diff options
Diffstat (limited to 'third_party/rust/mp4parse/src/lib.rs')
| -rw-r--r-- | third_party/rust/mp4parse/src/lib.rs | 5901 |
1 files changed, 5901 insertions, 0 deletions
diff --git a/third_party/rust/mp4parse/src/lib.rs b/third_party/rust/mp4parse/src/lib.rs new file mode 100644 index 0000000000..b5852dc7e6 --- /dev/null +++ b/third_party/rust/mp4parse/src/lib.rs @@ -0,0 +1,5901 @@ +//! Module for parsing ISO Base Media Format aka video/mp4 streams. + +// This Source Code Form is subject to the terms of the Mozilla Public +// License, v. 2.0. If a copy of the MPL was not distributed with this +// file, You can obtain one at https://mozilla.org/MPL/2.0/. + +// `clippy::upper_case_acronyms` is a nightly-only lint as of 2021-03-15, so we +// allow `clippy::unknown_clippy_lints` to ignore it on stable - but +// `clippy::unknown_clippy_lints` has been renamed in nightly, so we need to +// allow `renamed_and_removed_lints` to ignore a warning for that. +#![allow(renamed_and_removed_lints)] +#![allow(clippy::unknown_clippy_lints)] +#![allow(clippy::upper_case_acronyms)] + +#[macro_use] +extern crate log; + +extern crate bitreader; +extern crate byteorder; +extern crate fallible_collections; +extern crate num_traits; +use bitreader::{BitReader, ReadInto}; +use byteorder::{ReadBytesExt, WriteBytesExt}; + +use fallible_collections::TryRead; +use fallible_collections::TryReserveError; + +use num_traits::Num; +use std::convert::{TryFrom, TryInto as _}; +use std::fmt; +use std::io::Cursor; +use std::io::{Read, Take}; + +#[macro_use] +mod macros; + +mod boxes; +use boxes::{BoxType, FourCC}; + +// Unit tests. +#[cfg(test)] +mod tests; + +#[cfg(feature = "unstable-api")] +pub mod unstable; + +/// The HEIF image and image collection brand +/// The 'mif1' brand indicates structural requirements on files +/// See HEIF (ISO 23008-12:2017) § 10.2.1 +pub const MIF1_BRAND: FourCC = FourCC { value: *b"mif1" }; + +/// The HEIF image sequence brand +/// The 'msf1' brand indicates structural requirements on files +/// See HEIF (ISO 23008-12:2017) § 10.3.1 +pub const MSF1_BRAND: FourCC = FourCC { value: *b"msf1" }; + +/// The brand to identify AV1 image items +/// The 'avif' brand indicates structural requirements on files +/// See <https://aomediacodec.github.io/av1-avif/#image-and-image-collection-brand> +pub const AVIF_BRAND: FourCC = FourCC { value: *b"avif" }; + +/// The brand to identify AVIF image sequences +/// The 'avis' brand indicates structural requirements on files +/// See <https://aomediacodec.github.io/av1-avif/#image-and-image-collection-brand> +pub const AVIS_BRAND: FourCC = FourCC { value: *b"avis" }; + +/// A trait to indicate a type can be infallibly converted to `u64`. +/// This should only be implemented for infallible conversions, so only unsigned types are valid. +trait ToU64 { + fn to_u64(self) -> u64; +} + +/// Statically verify that the platform `usize` can fit within a `u64`. +/// If the size won't fit on the given platform, this will fail at compile time, but if a type +/// which can fail TryInto<usize> is used, it may panic. +impl ToU64 for usize { + fn to_u64(self) -> u64 { + static_assertions::const_assert!( + std::mem::size_of::<usize>() <= std::mem::size_of::<u64>() + ); + self.try_into().expect("usize -> u64 conversion failed") + } +} + +/// A trait to indicate a type can be infallibly converted to `usize`. +/// This should only be implemented for infallible conversions, so only unsigned types are valid. +pub trait ToUsize { + fn to_usize(self) -> usize; +} + +/// Statically verify that the given type can fit within a `usize`. +/// If the size won't fit on the given platform, this will fail at compile time, but if a type +/// which can fail TryInto<usize> is used, it may panic. +macro_rules! impl_to_usize_from { + ( $from_type:ty ) => { + impl ToUsize for $from_type { + fn to_usize(self) -> usize { + static_assertions::const_assert!( + std::mem::size_of::<$from_type>() <= std::mem::size_of::<usize>() + ); + self.try_into().expect(concat!( + stringify!($from_type), + " -> usize conversion failed" + )) + } + } + }; +} + +impl_to_usize_from!(u8); +impl_to_usize_from!(u16); +impl_to_usize_from!(u32); + +/// Indicate the current offset (i.e., bytes already read) in a reader +trait Offset { + fn offset(&self) -> u64; +} + +/// Wraps a reader to track the current offset +struct OffsetReader<'a, T: 'a> { + reader: &'a mut T, + offset: u64, +} + +impl<'a, T> OffsetReader<'a, T> { + fn new(reader: &'a mut T) -> Self { + Self { reader, offset: 0 } + } +} + +impl<'a, T> Offset for OffsetReader<'a, T> { + fn offset(&self) -> u64 { + self.offset + } +} + +impl<'a, T: Read> Read for OffsetReader<'a, T> { + fn read(&mut self, buf: &mut [u8]) -> std::io::Result<usize> { + let bytes_read = self.reader.read(buf)?; + self.offset = self + .offset + .checked_add(bytes_read.to_u64()) + .expect("total bytes read too large for offset type"); + Ok(bytes_read) + } +} + +pub type TryVec<T> = fallible_collections::TryVec<T>; +pub type TryString = fallible_collections::TryVec<u8>; +pub type TryHashMap<K, V> = fallible_collections::TryHashMap<K, V>; +pub type TryBox<T> = fallible_collections::TryBox<T>; + +// To ensure we don't use stdlib allocating types by accident +#[allow(dead_code)] +struct Vec; +#[allow(dead_code)] +struct Box; +#[allow(dead_code)] +struct HashMap; +#[allow(dead_code)] +struct String; + +/// The return value to the C API +/// Any detail that needs to be communicated to the caller must be encoded here +/// since the [`Error`] type's associated data is part of the FFI. +#[repr(C)] +#[derive(Clone, Copy, PartialEq, Debug)] +pub enum Status { + Ok = 0, + BadArg = 1, + Invalid = 2, + Unsupported = 3, + Eof = 4, + Io = 5, + Oom = 6, + MissingBrand, + FtypNotFirst, + NoImage, + MultipleMoov, + NoMoov, + LselNoEssential, + A1opNoEssential, + A1lxEssential, + TxformNoEssential, + NoPrimaryItem, + ImageItemType, + ItemTypeMissing, + ConstructionMethod, + ItemLocNotFound, + NoItemDataBox, +} + +#[repr(C)] +#[derive(Clone, Copy, Debug, PartialEq)] +pub enum Feature { + A1lx, + A1op, + Auxc, + Av1c, + Avis, + Clap, + Colr, + Grid, + Imir, + Ipro, + Irot, + Ispe, + Lsel, + Pasp, + Pixi, +} + +impl Feature { + fn supported(self) -> bool { + match self { + Self::Auxc + | Self::Av1c + | Self::Colr + | Self::Imir + | Self::Irot + | Self::Ispe + | Self::Pasp + | Self::Pixi => true, + Self::A1lx + | Self::A1op + | Self::Clap + | Self::Grid + | Self::Ipro + | Self::Lsel + | Self::Avis => false, + } + } +} + +impl TryFrom<&ItemProperty> for Feature { + type Error = Error; + + fn try_from(item_property: &ItemProperty) -> Result<Self, Self::Error> { + Ok(match item_property { + ItemProperty::AuxiliaryType(_) => Self::Auxc, + ItemProperty::AV1Config(_) => Self::Av1c, + ItemProperty::Channels(_) => Self::Pixi, + ItemProperty::CleanAperture => Self::Clap, + ItemProperty::Colour(_) => Self::Colr, + ItemProperty::ImageSpatialExtents(_) => Self::Ispe, + ItemProperty::LayeredImageIndexing => Self::A1lx, + ItemProperty::LayerSelection => Self::Lsel, + ItemProperty::Mirroring(_) => Self::Imir, + ItemProperty::OperatingPointSelector => Self::A1op, + ItemProperty::PixelAspectRatio(_) => Self::Pasp, + ItemProperty::Rotation(_) => Self::Irot, + item_property => { + error!("No known Feature variant for {:?}", item_property); + return Err(Error::Unsupported("missing Feature fox ItemProperty")); + } + }) + } +} + +/// A collection to indicate unsupported features that were encountered during +/// parsing. Since the default behavior for many such features is to ignore +/// them, this often not fatal and there may be several to report. +#[derive(Debug, Default)] +pub struct UnsupportedFeatures(u32); + +impl UnsupportedFeatures { + pub fn new() -> Self { + Self(0x0) + } + + pub fn into_bitfield(&self) -> u32 { + self.0 + } + + fn feature_to_bitfield(feature: Feature) -> u32 { + let index = feature as usize; + assert!( + u8::BITS.to_usize() * std::mem::size_of::<Self>() > index, + "You're gonna need a bigger bitfield" + ); + let bitfield = 1u32 << index; + assert_eq!(bitfield.count_ones(), 1); + bitfield + } + + pub fn insert(&mut self, feature: Feature) { + warn!("Unsupported feature: {:?}", feature); + self.0 |= Self::feature_to_bitfield(feature); + } + + pub fn contains(&self, feature: Feature) -> bool { + self.0 & Self::feature_to_bitfield(feature) != 0x0 + } + + pub fn is_empty(&self) -> bool { + self.0 == 0x0 + } +} + +impl<T> From<Status> for Result<T> { + /// A convenience method to enable shortcuts like + /// ``` + /// # extern crate mp4parse; + /// # use mp4parse::{Result,Status}; + /// # let _: Result<()> = + /// Status::MissingBrand.into(); + /// ``` + /// instead of + /// ``` + /// # extern crate mp4parse; + /// # use mp4parse::{Error,Result,Status}; + /// # let _: Result<()> = + /// Err(Error::from(Status::MissingBrand)); + /// ``` + /// Note that `Status::Ok` can't be supported this way and will panic. + fn from(parse_status: Status) -> Self { + match parse_status { + Status::Ok => panic!("Can't determine Ok(_) inner value from Status"), + err_status => Err(err_status.into()), + } + } +} + +/// For convenience of creating an error for an unsupported feature which we +/// want to communicate the specific feature back to the C API caller +impl From<Status> for Error { + fn from(parse_status: Status) -> Self { + match parse_status { + Status::Ok + | Status::BadArg + | Status::Invalid + | Status::Unsupported + | Status::Eof + | Status::Io + | Status::Oom => { + panic!("Status -> Error is only for Status:InvalidDataDetail errors") + } + Status::MissingBrand + | Status::FtypNotFirst + | Status::NoImage + | Status::MultipleMoov + | Status::NoMoov + | Status::LselNoEssential + | Status::A1opNoEssential + | Status::A1lxEssential + | Status::TxformNoEssential + | Status::NoPrimaryItem + | Status::ImageItemType + | Status::ItemTypeMissing + | Status::ConstructionMethod + | Status::ItemLocNotFound + | Status::NoItemDataBox => Self::InvalidDataDetail(parse_status), + } + } +} + +impl From<Status> for &str { + fn from(status: Status) -> Self { + match status { + Status::Ok + | Status::BadArg + | Status::Invalid + | Status::Unsupported + | Status::Eof + | Status::Io + | Status::Oom => { + panic!("Status -> Error is only for specific parsing errors") + } + Status::MissingBrand => { + "The file shall list 'avif' or 'avis' in the compatible_brands field + of the FileTypeBox \ + per https://aomediacodec.github.io/av1-avif/#file-constraints" + } + Status::FtypNotFirst => { + "The FileTypeBox shall be placed as early as possible in the file \ + per ISOBMFF (ISO 14496-12:2020) § 4.3.1" + } + Status::NoImage => "No primary image or image sequence found", + Status::NoMoov => { + "No moov box found; \ + files with avis or msf1 brands shall contain exactly one moov box \ + per ISOBMFF (ISO 14496-12:2020) § 8.2.1.1" + } + Status::MultipleMoov => { + "Multiple moov boxes found; \ + files with avis or msf1 brands shall contain exactly one moov box \ + per ISOBMFF (ISO 14496-12:2020) § 8.2.1.1" + } + Status::LselNoEssential => { + "LayerSelectorProperty (lsel) shall be marked as essential \ + per HEIF (ISO/IEC 23008-12:2017) § 6.5.11.1" + } + Status::A1opNoEssential => { + "OperatingPointSelectorProperty (a1op) shall be marked as essential \ + per https://aomediacodec.github.io/av1-avif/#operating-point-selector-property-description" + } + Status::A1lxEssential => { + "AV1LayeredImageIndexingProperty (a1lx) shall not be marked as essential \ + per https://aomediacodec.github.io/av1-avif/#layered-image-indexing-property-description" + } + Status::TxformNoEssential => { + "All transformative properties associated with coded and \ + derived images required or conditionally required by this \ + document shall be marked as essential \ + per MIAF (ISO 23000-22:2019) § 7.3.9" + } + Status::NoPrimaryItem => { + "Missing required PrimaryItemBox (pitm), required \ + per HEIF (ISO/IEC 23008-12:2017) § 10.2.1" + } + Status::ImageItemType => { + "Image item type is neither 'av01' nor 'grid'" + } + Status::ItemTypeMissing => { + "No ItemInfoEntry for item_ID" + } + Status::ConstructionMethod => { + "construction_method shall be 0 (file) or 1 (idat) per MIAF (ISO 23000-22:2019) § 7.2.1.7" + } + Status::ItemLocNotFound => { + "ItemLocationBox (iloc) contains an extent not present in any mdat or idat box" + } + Status::NoItemDataBox => { + "ItemLocationBox (iloc) construction_method indicates 1 (idat), \ + but no idat box is present." + } + } + } +} + +impl From<Error> for Status { + fn from(error: Error) -> Self { + match error { + Error::InvalidData(_) => Self::Invalid, + Error::Unsupported(_) => Self::Unsupported, + Error::InvalidDataDetail(parse_status) => parse_status, + Error::UnexpectedEOF => Self::Eof, + Error::Io(_) => { + // Getting std::io::ErrorKind::UnexpectedEof is normal + // but our From trait implementation should have converted + // those to our Error::UnexpectedEOF variant. + Self::Io + } + Error::NoMoov => Self::NoMoov, + Error::OutOfMemory => Self::Oom, + } + } +} + +impl From<Result<(), Status>> for Status { + fn from(result: Result<(), Status>) -> Self { + match result { + Ok(()) => Status::Ok, + Err(Status::Ok) => unreachable!(), + Err(e) => e, + } + } +} + +impl<T> From<Result<T>> for Status { + fn from(result: Result<T>) -> Self { + match result { + Ok(_) => Status::Ok, + Err(e) => Status::from(e), + } + } +} + +impl From<fallible_collections::TryReserveError> for Status { + fn from(_: fallible_collections::TryReserveError) -> Self { + Status::Oom + } +} + +impl From<std::io::Error> for Status { + fn from(_: std::io::Error) -> Self { + Status::Io + } +} + +/// Describes parser failures. +/// +/// This enum wraps the standard `io::Error` type, unified with +/// our own parser error states and those of crates we use. +#[derive(Debug)] +pub enum Error { + /// Parse error caused by corrupt or malformed data. + InvalidData(&'static str), + /// Similar to [`Self::InvalidData`], but for errors that have a specific + /// [`Status`] variant for communicating the detail across FFI. + /// See the helper [`From<Status> for Error`](enum.Error.html#impl-From<Status>) + InvalidDataDetail(Status), + /// Parse error caused by limited parser support rather than invalid data. + Unsupported(&'static str), + /// Reflect `std::io::ErrorKind::UnexpectedEof` for short data. + UnexpectedEOF, + /// Propagate underlying errors from `std::io`. + Io(std::io::Error), + /// read_mp4 terminated without detecting a moov box. + NoMoov, + /// Out of memory + OutOfMemory, +} + +impl std::fmt::Display for Error { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + write!(f, "{:?}", self) + } +} + +impl std::error::Error for Error {} + +impl From<bitreader::BitReaderError> for Error { + fn from(_: bitreader::BitReaderError) -> Error { + Error::InvalidData("invalid data") + } +} + +impl From<std::io::Error> for Error { + fn from(err: std::io::Error) -> Error { + match err.kind() { + std::io::ErrorKind::UnexpectedEof => Error::UnexpectedEOF, + _ => Error::Io(err), + } + } +} + +impl From<std::string::FromUtf8Error> for Error { + fn from(_: std::string::FromUtf8Error) -> Error { + Error::InvalidData("invalid utf8") + } +} + +impl From<std::str::Utf8Error> for Error { + fn from(_: std::str::Utf8Error) -> Error { + Error::InvalidData("invalid utf8") + } +} + +impl From<std::num::TryFromIntError> for Error { + fn from(_: std::num::TryFromIntError) -> Error { + Error::Unsupported("integer conversion failed") + } +} + +impl From<Error> for std::io::Error { + fn from(err: Error) -> Self { + let kind = match err { + Error::InvalidData(_) => std::io::ErrorKind::InvalidData, + Error::UnexpectedEOF => std::io::ErrorKind::UnexpectedEof, + Error::Io(io_err) => return io_err, + _ => std::io::ErrorKind::Other, + }; + Self::new(kind, err) + } +} + +impl From<TryReserveError> for Error { + fn from(_: TryReserveError) -> Error { + Error::OutOfMemory + } +} + +/// Result shorthand using our Error enum. +pub type Result<T, E = Error> = std::result::Result<T, E>; + +/// Basic ISO box structure. +/// +/// mp4 files are a sequence of possibly-nested 'box' structures. Each box +/// begins with a header describing the length of the box's data and a +/// four-byte box type which identifies the type of the box. Together these +/// are enough to interpret the contents of that section of the file. +/// +/// See ISOBMFF (ISO 14496-12:2020) § 4.2 +#[derive(Debug, Clone, Copy)] +struct BoxHeader { + /// Box type. + name: BoxType, + /// Size of the box in bytes. + size: u64, + /// Offset to the start of the contained data (or header size). + offset: u64, + /// Uuid for extended type. + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + uuid: Option<[u8; 16]>, +} + +impl BoxHeader { + const MIN_SIZE: u64 = 8; // 4-byte size + 4-byte type + const MIN_LARGE_SIZE: u64 = 16; // 4-byte size + 4-byte type + 16-byte size +} + +/// File type box 'ftyp'. +#[derive(Debug)] +struct FileTypeBox { + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + major_brand: FourCC, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + minor_version: u32, + compatible_brands: TryVec<FourCC>, +} + +impl FileTypeBox { + fn contains(&self, brand: &FourCC) -> bool { + self.compatible_brands.contains(brand) || self.major_brand == *brand + } +} + +/// Movie header box 'mvhd'. +#[derive(Debug)] +struct MovieHeaderBox { + pub timescale: u32, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + duration: u64, +} + +#[derive(Debug, Clone, Copy)] +pub struct Matrix { + pub a: i32, // 16.16 fix point + pub b: i32, // 16.16 fix point + pub u: i32, // 2.30 fix point + pub c: i32, // 16.16 fix point + pub d: i32, // 16.16 fix point + pub v: i32, // 2.30 fix point + pub x: i32, // 16.16 fix point + pub y: i32, // 16.16 fix point + pub w: i32, // 2.30 fix point +} + +/// Track header box 'tkhd' +#[derive(Debug, Clone)] +pub struct TrackHeaderBox { + track_id: u32, + pub disabled: bool, + pub duration: u64, + pub width: u32, + pub height: u32, + pub matrix: Matrix, +} + +/// Edit list box 'elst' +#[derive(Debug)] +struct EditListBox { + edits: TryVec<Edit>, +} + +#[derive(Debug)] +struct Edit { + segment_duration: u64, + media_time: i64, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + media_rate_integer: i16, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + media_rate_fraction: i16, +} + +/// Media header box 'mdhd' +#[derive(Debug)] +struct MediaHeaderBox { + timescale: u32, + duration: u64, +} + +// Chunk offset box 'stco' or 'co64' +#[derive(Debug)] +pub struct ChunkOffsetBox { + pub offsets: TryVec<u64>, +} + +// Sync sample box 'stss' +#[derive(Debug)] +pub struct SyncSampleBox { + pub samples: TryVec<u32>, +} + +// Sample to chunk box 'stsc' +#[derive(Debug)] +pub struct SampleToChunkBox { + pub samples: TryVec<SampleToChunk>, +} + +#[derive(Debug)] +pub struct SampleToChunk { + pub first_chunk: u32, + pub samples_per_chunk: u32, + pub sample_description_index: u32, +} + +// Sample size box 'stsz' +#[derive(Debug)] +pub struct SampleSizeBox { + pub sample_size: u32, + pub sample_sizes: TryVec<u32>, +} + +// Time to sample box 'stts' +#[derive(Debug)] +pub struct TimeToSampleBox { + pub samples: TryVec<Sample>, +} + +#[repr(C)] +#[derive(Debug)] +pub struct Sample { + pub sample_count: u32, + pub sample_delta: u32, +} + +#[derive(Debug, Clone, Copy)] +pub enum TimeOffsetVersion { + Version0(u32), + Version1(i32), +} + +#[derive(Debug, Clone)] +pub struct TimeOffset { + pub sample_count: u32, + pub time_offset: TimeOffsetVersion, +} + +#[derive(Debug)] +pub struct CompositionOffsetBox { + pub samples: TryVec<TimeOffset>, +} + +// Handler reference box 'hdlr' +#[derive(Debug)] +struct HandlerBox { + handler_type: FourCC, +} + +// Sample description box 'stsd' +#[derive(Debug)] +pub struct SampleDescriptionBox { + pub descriptions: TryVec<SampleEntry>, +} + +#[derive(Debug)] +pub enum SampleEntry { + Audio(AudioSampleEntry), + Video(VideoSampleEntry), + Unknown, +} + +/// An Elementary Stream Descriptor +/// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.5 +#[allow(non_camel_case_types)] +#[derive(Debug, Default)] +pub struct ES_Descriptor { + pub audio_codec: CodecType, + pub audio_object_type: Option<u16>, + pub extended_audio_object_type: Option<u16>, + pub audio_sample_rate: Option<u32>, + pub audio_channel_count: Option<u16>, + #[cfg(feature = "mp4v")] + pub video_codec: CodecType, + pub codec_esds: TryVec<u8>, + pub decoder_specific_data: TryVec<u8>, // Data in DECODER_SPECIFIC_TAG +} + +#[allow(non_camel_case_types)] +#[derive(Debug)] +pub enum AudioCodecSpecific { + ES_Descriptor(ES_Descriptor), + FLACSpecificBox(FLACSpecificBox), + OpusSpecificBox(OpusSpecificBox), + ALACSpecificBox(ALACSpecificBox), + MP3, + LPCM, + #[cfg(feature = "3gpp")] + AMRSpecificBox(TryVec<u8>), +} + +#[derive(Debug)] +pub struct AudioSampleEntry { + pub codec_type: CodecType, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + data_reference_index: u16, + pub channelcount: u32, + pub samplesize: u16, + pub samplerate: f64, + pub codec_specific: AudioCodecSpecific, + pub protection_info: TryVec<ProtectionSchemeInfoBox>, +} + +#[derive(Debug)] +pub enum VideoCodecSpecific { + AVCConfig(TryVec<u8>), + VPxConfig(VPxConfigBox), + AV1Config(AV1ConfigBox), + ESDSConfig(TryVec<u8>), + H263Config(TryVec<u8>), +} + +#[derive(Debug)] +pub struct VideoSampleEntry { + pub codec_type: CodecType, + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + data_reference_index: u16, + pub width: u16, + pub height: u16, + pub codec_specific: VideoCodecSpecific, + pub protection_info: TryVec<ProtectionSchemeInfoBox>, +} + +/// Represent a Video Partition Codec Configuration 'vpcC' box (aka vp9). The meaning of each +/// field is covered in detail in "VP Codec ISO Media File Format Binding". +#[derive(Debug)] +pub struct VPxConfigBox { + /// An integer that specifies the VP codec profile. + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + profile: u8, + /// An integer that specifies a VP codec level all samples conform to the following table. + /// For a description of the various levels, please refer to the VP9 Bitstream Specification. + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + level: u8, + /// An integer that specifies the bit depth of the luma and color components. Valid values + /// are 8, 10, and 12. + pub bit_depth: u8, + /// Really an enum defined by the "Colour primaries" section of ISO 23091-2:2019 § 8.1. + pub colour_primaries: u8, + /// Really an enum defined by "VP Codec ISO Media File Format Binding". + pub chroma_subsampling: u8, + /// Really an enum defined by the "Transfer characteristics" section of ISO 23091-2:2019 § 8.2. + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + transfer_characteristics: u8, + /// Really an enum defined by the "Matrix coefficients" section of ISO 23091-2:2019 § 8.3. + /// Available in 'VP Codec ISO Media File Format' version 1 only. + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + matrix_coefficients: Option<u8>, + /// Indicates the black level and range of the luma and chroma signals. 