalogued files map to (see
/// [`Session::sidecar_naming`]).
pub(crate) struct StemOwners(HashMap<String, PhotoId>);
impl StemOwners {
/// The owners of every shared stem in the library, each file's key made the long way: what
/// [`StemOwners::among`] must agree with for the photos it is asked about (this is how the
/// owners were worked out before, for every write).
#[cfg(test)]
pub(crate) fn of(cat: &lightcraft_catalog::Catalog) -> StemOwners {
let mut best: HashMap<String, (Rank, PhotoId, usize)> = HashMap::new();
for p in cat.photos().filter(|p| p.copy_of.is_none()) {
let Some(path) = file_path(p) else { continue };
let rank = rank_of(path, p.kind, p.id.0);
let e = best.entry(stem_key(path)).or_insert_with(|| (rank.clone(), p.id, 0));
e.2 += 1;
if rank < e.0 {
(e.0, e.1) = (rank, p.id);
}
}
StemOwners(best.into_iter().filter(|(_, v)| v.2 > 1).map(|(k, v)| (k, v.1)).collect())
}
/// The owners of the stems photos `ids` map to: all that naming those photos needs. The
/// library is passed over once, and only files of those stems are looked at closer, so a
/// sidecar written for one photo does not cost a table of the whole library.
pub(crate) fn among(cat: &lightcraft_catalog::Catalog, ids: &[PhotoId]) -> StemOwners {
let wanted: std::collections::HashSet<String> =
ids.iter().filter_map(|id| cat.photo(*id)).filter_map(|p| file_path(p)).map(stem_key).collect();
if wanted.is_empty() {
return StemOwners(HashMap::new());
}
Self::where_wanted(cat, &wanted)
}
/// The owners of the shared stems in `wanted`.
fn where_wanted(cat: &lightcraft_catalog::Catalog, wanted: &std::collections::HashSet<String>) -> StemOwners {
// per stem: (raw first, then file name, then id) of the best candidate; and how many
let mut best: HashMap<String, (Rank, PhotoId, usize)> = HashMap::new();
let mut key = String::new();
for p in cat.photos().filter(|p| p.copy_of.is_none()) {
let Some(path) = file_path(p) else { continue };
stem_key_into(&mut key, path);
if !wanted.contains(key.as_str()) {
continue;
}
let rank = rank_of(path, p.kind, p.id.0);
let e = best.entry(key.clone()).or_insert_with(|| (rank.clone(), p.id, 0));
e.2 += 1;
if rank < e.0 {
(e.0, e.1) = (rank, p.id);
}
}
StemOwners(best.into_iter().filter(|(_, v)| v.2 > 1).map(|(k, v)| (k, v.1)).collect())
}
/// Whether photo `p` may read the shared stem sidecar of its stem: no other file owns it.
fn owns(&self, p: &Photo) -> bool {
file_path(p).is_none_or(|path| self.0.get(&stem_key(path)).is_none_or(|owner| *owner == p.id))
}
/// `naming` for photo `p`, except that a stem sidecar another file owns becomes Full.
fn naming(&self, p: &Photo, naming: SidecarNaming) -> SidecarNaming {
match (naming, file_path(p)) {
(SidecarNaming::Stem, Some(path)) if self.0.get(&stem_key(path)).is_some_and(|owner| *owner != p.id) => SidecarNaming::Full,
_ => naming,
}
}
}
/// Which of the files an import is adding may read the shared stem sidecar (`IMG_0001.xmp`) of
/// their stem, and which read their own `<file>.xmp` (`IMG_0001.TIF.xmp`) instead (issue #1905:
/// before, a JPEG or a TIFF imported beside its raw read the raw's packet and took its rating,
/// metadata and develop settings). The order is the one [`StemOwners`] applies to writing, with
/// the arriving files ranked against the library and against each other: a raw first, then the
/// file name in lower case. A file that is not catalogued yet has no id, and the id it draws when
/// the import commits is above every id in the library, so it loses a tie against a catalogued
/// file; between arriving files the order they arrive in stands for it.
pub(crate) struct ImportStems {
/// The arriving files that own the stem sidecar of their stem, by path as it arrives.
owners: HashSet<String>,
}
impl ImportStems {
/// Work out [`ImportStems`] for the files `incoming`, in the order they arrive. `None` for a
/// library already looked over (a job prepared twice): there is nothing to rank against, so
/// every arriving file reads as it did before. The library is passed over once, and only the
/// stems the arriving files map to are looked at closer.
pub(crate) fn for_import<'a>(cat: Option<&lightcraft_catalog::Catalog>, incoming: impl Iterator<Item = (&'a str, MediaKind)>) -> ImportStems {
// per stem: the best of the library's files, and the best arriving file with its path
let mut best: HashMap<String, (Option<Rank>, Option<(Rank, &'a str)>)> = HashMap::new();
let mut arriving_paths = HashSet::new();
for (arriving, (path, kind)) in incoming.enumerate() {
arriving_paths.insert(path);
let e = best.entry(stem_key(path)).or_default();
let rank = rank_of(path, kind, arriving as u64);
if e.1.as_ref().is_none_or(|(b, _)| rank < *b) {
e.1 = Some((rank, path));
}
}
if let Some(cat) = cat {
let mut key = String::new();
for p in cat.photos().filter(|p| p.copy_of.is_none()) {
// the arriving file's own record (a trashed photo imported again) is not a rival
let Some(path) = file_path(p).filter(|path| !arriving_paths.contains(*path)) else { continue };
stem_key_into(&mut key, path);
let Some(e) = best.get_mut(key.as_str()) else { continue };
let rank = rank_of(path, p.kind, p.id.0);
if e.0.as_ref().is_none_or(|b| rank < *b) {
e.0 = Some(rank);
}
}
}
let owners = best
.into_values()
.filter_map(|(in_library, arriving)| {
let (rank, path) = arriving?;
let wins = in_library.as_ref().is_none_or(|b| (rank.0, rank.1.as_str()) < (b.0, b.1.as_str()));
wins.then(|| path.to_string())
})
.collect();
ImportStems { owners }
}
/// Whether the arriving file at `original` owns the stem sidecar of its stem.
pub(crate) fn owns(&self, original: &str) -> bool {
self.owners.contains(original)
}
}
/// How the sidecar of each of photos `ids` is named in library `cat`, whose naming is `naming`
/// ([`Session::sidecar_namings`], for a thread that has the library and not the session).
pub(crate) fn namings_in(cat: &lightcraft_catalog::Catalog, naming: SidecarNaming, ids: &[PhotoId]) -> Vec<SidecarNaming> {
lookups_in(cat, naming, ids).into_iter().map(|l| l.naming).collect()
}
/// [`namings_in`], with whether each photo may read the stem sidecar of its stem.
pub(crate) fn lookups_in(cat: &lightcraft_catalog::Catalog, naming: SidecarNaming, ids: &[PhotoId]) -> Vec<SidecarLookup> {
let owners = StemOwners::among(cat, ids);
ids.iter()
.map(|id| match cat.photo(*id) {
Some(p) => SidecarLookup { naming: owners.naming(p, naming), owns_stem: owners.owns(p) },
None => SidecarLookup { naming, owns_stem: true },
})
.collect()
}
impl Session {
/// The sidecar naming for photo `id`: the library's, except under Stem naming when another
/// catalogued file shares the stem (`IMG_0001.CR3` + `IMG_0001.JPG`) and owns the stem
/// sidecar — a raw first, else the first by file name. The others use Full naming
/// (`IMG_0001.JPG.xmp`), so their metadata never overwrites each other.