0 = legal range + /// (e.g. 16-235 for 8 bit sample depth); 1 = full range (e.g. 0-255 for 8-bit sample depth). + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + video_full_range_flag: bool, + /// This is not used for VP8 and VP9 . Intended for binary codec initialization data. + pub codec_init: TryVec<u8>, +} + +/// See [AV1-ISOBMFF § 2.3.3](https://aomediacodec.github.io/av1-isobmff/#av1codecconfigurationbox-syntax) +#[derive(Debug)] +pub struct AV1ConfigBox { + pub profile: u8, + pub level: u8, + pub tier: u8, + pub bit_depth: u8, + pub monochrome: bool, + pub chroma_subsampling_x: u8, + pub chroma_subsampling_y: u8, + pub chroma_sample_position: u8, + pub initial_presentation_delay_present: bool, + pub initial_presentation_delay_minus_one: u8, + // The raw config contained in the av1c box. Because some decoders accept this data as a binary + // blob, rather than as structured data, we store the blob here for convenience. + pub raw_config: TryVec<u8>, +} + +impl AV1ConfigBox { + const CONFIG_OBUS_OFFSET: usize = 4; + + pub fn config_obus(&self) -> &[u8] { + &self.raw_config[Self::CONFIG_OBUS_OFFSET..] + } +} + +#[derive(Debug)] +pub struct FLACMetadataBlock { + pub block_type: u8, + pub data: TryVec<u8>, +} + +/// Represents a FLACSpecificBox 'dfLa' +#[derive(Debug)] +pub struct FLACSpecificBox { + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + version: u8, + pub blocks: TryVec<FLACMetadataBlock>, +} + +#[derive(Debug)] +struct ChannelMappingTable { + stream_count: u8, + coupled_count: u8, + channel_mapping: TryVec<u8>, +} + +/// Represent an OpusSpecificBox 'dOps' +#[derive(Debug)] +pub struct OpusSpecificBox { + pub version: u8, + output_channel_count: u8, + pre_skip: u16, + input_sample_rate: u32, + output_gain: i16, + channel_mapping_family: u8, + channel_mapping_table: Option<ChannelMappingTable>, +} + +/// Represent an ALACSpecificBox 'alac' +#[derive(Debug)] +pub struct ALACSpecificBox { + #[allow(dead_code)] // See https://github.com/mozilla/mp4parse-rust/issues/340 + version: u8, + pub data: TryVec<u8>, +} + +#[derive(Debug)] +pub struct MovieExtendsBox { + pub fragment_duration: Option<MediaScaledTime>, +} + +pub type ByteData = TryVec<u8>; + +#[derive(Debug, Default)] +pub struct ProtectionSystemSpecificHeaderBox { + pub system_id: ByteData, + pub kid: TryVec<ByteData>, + pub data: ByteData, + + // The entire pssh box (include header) required by Gecko. + pub box_content: ByteData, +} + +#[derive(Debug, Default, Clone)] +pub struct SchemeTypeBox { + pub scheme_type: FourCC, + pub scheme_version: u32, +} + +#[derive(Debug, Default)] +pub struct TrackEncryptionBox { + pub is_encrypted: u8, + pub iv_size: u8, + pub kid: TryVec<u8>, + // Members for pattern encryption schemes + pub crypt_byte_block_count: Option<u8>, + pub skip_byte_block_count: Option<u8>, + pub constant_iv: Option<TryVec<u8>>, + // End pattern encryption scheme members +} + +#[derive(Debug, Default)] +pub struct ProtectionSchemeInfoBox { + pub original_format: FourCC, + pub scheme_type: Option<SchemeTypeBox>, + pub tenc: Option<TrackEncryptionBox>, +} + +/// Represents a userdata box 'udta'. +/// Currently, only the metadata atom 'meta' +/// is parsed. +#[derive(Debug, Default)] +pub struct UserdataBox { + pub meta: Option<MetadataBox>, +} + +/// Represents possible contents of the +/// ©gen or gnre atoms within a metadata box. +/// 'udta.meta.ilst' may only have either a +/// standard genre box 'gnre' or a custom +/// genre box '©gen', but never both at once. +#[derive(Debug, PartialEq)] +pub enum Genre { + /// A standard ID3v1 numbered genre. + StandardGenre(u8), + /// Any custom genre string. + CustomGenre(TryString), +} + +/// Represents the contents of a 'stik' +/// atom that indicates content types within +/// iTunes. +#[derive(Debug, Clone, Eq, PartialEq)] +pub enum MediaType { + /// Movie is stored as 0 in a 'stik' atom. + Movie, // 0 + /// Normal is stored as 1 in a 'stik' atom. + Normal, // 1 + /// AudioBook is stored as 2 in a 'stik' atom. + AudioBook, // 2 + /// WhackedBookmark is stored as 5 in a 'stik' atom. + WhackedBookmark, // 5 + /// MusicVideo is stored as 6 in a 'stik' atom. + MusicVideo, // 6 + /// ShortFilm is stored as 9 in a 'stik' atom. + ShortFilm, // 9 + /// TVShow is stored as 10 in a 'stik' atom. + TVShow, // 10 + /// Booklet is stored as 11 in a 'stik' atom. + Booklet, // 11 + /// An unknown 'stik' value. + Unknown(u8), +} + +/// Represents the parental advisory rating on the track, +/// stored within the 'rtng' atom. +#[derive(Debug, Clone, Eq, PartialEq)] +pub enum AdvisoryRating { + /// Clean is always stored as 2 in an 'rtng' atom. + Clean, // 2 + /// A value of 0 in an 'rtng' atom indicates 'Inoffensive' + Inoffensive, // 0 + /// Any non 2 or 0 value in 'rtng' indicates the track is explicit. + Explicit(u8), +} + +/// Represents the contents of 'ilst' atoms within +/// a metadata box 'meta', parsed as iTunes metadata using +/// the conventional tags. +#[derive(Debug, Default)] +pub struct MetadataBox { + /// The album name, '©alb' + pub album: Option<TryString>, + /// The artist name '©art' or '©ART' + pub artist: Option<TryString>, + /// The album artist 'aART' + pub album_artist: Option<TryString>, + /// Track comments '©cmt' + pub comment: Option<TryString>, + /// The date or year field '©day' + /// + /// This is stored as an arbitrary string, + /// and may not necessarily be in a valid date + /// format. + pub year: Option<TryString>, + /// The track title '©nam' + pub title: Option<TryString>, + /// The track genre '©gen' or 'gnre'. + pub genre: Option<Genre>, + /// The track number 'trkn'. + pub track_number: Option<u8>, + /// The disc number 'disk' + pub disc_number: Option<u8>, + /// The total number of tracks on the disc, + /// stored in 'trkn' + pub total_tracks: Option<u8>, + /// The total number of discs in the album, + /// stored in 'disk' + pub total_discs: Option<u8>, + /// The composer of the track '©wrt' + pub composer: Option<TryString>, + /// The encoder used to create this track '©too' + pub encoder: Option<TryString>, + /// The encoded-by settingo this track '©enc' + pub encoded_by: Option<TryString>, + /// The tempo or BPM of the track 'tmpo' + pub beats_per_minute: Option<u8>, + /// Copyright information of the track 'cprt' + pub copyright: Option<TryString>, + /// Whether or not this track is part of a compilation 'cpil' + pub compilation: Option<bool>, + /// The advisory rating of this track 'rtng' + pub advisory: Option<AdvisoryRating>, + /// The personal rating of this track, 'rate'. + /// + /// This is stored in the box as string data, but + /// the format is an integer percentage from 0 - 100, + /// where 100 is displayed as 5 stars out of 5. + pub rating: Option<TryString>, + /// The grouping this track belongs to '©grp' + pub grouping: Option<TryString>, + /// The media type of this track 'stik' + pub media_type: Option<MediaType>, // stik + /// Whether or not this track is a podcast 'pcst' + pub podcast: Option<bool>, + /// The category of ths track 'catg' + pub category: Option<TryString>, + /// The podcast keyword 'keyw' + pub keyword: Option<TryString>, + /// The podcast url 'purl' + pub podcast_url: Option<TryString>, + /// The podcast episode GUID 'egid' + pub podcast_guid: Option<TryString>, + /// The description of the track 'desc' + pub description: Option<TryString>, + /// The long description of the track 'ldes'. + /// + /// Unlike other string fields, the long description field + /// can be longer than 256 characters. + pub long_description: Option<TryString>, + /// The lyrics of the track '©lyr'. + /// + /// Unlike other string fields, the lyrics field + /// can be longer than 256 characters. + pub lyrics: Option<TryString>, + /// The name of the TV network this track aired on 'tvnn'. + pub tv_network_name: Option<TryString>, + /// The name of the TV Show for this track 'tvsh'. + pub tv_show_name: Option<TryString>, + /// The name of the TV Episode for this track 'tven'. + pub tv_episode_name: Option<TryString>, + /// The number of the TV Episode for this track 'tves'. + pub tv_episode_number: Option<u8>, + /// The season of the TV Episode of this track 'tvsn'. + pub tv_season: Option<u8>, + /// The date this track was purchased 'purd'. + pub purchase_date: Option<TryString>, + /// Whether or not this track supports gapless playback 'pgap' + pub gapless_playback: Option<bool>, + /// Any cover artwork attached to this track 'covr' + /// + /// 'covr' is unique in that it may contain multiple 'data' sub-entries, + /// each an image file. Here, each subentry's raw binary data is exposed, + /// which may contain image data in JPEG or PNG format. + pub cover_art: Option<TryVec<TryVec<u8>>>, + /// The owner of the track 'ownr' + pub owner: Option<TryString>, + /// Whether or not this track is HD Video 'hdvd' + pub hd_video: Option<bool>, + /// The name of the track to sort by 'sonm' + pub sort_name: Option<TryString>, + /// The name of the album to sort by 'soal' + pub sort_album: Option<TryString>, + /// The name of the artist to sort by 'soar' + pub sort_artist: Option<TryString>, + /// The name of the album artist to sort by 'soaa' + pub sort_album_artist: Option<TryString>, + /// The name of the composer to sort by 'soco' + pub sort_composer: Option<TryString>, + /// Metadata + #[cfg(feature = "meta-xml")] + pub xml: Option<XmlBox>, +} + +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.2.1 +#[cfg(feature = "meta-xml")] +#[derive(Debug)] +pub enum XmlBox { + /// XML metadata + StringXmlBox(TryString), + /// Binary XML metadata + BinaryXmlBox(TryVec<u8>), +} + +/// Internal data structures. +#[derive(Debug, Default)] +pub struct MediaContext { + pub timescale: Option<MediaTimeScale>, + /// Tracks found in the file. + pub tracks: TryVec<Track>, + pub mvex: Option<MovieExtendsBox>, + pub psshs: TryVec<ProtectionSystemSpecificHeaderBox>, + pub userdata: Option<Result<UserdataBox>>, + #[cfg(feature = "meta-xml")] + pub metadata: Option<Result<MetadataBox>>, +} + +/// An ISOBMFF item as described by an iloc box. For the sake of avoiding copies, +/// this can either be represented by the `Location` variant, which indicates +/// where the data exists within a `MediaDataBox` stored separately, or the +/// `Data` variant which owns the data. Unfortunately, it's not simple to +/// represent this as a [`std::borrow::Cow`], or other reference-based type, because +/// multiple instances may references different parts of the same [`MediaDataBox`] +/// and we want to avoid the copy that splitting the storage would entail. +enum IsobmffItem { + MdatLocation(Extent), + IdatLocation(Extent), + Data(TryVec<u8>), +} + +impl fmt::Debug for IsobmffItem { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + match &self { + IsobmffItem::MdatLocation(extent) | IsobmffItem::IdatLocation(extent) => f + .debug_struct("IsobmffItem::Location") + .field("0", &format_args!("{:?}", extent)) + .finish(), + IsobmffItem::Data(data) => f + .debug_struct("IsobmffItem::Data") + .field("0", &format_args!("{} bytes", data.len())) + .finish(), + } + } +} + +#[derive(Debug)] +struct AvifItem { + /// The `item_ID` from ISOBMFF (ISO 14496-12:2020) § 8.11.3 + /// + /// See [`read_iloc`] + id: ItemId, + + /// AV1 Image Item per <https://aomediacodec.github.io/av1-avif/#image-item> + image_data: IsobmffItem, +} + +impl AvifItem { + fn with_inline_data(id: ItemId) -> Self { + Self { + id, + image_data: IsobmffItem::Data(TryVec::new()), + } + } +} + +#[derive(Debug)] +pub struct AvifContext { + /// Level of deviation from the specification before failing the parse + strictness: ParseStrictness, + /// Referred to by the `IsobmffItem::*Location` variants of the `AvifItem`s in this struct + media_storage: TryVec<MediaDataBox>, + /// The item indicated by the `pitm` box, See ISOBMFF (ISO 14496-12:2020) § 8.11.4 + /// May be `None` in the pure image sequence case. + primary_item: Option<AvifItem>, + /// Associated alpha channel for the primary item, if any + alpha_item: Option<AvifItem>, + /// If true, divide RGB values by the alpha value. + /// See `prem` in MIAF (ISO 23000-22:2019) § 7.3.5.2 + pub premultiplied_alpha: bool, + /// All properties associated with `primary_item` or `alpha_item` + item_properties: ItemPropertiesBox, + /// Should probably only ever be [`AVIF_BRAND`] or [`AVIS_BRAND`], but other values + /// are legal as long as one of the two is the `compatible_brand` list. + pub major_brand: FourCC, + /// True if a `moov` box is present + pub has_sequence: bool, + /// A collection of unsupported features encountered during the parse + pub unsupported_features: UnsupportedFeatures, +} + +impl AvifContext { + pub fn primary_item_coded_data(&self) -> Option<&[u8]> { + self.primary_item + .as_ref() + .map(|item| self.item_as_slice(item)) + } + + pub fn primary_item_bits_per_channel(&self) -> Option<Result<&[u8]>> { + self.primary_item + .as_ref() + .map(|item| self.image_bits_per_channel(item.id)) + } + + pub fn alpha_item_coded_data(&self) -> Option<&[u8]> { + self.alpha_item + .as_ref() + .map(|item| self.item_as_slice(item)) + } + + pub fn alpha_item_bits_per_channel(&self) -> Option<Result<&[u8]>> { + self.alpha_item + .as_ref() + .map(|item| self.image_bits_per_channel(item.id)) + } + + fn image_bits_per_channel(&self, item_id: ItemId) -> Result<&[u8]> { + match self + .item_properties + .get(item_id, BoxType::PixelInformationBox)? + { + Some(ItemProperty::Channels(pixi)) => Ok(pixi.bits_per_channel.as_slice()), + Some(other_property) => panic!("property key mismatch: {:?}", other_property), + None => Ok(&[]), + } + } + + pub fn spatial_extents_ptr(&self) -> Result<*const ImageSpatialExtentsProperty> { + if let Some(primary_item) = &self.primary_item { + match self + .item_properties + .get(primary_item.id, BoxType::ImageSpatialExtentsProperty)? + { + Some(ItemProperty::ImageSpatialExtents(ispe)) => Ok(ispe), + Some(other_property) => panic!("property key mismatch: {:?}", other_property), + None => { + fail_if( + self.strictness != ParseStrictness::Permissive, + "ispe is a mandatory property", + )?; + Ok(std::ptr::null()) + } + } + } else { + Ok(std::ptr::null()) + } + } + + /// Returns None if there is no primary item or it has no associated NCLX colour boxes. + pub fn nclx_colour_information_ptr(&self) -> Option<Result<*const NclxColourInformation>> { + if let Some(primary_item) = &self.primary_item { + match self.item_properties.get_multiple(primary_item.id, |prop| { + matches!(prop, ItemProperty::Colour(ColourInformation::Nclx(_))) + }) { + Ok(nclx_colr_boxes) => match *nclx_colr_boxes.as_slice() { + [] => None, + [ItemProperty::Colour(ColourInformation::Nclx(nclx)), ..] => { + if nclx_colr_boxes.len() > 1 { + warn!("Multiple nclx colr boxes, using first"); + } + Some(Ok(nclx)) + } + _ => unreachable!("Expect only ColourInformation::Nclx(_) matches"), + }, + Err(e) => Some(Err(e)), + } + } else { + None + } + } + + /// Returns None if there is no primary item or it has no associated ICC colour boxes. + pub fn icc_colour_information(&self) -> Option<Result<&[u8]>> { + if let Some(primary_item) = &self.primary_item { + match self.item_properties.get_multiple(primary_item.id, |prop| { + matches!(prop, ItemProperty::Colour(ColourInformation::Icc(_, _))) + }) { + Ok(icc_colr_boxes) => match *icc_colr_boxes.as_slice() { + [] => None, + [ItemProperty::Colour(ColourInformation::Icc(icc, _)), ..] => { + if icc_colr_boxes.len() > 1 { + warn!("Multiple ICC profiles in colr boxes, using first"); + } + Some(Ok(icc.bytes.as_slice())) + } + _ => unreachable!("Expect only ColourInformation::Icc(_) matches"), + }, + Err(e) => Some(Err(e)), + } + } else { + None + } + } + + pub fn image_rotation(&self) -> Result<ImageRotation> { + if let Some(primary_item) = &self.primary_item { + match self + .item_properties + .get(primary_item.id, BoxType::ImageRotation)? + { + Some(ItemProperty::Rotation(irot)) => Ok(*irot), + Some(other_property) => panic!("property key mismatch: {:?}", other_property), + None => Ok(ImageRotation::D0), + } + } else { + Ok(ImageRotation::D0) + } + } + + pub fn image_mirror_ptr(&self) -> Result<*const ImageMirror> { + if let Some(primary_item) = &self.primary_item { + match self + .item_properties + .get(primary_item.id, BoxType::ImageMirror)? + { + Some(ItemProperty::Mirroring(imir)) => Ok(imir), + Some(other_property) => panic!("property key mismatch: {:?}", other_property), + None => Ok(std::ptr::null()), + } + } else { + Ok(std::ptr::null()) + } + } + + pub fn pixel_aspect_ratio_ptr(&self) -> Result<*const PixelAspectRatio> { + if let Some(primary_item) = &self.primary_item { + match self + .item_properties + .get(primary_item.id, BoxType::PixelAspectRatioBox)? + { + Some(ItemProperty::PixelAspectRatio(pasp)) => Ok(pasp), + Some(other_property) => panic!("property key mismatch: {:?}", other_property), + None => Ok(std::ptr::null()), + } + } else { + Ok(std::ptr::null()) + } + } + + /// A helper for the various `AvifItem`s to expose a reference to the + /// underlying data while avoiding copies. + fn item_as_slice<'a>(&'a self, item: &'a AvifItem) -> &'a [u8] { + match &item.image_data { + IsobmffItem::MdatLocation(extent) => { + for mdat in &self.media_storage { + if let Some(slice) = mdat.get(extent) { + return slice; + } + } + unreachable!( + "IsobmffItem::Location requires the location exists in AvifContext::media_storage" + ); + } + IsobmffItem::IdatLocation(_) => unimplemented!(), + IsobmffItem::Data(data) => data.as_slice(), + } + } +} + +struct AvifMeta { + item_references: TryVec<SingleItemTypeReferenceBox>, + item_properties: ItemPropertiesBox, + /// Required for AvifImageType::Primary, but optional otherwise + /// See HEIF (ISO/IEC 23008-12:2017) § 7.1, 10.2.1 + primary_item_id: Option<ItemId>, + item_infos: TryVec<ItemInfoEntry>, + iloc_items: TryHashMap<ItemId, ItemLocationBoxItem>, + item_data_box: Option<ItemDataBox>, +} + +/// An Item Data Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.11 +struct ItemDataBox { + data: TryVec<u8>, +} + +/// A Media Data Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.1.1 +struct MediaDataBox { + /// Offset of `data` from the beginning of the "file". See ConstructionMethod::File. + /// Note: the file may not be an actual file, read_avif supports any `&mut impl Read` + /// source for input. However we try to match the terminology used in the spec. + file_offset: u64, + data: TryVec<u8>, +} + +impl fmt::Debug for MediaDataBox { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + f.debug_struct("MediaDataBox") + .field("file_offset", &self.file_offset) + .field("data", &format_args!("{} bytes", self.data.len())) + .finish() + } +} + +fn u64_to_usize_logged(x: u64) -> Option<usize> { + match x.try_into() { + Ok(x) => Some(x), + Err(e) => { + error!("{:?} converting {:?}", e, x); + None + } + } +} + +/// Generalizes the different data boxes a [`ItemLocationBoxItem`] can refer to +trait DataBox { + fn data(&self) -> &[u8]; + + /// Convert an absolute offset to an offset relative to the beginning of the + /// slice [`DataBox::data`] returns. Returns None if the offset would be + /// negative or if the offset would overflow a `usize`. + fn start(&self, offset: u64) -> Option<usize>; + + /// Returns an appropriate variant of [`IsobmffItem`] to describe the extent + /// referencing data within this type of box. + fn location(&self, extent: &Extent) -> IsobmffItem; + + /// Return a slice from the DataBox specified by the provided `extent`. + /// Returns `None` if the extent isn't fully contained by the DataBox or if + /// either the offset or length (if the extent is bounded) of the slice + /// would overflow a `usize`. + fn get<'a>(&'a self, extent: &'a Extent) -> Option<&'a [u8]> { + match extent { + Extent::WithLength { offset, len } => { + let start = self.start(*offset)?; + let end = start.checked_add(*len); + if end.is_none() { + error!("Overflow adding {} + {}", start, len); + } + self.data().get(start..end?) + } + Extent::ToEnd { offset } => { + let start = self.start(*offset)?; + self.data().get(start..) + } + } + } +} + +impl DataBox for ItemDataBox { + fn data(&self) -> &[u8] { + &self.data + } + + fn start(&self, offset: u64) -> Option<usize> { + u64_to_usize_logged(offset) + } + + fn location(&self, extent: &Extent) -> IsobmffItem { + IsobmffItem::IdatLocation(extent.clone()) + } +} + +impl DataBox for MediaDataBox { + fn data(&self) -> &[u8] { + &self.data + } + + fn start(&self, offset: u64) -> Option<usize> { + let start = offset.checked_sub(self.file_offset); + if start.is_none() { + error!("Overflow subtracting {} + {}", offset, self.file_offset); + } + u64_to_usize_logged(start?) + } + + fn location(&self, extent: &Extent) -> IsobmffItem { + IsobmffItem::MdatLocation(extent.clone()) + } +} + +#[cfg(test)] +mod media_data_box_tests { + use super::*; + + impl MediaDataBox { + fn at_offset(file_offset: u64, data: std::vec::Vec<u8>) -> Self { + MediaDataBox { + file_offset, + data: data.into(), + } + } + } + + #[test] + fn extent_with_length_before_mdat_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { offset: 0, len: 2 }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_to_end_before_mdat_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::ToEnd { offset: 0 }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_with_length_crossing_front_mdat_boundary_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { offset: 99, len: 3 }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_with_length_which_is_subset_of_mdat() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { + offset: 101, + len: 2, + }; + + assert_eq!(mdat.get(&extent), Some(&[1, 1][..])); + } + + #[test] + fn extent_to_end_which_is_subset_of_mdat() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::ToEnd { offset: 101 }; + + assert_eq!(mdat.get(&extent), Some(&[1, 1, 1, 1][..])); + } + + #[test] + fn extent_with_length_which_is_all_of_mdat() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { + offset: 100, + len: 5, + }; + + assert_eq!(mdat.get(&extent), Some(mdat.data.as_slice())); + } + + #[test] + fn extent_to_end_which_is_all_of_mdat() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::ToEnd { offset: 100 }; + + assert_eq!(mdat.get(&extent), Some(mdat.data.as_slice())); + } + + #[test] + fn extent_with_length_crossing_back_mdat_boundary_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { + offset: 103, + len: 3, + }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_with_length_after_mdat_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::WithLength { + offset: 200, + len: 2, + }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_to_end_after_mdat_returns_none() { + let mdat = MediaDataBox::at_offset(100, vec![1; 5]); + let extent = Extent::ToEnd { offset: 200 }; + + assert!(mdat.get(&extent).is_none()); + } + + #[test] + fn extent_with_length_which_overflows_usize() { + let mdat = MediaDataBox::at_offset(std::u64::MAX - 1, vec![1; 5]); + let extent = Extent::WithLength { + offset: std::u64::MAX, + len: std::usize::MAX, + }; + + assert!(mdat.get(&extent).is_none()); + } + + // The end of the range would overflow `usize` if it were calculated, but + // because the range end is unbounded, we don't calculate it. + #[test] + fn extent_to_end_which_overflows_usize() { + let mdat = MediaDataBox::at_offset(std::u64::MAX - 1, vec![1; 5]); + let extent = Extent::ToEnd { + offset: std::u64::MAX, + }; + + assert_eq!(mdat.get(&extent), Some(&[1, 1, 1, 1][..])); + } +} + +#[derive(Clone, Copy, Debug, Eq, Hash, PartialEq)] +struct PropertyIndex(u16); +#[derive(Clone, Copy, Debug, Eq, Hash, PartialEq, PartialOrd)] +struct ItemId(u32); + +impl ItemId { + fn read(src: &mut impl ReadBytesExt, version: u8) -> Result<ItemId> { + Ok(ItemId(if version == 0 { + be_u16(src)?.into() + } else { + be_u32(src)? + })) + } +} + +/// Used for 'infe' boxes within 'iinf' boxes +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.6 +/// Only versions {2, 3} are supported +#[derive(Debug)] +struct ItemInfoEntry { + item_id: ItemId, + item_type: u32, +} + +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.12 +#[derive(Debug)] +struct SingleItemTypeReferenceBox { + item_type: FourCC, + from_item_id: ItemId, + to_item_id: ItemId, +} + +/// Potential sizes (in bytes) of variable-sized fields of the 'iloc' box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.3 +#[derive(Debug, Clone, Copy, PartialEq)] +enum IlocFieldSize { + Zero, + Four, + Eight, +} + +impl IlocFieldSize { + fn as_bits(&self) -> u8 { + match self { + IlocFieldSize::Zero => 0, + IlocFieldSize::Four => 32, + IlocFieldSize::Eight => 64, + } + } +} + +impl TryFrom<u8> for IlocFieldSize { + type Error = Error; + + fn try_from(value: u8) -> Result<Self> { + match value { + 0 => Ok(Self::Zero), + 4 => Ok(Self::Four), + 8 => Ok(Self::Eight), + _ => Err(Error::InvalidData("value must be in the set {0, 4, 8}")), + } + } +} + +#[derive(Debug, PartialEq)] +enum IlocVersion { + Zero, + One, + Two, +} + +impl TryFrom<u8> for IlocVersion { + type Error = Error; + + fn try_from(value: u8) -> Result<Self> { + match value { + 0 => Ok(Self::Zero), + 1 => Ok(Self::One), + 2 => Ok(Self::Two), + _ => Err(Error::Unsupported("unsupported version in 'iloc' box")), + } + } +} + +/// Used for 'iloc' boxes +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.3 +/// `base_offset` is omitted since it is integrated into the ranges in `extents` +/// `data_reference_index` is omitted, since only 0 (i.e., this file) is supported +#[derive(Debug)] +struct ItemLocationBoxItem { + construction_method: ConstructionMethod, + /// Unused for ConstructionMethod::Idat + extents: TryVec<Extent>, +} + +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.3 +/// +/// Note: per MIAF (ISO 23000-22:2019) § 7.2.1.7:<br /> +/// > MIAF image items are constrained as follows:<br /> +/// > — `construction_method` shall be equal to 0 for MIAF image items that are coded image items.<br /> +/// > — `construction_method` shall be equal to 0 or 1 for MIAF image items that are derived image items. +#[derive(Clone, Copy, Debug, PartialEq)] +enum ConstructionMethod { + File = 0, + Idat = 1, + Item = 2, +} + +/// Describes a region where a item specified by an `ItemLocationBoxItem` is stored. +/// The offset is `u64` since that's the maximum possible size and since the relative +/// nature of `MediaDataBox` means this can still possibly succeed even in the case +/// that the raw value exceeds std::usize::MAX on platforms where that type is smaller +/// than u64. However, `len` is stored as a `usize` since no value larger than +/// `std::usize::MAX` can be used in a successful indexing operation in rust. +/// `extent_index` is omitted since it's only used for ConstructionMethod::Item which +/// is currently not implemented. +#[derive(Clone, Debug)] +enum Extent { + WithLength { offset: u64, len: usize }, + ToEnd { offset: u64 }, +} + +#[derive(Debug, PartialEq)] +pub enum TrackType { + Audio, + Video, + Metadata, + Unknown, +} + +impl Default for TrackType { + fn default() -> Self { + TrackType::Unknown + } +} + +// This type is used by mp4parse_capi since it needs to be passed from FFI consumers +// The C-visible struct is renamed via mp4parse_capi/cbindgen.toml to match naming conventions +#[repr(C)] +#[derive(Clone, Copy, Debug, PartialEq)] +pub enum ParseStrictness { + Permissive, // Error only on ambiguous inputs + Normal, // Error on "shall" directives, log warnings for "should" + Strict, // Error on "should" directives +} + +/// Prefer [`fail_with_error_if`] so all the explanatory strings can be collected +/// in `From<Status> for &str`. +fn fail_if(violation: bool, message: &'static str) -> Result<()> { + if violation { + Err(Error::InvalidData(message)) + } else { + warn!("{}", message); + Ok(()) + } +} + +fn fail_with_error_if(violation: bool, error: Error) -> Result<()> { + if violation { + Err(error) + } else { + warn!("{:?}", error); + Ok(()) + } +} + +#[derive(Debug, Clone, Copy, PartialEq)] +pub enum CodecType { + Unknown, + MP3, + AAC, + FLAC, + Opus, + H264, // 14496-10 + MP4V, // 14496-2 + AV1, + VP9, + VP8, + EncryptedVideo, + EncryptedAudio, + LPCM, // QT + ALAC, + H263, + #[cfg(feature = "3gpp")] + AMRNB, + #[cfg(feature = "3gpp")] + AMRWB, +} + +impl Default for CodecType { + fn default() -> Self { + CodecType::Unknown + } +} + +/// The media's global (mvhd) timescale in units per second. +#[derive(Debug, Copy, Clone, PartialEq)] +pub struct MediaTimeScale(pub u64); + +/// A time to be scaled by the media's global (mvhd) timescale. +#[derive(Debug, Copy, Clone, PartialEq)] +pub struct MediaScaledTime(pub u64); + +/// The track's local (mdhd) timescale. +/// Members are timescale units per second and the track id. +#[derive(Debug, Copy, Clone, PartialEq)] +pub struct TrackTimeScale<T: Num>(pub T, pub usize); + +/// A time to be scaled by the track's local (mdhd) timescale. +/// Members are time in scale units and the track id. +#[derive(Debug, Copy, Clone, PartialEq)] +pub struct TrackScaledTime<T>(pub T, pub usize); + +impl<T> std::ops::Add for TrackScaledTime<T> +where + T: num_traits::CheckedAdd, +{ + type Output = Option<Self>; + + fn add(self, other: TrackScaledTime<T>) -> Self::Output { + self.0.checked_add(&other.0).map(|sum| Self(sum, self.1)) + } +} + +#[derive(Debug, Default)] +pub struct Track { + pub id: usize, + pub track_type: TrackType, + pub empty_duration: Option<MediaScaledTime>, + pub media_time: Option<TrackScaledTime<u64>>, + pub timescale: Option<TrackTimeScale<u64>>, + pub duration: Option<TrackScaledTime<u64>>, + pub track_id: Option<u32>, + pub tkhd: Option<TrackHeaderBox>, // TODO(kinetik): find a nicer way to export this. + pub stsd: Option<SampleDescriptionBox>, + pub stts: Option<TimeToSampleBox>, + pub stsc: Option<SampleToChunkBox>, + pub stsz: Option<SampleSizeBox>, + pub stco: Option<ChunkOffsetBox>, // It is for stco or co64. + pub stss: Option<SyncSampleBox>, + pub ctts: Option<CompositionOffsetBox>, +} + +impl Track { + fn new(id: usize) -> Track { + Track { + id, + ..Default::default() + } + } +} + +/// See ISOBMFF (ISO 14496-12:2020) § 4.2 +struct BMFFBox<'a, T: 'a> { + head: BoxHeader, + content: Take<&'a mut T>, +} + +struct BoxIter<'a, T: 'a> { + src: &'a mut T, +} + +impl<'a, T: Read> BoxIter<'a, T> { + fn new(src: &mut T) -> BoxIter<T> { + BoxIter { src } + } + + fn next_box(&mut self) -> Result<Option<BMFFBox<T>>> { + let r = read_box_header(self.src); + match r { + Ok(h) => Ok(Some(BMFFBox { + head: h, + content: self.src.take(h.size - h.offset), + })), + Err(Error::UnexpectedEOF) => Ok(None), + Err(e) => Err(e), + } + } +} + +impl<'a, T: Read> Read for BMFFBox<'a, T> { + fn read(&mut self, buf: &mut [u8]) -> std::io::Result<usize> { + self.content.read(buf) + } +} + +impl<'a, T: Read> TryRead for BMFFBox<'a, T> { + fn try_read_to_end(&mut self, buf: &mut TryVec<u8>) -> std::io::Result<usize> { + fallible_collections::try_read_up_to(self, self.bytes_left(), buf) + } +} + +impl<'a, T: Offset> Offset for BMFFBox<'a, T> { + fn offset(&self) -> u64 { + self.content.get_ref().offset() + } +} + +impl<'a, T: Read> BMFFBox<'a, T> { + fn bytes_left(&self) -> u64 { + self.content.limit() + } + + fn get_header(&self) -> &BoxHeader { + &self.head + } + + fn box_iter<'b>(&'b mut self) -> BoxIter<BMFFBox<'a, T>> { + BoxIter::new(self) + } +} + +impl<'a, T> Drop for BMFFBox<'a, T> { + fn drop(&mut self) { + if self.content.limit() > 0 { + let name: FourCC = From::from(self.head.name); + debug!("Dropping {} bytes in '{}'", self.content.limit(), name); + } + } +} + +/// Read and parse a box header. +/// +/// Call this first to determine the type of a particular mp4 box +/// and its length. Used internally for dispatching to specific +/// parsers for the internal content, or to get the length to +/// skip unknown or uninteresting boxes. +/// +/// See ISOBMFF (ISO 14496-12:2020) § 4.2 +fn read_box_header<T: ReadBytesExt>(src: &mut T) -> Result<BoxHeader> { + let size32 = be_u32(src)?; + let name = BoxType::from(be_u32(src)?); + let size = match size32 { + // valid only for top-level box and indicates it's the last box in the file. usually mdat. + 0 => return Err(Error::Unsupported("unknown sized box")), + 1 => { + let size64 = be_u64(src)?; + if size64 < BoxHeader::MIN_LARGE_SIZE { + return Err(Error::InvalidData("malformed wide size")); + } + size64 + } + _ => { + if u64::from(size32) < BoxHeader::MIN_SIZE { + return Err(Error::InvalidData("malformed size")); + } + u64::from(size32) + } + }; + let mut offset = match size32 { + 1 => BoxHeader::MIN_LARGE_SIZE, + _ => BoxHeader::MIN_SIZE, + }; + let uuid = if name == BoxType::UuidBox { + if size >= offset + 16 { + let mut buffer = [0u8; 16]; + let count = src.read(&mut buffer)?; + offset += count.to_u64(); + if count == 16 { + Some(buffer) + } else { + debug!("malformed uuid (short read), skipping"); + None + } + } else { + debug!("malformed uuid, skipping"); + None + } + } else { + None + }; + assert!(offset <= size); + Ok(BoxHeader { + name, + size, + offset, + uuid, + }) +} + +/// Parse the extra header fields for a full box. +fn read_fullbox_extra<T: ReadBytesExt>(src: &mut T) -> Result<(u8, u32)> { + let version = src.read_u8()?; + let flags_a = src.read_u8()?; + let flags_b = src.read_u8()?; + let flags_c = src.read_u8()?; + Ok(( + version, + u32::from(flags_a) << 16 | u32::from(flags_b) << 8 | u32::from(flags_c), + )) +} + +// Parse the extra fields for a full box whose flag fields must be zero. +fn read_fullbox_version_no_flags<T: ReadBytesExt>(src: &mut T) -> Result<u8> { + let (version, flags) = read_fullbox_extra(src)?; + + if flags != 0 { + return Err(Error::Unsupported("expected flags to be 0")); + } + + Ok(version) +} + +/// Skip over the entire contents of a box. +fn skip_box_content<T: Read>(src: &mut BMFFBox<T>) -> Result<()> { + // Skip the contents of unknown chunks. + let to_skip = { + let header = src.get_header(); + debug!("{:?} (skipped)", header); + header + .size + .checked_sub(header.offset) + .expect("header offset > size") + }; + assert_eq!(to_skip, src.bytes_left()); + skip(src, to_skip) +} + +/// Skip over the remain data of a box. +fn skip_box_remain<T: Read>(src: &mut BMFFBox<T>) -> Result<()> { + let remain = { + let header = src.get_header(); + let len = src.bytes_left(); + debug!("remain {} (skipped) in {:?}", len, header); + len + }; + skip(src, remain) +} + +#[derive(Debug)] +enum AvifImageType { + Primary, + Sequence, + Both, +} + +impl AvifImageType { + fn has_primary(&self) -> bool { + match self { + Self::Primary | Self::Both => true, + Self::Sequence => false, + } + } + + fn has_sequence(&self) -> bool { + match self { + Self::Primary => false, + Self::Sequence | Self::Both => true, + } + } +} + +/// Read the contents of an AVIF file +pub fn read_avif<T: Read>(f: &mut T, strictness: ParseStrictness) -> Result<AvifContext> { + let _ = env_logger::try_init(); + + debug!("read_avif(strictness: {:?})", strictness); + + let mut f = OffsetReader::new(f); + let mut iter = BoxIter::new(&mut f); + let expected_image_type; + let mut unsupported_features = UnsupportedFeatures::new(); + + // 'ftyp' box must occur first; see ISOBMFF (ISO 14496-12:2020) § 4.3.1 + let major_brand = if let Some(mut b) = iter.next_box()? { + if b.head.name == BoxType::FileTypeBox { + let ftyp = read_ftyp(&mut b)?; + + let has_avif_brand = ftyp.contains(&AVIF_BRAND); + let has_avis_brand = ftyp.contains(&AVIS_BRAND); + let has_mif1_brand = ftyp.contains(&MIF1_BRAND); + let has_msf1_brand = ftyp.contains(&MSF1_BRAND); + + let primary_image_expected = has_mif1_brand || has_avif_brand; + let image_sequence_expected = has_msf1_brand || has_avis_brand; + + expected_image_type = if primary_image_expected && image_sequence_expected { + AvifImageType::Both + } else if primary_image_expected { + AvifImageType::Primary + } else if image_sequence_expected { + AvifImageType::Sequence + } else { + return Status::NoImage.into(); + }; + debug!("expected_image_type: {:?}", expected_image_type); + + if primary_image_expected && !has_mif1_brand { + // This mandatory inclusion of this brand is in the process of being changed + // to optional. In anticipation of that, only give an error in strict mode + // See https://github.com/MPEGGroup/MIAF/issues/5 + // and https://github.com/MPEGGroup/FileFormat/issues/23 + fail_if( + strictness == ParseStrictness::Strict, + "The FileTypeBox should contain 'mif1' in the compatible_brands list \ + per MIAF (ISO 23000-22:2019) § 7.2.1.2", + )?; + } + + if !has_avif_brand && !has_avis_brand { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::MissingBrand.into(), + )?; + } + + ftyp.major_brand + } else { + return Status::FtypNotFirst.into(); + } + } else { + return Status::FtypNotFirst.into(); + }; + + if major_brand == AVIS_BRAND { + unsupported_features.insert(Feature::Avis); + } + + let mut meta = None; + let mut image_sequence = None; + let mut media_storage = TryVec::new(); + + while let Some(mut b) = iter.next_box()? { + trace!("read_avif parsing {:?