/// [`Session::sidecar_naming`] for many photos, their stems' owners worked out once (per
/// photo that is a pass over the whole catalog).
pub fn sidecar_namings(&self, ids: &[PhotoId]) -> Vec<SidecarNaming> {
namings_in(&self.catalog, self.xmp.naming, ids)
}
/// [`Session::sidecar_naming`] and the stem-sidecar ownership of photo `id`, for reading its sidecar.
pub fn sidecar_lookup(&self, id: PhotoId) -> SidecarLookup {
lookups_in(&self.catalog, self.xmp.naming, &[id]).into_iter().next().unwrap_or(SidecarLookup { naming: self.xmp.naming, owns_stem: true })
}
/// [`Session::sidecar_lookup`] for many photos, their stems' owners worked out once.
pub fn sidecar_lookups(&self, ids: &[PhotoId]) -> Vec<SidecarLookup> {
lookups_in(&self.catalog, self.xmp.naming, ids)
}
pub fn sidecar_naming(&self, id: PhotoId) -> SidecarNaming {
match self.catalog.photo(id) {
Some(p) if self.xmp.naming == SidecarNaming::Stem => StemOwners::among(&self.catalog, &[id]).naming(p, self.xmp.naming),
_ => self.xmp.naming,
}
}
/// Write the photo's XMP sidecar, merged into an existing one (see the module docs).
pub fn save_sidecar(&mut self, id: PhotoId) -> Result<SidecarSaved> {
self.save_sidecar_with(id, &StemOwners::among(&self.catalog, &[id]))
}
pub(crate) fn save_sidecar_with(&mut self, id: PhotoId, owners: &StemOwners) -> Result<SidecarSaved> {
let p = self.catalog.photo(id).ok_or(lightcraft_catalog::CatalogError::NoPhoto(id))?;
if p.copy_of.is_some() {
return Err(EngineError::Other(format!("{} is a virtual copy: its settings live only in the library", p.file_name)));
}
let orig = file_path(p).ok_or_else(|| EngineError::Other(format!("{} is not a file on disk", p.file_name)))?;
// A relative path would be taken from the working directory.
if Path::new(orig).is_relative() || !Path::new(orig).is_file() {
return Err(EngineError::Other(format!("{}: the original {orig} is missing; no XMP sidecar is written", p.file_name)));
}
let naming = owners.naming(p, self.xmp.naming);
let preferred = sidecar_path(orig, naming);
let other = if naming == SidecarNaming::Stem { SidecarNaming::Full } else { SidecarNaming::Stem };
let candidates: Vec<PathBuf> = [naming, other]
.into_iter()
.filter(|n| *n != SidecarNaming::Stem || owners.naming(p, SidecarNaming::Stem) == SidecarNaming::Stem)
.flat_map(|n| {
let path = sidecar_path(orig, n);
[path.clone(), path.with_extension("XMP")]
})
.collect();
// A separated packet is still the photo's metadata. Repair it before a fresh write
// occupies its destination; a failure preserves the queue and both existing packets.
self.repair_sidecars_for_paths(&candidates, &crate::rename::RealFs)
.map_err(|e| EngineError::Other(format!("{}: unfinished XMP repair: {e}", preferred.display())))?;
// Keep an existing packet's spelling/style so foreign properties survive this save and
// subsequent Undo/Redo writes. Shared stems owned by another photo were excluded above.
// On a case-insensitive volume `a.xmp` also answers for `a.XMP`: the spelling the folder
// lists comes first, so the atomic write doesn't rename the packet.
use crate::rename::MoveFs;
let existing: Vec<PathBuf> = candidates.into_iter().filter(|path| path.is_file()).collect();
let listed = existing.iter().find(|path| crate::rename::RealFs.listed(path)).cloned();
let path = listed.or_else(|| existing.into_iter().next()).unwrap_or(preferred);
self.validate_sidecar_packets_for_paths(std::slice::from_ref(&path), &crate::rename::RealFs)?;
let p = self.catalog.photo(id).ok_or(lightcraft_catalog::CatalogError::NoPhoto(id))?;
let err = |e: std::io::Error| EngineError::Other(format!("{}: {e}", path.display()));
let (data, merged, backup) = sidecar_contents(&path, sidecar_packet(p, &self.catalog), &(self.clock)()).map_err(err)?;
write_atomic(&path, data.as_bytes()).map_err(err)?;
self.note_sidecar_write(&path, data.as_bytes());
Ok(SidecarSaved { path, merged, backup })
}
/// The op that applies a photo's sidecar (or embedded XMP) to the catalog, if there is one.
pub fn read_sidecar_op(&self, id: PhotoId) -> Result<Option<(Op, PathBuf)>> {
self.read_sidecar_op_with_warnings(id).map(|result| result.map(|(op, from, _)| (op, from)))
}
/// Like [`Session::read_sidecar_op`], retaining transient diagnostics for the command report.
pub fn read_sidecar_op_with_warnings(&self, id: PhotoId) -> Result<Option<(Op, PathBuf, Vec<crate::import::ImportWarning>)>> {
let p = self.catalog.photo(id).ok_or(lightcraft_catalog::CatalogError::NoPhoto(id))?;
let Some(orig) = file_path(p) else { return Ok(None) };
let Some((packet, from)) = read_packet(orig, p.kind, self.sidecar_lookup(id)) else { return Ok(None) };
let sc = parse_sidecar(&packet, crate::crs::Target::for_photo(p)).map_err(|e| EngineError::Other(format!("{}: {e}", from.display())))?;
let sc = sc.with_file_orientation(orig);
let mut warnings = sc.warnings.clone();
for warning in &mut warnings {
warning.path = orig.to_string();
}
let op = self.sidecar_op(id, sc)?;
Ok(Some((op, from, warnings)))
}
/// The op that applies what a sidecar says (already read and parsed, e.g. on a worker
/// thread) to photo `id`: the sidecar wins. No file is read.
pub fn sidecar_op(&self, id: PhotoId, sc: SidecarData) -> Result<Op> {
let p = self.catalog.photo(id).ok_or(lightcraft_catalog::CatalogError::NoPhoto(id))?;
let sc = sc.resolve_label(&self.catalog);
let mut q = (**p).clone();
let develop_changed = merge_into(&mut q, &sc, &(self.clock)());
let mut ops = vec![
Op::SetRating { id, rating: q.rating },
Op::SetFlag { id, flag: q.flag },
Op::SetLabel { id, label: q.label },
Op::SetMeta { id, meta: Box::new(q.meta.clone()) },
];
if q.captured != p.captured {
ops.push(Op::SetCaptured { id, captured: q.captured.clone() });
}
if develop_changed {
ops.extend(self.develop_op(id, (*q.develop).clone(), "Read Metadata from File"));
}
Ok(Op::Batch { ops })
}
/// Auto-write: sidecars for photos changed by `ops` (errors are logged, not returned).
pub(crate) fn auto_write_sidecars(&mut self, ops: &[Op]) {
let mut ids = Vec::new();
ops.iter().for_each(|o| op_photos(o, &mut ids));
self.auto_write_sidecar_ids(ids);
}
/// Gather changed IDs before the saves mutate retained physical repair queues.
pub(crate) fn auto_write_sidecars_since(&mut self, log_start: usize) {
let mut ids = Vec::new();
if let Some(ops) = self.pending_log.get(log_start..) {
ops.iter().for_each(|o| op_photos(o, &mut ids));
}
self.auto_write_sidecar_ids(ids);
}
fn auto_write_sidecar_ids(&mut self, ids: Vec<PhotoId>) {
if ids.is_empty() {
return;
}
let owners = StemOwners::among(&self.catalog, &ids);
for id in ids {
if self.catalog.photo(id).is_some_and(|p| file_path(p).is_some() && p.copy_of.is_none()) {
match self.save_sidecar_with(id, &owners) {
Ok(_) => {
self.unsaved_sidecars.remove(&id);
}
Err(e) => {
// kept for a retry once the file is reachable: a network share that dropped out mustn't leave
// the sidecar silently behind the edit
log::warn!("auto-write XMP: {e}");
self.unsaved_sidecars.insert(id);
}
}
}
}
}
/// How many photos have an automatic sidecar write waiting (it failed: the folder offline, read-only or full).
pub fn unsaved_sidecars(&self) -> usize {
self.unsaved_sidecars.len()
}
/// Write the waiting sidecars whose originals are reachable again (the availability cache answers without
/// touching the disk for the others); photos that are gone, or with auto-write off, are dropped from the wait.
/// Returns how many were written.
pub fn retry_unsaved_sidecars(&mut self) -> usize {
if !self.xmp.auto_write {
self.unsaved_sidecars.clear();
return 0;
}
let mut ready = Vec::new();
for id in self.unsaved_sidecars.clone() {
match self.catalog.photo(id).filter(|p| p.copy_of.is_none()).and_then(|p| file_path(p).map(str::to_owned)) {
None => {
self.unsaved_sidecars.remove(&id);
}
Some(path) if self.media.availability.exists(&path) == Some(true) => ready.push(id),
Some(_) => {}
}
}
if ready.is_empty() {
return 0;
}
let owners = StemOwners::among(&self.catalog, &ready);
let mut written = 0;
for id in ready {
match self.save_sidecar_with(id, &owners) {
Ok(_) => {
self.unsaved_sidecars.remove(&id);
written += 1;
}
Err(e) => log::debug!("retry XMP: {e}"),
}
}
written
}
}
#[cfg(test)]
mod tests {
/// Issue #2674: a WebP keeps its EXIF and XMP chunks after the image data, a TIFF's first IFD can follow its strips
/// and a JPEG's Exif other long segments, so the orientation can be past the first [`super::ORIENTATION_SPAN`] of
/// the file. Every file gives what the import gives reading it whole: Exif over XMP wherever each is, a WebP's last
/// EXIF chunk and a JPEG's first Exif segment, and none for a file that states none.
#[test]
fn the_file_orientation_is_found_after_the_image_data() {
let dir = std::env::temp_dir().join(format!("lightcraft-file-orientation-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
let exif = lightcraft_meta::write_exif(&lightcraft_meta::Metadata {
orientation: Some(lightcraft_meta::Orientation::Rotate90),
..Default::default()
});
let span = super::ORIENTATION_SPAN as usize;
let (o6, o8) = (Some(lightcraft_meta::Orientation::Rotate90), Some(lightcraft_meta::Orientation::Rotate270));
// a WebP: VP8X, then `chunks`
let webp = |chunks: &[(&[u8], Vec<u8>)]| {
let mut w = b"RIFF\0\0\0\0WEBP".to_vec();
for (id, data) in [(&b"VP8X"[..], vec![0; 10])].iter().chain(chunks) {
w.extend_from_slice(id);
w.extend_from_slice(&(data.len() as u32).to_le_bytes());
w.extend_from_slice(data);
if data.len() % 2 == 1 {
w.push(0);
}
}
let size = u32::try_from(w.len() - 8).unwrap();
w[4..8].copy_from_slice(&size.to_le_bytes());
w
};
let image = |len: usize| (&b"VP8L"[..], vec![0; len]);
let exif_chunk = (&b"EXIF"[..], [&b"Exif\0\0"[..], &exif].concat());
// an EXIF chunk over the span (orientation 6 too), one stating no orientation, and an XMP packet stating 8
let huge_exif = (&b"EXIF"[..], [&b"Exif\0\0"[..], &exif, &vec![0; span]].concat());
let plain = lightcraft_meta::write_exif(&lightcraft_meta::Metadata::default());
let plain_chunk = (&b"EXIF"[..], [&b"Exif\0\0"[..], &plain].concat());
let packet = lightcraft_meta::write_xmp(&lightcraft_meta::Metadata { orientation: o8, ..Default::default() }, None);
let xmp_chunk = (&b"XMP "[..], packet.clone().into_bytes());
// the XMP packet padded with whitespace to `len` bytes
let padded_xmp = |len: usize| {
let mut p = packet.clone().into_bytes();
p.resize(len, b' ');
(&b"XMP "[..], p)
};
// an EXIF chunk holding only a TIFF header, whose IFD offset (16) points past the chunk's end into the next
// chunk's data, where an IFD stating orientation 6 sits
let stray_exif = (&b"EXIF"[..], b"II*\0\x10\0\0\0".to_vec());
let stray_ifd = (&b"JUNK"[..], vec![1, 0, 0x12, 1, 3, 0, 1, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0]);
// a TIFF whose first IFD follows more than the first MiB of image data: an orientation tag of `values` values
// whose field starts with `value` (with more than two, it is an offset), a tag pointing elsewhere, and an XMP
// packet (tag 700) stored after the IFD
let tiff = |le: bool, orientation: Option<(u32, u16)>, xmp: Option<&str>| {
let (u16_of, u32_of): (fn(u16) -> [u8; 2], fn(u32) -> [u8; 4]) =
if le { (u16::to_le_bytes, u32::to_le_bytes) } else { (u16::to_be_bytes, u32::to_be_bytes) };
let strips = span + 64;
let mut entries = Vec::new();
if let Some((values, value)) = orientation {
entries.push((274u16, 3u16, values, [u16_of(value), [0, 0]].concat()));
}
entries.push((305, 2, 64, u32_of(4).to_vec()));
let packet_at = 8 + strips + 2 + 12 * (entries.len() + usize::from(xmp.is_some())) + 4;
if let Some(x) = xmp {
entries.push((700, 7, x.len() as u32, u32_of(packet_at as u32).to_vec()));
}
let mut t = if le { b"II*\0".to_vec() } else { b"MM\0*".to_vec() };
t.extend_from_slice(&u32_of(8 + strips as u32));