} box", b.head.name); + match b.head.name { + BoxType::MetadataBox => { + if meta.is_some() { + return Err(Error::InvalidData( + "There should be zero or one meta boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.1.1", + )); + } + meta = Some(read_avif_meta( + &mut b, + strictness, + &mut unsupported_features, + )?); + } + BoxType::MovieBox if expected_image_type.has_sequence() => { + if image_sequence.is_some() { + return Status::MultipleMoov.into(); + } + image_sequence = Some(read_moov(&mut b, None)?); + } + BoxType::MediaDataBox => { + if b.bytes_left() > 0 { + let file_offset = b.offset(); + let data = b.read_into_try_vec()?; + media_storage.push(MediaDataBox { file_offset, data })?; + } + } + _ => skip_box_content(&mut b)?, + } + + check_parser_state!(b.content); + } + + let AvifMeta { + item_references, + item_properties, + primary_item_id, + item_infos, + iloc_items, + item_data_box, + } = meta.ok_or(Error::InvalidData("missing meta"))?; + + let (alpha_item_id, premultiplied_alpha) = if let Some(primary_item_id) = primary_item_id { + let mut alpha_item_ids = item_references + .iter() + // Auxiliary image for the primary image + .filter(|iref| { + iref.to_item_id == primary_item_id + && iref.from_item_id != primary_item_id + && iref.item_type == b"auxl" + }) + .map(|iref| iref.from_item_id) + // which has the alpha property + .filter(|&item_id| item_properties.is_alpha(item_id)); + let alpha_item_id = alpha_item_ids.next(); + if alpha_item_ids.next().is_some() { + return Err(Error::InvalidData("multiple alpha planes")); + } + + let premultiplied_alpha = alpha_item_id.map_or(false, |alpha_item_id| { + item_references.iter().any(|iref| { + iref.from_item_id == primary_item_id + && iref.to_item_id == alpha_item_id + && iref.item_type == b"prem" + }) + }); + + (alpha_item_id, premultiplied_alpha) + } else { + (None, false) + }; + + debug!("primary_item_id: {:?}", primary_item_id); + debug!("alpha_item_id: {:?}", alpha_item_id); + let mut primary_item = None; + let mut alpha_item = None; + + // store data or record location of relevant items + for (item_id, loc) in iloc_items { + let item = if Some(item_id) == primary_item_id { + &mut primary_item + } else if Some(item_id) == alpha_item_id { + &mut alpha_item + } else { + continue; + }; + + assert!(item.is_none()); + + // If our item is spread over multiple extents, we'll need to copy it + // into a contiguous buffer. Otherwise, we can just store the extent + // and return a pointer into the mdat/idat later to avoid the copy. + if loc.extents.len() > 1 { + *item = Some(AvifItem::with_inline_data(item_id)) + } + + trace!( + "{:?} construction_method: {:?}", + item_id, + loc.construction_method + ); + + // Generalize the process of connecting items to their data; returns + // true if the extent is successfully added to the AvifItem + let mut find_and_add_to_item = |extent: &Extent, dat: &dyn DataBox| -> Result<bool> { + if let Some(extent_slice) = dat.get(extent) { + match item { + None => { + trace!("Using IsobmffItem::Location"); + *item = Some(AvifItem { + id: item_id, + image_data: dat.location(extent), + }); + } + Some(AvifItem { + image_data: IsobmffItem::Data(bytes), + .. + }) => { + trace!("Using IsobmffItem::Data"); + // We could potentially optimize memory usage by trying to avoid reading + // or storing dat boxes which aren't used by our API, but for now it seems + // like unnecessary complexity + bytes.extend_from_slice(extent_slice)?; + } + _ => unreachable!(), + } + return Ok(true); + } + Ok(false) + }; + + match loc.construction_method { + ConstructionMethod::File => { + for extent in loc.extents { + let mut found = false; + // try to find an mdat which contains the extent + for mdat in media_storage.iter() { + if find_and_add_to_item(&extent, mdat)? { + found = true; + break; + } + } + + if !found { + return Status::ItemLocNotFound.into(); + } + } + } + ConstructionMethod::Idat => { + if let Some(idat) = &item_data_box { + for extent in loc.extents { + let found = find_and_add_to_item(&extent, idat)?; + if !found { + return Status::ItemLocNotFound.into(); + } + } + } else { + return Status::NoItemDataBox.into(); + } + } + ConstructionMethod::Item => { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::ConstructionMethod.into(), + )?; + } + } + + assert!(item.is_some()); + } + + assert!(primary_item.is_none() || primary_item_id.is_some()); + assert!(alpha_item.is_none() || alpha_item_id.is_some()); + + if expected_image_type.has_primary() && primary_item_id.is_none() { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::NoPrimaryItem.into(), + )?; + } + + // Lacking a brand that requires them, it's fine for moov boxes to exist in + // BMFF files; they're simply ignored + if expected_image_type.has_sequence() && image_sequence.is_none() { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::NoMoov.into(), + )?; + } + + // Returns true iff `id` is `Some` and there is no corresponding property for it + let missing_property_for = |id: Option<ItemId>, property: BoxType| -> bool { + id.map_or(false, |id| { + item_properties + .get(id, property) + .map_or(true, |opt| opt.is_none()) + }) + }; + + // TODO: add ispe check for alpha https://github.com/mozilla/mp4parse-rust/issues/353 + if missing_property_for(primary_item_id, BoxType::ImageSpatialExtentsProperty) { + fail_if( + strictness != ParseStrictness::Permissive, + "Missing 'ispe' property for primary item, required \ + per HEIF (ISO/IEC 23008-12:2017) § 6.5.3.1", + )?; + } + + // Generalize the property checks so we can apply them to primary and alpha items + let mut check_image_item = |item: &mut Option<AvifItem>| -> Result<()> { + let item_id = item.as_ref().map(|item| item.id); + let item_type = item_id.and_then(|item_id| { + item_infos + .iter() + .find(|item_info| item_id == item_info.item_id) + .map(|item_info| item_info.item_type) + }); + + match item_type.map(u32::to_be_bytes).as_ref() { + Some(b"av01") => { + if missing_property_for(item_id, BoxType::AV1CodecConfigurationBox) { + fail_if( + strictness != ParseStrictness::Permissive, + "One AV1 Item Configuration Property (av1C) is mandatory for an \ + image item of type 'av01' \ + per AVIF specification § 2.2.1", + )?; + } + + if missing_property_for(item_id, BoxType::PixelInformationBox) { + // The requirement to include pixi is in the process of being changed + // to allowing its omission to imply a default value. In anticipation + // of that, only give an error in strict mode + // See https://github.com/MPEGGroup/MIAF/issues/9 + fail_if( + if cfg!(feature = "missing-pixi-permitted") { + strictness == ParseStrictness::Strict + } else { + strictness != ParseStrictness::Permissive + }, + "The pixel information property shall be associated with every image \ + that is displayable (not hidden) \ + per MIAF (ISO/IEC 23000-22:2019) specification § 7.3.6.6", + )?; + } + } + Some(b"grid") => { + // TODO: https://github.com/mozilla/mp4parse-rust/issues/198 + unsupported_features.insert(Feature::Grid); + *item = None; + } + Some(_other_type) => return Status::ImageItemType.into(), + None => { + if item.is_some() { + return Status::ItemTypeMissing.into(); + } + } + } + + if let Some(AvifItem { id, .. }) = item { + if item_properties.forbidden_items.contains(id) { + error!("Not processing item id {:?} since it is associated with essential, but unsupported properties", id); + *item = None; + } + } + + Ok(()) + }; + + check_image_item(&mut primary_item)?; + check_image_item(&mut alpha_item)?; + + Ok(AvifContext { + strictness, + media_storage, + primary_item, + alpha_item, + premultiplied_alpha, + item_properties, + major_brand, + has_sequence: image_sequence.is_some(), + unsupported_features, + }) +} + +/// Parse a metadata box in the context of an AVIF +/// Currently requires the primary item to be an av01 item type and generates +/// an error otherwise. +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.1 +fn read_avif_meta<T: Read + Offset>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, + unsupported_features: &mut UnsupportedFeatures, +) -> Result<AvifMeta> { + let version = read_fullbox_version_no_flags(src)?; + + if version != 0 { + return Err(Error::Unsupported("unsupported meta version")); + } + + let mut read_handler_box = false; + let mut primary_item_id = None; + let mut item_infos = None; + let mut iloc_items = None; + let mut item_references = None; + let mut item_properties = None; + let mut item_data_box = None; + + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + trace!("read_avif_meta parsing {:?} box", b.head.name); + + if !read_handler_box && b.head.name != BoxType::HandlerBox { + fail_if( + strictness != ParseStrictness::Permissive, + "The HandlerBox shall be the first contained box within the MetaBox \ + per MIAF (ISO 23000-22:2019) § 7.2.1.5", + )?; + } + + match b.head.name { + BoxType::HandlerBox => { + if read_handler_box { + return Err(Error::InvalidData( + "There shall be exactly one hdlr box per ISOBMFF (ISO 14496-12:2020) § 8.4.3.1", + )); + } + let HandlerBox { handler_type } = read_hdlr(&mut b, strictness)?; + if handler_type != b"pict" { + fail_if( + strictness != ParseStrictness::Permissive, + "The HandlerBox handler_type must be 'pict' \ + per MIAF (ISO 23000-22:2019) § 7.2.1.5", + )?; + } + read_handler_box = true; + } + BoxType::ItemInfoBox => { + if item_infos.is_some() { + return Err(Error::InvalidData( + "There shall be zero or one iinf boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.6.1", + )); + } + item_infos = Some(read_iinf(&mut b, strictness, unsupported_features)?); + } + BoxType::ItemLocationBox => { + if iloc_items.is_some() { + return Err(Error::InvalidData( + "There shall be zero or one iloc boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.3.1", + )); + } + iloc_items = Some(read_iloc(&mut b)?); + } + BoxType::PrimaryItemBox => { + if primary_item_id.is_some() { + return Err(Error::InvalidData( + "There shall be zero or one pitm boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.4.1", + )); + } + primary_item_id = Some(read_pitm(&mut b)?); + } + BoxType::ItemReferenceBox => { + if item_references.is_some() { + return Err(Error::InvalidData("There shall be zero or one iref boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.12.1")); + } + item_references = Some(read_iref(&mut b)?); + } + BoxType::ItemPropertiesBox => { + if item_properties.is_some() { + return Err(Error::InvalidData("There shall be zero or one iprp boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.14.1")); + } + item_properties = Some(read_iprp( + &mut b, + MIF1_BRAND, + strictness, + unsupported_features, + )?); + } + BoxType::ItemDataBox => { + if item_data_box.is_some() { + return Err(Error::InvalidData( + "There shall be zero or one idat boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.11", + )); + } + item_data_box = Some(ItemDataBox { + data: b.read_into_try_vec()?, + }); + } + _ => skip_box_content(&mut b)?, + } + + check_parser_state!(b.content); + } + + Ok(AvifMeta { + item_properties: item_properties.unwrap_or_default(), + item_references: item_references.unwrap_or_default(), + primary_item_id, + item_infos: item_infos.unwrap_or_default(), + iloc_items: iloc_items.ok_or(Error::InvalidData("iloc missing"))?, + item_data_box, + }) +} + +/// Parse a Primary Item Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.4 +fn read_pitm<T: Read>(src: &mut BMFFBox<T>) -> Result<ItemId> { + let version = read_fullbox_version_no_flags(src)?; + + let item_id = ItemId(match version { + 0 => be_u16(src)?.into(), + 1 => be_u32(src)?, + _ => return Err(Error::Unsupported("unsupported pitm version")), + }); + + Ok(item_id) +} + +/// Parse an Item Information Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.6 +fn read_iinf<T: Read>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, + unsupported_features: &mut UnsupportedFeatures, +) -> Result<TryVec<ItemInfoEntry>> { + let version = read_fullbox_version_no_flags(src)?; + + match version { + 0 | 1 => (), + _ => return Err(Error::Unsupported("unsupported iinf version")), + } + + let entry_count = if version == 0 { + be_u16(src)?.to_usize() + } else { + be_u32(src)?.to_usize() + }; + let mut item_infos = TryVec::with_capacity(entry_count)?; + + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + if b.head.name != BoxType::ItemInfoEntry { + return Err(Error::InvalidData( + "iinf box shall contain only infe boxes per ISOBMFF (ISO 14496-12:2020) § 8.11.6.2", + )); + } + + if let Some(infe) = read_infe(&mut b, strictness, unsupported_features)? { + item_infos.push(infe)?; + } + + check_parser_state!(b.content); + } + + Ok(item_infos) +} + +/// A simple wrapper to interpret a u32 as a 4-byte string in big-endian +/// order without requiring any allocation. +struct U32BE(u32); + +impl std::fmt::Display for U32BE { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + match std::str::from_utf8(&self.0.to_be_bytes()) { + Ok(s) => f.write_str(s), + Err(_) => write!(f, "{:x?}", self.0), + } + } +} + +/// Parse an Item Info Entry +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.6.2 +fn read_infe<T: Read>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, + unsupported_features: &mut UnsupportedFeatures, +) -> Result<Option<ItemInfoEntry>> { + let (version, flags) = read_fullbox_extra(src)?; + + // According to the standard, it seems the flags field shall be 0, but at + // least one sample AVIF image has a nonzero value. + // See https://github.com/AOMediaCodec/av1-avif/issues/146 + if flags != 0 { + fail_if( + strictness == ParseStrictness::Strict, + "'infe' flags field shall be 0 \ + per ISOBMFF (ISO 14496-12:2020) § 8.11.6.2", + )?; + } + + // mif1 brand (see HEIF (ISO 23008-12:2017) § 10.2.1) only requires v2 and 3 + let item_id = ItemId(match version { + 2 => be_u16(src)?.into(), + 3 => be_u32(src)?, + _ => return Err(Error::Unsupported("unsupported version in 'infe' box")), + }); + + let item_protection_index = be_u16(src)?; + + let item_type = be_u32(src)?; + debug!("infe {:?} item_type: {}", item_id, U32BE(item_type)); + + // There are some additional fields here, but they're not of interest to us + skip_box_remain(src)?; + + if item_protection_index != 0 { + unsupported_features.insert(Feature::Ipro); + Ok(None) + } else { + Ok(Some(ItemInfoEntry { item_id, item_type })) + } +} + +/// Parse an Item Reference Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.12 +fn read_iref<T: Read>(src: &mut BMFFBox<T>) -> Result<TryVec<SingleItemTypeReferenceBox>> { + let mut item_references = TryVec::new(); + let version = read_fullbox_version_no_flags(src)?; + if version > 1 { + return Err(Error::Unsupported("iref version")); + } + + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + trace!("read_iref parsing {:?} referenceType", b.head.name); + let from_item_id = ItemId::read(&mut b, version)?; + let reference_count = be_u16(&mut b)?; + item_references.reserve(reference_count.to_usize())?; + for _ in 0..reference_count { + let to_item_id = ItemId::read(&mut b, version)?; + if from_item_id == to_item_id { + return Err(Error::InvalidData( + "from_item_id and to_item_id must be different", + )); + } + item_references.push(SingleItemTypeReferenceBox { + item_type: b.head.name.into(), + from_item_id, + to_item_id, + })?; + } + check_parser_state!(b.content); + } + + trace!("read_iref -> {:#?}", item_references); + + Ok(item_references) +} + +/// Parse an Item Properties Box +/// +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14) +/// +/// Note: HEIF (ISO 23008-12:2017) § 9.3.1 also defines the `iprp` box and +/// related types, but lacks additional requirements specified in 14496-12:2020. +/// +/// Note: Currently HEIF (ISO 23008-12:2017) § 6.5.5.1 specifies "At most one" +/// `colr` box per item, but this is being amended in [DIS 23008-12](https://www.iso.org/standard/83650.html). +/// The new text is likely to be "At most one for a given value of `colour_type`", +/// so this implementation adheres to that language for forward compatibility. +fn read_iprp<T: Read>( + src: &mut BMFFBox<T>, + brand: FourCC, + strictness: ParseStrictness, + unsupported_features: &mut UnsupportedFeatures, +) -> Result<ItemPropertiesBox> { + let mut iter = src.box_iter(); + + let properties = match iter.next_box()? { + Some(mut b) if b.head.name == BoxType::ItemPropertyContainerBox => { + read_ipco(&mut b, strictness) + } + Some(_) => Err(Error::InvalidData("unexpected iprp child")), + None => Err(Error::UnexpectedEOF), + }?; + + let mut ipma_version_and_flag_values_seen = TryVec::with_capacity(1)?; + let mut association_entries = TryVec::<ItemPropertyAssociationEntry>::new(); + let mut forbidden_items = TryVec::new(); + + while let Some(mut b) = iter.next_box()? { + if b.head.name != BoxType::ItemPropertyAssociationBox { + return Err(Error::InvalidData("unexpected iprp child")); + } + + let (version, flags) = read_fullbox_extra(&mut b)?; + if ipma_version_and_flag_values_seen.contains(&(version, flags)) { + fail_if( + strictness != ParseStrictness::Permissive, + "There shall be at most one ItemPropertyAssociationbox with a given pair of \ + values of version and flags \ + per ISOBMFF (ISO 14496-12:2020 § 8.11.14.1", + )?; + } + if flags != 0 && properties.len() <= 127 { + fail_if( + strictness == ParseStrictness::Strict, + "Unless there are more than 127 properties in the ItemPropertyContainerBox, \ + flags should be equal to 0 \ + per ISOBMFF (ISO 14496-12:2020 § 8.11.14.1", + )?; + } + ipma_version_and_flag_values_seen.push((version, flags))?; + for association_entry in read_ipma(&mut b, strictness, version, flags)? { + if forbidden_items.contains(&association_entry.item_id) { + warn!( + "Skipping {:?} since the item referenced shall not be processed", + association_entry + ); + } + + if let Some(previous_entry) = association_entries + .iter() + .find(|e| association_entry.item_id == e.item_id) + { + error!( + "Duplicate ipma entries for item_id\n1: {:?}\n2: {:?}", + previous_entry, association_entry + ); + // It's technically possible to make sense of this situation by merging ipma + // boxes, but this is a "shall" requirement, so we'd only do it in + // ParseStrictness::Permissive mode, and this hasn't shown up in the wild + return Err(Error::InvalidData( + "There shall be at most one occurrence of a given item_ID, \ + in the set of ItemPropertyAssociationBox boxes \ + per ISOBMFF (ISO 14496-12:2020) § 8.11.14.1", + )); + } + + const TRANSFORM_ORDER_ERROR: &str = + "These properties, if used, shall be indicated to be applied \ + in the following order: clean aperture first, then rotation, \ + then mirror. \ + per MIAF (ISO/IEC 23000-22:2019) § 7.3.6.7"; + const TRANSFORM_ORDER: &[BoxType] = &[ + BoxType::CleanApertureBox, + BoxType::ImageRotation, + BoxType::ImageMirror, + ]; + let mut prev_transform_index = None; + // Realistically, there should only ever be 1 nclx and 1 icc + let mut colour_type_indexes: TryHashMap<FourCC, PropertyIndex> = + TryHashMap::with_capacity(2)?; + + for a in &association_entry.associations { + if a.property_index == PropertyIndex(0) { + if a.essential { + fail_if( + strictness != ParseStrictness::Permissive, + "the essential indicator shall be 0 for property index 0 \ + per ISOBMFF (ISO 14496-12:2020 § 8.11.14.3", + )?; + } + continue; + } + + if let Some(property) = properties.get(&a.property_index) { + assert!(brand == MIF1_BRAND); + + let feature = Feature::try_from(property); + let property_supported = match feature { + Ok(feature) => { + if feature.supported() { + true + } else { + unsupported_features.insert(feature); + false + } + } + Err(_) => false, + }; + + if !property_supported { + if a.essential && strictness != ParseStrictness::Permissive { + error!("Unsupported essential property {:?}", property); + forbidden_items.push(association_entry.item_id)?; + } else { + debug!( + "Ignoring unknown {} property {:?}", + if a.essential { + "essential" + } else { + "non-essential" + }, + property + ); + } + } + + // Check additional requirements on specific properties + match property { + ItemProperty::AV1Config(_) + | ItemProperty::CleanAperture + | ItemProperty::Mirroring(_) + | ItemProperty::Rotation(_) => { + if !a.essential { + warn!("{:?