t.resize(8 + strips, 0);
t.extend_from_slice(&u16_of(entries.len() as u16));
for (tag, kind, count, value) in entries {
t.extend_from_slice(&u16_of(tag));
t.extend_from_slice(&u16_of(kind));
t.extend_from_slice(&u32_of(count));
t.extend_from_slice(&value);
}
t.extend_from_slice(&[0; 4]);
t.extend_from_slice(xmp.unwrap_or_default().as_bytes());
t
};
// a JPEG: the APP1 payloads `early`, more than the first MiB of other segments, `late`, the image data
let jpeg = |early: &[Vec<u8>], late: &[Vec<u8>]| {
let app1 = |j: &mut Vec<u8>, payload: &Vec<u8>| {
j.extend_from_slice(&[0xff, 0xe1]);
j.extend_from_slice(&(payload.len() as u16 + 2).to_be_bytes());
j.extend_from_slice(payload);
};
let mut j = vec![0xff, 0xd8];
early.iter().for_each(|p| app1(&mut j, p));
for _ in 0..17 {
j.extend_from_slice(&[0xff, 0xef, 0xff, 0xff]);
j.resize(j.len() + 0xfffd, 0);
}
late.iter().for_each(|p| app1(&mut j, p));
j.extend_from_slice(&[0xff, 0xda, 0, 2, 0xff, 0xd9]);
j
};
let exif_app1 = [&b"Exif\0\0"[..], &exif].concat();
let plain_app1 = [&b"Exif\0\0"[..], &plain].concat();
let xmp_app1 = [&b"http://ns.adobe.com/xap/1.0/\0"[..], packet.as_bytes()].concat();
for (name, bytes, want) in [
("small.webp", webp(&[image(64), exif_chunk.clone()]), o6),
("large.webp", webp(&[image(span + 64), exif_chunk.clone(), xmp_chunk.clone()]), o6),
// EXIF wins in whichever order the chunks come, and an XMP in the first MiB doesn't stop the search
("reversed.webp", webp(&[image(span + 64), xmp_chunk.clone(), exif_chunk.clone()]), o6),
("xmp-first.webp", webp(&[xmp_chunk.clone(), image(span + 64), exif_chunk.clone()]), o6),
("none.webp", webp(&[image(span + 64), exif_chunk.clone()])[..span + 100].to_vec(), None),
("xmp.webp", webp(&[image(span + 64), xmp_chunk.clone()]), o8),
// an EXIF chunk of any size is read by its IFD: one over the span, one (without `Exif\0\0`, big-endian)
// whose IFD is more than the span into it
("huge.webp", webp(&[image(span + 64), huge_exif, xmp_chunk.clone()]), o6),
("late-ifd.webp", webp(&[image(span + 64), (&b"EXIF"[..], tiff(false, Some((1, 6)), None)), xmp_chunk.clone()]), o6),
// nothing outside the EXIF chunk is read for its IFD
("outside.webp", webp(&[image(span + 64), stray_exif, stray_ifd, xmp_chunk.clone()]), o8),
// the last EXIF chunk counts, here a late one stating no orientation after one in the first MiB
("last-exif.webp", webp(&[exif_chunk, image(span + 64), plain_chunk, xmp_chunk]), o8),
// an XMP chunk of the span is read
("span-xmp.webp", webp(&[image(64), padded_xmp(span)]), o8),
("le.tif", tiff(true, Some((1, 8)), None), o8),
("be.tif", tiff(false, Some((1, 8)), None), o8),
("offset.tif", tiff(true, Some((3, 8)), None), None),
// the tag wins over the IFD's XMP; XMP fills in for no tag or an invalid one
("both.tif", tiff(true, Some((1, 6)), Some(&packet)), o6),
("xmp.tif", tiff(false, None, Some(&packet)), o8),
("invalid.tif", tiff(true, Some((1, 0)), Some(&packet)), o8),
("none.jpg", jpeg(&[], &[]), None),
("xmp-first.jpg", jpeg(std::slice::from_ref(&xmp_app1), std::slice::from_ref(&exif_app1)), o6),
("xmp-only.jpg", jpeg(&[], std::slice::from_ref(&xmp_app1)), o8),
// the first Exif segment counts, though it states no orientation
("two-exif.jpg", jpeg(&[plain_app1], &[exif_app1, xmp_app1]), o8),
] {
assert_eq!(lightcraft_meta::extract(&bytes).orientation, want, "import: {name}");
let path = dir.join(name);
std::fs::write(&path, bytes).unwrap();
assert_eq!(super::file_orientation(&path.to_string_lossy()), want, "{name}");
}
// a larger XMP chunk isn't read, where the import (reading the whole file) takes its orientation
let over = webp(&[image(64), padded_xmp(span + 1)]);
assert_eq!(lightcraft_meta::extract(&over).orientation, o8);
std::fs::write(dir.join("over-xmp.webp"), over).unwrap();
assert_eq!(super::file_orientation(&dir.join("over-xmp.webp").to_string_lossy()), None);
assert_eq!(super::file_orientation(&dir.join("missing.webp").to_string_lossy()), None);
std::fs::remove_dir_all(dir).unwrap();
}
/// Issue #2674: the same for a PNG, whose `eXIf` and XMP `iTXt` chunks can follow the image data, and a PSD, whose
/// image resources can outgrow the first [`super::ORIENTATION_SPAN`].
#[test]
fn a_png_or_psd_gives_the_orientation_the_import_gives() {
let dir = std::env::temp_dir().join(format!("lightcraft-png-psd-orientation-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
let span = super::ORIENTATION_SPAN as usize;
let (o6, o8) = (Some(lightcraft_meta::Orientation::Rotate90), Some(lightcraft_meta::Orientation::Rotate270));
let exif = lightcraft_meta::write_exif(&lightcraft_meta::Metadata { orientation: o6, ..Default::default() });
let plain = lightcraft_meta::write_exif(&lightcraft_meta::Metadata::default());
let packet = lightcraft_meta::write_xmp(&lightcraft_meta::Metadata { orientation: o8, ..Default::default() }, None);
// a PNG: its IHDR, then `chunks`, each with a CRC (nothing checks it)
let png = |chunks: &[(&[u8], Vec<u8>)]| {
let mut p = b"\x89PNG\r\n\x1a\n".to_vec();
for (id, data) in [(&b"IHDR"[..], vec![0; 13])].iter().chain(chunks) {
p.extend_from_slice(&(data.len() as u32).to_be_bytes());
p.extend_from_slice(id);
p.extend_from_slice(data);
p.extend_from_slice(&[0; 4]);
}
p
};
let idat = |len: usize| (&b"IDAT"[..], vec![0; len]);
let iend = (&b"IEND"[..], Vec::new());
let exif_chunk = (&b"eXIf"[..], exif.clone());
let plain_chunk = (&b"eXIf"[..], plain.clone());
// an `iTXt` chunk: the keyword, compression flag `flag`, method, no language, no translated keyword, the text
let itxt = |keyword: &[u8], flag: u8, text: &[u8]| (&b"iTXt"[..], [keyword, &[0, flag, 0, 0, 0], text].concat());
let xmp_key = &b"XML:com.adobe.xmp"[..];
let xmp_chunk = itxt(xmp_key, 0, packet.as_bytes());
// the XMP packet padded with whitespace, in a chunk of `len` bytes
let padded_xmp = |len: usize| {
let mut p = packet.clone().into_bytes();
p.resize(len - 22, b' ');
itxt(xmp_key, 0, &p)
};
// a big-endian TIFF stream whose IFD, stating orientation 6, is more than the span into it
let ifd = [0u8, 1, 1, 0x12, 0, 3, 0, 0, 0, 1, 0, 6, 0, 0, 0, 0, 0, 0];
let late_ifd = [&b"MM\0*"[..], &(8 + span as u32).to_be_bytes(), &vec![0; span], &ifd].concat();
// an `eXIf` holding only a TIFF header, whose IFD offset (20) points past the chunk's end into the next chunk's
// data, where an IFD stating orientation 6 sits
let stray_exif = (&b"eXIf"[..], b"II*\0\x14\0\0\0".to_vec());
let stray_ifd = (&b"juNK"[..], vec![1, 0, 0x12, 1, 3, 0, 1, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0]);
// a PSD: its header, `colour` bytes of colour-mode data, the image resources (id, data), `pixels` other bytes
let psd = |colour: usize, resources: &[(u16, &[u8])], pixels: usize| {
let mut section = Vec::new();
for (id, data) in resources {
section.extend_from_slice(b"8BIM");
section.extend_from_slice(&id.to_be_bytes());
section.extend_from_slice(&[0, 0]);
section.extend_from_slice(&(data.len() as u32).to_be_bytes());
section.extend_from_slice(data);
if data.len() % 2 == 1 {
section.push(0);
}
}
// a valid header: one RGB pixel, 8 bits
let mut p = b"8BPS\0\x01".to_vec();
p.resize(12, 0);
p.extend_from_slice(&[0, 3, 0, 0, 0, 1, 0, 0, 0, 1, 0, 8, 0, 3]);
p.extend_from_slice(&(colour as u32).to_be_bytes());
p.resize(p.len() + colour, 0);
p.extend_from_slice(&(section.len() as u32).to_be_bytes());
p.extend_from_slice(§ion);
p.resize(p.len() + pixels, 0);
p
};
let long = vec![0; span + 64];
let (exif_res, plain_res, xmp_res, long_res) = ((1058, &exif[..]), (1058, &plain[..]), (1060, packet.as_bytes()), (1039, &long[..]));
let late_png = png(&[idat(span + 64), exif_chunk.clone()]);
let late_psd = psd(0, &[long_res, exif_res], 0);
for (name, bytes, want) in [
("early.png", png(&[exif_chunk.clone(), idat(span + 64), iend.clone()]), o6),
("late.png", png(&[idat(span + 64), exif_chunk.clone(), xmp_chunk.clone(), iend.clone()]), o6),
// the `eXIf` wins in whichever order the chunks come, and an XMP in the first MiB doesn't stop the search
("reversed.png", png(&[idat(span + 64), xmp_chunk.clone(), exif_chunk.clone(), iend.clone()]), o6),
("xmp-first.png", png(&[xmp_chunk.clone(), idat(span + 64), exif_chunk.clone(), iend.clone()]), o6),
("xmp.png", png(&[idat(span + 64), xmp_chunk.clone(), iend.clone()]), o8),
("none.png", png(&[idat(span + 64), iend.clone()]), None),
// the `eXIf` is cut short, or comes after the end of the image
("cut.png", late_png[..late_png.len() - 6].to_vec(), None),
("after-iend.png", png(&[idat(span + 64), iend.clone(), exif_chunk.clone()]), None),
// an `eXIf` of any size is read by its IFD: with `Exif\0\0`, over the span, and with its IFD past the span
("prefixed.png", png(&[idat(span + 64), (&b"eXIf"[..], [&b"Exif\0\0"[..], &exif].concat())]), o6),
("huge.png", png(&[idat(64), (&b"eXIf"[..], [&exif[..], &vec![0; span]].concat()), xmp_chunk.clone()]), o6),
("late-ifd.png", png(&[idat(span + 64), (&b"eXIf"[..], late_ifd), xmp_chunk.clone()]), o6),
// nothing outside the `eXIf` is read for its IFD
("outside.png", png(&[idat(span + 64), stray_exif, stray_ifd, xmp_chunk.clone()]), o8),
// the last `eXIf` counts, here a late one stating no orientation after one in the first MiB
("last-exif.png", png(&[exif_chunk.clone(), idat(span + 64), plain_chunk, xmp_chunk.clone()]), o8),
// a compressed packet or another keyword isn't the XMP; a chunk of the span is read
("compressed.png", png(&[idat(span + 64), xmp_chunk.clone(), itxt(xmp_key, 1, packet.as_bytes())]), o8),
("only-compressed.png", png(&[idat(span + 64), itxt(xmp_key, 1, packet.as_bytes())]), None),
("other-text.png", png(&[idat(span + 64), itxt(b"Comment", 0, packet.as_bytes())]), None),
("short-text.png", png(&[idat(span + 64), xmp_chunk.clone(), (&b"iTXt"[..], b"XML:com.adobe.xmp\0\0".to_vec())]), o8),
("span-xmp.png", png(&[idat(64), padded_xmp(span)]), o8),
("late.psd", late_psd.clone(), o6),
("early.psd", psd(0, &[exif_res], span + 64), o6),
// Exif wins in whichever order the resources come; XMP fills in for none, or none stating an orientation
("both.psd", psd(0, &[xmp_res, long_res, exif_res], 0), o6),
("xmp.psd", psd(768, &[long_res, xmp_res], 0), o8),
("plain.psd", psd(0, &[long_res, plain_res, xmp_res], 0), o8),
("none.psd", psd(0, &[long_res], span), None),
// as at import, nothing is taken from resources that are cut short or state their Exif twice
("cut.psd", late_psd[..late_psd.len() - 1].to_vec(), None),
("two-exif.psd", psd(0, &[long_res, exif_res, exif_res], 0), None),
] {
assert_eq!(lightcraft_meta::extract(&bytes).orientation, want, "import: {name}");
let path = dir.join(name);
std::fs::write(&path, bytes).unwrap();
assert_eq!(super::file_orientation(&path.to_string_lossy()), want, "{name}");
}
// a larger XMP chunk isn't read, where the import (reading the whole file) takes its orientation
let over = png(&[idat(64), padded_xmp(span + 1)]);
assert_eq!(lightcraft_meta::extract(&over).orientation, o8);
std::fs::write(dir.join("over-xmp.png"), over).unwrap();
assert_eq!(super::file_orientation(&dir.join("over-xmp.png").to_string_lossy()), None);
std::fs::remove_dir_all(dir).unwrap();
// the `eXIf` as the last chunk walked, and as the first one not: IHDR, `chunks` of empty `IDAT`, the `eXIf`
let steps = super::png_chunk_steps;
let split = |chunks: usize| png(&[vec![idat(0); chunks], vec![exif_chunk.clone()]].concat());
let len = |chunks: usize| (8 + 25 + 12 * chunks + 12 + exif.len()) as u64;
let walked = (0..).find(|&n| n as u64 + 2 == steps(len(n)) && n as u64 + 3 > steps(len(n + 1))).unwrap();
let (last, past) = (split(walked), split(walked + 1));
assert_eq!(last.len() as u64, len(walked));
assert_eq!(super::png_orientation(&mut std::io::Cursor::new(&last)), o6);