} is missing required 'essential' bit", property); + // This is a "shall", but it is likely to change, so only + // fail if using strict parsing. + // See https://github.com/mozilla/mp4parse-rust/issues/284 + fail_with_error_if( + strictness == ParseStrictness::Strict, + Status::TxformNoEssential.into(), + )?; + } + } + + // NOTE: this is contrary to the published specification; see doc comment + // at the beginning of this function for more details + ItemProperty::Colour(colr) => { + let colour_type = colr.colour_type(); + if let Some(prev_colr_index) = colour_type_indexes.get(&colour_type) { + warn!( + "Multiple '{}' type colr associations with {:?}: {:?} and {:?}", + colour_type, + association_entry.item_id, + a.property_index, + prev_colr_index + ); + fail_if( + strictness != ParseStrictness::Permissive, + "Each item shall have at most one property association with a + ColourInformationBox (colr) for a given value of colour_type \ + per HEIF (ISO/IEC DIS 23008-12) § 6.5.5.1", + )?; + } else { + colour_type_indexes.insert(colour_type, a.property_index)?; + } + } + + // The following properties are unsupported, but we still enforce that + // they've been correctly marked as essential or not. + ItemProperty::LayeredImageIndexing => { + assert!(feature.is_ok() && unsupported_features.contains(feature?)); + if a.essential { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::A1lxEssential.into(), + )?; + } + } + + ItemProperty::LayerSelection => { + assert!(feature.is_ok() && unsupported_features.contains(feature?)); + if a.essential { + assert!( + forbidden_items.contains(&association_entry.item_id) + || strictness == ParseStrictness::Permissive + ); + } else { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::LselNoEssential.into(), + )?; + } + } + + ItemProperty::OperatingPointSelector => { + assert!(feature.is_ok() && unsupported_features.contains(feature?)); + if a.essential { + assert!( + forbidden_items.contains(&association_entry.item_id) + || strictness == ParseStrictness::Permissive + ); + } else { + fail_with_error_if( + strictness != ParseStrictness::Permissive, + Status::A1opNoEssential.into(), + )?; + } + } + + other_property => { + trace!("No additional checks for {:?}", other_property); + } + } + + if let Some(transform_index) = TRANSFORM_ORDER + .iter() + .position(|t| *t == BoxType::from(property)) + { + if let Some(prev) = prev_transform_index { + if prev >= transform_index { + error!( + "{:?} after {:?}", + TRANSFORM_ORDER[transform_index], TRANSFORM_ORDER[prev] + ); + return Err(Error::InvalidData(TRANSFORM_ORDER_ERROR)); + } + } + prev_transform_index = Some(transform_index); + } + } else { + error!( + "Missing property at {:?} for {:?}", + a.property_index, association_entry.item_id + ); + fail_if( + strictness != ParseStrictness::Permissive, + "Invalid property index in ipma", + )?; + } + } + association_entries.push(association_entry)? + } + + check_parser_state!(b.content); + } + + let iprp = ItemPropertiesBox { + properties, + association_entries, + forbidden_items, + }; + trace!("read_iprp -> {:#?}", iprp); + Ok(iprp) +} + +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +/// Variants with no associated data are recognized but not necessarily supported. +/// See [`Feature`] to determine support. +#[derive(Debug)] +pub enum ItemProperty { + AuxiliaryType(AuxiliaryTypeProperty), + AV1Config(AV1ConfigBox), + Channels(PixelInformation), + CleanAperture, + Colour(ColourInformation), + ImageSpatialExtents(ImageSpatialExtentsProperty), + LayeredImageIndexing, + LayerSelection, + Mirroring(ImageMirror), + OperatingPointSelector, + PixelAspectRatio(PixelAspectRatio), + Rotation(ImageRotation), + /// Necessary to validate property indices in read_iprp + Unsupported(BoxType), +} + +impl From<&ItemProperty> for BoxType { + fn from(item_property: &ItemProperty) -> Self { + match item_property { + ItemProperty::AuxiliaryType(_) => BoxType::AuxiliaryTypeProperty, + ItemProperty::AV1Config(_) => BoxType::AV1CodecConfigurationBox, + ItemProperty::CleanAperture => BoxType::CleanApertureBox, + ItemProperty::Colour(_) => BoxType::ColourInformationBox, + ItemProperty::LayeredImageIndexing => BoxType::AV1LayeredImageIndexingProperty, + ItemProperty::LayerSelection => BoxType::LayerSelectorProperty, + ItemProperty::Mirroring(_) => BoxType::ImageMirror, + ItemProperty::OperatingPointSelector => BoxType::OperatingPointSelectorProperty, + ItemProperty::PixelAspectRatio(_) => BoxType::PixelAspectRatioBox, + ItemProperty::Rotation(_) => BoxType::ImageRotation, + ItemProperty::ImageSpatialExtents(_) => BoxType::ImageSpatialExtentsProperty, + ItemProperty::Channels(_) => BoxType::PixelInformationBox, + ItemProperty::Unsupported(box_type) => *box_type, + } + } +} + +#[derive(Debug)] +struct ItemPropertyAssociationEntry { + item_id: ItemId, + associations: TryVec<Association>, +} + +/// For storing ItemPropertyAssociation data +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +#[derive(Debug)] +struct Association { + essential: bool, + property_index: PropertyIndex, +} + +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +/// +/// The properties themselves are stored in `properties`, but the items they're +/// associated with are stored in `association_entries`. It's necessary to +/// maintain this indirection because multiple items can reference the same +/// property. For example, both the primary item and alpha item can share the +/// same [`ImageSpatialExtentsProperty`]. +#[derive(Debug, Default)] +pub struct ItemPropertiesBox { + /// `ItemPropertyContainerBox property_container` in the spec + properties: TryHashMap<PropertyIndex, ItemProperty>, + /// `ItemPropertyAssociationBox association[]` in the spec + association_entries: TryVec<ItemPropertyAssociationEntry>, + /// Items that shall not be processed due to unsupported properties that + /// have been marked essential. + /// See HEIF (ISO/IEC 23008-12:2017) § 9.3.1 + forbidden_items: TryVec<ItemId>, +} + +impl ItemPropertiesBox { + /// For displayable images `av1C`, `pixi` and `ispe` are mandatory, `colr` + /// is typically included too, so we might as well use an even power of 2. + const MIN_PROPERTIES: usize = 4; + + fn is_alpha(&self, item_id: ItemId) -> bool { + match self.get(item_id, BoxType::AuxiliaryTypeProperty) { + Ok(Some(ItemProperty::AuxiliaryType(urn))) => { + urn.aux_type.as_slice() == "urn:mpeg:mpegB:cicp:systems:auxiliary:alpha".as_bytes() + } + Ok(Some(other_property)) => panic!("property key mismatch: {:?}", other_property), + Ok(None) => false, + Err(e) => { + error!( + "is_alpha: Error checking AuxiliaryTypeProperty ({}), returning false", + e + ); + false + } + } + } + + fn get(&self, item_id: ItemId, property_type: BoxType) -> Result<Option<&ItemProperty>> { + match self + .get_multiple(item_id, |prop| BoxType::from(prop) == property_type)? + .as_slice() + { + &[] => Ok(None), + &[single_value] => Ok(Some(single_value)), + multiple_values => { + error!( + "Multiple values for {:?}: {:?}", + property_type, multiple_values + ); + // TODO: add test + Err(Error::InvalidData("conflicting item property values")) + } + } + } + + fn get_multiple( + &self, + item_id: ItemId, + filter: impl Fn(&ItemProperty) -> bool, + ) -> Result<TryVec<&ItemProperty>> { + let mut values = TryVec::new(); + for entry in &self.association_entries { + for a in &entry.associations { + if entry.item_id == item_id { + match self.properties.get(&a.property_index) { + Some(ItemProperty::Unsupported(_)) => {} + Some(property) if filter(property) => values.push(property)?, + _ => {} + } + } + } + } + + Ok(values) + } +} + +/// An upper bound which can be used to check overflow at compile time +trait UpperBounded { + const MAX: u64; +} + +/// Implement type $name as a newtype wrapper around an unsigned int which +/// implements the UpperBounded trait. +macro_rules! impl_bounded { + ( $name:ident, $inner:ty ) => { + #[derive(Clone, Copy)] + pub struct $name($inner); + + impl $name { + pub const fn new(n: $inner) -> Self { + Self(n) + } + + #[allow(dead_code)] + pub fn get(self) -> $inner { + self.0 + } + } + + impl UpperBounded for $name { + const MAX: u64 = <$inner>::MAX as u64; + } + }; +} + +/// Implement type $name as a type representing the product of two unsigned ints +/// which implements the UpperBounded trait. +macro_rules! impl_bounded_product { + ( $name:ident, $multiplier:ty, $multiplicand:ty, $inner:ty) => { + #[derive(Clone, Copy)] + pub struct $name($inner); + + impl $name { + pub fn new(value: $inner) -> Self { + assert!(value <= Self::MAX); + Self(value) + } + + pub fn get(self) -> $inner { + self.0 + } + } + + impl UpperBounded for $name { + const MAX: u64 = <$multiplier>::MAX * <$multiplicand>::MAX; + } + }; +} + +mod bounded_uints { + use UpperBounded; + + impl_bounded!(U8, u8); + impl_bounded!(U16, u16); + impl_bounded!(U32, u32); + impl_bounded!(U64, u64); + + impl_bounded_product!(U32MulU8, U32, U8, u64); + impl_bounded_product!(U32MulU16, U32, U16, u64); + + impl UpperBounded for std::num::NonZeroU8 { + const MAX: u64 = u8::MAX as u64; + } +} + +use bounded_uints::*; + +/// Implement the multiplication operator for $lhs * $rhs giving $output, which +/// is internally represented as $inner. The operation is statically checked +/// to ensure the product won't overflow $inner, nor exceed <$output>::MAX. +macro_rules! impl_mul { + ( ($lhs:ty , $rhs:ty) => ($output:ty, $inner:ty) ) => { + impl std::ops::Mul<$rhs> for $lhs { + type Output = $output; + + fn mul(self, rhs: $rhs) -> Self::Output { + static_assertions::const_assert!(<$output>::MAX <= <$inner>::MAX as u64); + static_assertions::const_assert!(<$lhs>::MAX * <$rhs>::MAX <= <$output>::MAX); + + let lhs: $inner = self.get().into(); + let rhs: $inner = rhs.get().into(); + Self::Output::new(lhs.checked_mul(rhs).expect("infallible")) + } + } + }; +} + +impl_mul!((U8, std::num::NonZeroU8) => (U16, u16)); +impl_mul!((U32, std::num::NonZeroU8) => (U32MulU8, u64)); +impl_mul!((U32, U16) => (U32MulU16, u64)); + +impl std::ops::Add<U32MulU16> for U32MulU8 { + type Output = U64; + + fn add(self, rhs: U32MulU16) -> Self::Output { + static_assertions::const_assert!(U32MulU8::MAX + U32MulU16::MAX < U64::MAX); + let lhs: u64 = self.get(); + let rhs: u64 = rhs.get(); + Self::Output::new(lhs.checked_add(rhs).expect("infallible")) + } +} + +const MAX_IPMA_ASSOCIATION_COUNT: U8 = U8::new(u8::MAX); + +/// After reading only the `entry_count` field of an ipma box, we can check its +/// basic validity and calculate (assuming validity) the number of associations +/// which will be contained (allowing preallocation of the storage). +/// All the arithmetic is compile-time verified to not overflow via supporting +/// types implementing the UpperBounded trait. Types are declared explicitly to +/// show there isn't any accidental inference to primitive types. +/// +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +fn calculate_ipma_total_associations( + version: u8, + bytes_left: u64, + entry_count: U32, + num_association_bytes: std::num::NonZeroU8, +) -> Result<usize> { + let min_entry_bytes = + std::num::NonZeroU8::new(1 /* association_count */ + if version == 0 { 2 } else { 4 }) + .unwrap(); + + let total_non_association_bytes: U32MulU8 = entry_count * min_entry_bytes; + let total_association_bytes: u64 = + if let Some(difference) = bytes_left.checked_sub(total_non_association_bytes.get()) { + // All the storage for the `essential` and `property_index` parts (assuming a valid ipma box size) + difference + } else { + return Err(Error::InvalidData( + "ipma box below minimum size for entry_count", + )); + }; + + let max_association_bytes_per_entry: U16 = MAX_IPMA_ASSOCIATION_COUNT * num_association_bytes; + let max_total_association_bytes: U32MulU16 = entry_count * max_association_bytes_per_entry; + let max_bytes_left: U64 = total_non_association_bytes + max_total_association_bytes; + + if bytes_left > max_bytes_left.get() { + return Err(Error::InvalidData( + "ipma box exceeds maximum size for entry_count", + )); + } + + let total_associations: u64 = total_association_bytes / u64::from(num_association_bytes.get()); + + Ok(total_associations.try_into()?) +} + +/// Parse an ItemPropertyAssociation box +/// +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +fn read_ipma<T: Read>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, + version: u8, + flags: u32, +) -> Result<TryVec<ItemPropertyAssociationEntry>> { + let entry_count = be_u32(src)?; + let num_association_bytes = + std::num::NonZeroU8::new(if flags & 1 == 1 { 2 } else { 1 }).unwrap(); + + let total_associations = calculate_ipma_total_associations( + version, + src.bytes_left(), + U32::new(entry_count), + num_association_bytes, + )?; + // Assuming most items will have at least `MIN_PROPERTIES` and knowing the + // total number of item -> property associations (`total_associations`), + // we can provide a good estimate for how many elements we'll need in this + // vector, even though we don't know precisely how many items there will be + // properties for. + let mut entries = TryVec::<ItemPropertyAssociationEntry>::with_capacity( + total_associations / ItemPropertiesBox::MIN_PROPERTIES, + )?; + + for _ in 0..entry_count { + let item_id = ItemId::read(src, version)?; + + if let Some(previous_association) = entries.last() { + #[allow(clippy::comparison_chain)] + if previous_association.item_id > item_id { + return Err(Error::InvalidData( + "Each ItemPropertyAssociation box shall be ordered by increasing item_ID", + )); + } else if previous_association.item_id == item_id { + return Err(Error::InvalidData("There shall be at most one association box for each item_ID, in any ItemPropertyAssociation box")); + } + } + + let association_count = src.read_u8()?; + let mut associations = TryVec::with_capacity(association_count.to_usize())?; + for _ in 0..association_count { + let association = src + .take(num_association_bytes.get().into()) + .read_into_try_vec()?; + let mut association = BitReader::new(association.as_slice()); + let essential = association.read_bool()?; + let property_index = + PropertyIndex(association.read_u16(association.remaining().try_into()?)?); + associations.push(Association { + essential, + property_index, + })?; + } + + entries.push(ItemPropertyAssociationEntry { + item_id, + associations, + })?; + } + + check_parser_state!(src.content); + + if version != 0 { + if let Some(ItemPropertyAssociationEntry { + item_id: max_item_id, + .. + }) = entries.last() + { + if *max_item_id <= ItemId(u16::MAX.into()) { + fail_if( + strictness == ParseStrictness::Strict, + "The ipma version 0 should be used unless 32-bit item_ID values are needed \ + per ISOBMFF (ISO 14496-12:2020 § 8.11.14.1", + )?; + } + } + } + + trace!("read_ipma -> {:#?}", entries); + + Ok(entries) +} + +/// Parse an ItemPropertyContainerBox +/// +/// See ISOBMFF (ISO 14496-12:2020 § 8.11.14.1 +fn read_ipco<T: Read>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, +) -> Result<TryHashMap<PropertyIndex, ItemProperty>> { + let mut properties = TryHashMap::with_capacity(ItemPropertiesBox::MIN_PROPERTIES)?; + + let mut index = PropertyIndex(1); // ipma uses 1-based indexing + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + let property = match b.head.name { + BoxType::AuxiliaryTypeProperty => ItemProperty::AuxiliaryType(read_auxc(&mut b)?), + BoxType::AV1CodecConfigurationBox => ItemProperty::AV1Config(read_av1c(&mut b)?), + BoxType::ColourInformationBox => ItemProperty::Colour(read_colr(&mut b, strictness)?), + BoxType::ImageMirror => ItemProperty::Mirroring(read_imir(&mut b)?), + BoxType::ImageRotation => ItemProperty::Rotation(read_irot(&mut b)?), + BoxType::ImageSpatialExtentsProperty => { + ItemProperty::ImageSpatialExtents(read_ispe(&mut b)?) + } + BoxType::PixelAspectRatioBox => ItemProperty::PixelAspectRatio(read_pasp(&mut b)?), + BoxType::PixelInformationBox => ItemProperty::Channels(read_pixi(&mut b)?), + + other_box_type => { + // Even if we didn't do anything with other property types, we still store + // a record at the index to identify invalid indices in ipma boxes + skip_box_remain(&mut b)?; + let item_property = match other_box_type { + BoxType::AV1LayeredImageIndexingProperty => ItemProperty::LayeredImageIndexing, + BoxType::CleanApertureBox => ItemProperty::CleanAperture, + BoxType::LayerSelectorProperty => ItemProperty::LayerSelection, + BoxType::OperatingPointSelectorProperty => ItemProperty::OperatingPointSelector, + _ => { + warn!("No ItemProperty variant for {:?}", other_box_type); + ItemProperty::Unsupported(other_box_type) + } + }; + debug!("Storing empty record {:?}", item_property); + item_property + } + }; + properties.insert(index, property)?; + + index = PropertyIndex( + index + .0 + .checked_add(1) // must include ignored properties to have correct indexes + .ok_or(Error::InvalidData("ipco index overflow"))?, + ); + + check_parser_state!(b.content); + } + + Ok(properties) +} + +#[repr(C)] +#[derive(Clone, Copy, Debug, PartialEq)] +pub struct ImageSpatialExtentsProperty { + image_width: u32, + image_height: u32, +} + +/// Parse image spatial extents property +/// +/// See HEIF (ISO 23008-12:2017) § 6.5.3.1 +fn read_ispe<T: Read>(src: &mut BMFFBox<T>) -> Result<ImageSpatialExtentsProperty> { + if read_fullbox_version_no_flags(src)? != 0 { + return Err(Error::Unsupported("ispe version")); + } + + let image_width = be_u32(src)?; + let image_height = be_u32(src)?; + + Ok(ImageSpatialExtentsProperty { + image_width, + image_height, + }) +} + +#[repr(C)] +#[derive(Debug)] +pub struct PixelAspectRatio { + h_spacing: u32, + v_spacing: u32, +} + +/// Parse pixel aspect ratio property +/// +/// See HEIF (ISO 23008-12:2017) § 6.5.4.1 +/// See ISOBMFF (ISO 14496-12:2020) § 12.1.4.2 +fn read_pasp<T: Read>(src: &mut BMFFBox<T>) -> Result<PixelAspectRatio> { + let h_spacing = be_u32(src)?; + let v_spacing = be_u32(src)?; + + Ok(PixelAspectRatio { + h_spacing, + v_spacing, + }) +} + +#[derive(Debug)] +pub struct PixelInformation { + bits_per_channel: TryVec<u8>, +} + +/// Parse pixel information +/// See HEIF (ISO 23008-12:2017) § 6.5.6 +fn read_pixi<T: Read>(src: &mut BMFFBox<T>) -> Result<PixelInformation> { + let version = read_fullbox_version_no_flags(src)?; + if version != 0 { + return Err(Error::Unsupported("pixi version")); + } + + let num_channels = src.read_u8()?; + let mut bits_per_channel = TryVec::with_capacity(num_channels.to_usize())?; + let num_channels_read = src.try_read_to_end(&mut bits_per_channel)?; + + if u8::try_from(num_channels_read)? != num_channels { + return Err(Error::InvalidData("invalid num_channels")); + } + + check_parser_state!(src.content); + Ok(PixelInformation { bits_per_channel }) +} + +/// Despite [Rec. ITU-T H.273] (12/2016) defining the CICP fields as having a +/// range of 0-255, and only a small fraction of those values being used, +/// ISOBMFF (ISO 14496-12:2020) § 12.1.5 defines them as 16-bit values in the +/// `colr` box. Since we have no use for the additional range, and it would +/// complicate matters later, we fallibly convert before storing the input. +/// +/// [Rec. ITU-T H.273]: https://www.itu.int/rec/T-REC-H.273-201612-I/en +#[repr(C)] +#[derive(Debug)] +pub struct NclxColourInformation { + colour_primaries: u8, + transfer_characteristics: u8, + matrix_coefficients: u8, + full_range_flag: bool, +} + +/// The raw bytes of the ICC profile +#[repr(C)] +pub struct IccColourInformation { + bytes: TryVec<u8>, +} + +impl fmt::Debug for IccColourInformation { + fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { + f.debug_struct("IccColourInformation") + .field("data", &format_args!("{} bytes", self.bytes.len())) + .finish() + } +} + +#[repr(C)] +#[derive(Debug)] +pub enum ColourInformation { + Nclx(NclxColourInformation), + Icc(IccColourInformation, FourCC), +} + +impl ColourInformation { + fn colour_type(&self) -> FourCC { + match self { + Self::Nclx(_) => FourCC::from(*b"nclx"), + Self::Icc(_, colour_type) => colour_type.clone(), + } + } +} + +/// Parse colour information +/// See ISOBMFF (ISO 14496-12:2020) § 12.1.5 +fn read_colr<T: Read>( + src: &mut BMFFBox<T>, + strictness: ParseStrictness, +) -> Result<ColourInformation> { + let colour_type = be_u32(src)?.to_be_bytes(); + + match &colour_type { + b"nclx" => { + const NUM_RESERVED_BITS: u8 = 7; + let colour_primaries = be_u16(src)?.try_into()?; + let transfer_characteristics = be_u16(src)?.try_into()?; + let matrix_coefficients = be_u16(src)?.try_into()?; + let bytes = src.read_into_try_vec()?; + let mut bit_reader = BitReader::new(&bytes); + let full_range_flag = bit_reader.read_bool()?; + if bit_reader.remaining() != NUM_RESERVED_BITS.into() { + error!