assert_eq!(lightcraft_meta::extract(&past).orientation, o6);
assert_eq!(super::png_orientation(&mut std::io::Cursor::new(&past)), None);
// image data split into chunks of a KiB is walked whatever its length, up to the limit
let kib_chunks = png(&[vec![idat(1012); 6000], vec![exif_chunk.clone()]].concat());
assert_eq!(super::png_orientation(&mut std::io::Cursor::new(&kib_chunks)), o6);
assert_eq!(steps(u64::MAX), 1 << 20);
// image resources that end exactly at the limit, and a byte past it
let limit = super::PSD_RESOURCES_SPAN as usize;
let padding = vec![0; limit - 58 - exif.len().next_multiple_of(2)];
let within = psd(0, &[(1039, &padding[..]), exif_res], 0);
assert_eq!(within.len(), limit);
assert_eq!(super::psd_orientation(&mut std::io::Cursor::new(&within)), o6);
let past = psd(1, &[(1039, &padding[..]), exif_res], 0);
assert_eq!(lightcraft_meta::extract(&past).orientation, o6);
assert_eq!(super::psd_orientation(&mut std::io::Cursor::new(&past)), None);
}
#[test]
fn sidecars_are_written_only_beside_an_existing_original() {
let dir = std::env::temp_dir().join(format!("lightcraft-sidecar-local-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
std::fs::write(dir.join("IMG_1.jpg"), b"original").unwrap();
let originals = [
(r"\\192.0.2.1\share\IMG_1.jpg".to_string(), false),
(dir.join("IMG_2.jpg").to_string_lossy().to_string(), false),
(dir.join("IMG_1.jpg").to_string_lossy().to_string(), true),
];
for (original, written) in originals {
let _ = std::fs::remove_file(dir.join("IMG_1.xmp"));
let mut s = Session::new();
let p = Photo::new(PhotoId(1), Source::File { path: original.clone() }, "IMG_1.jpg", "JPEG", 4, 3, "2026-01-01T00:00:00");
s.catalog.apply(Op::AddPhoto { photo: Box::new(p) }).unwrap();
assert_eq!(s.save_sidecar(PhotoId(1)).is_ok(), written, "{original}");
assert_eq!(dir.join("IMG_1.xmp").is_file(), written, "{original}");
assert!(!dir.join("IMG_2.xmp").exists(), "{original}");
assert!(!Path::new(r"\\192.0.2.1\share\IMG_1.xmp").exists(), "{original}");
}
std::fs::remove_dir_all(dir).unwrap();
}
/// A readable sidecar whose merged packet fails read-back (too many namespace declarations in scope) is kept: the
/// save fails instead of moving it to a backup and writing fresh content only.
#[test]
fn a_readable_sidecar_that_cannot_be_merged_is_kept() {
let dir = std::env::temp_dir().join(format!("lightcraft-sidecar-keep-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
let decls: String = (0..500).map(|i| format!(" xmlns:n{i}=\"urn:n:{i}\"")).collect();
let old = format!(
"<x:xmpmeta xmlns:x=\"adobe:ns:meta/\"><rdf:RDF xmlns:rdf=\"http://www.w3.org/1999/02/22-rdf-syntax-ns#\"{decls}><rdf:Description rdf:about=\"\"/></rdf:RDF></x:xmpmeta>"
);
assert!(lightcraft_meta::parse_xmp(&old).is_ok(), "the existing sidecar is readable");
let path = dir.join("a.xmp");
std::fs::write(&path, &old).unwrap();
let fresh = lightcraft_meta::write_xmp_lc(&lightcraft_meta::Metadata { rating: Some(3), ..Default::default() }, &[("settings", "{}")]);
assert!(super::sidecar_contents(&path, fresh, "t").is_err());
assert_eq!(std::fs::read_to_string(&path).unwrap(), old);
assert_eq!(std::fs::read_dir(&dir).unwrap().count(), 1, "no backup was made");
std::fs::remove_dir_all(dir).unwrap();
}
/// Text that is not XMP at all (it parses as empty metadata, but has no `rdf:RDF` element) is backed up and replaced.
#[test]
fn a_sidecar_with_no_rdf_element_is_backed_up_and_replaced() {
let dir = std::env::temp_dir().join(format!("lightcraft-sidecar-plain-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
let path = dir.join("a.xmp");
std::fs::write(&path, "not xml at all").unwrap();
let fresh = lightcraft_meta::write_xmp_lc(&lightcraft_meta::Metadata { rating: Some(3), ..Default::default() }, &[("settings", "{}")]);
let (data, merged, backup) = super::sidecar_contents(&path, fresh.clone(), "t").unwrap();
assert!(!merged && data == fresh);
assert!(backup.is_some_and(|b| std::fs::read_to_string(b).unwrap() == "not xml at all"));
std::fs::remove_dir_all(dir).unwrap();
}
/// A sidecar over the parser's namespace-declaration limit cannot be read or merged: the save fails and the file stays.
#[test]
fn a_sidecar_over_the_complexity_limit_is_kept() {
let dir = std::env::temp_dir().join(format!("lightcraft-sidecar-complex-{}", std::process::id()));
let _ = std::fs::remove_dir_all(&dir);
std::fs::create_dir_all(&dir).unwrap();
let decls: String = (0..600).map(|i| format!(" xmlns:n{i}=\"urn:n:{i}\"")).collect();
let old = format!(
"<x:xmpmeta xmlns:x=\"adobe:ns:meta/\"><rdf:RDF xmlns:rdf=\"http://www.w3.org/1999/02/22-rdf-syntax-ns#\"{decls}><rdf:Description rdf:about=\"\"/></rdf:RDF></x:xmpmeta>"
);
assert_eq!(lightcraft_meta::parse_xmp(&old).unwrap_err(), lightcraft_meta::XmpError::TooComplex);
let path = dir.join("a.xmp");
std::fs::write(&path, &old).unwrap();
let fresh = lightcraft_meta::write_xmp_lc(&lightcraft_meta::Metadata { rating: Some(3), ..Default::default() }, &[("settings", "{}")]);
assert!(super::sidecar_contents(&path, fresh, "t").is_err());
assert_eq!(std::fs::read_to_string(&path).unwrap(), old);
assert_eq!(std::fs::read_dir(&dir).unwrap().count(), 1, "no backup was made");
std::fs::remove_dir_all(dir).unwrap();
}
/// A sidecar written by Lightroom Classic (names flat, paths in `lr:hierarchicalSubject`)
/// gives the photo its keywords as paths.
#[test]
fn sidecar_keywords_keep_their_hierarchy() {
let m = lightcraft_meta::Metadata {
keywords: vec!["Lisbon".into(), "Places".into()],
hierarchical_keywords: vec!["Places|Lisbon".into()],
..Default::default()
};
let sc = super::parse_sidecar(&lightcraft_meta::write_xmp(&m, None), crate::crs::Target::Rendered).unwrap();
assert_eq!(sc.keywords, Some(vec!["Places|Lisbon".to_string()]));
}
/// Issue #2556: a sidecar's `lc:settings` colour-range mask keeps at most `MAX_COLOR_RANGE_SAMPLES` samples.
#[test]
fn sidecar_color_range_samples_are_capped() {
use lightcraft_develop::{MAX_COLOR_RANGE_SAMPLES, Mask, MaskComponent, MaskOp, MaskShape};
let samples: Vec<[f64; 3]> = (0..1000).map(|i| [i as f64 / 1000.0, 0.1, -0.1]).collect();
let mut d = DevelopSettings::default();
let shape = MaskShape::ColorRange { samples: samples.clone(), refine: 50.0 };
d.masks.push(Mask { id: 1, components: vec![MaskComponent { name: None, op: MaskOp::Add, invert: false, shape }], ..Default::default() });
let xmp = lightcraft_meta::write_xmp(&lightcraft_meta::Metadata::default(), Some(&d.to_json().to_string()));
let sc = super::parse_sidecar(&xmp, crate::crs::Target::Rendered).unwrap();
let Some(DevelopPatch::Full(s)) = sc.develop else { panic!("expected full settings") };
let Some(MaskShape::ColorRange { samples: got, .. }) = s.masks.first().and_then(|m| m.components.first()).map(|c| &c.shape) else {
panic!("expected a colour range")
};
assert_eq!(got.as_slice(), &samples[..MAX_COLOR_RANGE_SAMPLES]);
}
use super::*;
use lightcraft_catalog::Meta;
fn photo() -> Photo {
let mut p = Photo::new(PhotoId(7), Source::File { path: "/x/IMG_1.jpg".into() }, "IMG_1.jpg", "JPEG", 40, 30, "2026-01-01T00:00:00");
p.rating = 4;
p.flag = Flag::Pick;
p.label = Some(ColorLabel::Purple);
p.captured = Some("2025-06-07T08:09:10".into());
p.meta = Meta {
title: "Dune & sky".into(),
caption: "Line <one>\nline two".into(),
copyright: "© 2026 Me".into(),
creator: "Ann Lee".into(),
location: "Namib".into(),
keywords: vec!["desert".into(), "red".into()],
gps: Some((-24.75, 15.3)),
..Default::default()
};
let mut d = DevelopSettings::default();
d.light.exposure = 0.75;
d.effects.clarity = 22.0;
d.crop.geometry.angle = 2.5;
d.masks.push(lightcraft_develop::Mask { id: 1, ..Default::default() });
p.develop = Arc::new(d);
p
}
#[test]
fn naming() {
assert_eq!(sidecar_path("/a/b/IMG_1.CR3", SidecarNaming::Stem), PathBuf::from("/a/b/IMG_1.xmp"));
assert_eq!(sidecar_path("/a/b/IMG_1.CR3", SidecarNaming::Full), PathBuf::from("/a/b/IMG_1.CR3.xmp"));
}
#[test]
fn packet_roundtrip_restores_everything() {
let p = photo();
let x = sidecar_packet(&p, &lightcraft_catalog::Catalog::new());
let sc = parse_sidecar(&x, crate::crs::Target::Rendered).unwrap();
assert_eq!(sc.rating, Some(4));
assert_eq!(sc.flag, Some(Flag::Pick));
assert_eq!(sc.label, Some(Some(ColorLabel::Purple)));
let mut q = Photo::new(PhotoId(7), p.source.clone(), "IMG_1.jpg", "JPEG", 40, 30, "2026-01-01T00:00:00");
q.captured = p.captured.clone();
q.meta.gps = p.meta.gps;
assert!(merge_into(&mut q, &sc, "now"));
assert_eq!(q.develop, p.develop);
assert_eq!(q.meta, p.meta);
assert_eq!((q.rating, q.flag, q.label), (p.rating, p.flag, p.label));
assert_eq!(q.edited.as_deref(), Some("now"));
}
#[test]
fn foreign_sidecar_rating_reject_label_and_crs() {
let x = r#"<x:xmpmeta xmlns:x="adobe:ns:meta/"><rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
<rdf:Description rdf:about="" xmlns:xmp="http://ns.adobe.com/xap/1.0/" xmlns:crs="http://ns.adobe.com/camera-raw-settings/1.0/"
xmp:Rating="-1" xmp:Label="Green" crs:Exposure2012="-0.40" crs:Vibrance="+12"/></rdf:RDF></x:xmpmeta>"#;
let sc = parse_sidecar(x, crate::crs::Target::RawAbsolute).unwrap();
assert_eq!((sc.rating, sc.flag, sc.label), (Some(0), Some(Flag::Reject), Some(Some(ColorLabel::Green))));
let mut p = photo();
let before = p.develop.clone();
assert!(merge_into(&mut p, &sc, "t"));
assert_eq!(p.develop.light.exposure, -0.4);
assert_eq!(p.develop.color.vibrance, 12.0);
assert_eq!(p.develop.effects.clarity, before.effects.clarity, "crs partial keeps unrelated settings");
assert_eq!(p.meta.title, "Dune & sky", "fields the sidecar doesn't state are kept");
}
#[test]
fn op_photo_collection() {
let mut v = Vec::new();
op_photos(
&Op::Batch {
ops: vec![
Op::SetRating { id: PhotoId(1), rating: 2 },
Op::SetFlag { id: PhotoId(1), flag: Flag::Pick },
Op::SetLabel { id: PhotoId(3), label: None },
],
},
&mut v,
);
assert_eq!(v, vec![PhotoId(1), PhotoId(3)]);
}
}
#[cfg(test)]
#[path = "sidecar/tests_repair_saves.rs"]
mod tests_repair_saves;
#[cfg(test)]
#[path = "sidecar/tests_repair_companions.rs"]
mod tests_repair_companions;
#[cfg(test)]
#[path = "sidecar/tests_repair_provenance.rs"]
mod tests_repair_provenance;
#[cfg(test)]
#[path = "sidecar/tests_stem_owners.rs"]
mod tests_stem_owners;
#[cfg(test)]
mod jpeg_header_tests {
use super::{JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS, read_jpeg_xmp};
use std::io::{self, Cursor, Read, Seek, SeekFrom};
struct CountedReader {
cursor: Cursor<Vec<u8>>,
bytes_read: usize,
largest_read: usize,
seeks: usize,
chunk: usize,
read_error_at: Option<u64>,
seek_error: bool,
}
impl CountedReader {
fn new(bytes: Vec<u8>) -> Self {
Self { cursor: Cursor::new(bytes), bytes_read: 0, largest_read: 0, seeks: 0, chunk: usize::MAX, read_error_at: None, seek_error: false }
}
}
impl Read for CountedReader {
fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
self.largest_read = self.largest_read.max(buf.len());
if self.read_error_at.is_some_and(|at| self.cursor.position() >= at) {
return Err(io::Error::other("injected header read failure"));
}
let len = buf.len().min(self.chunk);
let read = self.cursor.read(&mut buf[..len])?;
self.bytes_read += read;
Ok(read)
}
}
impl Seek for CountedReader {
fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> {
self.seeks += 1;
if self.seek_error {
return Err(io::Error::other("injected header seek failure"));
}
self.cursor.seek(pos)
}
}
fn segment(marker: u8, payload: &[u8]) -> Vec<u8> {
let len = u16::try_from(payload.len() + 2).unwrap();
let mut bytes = vec![0xff, marker];
bytes.extend_from_slice(&len.to_be_bytes());
bytes.extend_from_slice(payload);
bytes
}
fn packet(rating: i8) -> String {
lightcraft_meta::write_xmp(&lightcraft_meta::Metadata { rating: Some(rating), ..Default::default() }, None)
}
fn xmp_segment(packet: &str) -> Vec<u8> {
let mut payload = b"http://ns.adobe.com/xap/1.0/\0".to_vec();
payload.extend_from_slice(packet.as_bytes());
segment(0xe1, &payload)