( + "read_colr expected {} reserved bits, found {}", + NUM_RESERVED_BITS, + bit_reader.remaining() + ); + return Err(Error::InvalidData("Unexpected size for colr box")); + } + if bit_reader.read_u8(NUM_RESERVED_BITS)? != 0 { + fail_if( + strictness != ParseStrictness::Permissive, + "The 7 reserved bits at the end of the ColourInformationBox \ + for colour_type == 'nclx' must be 0 \ + per ISOBMFF (ISO 14496-12:2020) § 12.1.5.2", + )?; + } + + Ok(ColourInformation::Nclx(NclxColourInformation { + colour_primaries, + transfer_characteristics, + matrix_coefficients, + full_range_flag, + })) + } + b"rICC" | b"prof" => Ok(ColourInformation::Icc( + IccColourInformation { + bytes: src.read_into_try_vec()?, + }, + FourCC::from(colour_type), + )), + _ => { + error!("read_colr colour_type: {:?}", colour_type); + Err(Error::InvalidData( + "Unsupported colour_type for ColourInformationBox", + )) + } + } +} + +#[repr(C)] +#[derive(Clone, Copy, Debug)] +/// Rotation in the positive (that is, anticlockwise) direction +/// Visualized in terms of starting with (⥠) UPWARDS HARPOON WITH BARB LEFT FROM BAR +/// similar to a DIGIT ONE (1) +pub enum ImageRotation { + /// ⥠ UPWARDS HARPOON WITH BARB LEFT FROM BAR + D0, + /// ⥞ LEFTWARDS HARPOON WITH BARB DOWN FROM BAR + D90, + /// ⥝ DOWNWARDS HARPOON WITH BARB RIGHT FROM BAR + D180, + /// ⥛ RIGHTWARDS HARPOON WITH BARB UP FROM BAR + D270, +} + +/// Parse image rotation box +/// See HEIF (ISO 23008-12:2017) § 6.5.10 +fn read_irot<T: Read>(src: &mut BMFFBox<T>) -> Result<ImageRotation> { + let irot = src.read_into_try_vec()?; + let mut irot = BitReader::new(&irot); + let _reserved = irot.read_u8(6)?; + let image_rotation = match irot.read_u8(2)? { + 0 => ImageRotation::D0, + 1 => ImageRotation::D90, + 2 => ImageRotation::D180, + 3 => ImageRotation::D270, + _ => unreachable!(), + }; + + check_parser_state!(src.content); + + Ok(image_rotation) +} + +/// The axis about which the image is mirrored (opposite of flip) +/// Visualized in terms of starting with (⥠) UPWARDS HARPOON WITH BARB LEFT FROM BAR +/// similar to a DIGIT ONE (1) +#[repr(C)] +#[derive(Debug)] +pub enum ImageMirror { + /// top and bottom parts exchanged + /// ⥡ DOWNWARDS HARPOON WITH BARB LEFT FROM BAR + TopBottom, + /// left and right parts exchanged + /// ⥜ UPWARDS HARPOON WITH BARB RIGHT FROM BAR + LeftRight, +} + +/// Parse image mirroring box +/// See HEIF (ISO 23008-12:2017) § 6.5.12<br /> +/// Note: [ISO/IEC 23008-12:2017/DAmd 2](https://www.iso.org/standard/81688.html) +/// reverses the interpretation of the 'imir' box in § 6.5.12.3: +/// > `axis` specifies a vertical (`axis` = 0) or horizontal (`axis` = 1) axis +/// > for the mirroring operation. +/// +/// is replaced with: +/// > `mode` specifies how the mirroring is performed: 0 indicates that the top +/// > and bottom parts of the image are exchanged; 1 specifies that the left and +/// > right parts are exchanged. +/// > +/// > NOTE: In Exif, orientation tag can be used to signal mirroring operations. +/// > Exif orientation tag 4 corresponds to `mode` = 0 of `ImageMirror`, and +/// > Exif orientation tag 2 corresponds to `mode` = 1 accordingly. +/// +/// This implementation conforms to the text in Draft Amendment 2, which is the +/// opposite of the published standard as of 4 June 2021. +fn read_imir<T: Read>(src: &mut BMFFBox<T>) -> Result<ImageMirror> { + let imir = src.read_into_try_vec()?; + let mut imir = BitReader::new(&imir); + let _reserved = imir.read_u8(7)?; + let image_mirror = match imir.read_u8(1)? { + 0 => ImageMirror::TopBottom, + 1 => ImageMirror::LeftRight, + _ => unreachable!(), + }; + + check_parser_state!(src.content); + + Ok(image_mirror) +} + +/// See HEIF (ISO 23008-12:2017) § 6.5.8 +#[derive(Debug, PartialEq)] +pub struct AuxiliaryTypeProperty { + aux_type: TryString, + aux_subtype: TryString, +} + +/// Parse image properties for auxiliary images +/// See HEIF (ISO 23008-12:2017) § 6.5.8 +fn read_auxc<T: Read>(src: &mut BMFFBox<T>) -> Result<AuxiliaryTypeProperty> { + let version = read_fullbox_version_no_flags(src)?; + if version != 0 { + return Err(Error::Unsupported("auxC version")); + } + + let mut aux = TryString::new(); + src.try_read_to_end(&mut aux)?; + + let (aux_type, aux_subtype): (TryString, TryVec<u8>); + if let Some(nul_byte_pos) = aux.iter().position(|&b| b == b'\0') { + let (a, b) = aux.as_slice().split_at(nul_byte_pos); + aux_type = a.try_into()?; + aux_subtype = (&b[1..]).try_into()?; + } else { + aux_type = aux; + aux_subtype = TryVec::new(); + } + + Ok(AuxiliaryTypeProperty { + aux_type, + aux_subtype, + }) +} + +/// Parse an item location box inside a meta box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.3 +fn read_iloc<T: Read>(src: &mut BMFFBox<T>) -> Result<TryHashMap<ItemId, ItemLocationBoxItem>> { + let version: IlocVersion = read_fullbox_version_no_flags(src)?.try_into()?; + + let iloc = src.read_into_try_vec()?; + let mut iloc = BitReader::new(&iloc); + + let offset_size: IlocFieldSize = iloc.read_u8(4)?.try_into()?; + let length_size: IlocFieldSize = iloc.read_u8(4)?.try_into()?; + let base_offset_size: IlocFieldSize = iloc.read_u8(4)?.try_into()?; + + let index_size: Option<IlocFieldSize> = match version { + IlocVersion::One | IlocVersion::Two => Some(iloc.read_u8(4)?.try_into()?), + IlocVersion::Zero => { + let _reserved = iloc.read_u8(4)?; + None + } + }; + + let item_count = match version { + IlocVersion::Zero | IlocVersion::One => iloc.read_u32(16)?, + IlocVersion::Two => iloc.read_u32(32)?, + }; + + let mut items = TryHashMap::with_capacity(item_count.to_usize())?; + + for _ in 0..item_count { + let item_id = ItemId(match version { + IlocVersion::Zero | IlocVersion::One => iloc.read_u32(16)?, + IlocVersion::Two => iloc.read_u32(32)?, + }); + + // The spec isn't entirely clear how an `iloc` should be interpreted for version 0, + // which has no `construction_method` field. It does say: + // "For maximum compatibility, version 0 of this box should be used in preference to + // version 1 with `construction_method==0`, or version 2 when possible." + // We take this to imply version 0 can be interpreted as using file offsets. + let construction_method = match version { + IlocVersion::Zero => ConstructionMethod::File, + IlocVersion::One | IlocVersion::Two => { + let _reserved = iloc.read_u16(12)?; + match iloc.read_u16(4)? { + 0 => ConstructionMethod::File, + 1 => ConstructionMethod::Idat, + 2 => ConstructionMethod::Item, + _ => return Err(Error::InvalidData("construction_method is taken from the set 0, 1 or 2 per ISOBMFF (ISO 14496-12:2020) § 8.11.3.3")) + } + } + }; + + let data_reference_index = iloc.read_u16(16)?; + if data_reference_index != 0 { + return Err(Error::Unsupported( + "external file references (iloc.data_reference_index != 0) are not supported", + )); + } + let base_offset = iloc.read_u64(base_offset_size.as_bits())?; + let extent_count = iloc.read_u16(16)?; + + if extent_count < 1 { + return Err(Error::InvalidData( + "extent_count must have a value 1 or greater per ISOBMFF (ISO 14496-12:2020) § 8.11.3.3", + )); + } + + // "If only one extent is used (extent_count = 1) then either or both of the + // offset and length may be implied" + if extent_count != 1 + && (offset_size == IlocFieldSize::Zero || length_size == IlocFieldSize::Zero) + { + return Err(Error::InvalidData( + "extent_count != 1 requires explicit offset and length per ISOBMFF (ISO 14496-12:2020) § 8.11.3.3", + )); + } + + let mut extents = TryVec::with_capacity(extent_count.to_usize())?; + + for _ in 0..extent_count { + // Parsed but currently ignored, see `Extent` + let _extent_index = match &index_size { + None | Some(IlocFieldSize::Zero) => None, + Some(index_size) => { + debug_assert!(version == IlocVersion::One || version == IlocVersion::Two); + Some(iloc.read_u64(index_size.as_bits())?) + } + }; + + // Per ISOBMFF (ISO 14496-12:2020) § 8.11.3.1: + // "If the offset is not identified (the field has a length of zero), then the + // beginning of the source (offset 0) is implied" + // This behavior will follow from BitReader::read_u64(0) -> 0. + let extent_offset = iloc.read_u64(offset_size.as_bits())?; + let extent_length = iloc.read_u64(length_size.as_bits())?.try_into()?; + + // "If the length is not specified, or specified as zero, then the entire length of + // the source is implied" (ibid) + let offset = base_offset + .checked_add(extent_offset) + .ok_or(Error::InvalidData("offset calculation overflow"))?; + let extent = if extent_length == 0 { + Extent::ToEnd { offset } + } else { + Extent::WithLength { + offset, + len: extent_length, + } + }; + + extents.push(extent)?; + } + + let loc = ItemLocationBoxItem { + construction_method, + extents, + }; + + if items.insert(item_id, loc)?.is_some() { + return Err(Error::InvalidData("duplicate item_ID in iloc")); + } + } + + if iloc.remaining() == 0 { + Ok(items) + } else { + Err(Error::InvalidData("invalid iloc size")) + } +} + +/// Read the contents of a box, including sub boxes. +pub fn read_mp4<T: Read>(f: &mut T) -> Result<MediaContext> { + let mut context = None; + let mut found_ftyp = false; + // TODO(kinetik): Top-level parsing should handle zero-sized boxes + // rather than throwing an error. + let mut iter = BoxIter::new(f); + while let Some(mut b) = iter.next_box()? { + // box ordering: ftyp before any variable length box (inc. moov), + // but may not be first box in file if file signatures etc. present + // fragmented mp4 order: ftyp, moov, pairs of moof/mdat (1-multiple), mfra + + // "special": uuid, wide (= 8 bytes) + // isom: moov, mdat, free, skip, udta, ftyp, moof, mfra + // iso2: pdin, meta + // iso3: meco + // iso5: styp, sidx, ssix, prft + // unknown, maybe: id32 + + // qt: pnot + + // possibly allow anything where all printable and/or all lowercase printable + // "four printable characters from the ISO 8859-1 character set" + match b.head.name { + BoxType::FileTypeBox => { + let ftyp = read_ftyp(&mut b)?; + found_ftyp = true; + debug!("{:?}", ftyp); + } + BoxType::MovieBox => { + context = Some(read_moov(&mut b, context)?); + } + #[cfg(feature = "meta-xml")] + BoxType::MetadataBox => { + if let Some(ctx) = &mut context { + ctx.metadata = Some(read_meta(&mut b)); + } + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + if context.is_some() { + debug!( + "found moov {}, could stop pure 'moov' parser now", + if found_ftyp { + "and ftyp" + } else { + "but no ftyp" + } + ); + } + } + + // XXX(kinetik): This isn't perfect, as a "moov" with no contents is + // treated as okay but we haven't found anything useful. Needs more + // thought for clearer behaviour here. + context.ok_or(Error::NoMoov) +} + +/// Parse a Movie Header Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.2.2 +fn parse_mvhd<T: Read>(f: &mut BMFFBox<T>) -> Result<Option<MediaTimeScale>> { + let mvhd = read_mvhd(f)?; + debug!("{:?}", mvhd); + if mvhd.timescale == 0 { + return Err(Error::InvalidData("zero timescale in mdhd")); + } + let timescale = Some(MediaTimeScale(u64::from(mvhd.timescale))); + Ok(timescale) +} + +/// Parse a Movie Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.2.1 +/// Note that despite the spec indicating "exactly one" moov box should exist at +/// the file container level, we support reading and merging multiple moov boxes +/// such as with tests/test_case_1185230.mp4. +fn read_moov<T: Read>(f: &mut BMFFBox<T>, context: Option<MediaContext>) -> Result<MediaContext> { + let MediaContext { + mut timescale, + mut tracks, + mut mvex, + mut psshs, + mut userdata, + #[cfg(feature = "meta-xml")] + metadata, + } = context.unwrap_or_default(); + + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MovieHeaderBox => { + timescale = parse_mvhd(&mut b)?; + } + BoxType::TrackBox => { + let mut track = Track::new(tracks.len()); + read_trak(&mut b, &mut track)?; + tracks.push(track)?; + } + BoxType::MovieExtendsBox => { + mvex = Some(read_mvex(&mut b)?); + debug!("{:?}", mvex); + } + BoxType::ProtectionSystemSpecificHeaderBox => { + let pssh = read_pssh(&mut b)?; + debug!("{:?}", pssh); + psshs.push(pssh)?; + } + BoxType::UserdataBox => { + userdata = Some(read_udta(&mut b)); + debug!("{:?}", userdata); + if let Some(Err(_)) = userdata { + // There was an error parsing userdata. Such failures are not fatal to overall + // parsing, just skip the rest of the box. + skip_box_remain(&mut b)?; + } + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + + Ok(MediaContext { + timescale, + tracks, + mvex, + psshs, + userdata, + #[cfg(feature = "meta-xml")] + metadata, + }) +} + +fn read_pssh<T: Read>(src: &mut BMFFBox<T>) -> Result<ProtectionSystemSpecificHeaderBox> { + let len = src.bytes_left(); + let mut box_content = read_buf(src, len)?; + let (system_id, kid, data) = { + let pssh = &mut Cursor::new(&box_content); + + let (version, _) = read_fullbox_extra(pssh)?; + + let system_id = read_buf(pssh, 16)?; + + let mut kid = TryVec::<ByteData>::new(); + if version > 0 { + let count = be_u32(pssh)?; + for _ in 0..count { + let item = read_buf(pssh, 16)?; + kid.push(item)?; + } + } + + let data_size = be_u32(pssh)?; + let data = read_buf(pssh, data_size.into())?; + + (system_id, kid, data) + }; + + let mut pssh_box = TryVec::new(); + write_be_u32(&mut pssh_box, src.head.size.try_into()?)?; + pssh_box.extend_from_slice(b"pssh")?; + pssh_box.append(&mut box_content)?; + + Ok(ProtectionSystemSpecificHeaderBox { + system_id, + kid, + data, + box_content: pssh_box, + }) +} + +/// Parse a Movie Extends Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.8.1 +fn read_mvex<T: Read>(src: &mut BMFFBox<T>) -> Result<MovieExtendsBox> { + let mut iter = src.box_iter(); + let mut fragment_duration = None; + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MovieExtendsHeaderBox => { + let duration = read_mehd(&mut b)?; + fragment_duration = Some(duration); + } + _ => skip_box_content(&mut b)?, + } + } + Ok(MovieExtendsBox { fragment_duration }) +} + +fn read_mehd<T: Read>(src: &mut BMFFBox<T>) -> Result<MediaScaledTime> { + let (version, _) = read_fullbox_extra(src)?; + let fragment_duration = match version { + 1 => be_u64(src)?, + 0 => u64::from(be_u32(src)?), + _ => return Err(Error::InvalidData("unhandled mehd version")), + }; + Ok(MediaScaledTime(fragment_duration)) +} + +/// Parse a Track Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.3.1. +fn read_trak<T: Read>(f: &mut BMFFBox<T>, track: &mut Track) -> Result<()> { + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::TrackHeaderBox => { + let tkhd = read_tkhd(&mut b)?; + track.track_id = Some(tkhd.track_id); + track.tkhd = Some(tkhd.clone()); + debug!("{:?}", tkhd); + } + BoxType::EditBox => read_edts(&mut b, track)?, + BoxType::MediaBox => read_mdia(&mut b, track)?, + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +fn read_edts<T: Read>(f: &mut BMFFBox<T>, track: &mut Track) -> Result<()> { + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::EditListBox => { + let elst = read_elst(&mut b)?; + if elst.edits.is_empty() { + debug!("empty edit list"); + continue; + } + let mut empty_duration = 0; + let mut idx = 0; + if elst.edits[idx].media_time == -1 { + if elst.edits.len() < 2 { + debug!("expected additional edit, ignoring edit list"); + continue; + } + empty_duration = elst.edits[idx].segment_duration; + idx += 1; + } + track.empty_duration = Some(MediaScaledTime(empty_duration)); + let media_time = elst.edits[idx].media_time; + if media_time < 0 { + debug!("unexpected negative media time in edit"); + } + track.media_time = Some(TrackScaledTime::<u64>( + std::cmp::max(0, media_time) as u64, + track.id, + )); + if elst.edits.len() > 2 { + debug!("ignoring edit list with {} entries", elst.edits.len()); + } + debug!("{:?}", elst); + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +#[allow(clippy::type_complexity)] // Allow the complex return, maybe rework in future +fn parse_mdhd<T: Read>( + f: &mut BMFFBox<T>, + track: &mut Track, +) -> Result<( + MediaHeaderBox, + Option<TrackScaledTime<u64>>, + Option<TrackTimeScale<u64>>, +)> { + let mdhd = read_mdhd(f)?; + let duration = match mdhd.duration { + std::u64::MAX => None, + duration => Some(TrackScaledTime::<u64>(duration, track.id)), + }; + if mdhd.timescale == 0 { + return Err(Error::InvalidData("zero timescale in mdhd")); + } + let timescale = Some(TrackTimeScale::<u64>(u64::from(mdhd.timescale), track.id)); + Ok((mdhd, duration, timescale)) +} + +fn read_mdia<T: Read>(f: &mut BMFFBox<T>, track: &mut Track) -> Result<()> { + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MediaHeaderBox => { + let (mdhd, duration, timescale) = parse_mdhd(&mut b, track)?; + track.duration = duration; + track.timescale = timescale; + debug!("{:?}", mdhd); + } + BoxType::HandlerBox => { + let hdlr = read_hdlr(&mut b, ParseStrictness::Permissive)?; + + match hdlr.handler_type.value.as_ref() { + b"vide" => track.track_type = TrackType::Video, + b"soun" => track.track_type = TrackType::Audio, + b"meta" => track.track_type = TrackType::Metadata, + _ => (), + } + debug!("{:?}", hdlr); + } + BoxType::MediaInformationBox => read_minf(&mut b, track)?, + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +fn read_minf<T: Read>(f: &mut BMFFBox<T>, track: &mut Track) -> Result<()> { + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::SampleTableBox => read_stbl(&mut b, track)?, + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +fn read_stbl<T: Read>(f: &mut BMFFBox<T>, track: &mut Track) -> Result<()> { + let mut iter = f.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::SampleDescriptionBox => { + let stsd = read_stsd(&mut b, track)?; + debug!("{:?}", stsd); + track.stsd = Some(stsd); + } + BoxType::TimeToSampleBox => { + let stts = read_stts(&mut b)?; + debug!("{:?}", stts); + track.stts = Some(stts); + } + BoxType::SampleToChunkBox => { + let stsc = read_stsc(&mut b)?; + debug!("{:?}", stsc); + track.stsc = Some(stsc); + } + BoxType::SampleSizeBox => { + let stsz = read_stsz(&mut b)?; + debug!("{:?}", stsz); + track.stsz = Some(stsz); + } + BoxType::ChunkOffsetBox => { + let stco = read_stco(&mut b)?; + debug!("{:?}", stco); + track.stco = Some(stco); + } + BoxType::ChunkLargeOffsetBox => { + let co64 = read_co64(&mut b)?; + debug!("{:?}", co64); + track.stco = Some(co64); + } + BoxType::SyncSampleBox => { + let stss = read_stss(&mut b)?; + debug!("{:?}", stss); + track.stss = Some(stss); + } + BoxType::CompositionOffsetBox => { + let ctts = read_ctts(&mut b)?; + debug!("{:?}", ctts); + track.ctts = Some(ctts); + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +/// Parse an ftyp box. +/// See ISOBMFF (ISO 14496-12:2020) § 4.3 +fn read_ftyp<T: Read>(src: &mut BMFFBox<T>) -> Result<FileTypeBox> { + let major = be_u32(src)?; + let minor = be_u32(src)?; + let bytes_left = src.bytes_left(); + if bytes_left % 4 != 0 { + return Err(Error::InvalidData("invalid ftyp size")); + } + // Is a brand_count of zero valid? + let brand_count = bytes_left / 4; + let mut brands = TryVec::with_capacity(brand_count.try_into()?)