}
fn jpeg(packet: &str) -> Vec<u8> {
let mut bytes = vec![0xff, 0xd8];
bytes.extend(xmp_segment(packet));
bytes
}
#[test]
fn first_standard_packet_skips_other_segments_and_stops_before_tail() {
let first = packet(2);
let first_payload_len = xmp_segment(&first).len() - 4;
let mut bytes = vec![0xff, 0xd8];
bytes.extend(segment(0xe0, &[7; 17]));
bytes.extend(segment(0xe1, b"Exif\0\0tiny"));
let mut extended = b"http://ns.adobe.com/xmp/extension/\0".to_vec();
extended.extend_from_slice(&[0; 40]);
extended.extend_from_slice(b"not a standard packet");
bytes.extend(segment(0xe1, &extended));
bytes.extend(xmp_segment(&first));
let end = bytes.len();
bytes.extend(xmp_segment(&packet(5)));
bytes.extend_from_slice(&[0xff, 0xda, 0, 2]);
bytes.extend_from_slice(&[3; 128]);
let mut reader = CountedReader::new(bytes);
reader.read_error_at = Some(end as u64);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), Some(first));
assert_eq!(reader.cursor.position(), end as u64, "later packets and image bytes are not read");
assert_eq!(reader.bytes_read, end - 17, "APP0 payload is skipped, not buffered");
assert_eq!(reader.seeks, 1);
assert_eq!(reader.largest_read, first_payload_len, "reads at most one complete APP1 payload at a time");
}
#[test]
fn stops_at_sos_or_eoi_without_reading_later_xmp() {
for stop in [0xda, 0xd9] {
let mut bytes = vec![0xff, 0xd8, 0xff, stop];
bytes.extend(xmp_segment(&packet(3)));
let mut reader = CountedReader::new(bytes);
reader.read_error_at = Some(4);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), None);
assert_eq!((reader.bytes_read, reader.cursor.position(), reader.seeks), (4, 4, 0));
}
}
#[test]
fn rejects_incomplete_app1_even_when_partial_xmp_parses() {
let expected = packet(4);
let mut bytes = jpeg(&expected);
let declared = u16::from_be_bytes([bytes[4], bytes[5]]) + 7;
bytes[4..6].copy_from_slice(&declared.to_be_bytes());
let partial = lightcraft_meta::jpeg_segments(&bytes).xmp.unwrap();
assert_eq!(partial, expected, "the shared extractor tolerates a truncated declared payload");
assert_eq!(super::parse_sidecar(&partial, crate::crs::Target::Rendered).unwrap().rating, Some(4));
let len = bytes.len();
let mut reader = CountedReader::new(bytes);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), None);
assert_eq!(reader.bytes_read, len);
}
#[test]
fn span_budget_counts_soi_headers_and_skipped_payloads() {
let expected = packet(3);
let mut bytes = vec![0xff, 0xd8];
bytes.extend(segment(0xe0, &[9; 23]));
bytes.extend(xmp_segment(&expected));
let end = bytes.len();
let mut exact = CountedReader::new(bytes.clone());
assert_eq!(read_jpeg_xmp(&mut exact, end as u64, JPEG_XMP_HEADER_STEPS), Some(expected));
assert_eq!(exact.cursor.position(), end as u64, "a complete early APP1 does not require EOI");
assert_eq!(exact.bytes_read, end - 23);
let mut short = CountedReader::new(bytes);
assert_eq!(read_jpeg_xmp(&mut short, (end - 1) as u64, JPEG_XMP_HEADER_STEPS), None);
assert_eq!((short.bytes_read, short.cursor.position(), short.seeks), (10, 33, 1), "reject before APP1 payload read");
assert_eq!(short.largest_read, 2);
}
#[test]
fn span_budget_rejects_skip_before_seeking() {
let mut bytes = vec![0xff, 0xd8];
bytes.extend(segment(0xe0, &[8; 18]));
bytes.extend(xmp_segment(&packet(1)));
let mut reader = CountedReader::new(bytes);
assert_eq!(read_jpeg_xmp(&mut reader, 6, JPEG_XMP_HEADER_STEPS), None);
assert_eq!((reader.bytes_read, reader.cursor.position(), reader.seeks), (6, 6, 0));
}
#[test]
fn marker_budget_counts_fill_and_standalone_markers() {
let expected = packet(1);
// Two fill steps, SOI, TEM, RST3, then APP1: six steps after the initial SOI.
let mut bytes = vec![0xff, 0xd8, 0xff, 0xff, 0xff, 0xd8, 0xff, 0x01, 0xff, 0xd3];
bytes.extend(xmp_segment(&expected));
for (steps, end) in [(0, 2), (5, 10)] {
let mut reader = CountedReader::new(bytes.clone());
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, steps), None);
assert_eq!((reader.bytes_read, reader.cursor.position()), (end, end as u64));
}
let mut reader = CountedReader::new(bytes);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, 6), Some(expected));
}
#[test]
fn rejects_invalid_headers_and_segment_lengths() {
for bytes in [
vec![],
vec![0xff],
b"\x89PNG\r\n\x1a\n".to_vec(),
vec![0xff, 0xd8, 0, 0xe1, 0, 2],
vec![0xff, 0xd8, 0xff, 0xe1, 0],
vec![0xff, 0xd8, 0xff, 0xe1, 0, 0],
vec![0xff, 0xd8, 0xff, 0xe1, 0, 1],
vec![0xff, 0xd8, 0xff, 0xe1, 0, 3],
] {
let len = bytes.len();
let mut reader = CountedReader::new(bytes);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), None);
assert!(reader.bytes_read <= len);
}
}
#[test]
fn propagates_read_and_seek_failures() {
for at in [0, 6] {
let mut reader = CountedReader::new(jpeg(&packet(4)));
reader.read_error_at = Some(at);
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), None);
assert_eq!(reader.bytes_read as u64, at);
}
let mut bytes = vec![0xff, 0xd8];
bytes.extend(segment(0xe0, &[1]));
bytes.extend(xmp_segment(&packet(4)));
let mut reader = CountedReader::new(bytes);
reader.seek_error = true;
assert_eq!(read_jpeg_xmp(&mut reader, JPEG_XMP_HEADER_SPAN, JPEG_XMP_HEADER_STEPS), None);
assert_eq!((reader.bytes_read, reader.seeks), (6, 1));
}
#[test]
fn handles_short_reads_without_bypassing_span_limit() {
let expected = packet(2);
let bytes = jpeg(&expected);
let end = bytes.len();
let mut exact = CountedReader::new(bytes.clone());
exact.chunk = 1;
assert_eq!(read_jpeg_xmp(&mut exact, end as u64, JPEG_XMP_HEADER_STEPS), Some(expected));
assert_eq!(exact.bytes_read, end);
for (span, read) in [(0, 0), (1, 1), (end - 1, 6)] {
let mut reader = CountedReader::new(bytes.clone());
reader.chunk = 1;
assert_eq!(read_jpeg_xmp(&mut reader, span as u64, JPEG_XMP_HEADER_STEPS), None);
assert_eq!(reader.bytes_read, read);
assert!(reader.cursor.position() <= span as u64);
}
}
}
No file is open. Choose one in the tree, or press Ctrl+O (⌘O on a Mac) to find one by name.
Syntax highlighting could not load. The unmodified read-only text remains available above.