?; + for _ in 0..brand_count { + brands.push(be_u32(src)?.into())?; + } + Ok(FileTypeBox { + major_brand: From::from(major), + minor_version: minor, + compatible_brands: brands, + }) +} + +/// Parse an mvhd box. +fn read_mvhd<T: Read>(src: &mut BMFFBox<T>) -> Result<MovieHeaderBox> { + let (version, _) = read_fullbox_extra(src)?; + match version { + // 64 bit creation and modification times. + 1 => { + skip(src, 16)?; + } + // 32 bit creation and modification times. + 0 => { + skip(src, 8)?; + } + _ => return Err(Error::InvalidData("unhandled mvhd version")), + } + let timescale = be_u32(src)?; + let duration = match version { + 1 => be_u64(src)?, + 0 => { + let d = be_u32(src)?; + if d == std::u32::MAX { + std::u64::MAX + } else { + u64::from(d) + } + } + _ => return Err(Error::InvalidData("unhandled mvhd version")), + }; + // Skip remaining fields. + skip(src, 80)?; + Ok(MovieHeaderBox { + timescale, + duration, + }) +} + +/// Parse a tkhd box. +fn read_tkhd<T: Read>(src: &mut BMFFBox<T>) -> Result<TrackHeaderBox> { + let (version, flags) = read_fullbox_extra(src)?; + let disabled = flags & 0x1u32 == 0 || flags & 0x2u32 == 0; + match version { + // 64 bit creation and modification times. + 1 => { + skip(src, 16)?; + } + // 32 bit creation and modification times. + 0 => { + skip(src, 8)?; + } + _ => return Err(Error::InvalidData("unhandled tkhd version")), + } + let track_id = be_u32(src)?; + skip(src, 4)?; + let duration = match version { + 1 => be_u64(src)?, + 0 => u64::from(be_u32(src)?), + _ => return Err(Error::InvalidData("unhandled tkhd version")), + }; + // Skip uninteresting fields. + skip(src, 16)?; + + let matrix = Matrix { + a: be_i32(src)?, + b: be_i32(src)?, + u: be_i32(src)?, + c: be_i32(src)?, + d: be_i32(src)?, + v: be_i32(src)?, + x: be_i32(src)?, + y: be_i32(src)?, + w: be_i32(src)?, + }; + + let width = be_u32(src)?; + let height = be_u32(src)?; + Ok(TrackHeaderBox { + track_id, + disabled, + duration, + width, + height, + matrix, + }) +} + +/// Parse a elst box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.6.6 +fn read_elst<T: Read>(src: &mut BMFFBox<T>) -> Result<EditListBox> { + let (version, _) = read_fullbox_extra(src)?; + let edit_count = be_u32(src)?; + let mut edits = TryVec::with_capacity(edit_count.to_usize())?; + for _ in 0..edit_count { + let (segment_duration, media_time) = match version { + 1 => { + // 64 bit segment duration and media times. + (be_u64(src)?, be_i64(src)?) + } + 0 => { + // 32 bit segment duration and media times. + (u64::from(be_u32(src)?), i64::from(be_i32(src)?)) + } + _ => return Err(Error::InvalidData("unhandled elst version")), + }; + let media_rate_integer = be_i16(src)?; + let media_rate_fraction = be_i16(src)?; + edits.push(Edit { + segment_duration, + media_time, + media_rate_integer, + media_rate_fraction, + })?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(EditListBox { edits }) +} + +/// Parse a mdhd box. +fn read_mdhd<T: Read>(src: &mut BMFFBox<T>) -> Result<MediaHeaderBox> { + let (version, _) = read_fullbox_extra(src)?; + let (timescale, duration) = match version { + 1 => { + // Skip 64-bit creation and modification times. + skip(src, 16)?; + + // 64 bit duration. + (be_u32(src)?, be_u64(src)?) + } + 0 => { + // Skip 32-bit creation and modification times. + skip(src, 8)?; + + // 32 bit duration. + let timescale = be_u32(src)?; + let duration = { + // Since we convert the 32-bit duration to 64-bit by + // upcasting, we need to preserve the special all-1s + // ("unknown") case by hand. + let d = be_u32(src)?; + if d == std::u32::MAX { + std::u64::MAX + } else { + u64::from(d) + } + }; + (timescale, duration) + } + _ => return Err(Error::InvalidData("unhandled mdhd version")), + }; + + // Skip uninteresting fields. + skip(src, 4)?; + + Ok(MediaHeaderBox { + timescale, + duration, + }) +} + +/// Parse a stco box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.7.5 +fn read_stco<T: Read>(src: &mut BMFFBox<T>) -> Result<ChunkOffsetBox> { + let (_, _) = read_fullbox_extra(src)?; + let offset_count = be_u32(src)?; + let mut offsets = TryVec::with_capacity(offset_count.to_usize())?; + for _ in 0..offset_count { + offsets.push(be_u32(src)?.into())?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(ChunkOffsetBox { offsets }) +} + +/// Parse a co64 box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.7.5 +fn read_co64<T: Read>(src: &mut BMFFBox<T>) -> Result<ChunkOffsetBox> { + let (_, _) = read_fullbox_extra(src)?; + let offset_count = be_u32(src)?; + let mut offsets = TryVec::with_capacity(offset_count.to_usize())?; + for _ in 0..offset_count { + offsets.push(be_u64(src)?)?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(ChunkOffsetBox { offsets }) +} + +/// Parse a stss box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.6.2 +fn read_stss<T: Read>(src: &mut BMFFBox<T>) -> Result<SyncSampleBox> { + let (_, _) = read_fullbox_extra(src)?; + let sample_count = be_u32(src)?; + let mut samples = TryVec::with_capacity(sample_count.to_usize())?; + for _ in 0..sample_count { + samples.push(be_u32(src)?)?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(SyncSampleBox { samples }) +} + +/// Parse a stsc box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.7.4 +fn read_stsc<T: Read>(src: &mut BMFFBox<T>) -> Result<SampleToChunkBox> { + let (_, _) = read_fullbox_extra(src)?; + let sample_count = be_u32(src)?; + let mut samples = TryVec::with_capacity(sample_count.to_usize())?; + for _ in 0..sample_count { + let first_chunk = be_u32(src)?; + let samples_per_chunk = be_u32(src)?; + let sample_description_index = be_u32(src)?; + samples.push(SampleToChunk { + first_chunk, + samples_per_chunk, + sample_description_index, + })?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(SampleToChunkBox { samples }) +} + +/// Parse a Composition Time to Sample Box +/// See ISOBMFF (ISO 14496-12:2020) § 8.6.1.3 +fn read_ctts<T: Read>(src: &mut BMFFBox<T>) -> Result<CompositionOffsetBox> { + let (version, _) = read_fullbox_extra(src)?; + + let counts = be_u32(src)?; + + if counts + .checked_mul(8) + .map_or(true, |bytes| u64::from(bytes) > src.bytes_left()) + { + return Err(Error::InvalidData("insufficient data in 'ctts' box")); + } + + let mut offsets = TryVec::with_capacity(counts.to_usize())?; + for _ in 0..counts { + let (sample_count, time_offset) = match version { + // According to spec, Version0 shoule be used when version == 0; + // however, some buggy contents have negative value when version == 0. + // So we always use Version1 here. + 0..=1 => { + let count = be_u32(src)?; + let offset = TimeOffsetVersion::Version1(be_i32(src)?); + (count, offset) + } + _ => { + return Err(Error::InvalidData("unsupported version in 'ctts' box")); + } + }; + offsets.push(TimeOffset { + sample_count, + time_offset, + })?; + } + + check_parser_state!(src.content); + + Ok(CompositionOffsetBox { samples: offsets }) +} + +/// Parse a stsz box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.7.3.2 +fn read_stsz<T: Read>(src: &mut BMFFBox<T>) -> Result<SampleSizeBox> { + let (_, _) = read_fullbox_extra(src)?; + let sample_size = be_u32(src)?; + let sample_count = be_u32(src)?; + let mut sample_sizes = TryVec::new(); + if sample_size == 0 { + sample_sizes.reserve(sample_count.to_usize())?; + for _ in 0..sample_count { + sample_sizes.push(be_u32(src)?)?; + } + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(SampleSizeBox { + sample_size, + sample_sizes, + }) +} + +/// Parse a stts box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.6.1.2 +fn read_stts<T: Read>(src: &mut BMFFBox<T>) -> Result<TimeToSampleBox> { + let (_, _) = read_fullbox_extra(src)?; + let sample_count = be_u32(src)?; + let mut samples = TryVec::with_capacity(sample_count.to_usize())?; + for _ in 0..sample_count { + let sample_count = be_u32(src)?; + let sample_delta = be_u32(src)?; + samples.push(Sample { + sample_count, + sample_delta, + })?; + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(TimeToSampleBox { samples }) +} + +/// Parse a VPx Config Box. +fn read_vpcc<T: Read>(src: &mut BMFFBox<T>) -> Result<VPxConfigBox> { + let (version, _) = read_fullbox_extra(src)?; + let supported_versions = [0, 1]; + if !supported_versions.contains(&version) { + return Err(Error::Unsupported("unknown vpcC version")); + } + + let profile = src.read_u8()?; + let level = src.read_u8()?; + let ( + bit_depth, + colour_primaries, + chroma_subsampling, + transfer_characteristics, + matrix_coefficients, + video_full_range_flag, + ) = if version == 0 { + let (bit_depth, colour_primaries) = { + let byte = src.read_u8()?; + ((byte >> 4) & 0x0f, byte & 0x0f) + }; + // Note, transfer_characteristics was known as transfer_function in v0 + let (chroma_subsampling, transfer_characteristics, video_full_range_flag) = { + let byte = src.read_u8()?; + ((byte >> 4) & 0x0f, (byte >> 1) & 0x07, (byte & 1) == 1) + }; + ( + bit_depth, + colour_primaries, + chroma_subsampling, + transfer_characteristics, + None, + video_full_range_flag, + ) + } else { + let (bit_depth, chroma_subsampling, video_full_range_flag) = { + let byte = src.read_u8()?; + ((byte >> 4) & 0x0f, (byte >> 1) & 0x07, (byte & 1) == 1) + }; + let colour_primaries = src.read_u8()?; + let transfer_characteristics = src.read_u8()?; + let matrix_coefficients = src.read_u8()?; + + ( + bit_depth, + colour_primaries, + chroma_subsampling, + transfer_characteristics, + Some(matrix_coefficients), + video_full_range_flag, + ) + }; + + let codec_init_size = be_u16(src)?; + let codec_init = read_buf(src, codec_init_size.into())?; + + // TODO(rillian): validate field value ranges. + Ok(VPxConfigBox { + profile, + level, + bit_depth, + colour_primaries, + chroma_subsampling, + transfer_characteristics, + matrix_coefficients, + video_full_range_flag, + codec_init, + }) +} + +/// See [AV1-ISOBMFF § 2.3.3](https://aomediacodec.github.io/av1-isobmff/#av1codecconfigurationbox-syntax) +fn read_av1c<T: Read>(src: &mut BMFFBox<T>) -> Result<AV1ConfigBox> { + // We want to store the raw config as well as a structured (parsed) config, so create a copy of + // the raw config so we have it later, and then parse the structured data from that. + let raw_config = src.read_into_try_vec()?; + let mut raw_config_slice = raw_config.as_slice(); + let marker_byte = raw_config_slice.read_u8()?; + if marker_byte & 0x80 != 0x80 { + return Err(Error::Unsupported("missing av1C marker bit")); + } + if marker_byte & 0x7f != 0x01 { + return Err(Error::Unsupported("missing av1C marker bit")); + } + let profile_byte = raw_config_slice.read_u8()?; + let profile = (profile_byte & 0xe0) >> 5; + let level = profile_byte & 0x1f; + let flags_byte = raw_config_slice.read_u8()?; + let tier = (flags_byte & 0x80) >> 7; + let bit_depth = match flags_byte & 0x60 { + 0x60 => 12, + 0x40 => 10, + _ => 8, + }; + let monochrome = flags_byte & 0x10 == 0x10; + let chroma_subsampling_x = (flags_byte & 0x08) >> 3; + let chroma_subsampling_y = (flags_byte & 0x04) >> 2; + let chroma_sample_position = flags_byte & 0x03; + let delay_byte = raw_config_slice.read_u8()?; + let initial_presentation_delay_present = (delay_byte & 0x10) == 0x10; + let initial_presentation_delay_minus_one = if initial_presentation_delay_present { + delay_byte & 0x0f + } else { + 0 + }; + + Ok(AV1ConfigBox { + profile, + level, + tier, + bit_depth, + monochrome, + chroma_subsampling_x, + chroma_subsampling_y, + chroma_sample_position, + initial_presentation_delay_present, + initial_presentation_delay_minus_one, + raw_config, + }) +} + +fn read_flac_metadata<T: Read>(src: &mut BMFFBox<T>) -> Result<FLACMetadataBlock> { + let temp = src.read_u8()?; + let block_type = temp & 0x7f; + let length = be_u24(src)?.into(); + if length > src.bytes_left() { + return Err(Error::InvalidData( + "FLACMetadataBlock larger than parent box", + )); + } + let data = read_buf(src, length)?; + Ok(FLACMetadataBlock { block_type, data }) +} + +/// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.5 +fn find_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> { + // Tags for elementary stream description + const ESDESCR_TAG: u8 = 0x03; + const DECODER_CONFIG_TAG: u8 = 0x04; + const DECODER_SPECIFIC_TAG: u8 = 0x05; + + let mut remains = data; + + // Descriptor length should be more than 2 bytes. + while remains.len() > 2 { + let des = &mut Cursor::new(remains); + let tag = des.read_u8()?; + + // See MPEG-4 Systems (ISO 14496-1:2010) § 8.3.3 for interpreting size of expandable classes + + let mut end: u32 = 0; // It's u8 without declaration type that is incorrect. + // MSB of extend_or_len indicates more bytes, up to 4 bytes. + for _ in 0..4 { + if des.position() == remains.len().to_u64() { + // There's nothing more to read, the 0x80 was actually part of + // the content, and not an extension size. + end = des.position() as u32; + break; + } + let extend_or_len = des.read_u8()?; + end = (end << 7) + u32::from(extend_or_len & 0x7F); + if (extend_or_len & 0b1000_0000) == 0 { + end += des.position() as u32; + break; + } + } + + if end.to_usize() > remains.len() || u64::from(end) < des.position() { + return Err(Error::InvalidData("Invalid descriptor.")); + } + + let descriptor = &remains[des.position().try_into()?..end.to_usize()]; + + match tag { + ESDESCR_TAG => { + read_es_descriptor(descriptor, esds)?; + } + DECODER_CONFIG_TAG => { + read_dc_descriptor(descriptor, esds)?; + } + DECODER_SPECIFIC_TAG => { + read_ds_descriptor(descriptor, esds)?; + } + _ => { + debug!("Unsupported descriptor, tag {}", tag); + } + } + + remains = &remains[end.to_usize()..remains.len()]; + debug!("remains.len(): {}", remains.len()); + } + + Ok(()) +} + +fn get_audio_object_type(bit_reader: &mut BitReader) -> Result<u16> { + let mut audio_object_type: u16 = ReadInto::read(bit_reader, 5)?; + + // Extend audio object type, for example, HE-AAC. + if audio_object_type == 31 { + let audio_object_type_ext: u16 = ReadInto::read(bit_reader, 6)?; + audio_object_type = 32 + audio_object_type_ext; + } + Ok(audio_object_type) +} + +/// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.7 and probably 14496-3 somewhere? +fn read_ds_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> { + #[cfg(feature = "mp4v")] + // Check if we are in a Visual esda Box. + if esds.video_codec != CodecType::Unknown { + esds.decoder_specific_data.extend_from_slice(data)?; + return Ok(()); + } + + // We are in an Audio esda Box. + let frequency_table = vec![ + (0x0, 96000), + (0x1, 88200), + (0x2, 64000), + (0x3, 48000), + (0x4, 44100), + (0x5, 32000), + (0x6, 24000), + (0x7, 22050), + (0x8, 16000), + (0x9, 12000), + (0xa, 11025), + (0xb, 8000), + (0xc, 7350), + ]; + + let bit_reader = &mut BitReader::new(data); + + let mut audio_object_type = get_audio_object_type(bit_reader)?; + + let sample_index: u32 = ReadInto::read(bit_reader, 4)?; + + // Sample frequency could be from table, or retrieved from stream directly + // if index is 0x0f. + let sample_frequency = match sample_index { + 0x0F => Some(ReadInto::read(bit_reader, 24)?), + _ => frequency_table + .iter() + .find(|item| item.0 == sample_index) + .map(|x| x.1), + }; + + let channel_configuration: u16 = ReadInto::read(bit_reader, 4)?; + + let extended_audio_object_type = match audio_object_type { + 5 | 29 => Some(5), + _ => None, + }; + + if audio_object_type == 5 || audio_object_type == 29 { + // We have an explicit signaling for BSAC extension, should the decoder + // decode the BSAC extension (all Gecko's AAC decoders do), then this is + // what the stream will actually look like once decoded. + let _extended_sample_index = ReadInto::read(bit_reader, 4)?; + let _extended_sample_frequency: Option<u32> = match _extended_sample_index { + 0x0F => Some(ReadInto::read(bit_reader, 24)?), + _ => frequency_table + .iter() + .find(|item| item.0 == sample_index) + .map(|x| x.1), + }; + audio_object_type = get_audio_object_type(bit_reader)?; + let _extended_channel_configuration = match audio_object_type { + 22 => ReadInto::read(bit_reader, 4)?, + _ => channel_configuration, + }; + }; + + match audio_object_type { + 1..=4 | 6 | 7 | 17 | 19..=23 => { + if sample_frequency.is_none() { + return Err(Error::Unsupported("unknown frequency")); + } + + // parsing GASpecificConfig + + // If the sampling rate is not one of the rates listed in the right + // column in Table 4.82, the sampling frequency dependent tables + // (code tables, scale factor band tables etc.) must be deduced in + // order for the bitstream payload to be parsed. Since a given + // sampling frequency is associated with only one sampling frequency + // table, and since maximum flexibility is desired in the range of + // possible sampling frequencies, the following table shall be used + // to associate an implied sampling frequency with the desired + // sampling frequency dependent tables. + let sample_frequency_value = match sample_frequency.unwrap() { + 0..=9390 => 8000, + 9391..=11501 => 11025, + 11502..=13855 => 12000, + 13856..=18782 => 16000, + 18783..=23003 => 22050, + 23004..=27712 => 24000, + 27713..=37565 => 32000, + 37566..=46008 => 44100, + 46009..=55425 => 48000, + 55426..=75131 => 64000, + 75132..=92016 => 88200, + _ => 96000, + }; + + bit_reader.skip(1)?; // frameLengthFlag + let depend_on_core_order: u8 = ReadInto::read(bit_reader, 1)?; + if depend_on_core_order > 0 { + bit_reader.skip(14)?; // codeCoderDelay + } + bit_reader.skip(1)?; // extensionFlag + + let channel_counts = match channel_configuration { + 0 => { + debug!("Parsing program_config_element for channel counts"); + + bit_reader.skip(4)?; // element_instance_tag + bit_reader.skip(2)?; // object_type + bit_reader.skip(4)?; // sampling_frequency_index + let num_front_channel: u8 = ReadInto::read(bit_reader, 4)?; + let num_side_channel: u8 = ReadInto::read(bit_reader, 4)?; + let num_back_channel: u8 = ReadInto::read(bit_reader, 4)?; + let num_lfe_channel: u8 = ReadInto::read(bit_reader, 2)?; + bit_reader.skip(3)?; // num_assoc_data + bit_reader.skip(4)?; // num_valid_cc + + let mono_mixdown_present: bool = ReadInto::read(bit_reader, 1)?; + if mono_mixdown_present { + bit_reader.skip(4)?; // mono_mixdown_element_number + } + + let stereo_mixdown_present: bool = ReadInto::read(bit_reader, 1)?; + if stereo_mixdown_present { + bit_reader.skip(4)?; // stereo_mixdown_element_number + } + + let matrix_mixdown_idx_present: bool = ReadInto::read(bit_reader, 1)?; + if matrix_mixdown_idx_present { + bit_reader.skip(2)?; // matrix_mixdown_idx + bit_reader.skip(1)?; // pseudo_surround_enable + } + let mut _channel_counts = 0; + _channel_counts += read_surround_channel_count(bit_reader, num_front_channel)?; + _channel_counts += read_surround_channel_count(bit_reader, num_side_channel)?; + _channel_counts += read_surround_channel_count(bit_reader, num_back_channel)?; + _channel_counts += read_surround_channel_count(bit_reader, num_lfe_channel)?; + _channel_counts + } + 1..=7 => channel_configuration, + // Amendment 4 of the AAC standard in 2013 below + 11 => 7, // 6.1 Amendment 4 of the AAC standard in 2013 + 12 | 14 => 8, // 7.1 (a/d) of ITU BS.2159 + _ => { + return Err(Error::Unsupported("invalid channel configuration")); + } + }; + + esds.audio_object_type = Some(audio_object_type); + esds.extended_audio_object_type = extended_audio_object_type; + esds.audio_sample_rate = Some(sample_frequency_value); + esds.audio_channel_count = Some(channel_counts); + if !esds.decoder_specific_data.is_empty() { + return Err(Error::InvalidData( + "There can be only one DecSpecificInfoTag descriptor", + )); + } + esds.decoder_specific_data.extend_from_slice(data)?; + + Ok(()) + } + _ => Err(Error::Unsupported("unknown aac audio object type")), + } +} + +fn read_surround_channel_count(bit_reader: &mut BitReader, channels: u8) -> Result<u16> { + let mut count = 0; + for _ in 0..channels { + let is_cpe: bool = ReadInto::read(bit_reader, 1)?; + count += if is_cpe { 2 } else { 1 }; + bit_reader.skip(4)?; + } + Ok(count) +} + +/// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.6 +fn read_dc_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> { + let des = &mut Cursor::new(data); + let object_profile = des.read_u8()?; + + #[cfg(feature = "mp4v")] + { + esds.video_codec = match object_profile { + 0x20..=0x24 => CodecType::MP4V, + _ => CodecType::Unknown, + }; + } + + // Skip uninteresting fields. + skip(des, 12)?; + + if data.len().to_u64() > des.position() { + find_descriptor(&data[des.position().try_into()?..data.len()], esds)?; + } + + esds.audio_codec = match object_profile { + 0x40 | 0x66 | 0x67 => CodecType::AAC, + 0x69 | 0x6B => CodecType::MP3, + _ => CodecType::Unknown, + }; + + debug!( + "read_dc_descriptor: esds.audio_codec = {:?}", + esds.audio_codec + ); + + Ok(()) +} + +/// See MPEG-4 Systems (ISO 14496-1:2010) § 7.2.6.5 +fn read_es_descriptor(data: &[u8], esds: &mut ES_Descriptor) -> Result<()> { + let des = &mut Cursor::new(data); + + skip(des, 2)?; + + let esds_flags = des.read_u8()?; + + // Stream dependency flag, first bit from left most. + if esds_flags & 0x80 > 0 { + // Skip uninteresting fields. + skip(des, 2)?; + } + + // Url flag, second bit from left most. + if esds_flags & 0x40 > 0 { + // Skip uninteresting fields. + let skip_es_len = u64::from(des.read_u8()?) + 2; + skip(des, skip_es_len)?; + } + + if data.len().to_u64() > des.position() { + find_descriptor(&data[des.position().try_into()?..data.len()], esds)?; + } + + Ok(()) +} + +/// See MP4 (ISO 14496-14:2020) § 6.7.2 +fn read_esds<T: Read>(src: &mut BMFFBox<T>) -> Result<ES_Descriptor> { + let (_, _) = read_fullbox_extra(src)?; + + let esds_array = read_buf(src, src.bytes_left())?; + + let mut es_data = ES_Descriptor::default(); + find_descriptor(&esds_array, &mut es_data)?; + + es_data.codec_esds = esds_array; + + Ok(es_data) +} + +/// Parse `FLACSpecificBox`. +/// See [Encapsulation of FLAC in ISO Base Media File Format](https://github.com/xiph/flac/blob/master/doc/isoflac.txt) § 3.3.2 +fn read_dfla<T: Read>(src: &mut BMFFBox<T>) -> Result<FLACSpecificBox> { + let (version, flags) = read_fullbox_extra(src)?; + if version != 0 { + return Err(Error::Unsupported("unknown dfLa (FLAC) version")); + } + if flags != 0 { + return Err(Error::InvalidData("no-zero dfLa (FLAC) flags")); + } + let mut blocks = TryVec::new(); + while src.bytes_left() > 0 { + let block = read_flac_metadata(src)?; + blocks.push(block)?; + } + // The box must have at least one meta block, and the first block + // must be the METADATA_BLOCK_STREAMINFO + if blocks.is_empty() { + return Err(Error::InvalidData("FLACSpecificBox missing metadata")); + } else if blocks[0].block_type != 0 { + return Err(Error::InvalidData( + "FLACSpecificBox must have STREAMINFO metadata first", + )); + } else if blocks[0].data.len() != 34 { + return Err(Error::InvalidData( + "FLACSpecificBox STREAMINFO block is the wrong size", + )); + } + Ok(FLACSpecificBox { version, blocks }) +} + +/// Parse `OpusSpecificBox`. +fn read_dops<T: Read>(src: &mut BMFFBox<T>) -> Result<OpusSpecificBox> { + let version = src.read_u8()?; + if version != 0 { + return Err(Error::Unsupported("unknown dOps (Opus) version")); + } + + let output_channel_count = src.read_u8()?; + let pre_skip = be_u16(src)?; + let input_sample_rate = be_u32(src)?; + let output_gain = be_i16(src)?; + let channel_mapping_family = src.read_u8()?; + + let channel_mapping_table = if channel_mapping_family == 0 { + None + } else { + let stream_count = src.read_u8()?; + let coupled_count = src.read_u8()?; + let channel_mapping = read_buf(src, output_channel_count.into())?; + + Some(ChannelMappingTable { + stream_count, + coupled_count, + channel_mapping, + }) + }; + + // TODO(kinetik): validate field value ranges. + Ok(OpusSpecificBox { + version, + output_channel_count, + pre_skip, + input_sample_rate, + output_gain, + channel_mapping_family, + channel_mapping_table, + }) +} + +/// Re-serialize the Opus codec-specific config data as an `OpusHead` packet. +/// +/// Some decoders expect the initialization data in the format used by the +/// Ogg and WebM encapsulations. To support this we prepend the `OpusHead` +/// tag and byte-swap the data from big- to little-endian relative to the +/// dOps box. +pub fn serialize_opus_header<W: byteorder::WriteBytesExt + std::io::Write>( + opus: &OpusSpecificBox, + dst: &mut W, +) -> Result<()> { + match dst.write(b"OpusHead") { + Err(e) => return Err(Error::from(e)), + Ok(bytes) => { + if bytes != 8 { + return Err(Error::InvalidData("Couldn't write OpusHead tag.")); + } + } + } + // In mp4 encapsulation, the version field is 0, but in ogg + // it is 1. While decoders generally accept zero as well, write + // out the version of the header we're supporting rather than + // whatever we parsed out of mp4. + dst.write_u8(1)?; + dst.write_u8(opus.output_channel_count)?; + dst.write_u16::<byteorder::LittleEndian>(opus.pre_skip)?; + dst.write_u32::<byteorder::LittleEndian>(opus.input_sample_rate)?; + dst.write_i16::<byteorder::LittleEndian>(opus.output_gain)?; + dst.write_u8(opus.channel_mapping_family)?; + match opus.channel_mapping_table { + None => {} + Some(ref table) => { + dst.write_u8(table.stream_count)?; + dst.write_u8(table.coupled_count)?; + match dst.write(&table.channel_mapping) { + Err(e) => return Err(Error::from(e)), + Ok(bytes) => { + if bytes != table.channel_mapping.len() { + return Err(Error::InvalidData( + "Couldn't write channel mapping table data.", + )); + } + } + } + } + }; + Ok(()) +} + +/// Parse `ALACSpecificBox`. +fn read_alac<T: Read>(src: &mut BMFFBox<T>) -> Result<ALACSpecificBox> { + let (version, flags) = read_fullbox_extra(src)?; + if version != 0 { + return Err(Error::Unsupported("unknown alac (ALAC) version")); + } + if flags != 0 { + return Err(Error::InvalidData("no-zero alac (ALAC) flags")); + } + + let length = match src.bytes_left() { + x @ 24 | x @ 48 => x, + _ => { + return Err(Error::InvalidData( + "ALACSpecificBox magic cookie is the wrong size", + )) + } + }; + let data = read_buf(src, length)?; + + Ok(ALACSpecificBox { version, data }) +} + +/// Parse a Handler Reference Box.<br /> +/// See ISOBMFF (ISO 14496-12:2020) § 8.4.3<br /> +/// See [\[ISOBMFF\]: reserved (field = 0;) handling is ambiguous](https://github.com/MPEGGroup/FileFormat/issues/36) +fn read_hdlr<T: Read>(src: &mut BMFFBox<T>, strictness: ParseStrictness) -> Result<HandlerBox> { + if read_fullbox_version_no_flags(src)? != 0 { + return Err(Error::Unsupported("hdlr version")); + } + + let pre_defined = be_u32(src)?; + if pre_defined != 0 { + fail_if( + strictness == ParseStrictness::Strict, + "The HandlerBox 'pre_defined' field shall be 0 \ + per ISOBMFF (ISO 14496-12:2020) § 8.4.3.2", + )?; + } + + let handler_type = FourCC::from(be_u32(src)?); + + for _ in 1..=3 { + let reserved = be_u32(src)?; + if reserved != 0 { + fail_if( + strictness == ParseStrictness::Strict, + "The HandlerBox 'reserved' fields shall be 0 \ + per ISOBMFF (ISO 14496-12:2020) § 8.4.3.2", + )?; + } + } + + match std::str::from_utf8(src.read_into_try_vec()?.as_slice()) { + Ok(name) => { + match name.bytes().filter(|&b| b == b'\0').count() { + 0 => fail_if( + strictness != ParseStrictness::Permissive, + "The HandlerBox 'name' field shall be null-terminated \ + per ISOBMFF (ISO 14496-12:2020) § 8.4.3.2", + )?, + 1 => (), + n => + // See https://github.com/MPEGGroup/FileFormat/issues/35 + { + error!("Found {} nul bytes in {:x?}", n, name); + fail_if( + strictness == ParseStrictness::Strict, + "The HandlerBox 'name' field shall have a NUL byte \ + only in the final position \ + per ISOBMFF (ISO 14496-12:2020) § 8.4.3.2", + )? + } + } + } + Err(_) => fail_if( + strictness != ParseStrictness::Permissive, + "The HandlerBox 'name' field shall be valid utf8 \ + per ISOBMFF (ISO 14496-12:2020) § 8.4.3.2", + )?, + } + + Ok(HandlerBox { handler_type }) +} + +/// Parse an video description inside an stsd box. +fn read_video_sample_entry<T: Read>(src: &mut BMFFBox<T>) -> Result<SampleEntry> { + let name = src.get_header().name; + let codec_type = match name { + BoxType::AVCSampleEntry | BoxType::AVC3SampleEntry => CodecType::H264, + BoxType::MP4VideoSampleEntry => CodecType::MP4V, + BoxType::VP8SampleEntry => CodecType::VP8, + BoxType::VP9SampleEntry => CodecType::VP9, + BoxType::AV1SampleEntry => CodecType::AV1, + BoxType::ProtectedVisualSampleEntry => CodecType::EncryptedVideo, + BoxType::H263SampleEntry => CodecType::H263, + _ => { + debug!("Unsupported video codec, box {:?} found", name); + CodecType::Unknown + } + }; + + // Skip uninteresting fields. + skip(src, 6)?; + + let data_reference_index = be_u16(src)?; + + // Skip uninteresting fields. + skip(src, 16)?; + + let width = be_u16(src)?; + let height = be_u16(src)?; + + // Skip uninteresting fields. + skip(src, 50)?; + + // Skip clap/pasp/etc. for now. + let mut codec_specific = None; + let mut protection_info = TryVec::new(); + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::AVCConfigurationBox => { + if (name != BoxType::AVCSampleEntry + && name != BoxType::AVC3SampleEntry + && name != BoxType::ProtectedVisualSampleEntry) + || codec_specific.is_some() + { + return Err(Error::InvalidData("malformed video sample entry")); + } + let avcc_size = b + .head + .size + .checked_sub(b.head.offset) + .expect("offset invalid"); + let avcc = read_buf(&mut b.content, avcc_size)?; + debug!("{:?} (avcc)", avcc); + // TODO(kinetik): Parse avcC box? For now we just stash the data. + codec_specific = Some(VideoCodecSpecific::AVCConfig(avcc)); + } + BoxType::H263SpecificBox => { + if (name != BoxType::H263SampleEntry) || codec_specific.is_some() { + return Err(Error::InvalidData("malformed video sample entry")); + } + let h263_dec_spec_struc_size = b + .head + .size + .checked_sub(b.head.offset) + .expect("offset invalid"); + let h263_dec_spec_struc = read_buf(&mut b.content, h263_dec_spec_struc_size)?; + debug!("{:?} (h263DecSpecStruc)", h263_dec_spec_struc); + + codec_specific = Some(VideoCodecSpecific::H263Config(h263_dec_spec_struc)); + } + BoxType::VPCodecConfigurationBox => { + // vpcC + if (name != BoxType::VP8SampleEntry + && name != BoxType::VP9SampleEntry + && name != BoxType::ProtectedVisualSampleEntry) + || codec_specific.is_some() + { + return Err(Error::InvalidData("malformed video sample entry")); + } + let vpcc = read_vpcc(&mut b)?; + codec_specific = Some(VideoCodecSpecific::VPxConfig(vpcc)); + } + BoxType::AV1CodecConfigurationBox => { + if name != BoxType::AV1SampleEntry && name != BoxType::ProtectedVisualSampleEntry { + return Err(Error::InvalidData("malformed video sample entry")); + } + let av1c = read_av1c(&mut b)?; + codec_specific = Some(VideoCodecSpecific::AV1Config(av1c)); + } + BoxType::ESDBox => { + if name != BoxType::MP4VideoSampleEntry || codec_specific.is_some() { + return Err(Error::InvalidData("malformed video sample entry")); + } + #[cfg(not(feature = "mp4v"))] + { + let (_, _) = read_fullbox_extra(&mut b.content)?; + // Subtract 4 extra to offset the members of fullbox not + // accounted for in head.offset + let esds_size = b + .head + .size + .checked_sub(b.head.offset + 4) + .expect("offset invalid"); + let esds = read_buf(&mut b.content, esds_size)?; + codec_specific = Some(VideoCodecSpecific::ESDSConfig(esds)); + } + #[cfg(feature = "mp4v")] + { + // Read ES_Descriptor inside an esds box. + // See ISOBMFF (ISO 14496-1:2010 §7.2.6.5) + let esds = read_esds(&mut b)?; + codec_specific = + Some(VideoCodecSpecific::ESDSConfig(esds.decoder_specific_data)); + } + } + BoxType::ProtectionSchemeInfoBox => { + if name != BoxType::ProtectedVisualSampleEntry { + return Err(Error::InvalidData("malformed video sample entry")); + } + let sinf = read_sinf(&mut b)?; + debug!("{:?} (sinf)", sinf); + protection_info.push(sinf)?; + } + _ => { + debug!("Unsupported video codec, box {:?} found", b.head.name); + skip_box_content(&mut b)?; + } + } + check_parser_state!(b.content); + } + + Ok( + codec_specific.map_or(SampleEntry::Unknown, |codec_specific| { + SampleEntry::Video(VideoSampleEntry { + codec_type, + data_reference_index, + width, + height, + codec_specific, + protection_info, + }) + }), + ) +} + +fn read_qt_wave_atom<T: Read>(src: &mut BMFFBox<T>) -> Result<ES_Descriptor> { + let mut codec_specific = None; + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::ESDBox => { + let esds = read_esds(&mut b)?; + codec_specific = Some(esds); + } + _ => skip_box_content(&mut b)?, + } + } + + codec_specific.ok_or(Error::InvalidData("malformed audio sample entry")) +} + +/// Parse an audio description inside an stsd box. +/// See ISOBMFF (ISO 14496-12:2020) § 12.2.3 +fn read_audio_sample_entry<T: Read>(src: &mut BMFFBox<T>) -> Result<SampleEntry> { + let name = src.get_header().name; + + // Skip uninteresting fields. + skip(src, 6)?; + + let data_reference_index = be_u16(src)?; + + // XXX(kinetik): This is "reserved" in BMFF, but some old QT MOV variant + // uses it, need to work out if we have to support it. Without checking + // here and reading extra fields after samplerate (or bailing with an + // error), the parser loses sync completely. + let version = be_u16(src)?; + + // Skip uninteresting fields. + skip(src, 6)?; + + let mut channelcount = u32::from(be_u16(src)?); + let samplesize = be_u16(src)?; + + // Skip uninteresting fields. + skip(src, 4)?; + + let mut samplerate = f64::from(be_u32(src)? >> 16); // 16.16 fixed point; + + match version { + 0 => (), + 1 => { + // Quicktime sound sample description version 1. + // Skip uninteresting fields. + skip(src, 16)?; + } + 2 => { + // Quicktime sound sample description version 2. + skip(src, 4)?; + samplerate = f64::from_bits(be_u64(src)?); + channelcount = be_u32(src)?; + skip(src, 20)?; + } + _ => { + return Err(Error::Unsupported( + "unsupported non-isom audio sample entry", + )) + } + } + + let (mut codec_type, mut codec_specific) = match name { + BoxType::MP3AudioSampleEntry => (CodecType::MP3, Some(AudioCodecSpecific::MP3)), + BoxType::LPCMAudioSampleEntry => (CodecType::LPCM, Some(AudioCodecSpecific::LPCM)), + // Some mp4 file with AMR doesn't have AMRSpecificBox "damr" in followed while loop, + // we use empty box by default. + #[cfg(feature = "3gpp")] + BoxType::AMRNBSampleEntry => ( + CodecType::AMRNB, + Some(AudioCodecSpecific::AMRSpecificBox(Default::default())), + ), + #[cfg(feature = "3gpp")] + BoxType::AMRWBSampleEntry => ( + CodecType::AMRWB, + Some(AudioCodecSpecific::AMRSpecificBox(Default::default())), + ), + _ => (CodecType::Unknown, None), + }; + let mut protection_info = TryVec::new(); + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::ESDBox => { + if (name != BoxType::MP4AudioSampleEntry + && name != BoxType::ProtectedAudioSampleEntry) + || codec_specific.is_some() + { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let esds = read_esds(&mut b)?; + codec_type = esds.audio_codec; + codec_specific = Some(AudioCodecSpecific::ES_Descriptor(esds)); + } + BoxType::FLACSpecificBox => { + if (name != BoxType::FLACSampleEntry && name != BoxType::ProtectedAudioSampleEntry) + || codec_specific.is_some() + { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let dfla = read_dfla(&mut b)?; + codec_type = CodecType::FLAC; + codec_specific = Some(AudioCodecSpecific::FLACSpecificBox(dfla)); + } + BoxType::OpusSpecificBox => { + if (name != BoxType::OpusSampleEntry && name != BoxType::ProtectedAudioSampleEntry) + || codec_specific.is_some() + { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let dops = read_dops(&mut b)?; + codec_type = CodecType::Opus; + codec_specific = Some(AudioCodecSpecific::OpusSpecificBox(dops)); + } + BoxType::ALACSpecificBox => { + if name != BoxType::ALACSpecificBox || codec_specific.is_some() { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let alac = read_alac(&mut b)?; + codec_type = CodecType::ALAC; + codec_specific = Some(AudioCodecSpecific::ALACSpecificBox(alac)); + } + BoxType::QTWaveAtom => { + let qt_esds = read_qt_wave_atom(&mut b)?; + codec_type = qt_esds.audio_codec; + codec_specific = Some(AudioCodecSpecific::ES_Descriptor(qt_esds)); + } + BoxType::ProtectionSchemeInfoBox => { + if name != BoxType::ProtectedAudioSampleEntry { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let sinf = read_sinf(&mut b)?; + debug!("{:?} (sinf)", sinf); + codec_type = CodecType::EncryptedAudio; + protection_info.push(sinf)?; + } + #[cfg(feature = "3gpp")] + BoxType::AMRSpecificBox => { + if codec_type != CodecType::AMRNB && codec_type != CodecType::AMRWB { + return Err(Error::InvalidData("malformed audio sample entry")); + } + let amr_dec_spec_struc_size = b + .head + .size + .checked_sub(b.head.offset) + .expect("offset invalid"); + let amr_dec_spec_struc = read_buf(&mut b.content, amr_dec_spec_struc_size)?; + debug!("{:?} (AMRDecSpecStruc)", amr_dec_spec_struc); + codec_specific = Some(AudioCodecSpecific::AMRSpecificBox(amr_dec_spec_struc)); + } + _ => { + debug!("Unsupported audio codec, box {:?} found", b.head.name); + skip_box_content(&mut b)?; + } + } + check_parser_state!(b.content); + } + + Ok( + codec_specific.map_or(SampleEntry::Unknown, |codec_specific| { + SampleEntry::Audio(AudioSampleEntry { + codec_type, + data_reference_index, + channelcount, + samplesize, + samplerate, + codec_specific, + protection_info, + }) + }), + ) +} + +/// Parse a stsd box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.5.2 +/// See MP4 (ISO 14496-14:2020) § 6.7.2 +fn read_stsd<T: Read>(src: &mut BMFFBox<T>, track: &mut Track) -> Result<SampleDescriptionBox> { + let (_, _) = read_fullbox_extra(src)?; + + let description_count = be_u32(src)?; + let mut descriptions = TryVec::new(); + + { + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + let description = match track.track_type { + TrackType::Video => read_video_sample_entry(&mut b), + TrackType::Audio => read_audio_sample_entry(&mut b), + TrackType::Metadata => Err(Error::Unsupported("metadata track")), + TrackType::Unknown => Err(Error::Unsupported("unknown track type")), + }; + let description = match description { + Ok(desc) => desc, + Err(Error::Unsupported(_)) => { + // read_{audio,video}_desc may have returned Unsupported + // after partially reading the box content, so we can't + // simply use skip_box_content here. + let to_skip = b.bytes_left(); + skip(&mut b, to_skip)?; + SampleEntry::Unknown + } + Err(e) => return Err(e), + }; + descriptions.push(description)?; + check_parser_state!(b.content); + if descriptions.len() == description_count.to_usize() { + break; + } + } + } + + // Padding could be added in some contents. + skip_box_remain(src)?; + + Ok(SampleDescriptionBox { descriptions }) +} + +fn read_sinf<T: Read>(src: &mut BMFFBox<T>) -> Result<ProtectionSchemeInfoBox> { + let mut sinf = ProtectionSchemeInfoBox::default(); + + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::OriginalFormatBox => { + sinf.original_format = FourCC::from(be_u32(&mut b)?); + } + BoxType::SchemeTypeBox => { + sinf.scheme_type = Some(read_schm(&mut b)?); + } + BoxType::SchemeInformationBox => { + // We only need tenc box in schi box so far. + sinf.tenc = read_schi(&mut b)?; + } + _ => skip_box_content(&mut b)?, + } + check_parser_state!(b.content); + } + + Ok(sinf) +} + +fn read_schi<T: Read>(src: &mut BMFFBox<T>) -> Result<Option<TrackEncryptionBox>> { + let mut tenc = None; + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::TrackEncryptionBox => { + if tenc.is_some() { + return Err(Error::InvalidData( + "tenc box should be only one at most in sinf box", + )); + } + tenc = Some(read_tenc(&mut b)?); + } + _ => skip_box_content(&mut b)?, + } + } + + Ok(tenc) +} + +fn read_tenc<T: Read>(src: &mut BMFFBox<T>) -> Result<TrackEncryptionBox> { + let (version, _) = read_fullbox_extra(src)?; + + // reserved byte + skip(src, 1)?; + // the next byte is used to signal the default pattern in version >= 1 + let (default_crypt_byte_block, default_skip_byte_block) = match version { + 0 => { + skip(src, 1)?; + (None, None) + } + _ => { + let pattern_byte = src.read_u8()?; + let crypt_bytes = pattern_byte >> 4; + let skip_bytes = pattern_byte & 0x0f; + (Some(crypt_bytes), Some(skip_bytes)) + } + }; + let default_is_encrypted = src.read_u8()?; + let default_iv_size = src.read_u8()?; + let default_kid = read_buf(src, 16)?; + // If default_is_encrypted == 1 && default_iv_size == 0 we expect a default_constant_iv + let default_constant_iv = match (default_is_encrypted, default_iv_size) { + (1, 0) => { + let default_constant_iv_size = src.read_u8()?; + Some(read_buf(src, default_constant_iv_size.into())?) + } + _ => None, + }; + + Ok(TrackEncryptionBox { + is_encrypted: default_is_encrypted, + iv_size: default_iv_size, + kid: default_kid, + crypt_byte_block_count: default_crypt_byte_block, + skip_byte_block_count: default_skip_byte_block, + constant_iv: default_constant_iv, + }) +} + +fn read_schm<T: Read>(src: &mut BMFFBox<T>) -> Result<SchemeTypeBox> { + // Flags can be used to signal presence of URI in the box, but we don't + // use the URI so don't bother storing the flags. + let (_, _) = read_fullbox_extra(src)?; + let scheme_type = FourCC::from(be_u32(src)?); + let scheme_version = be_u32(src)?; + // Null terminated scheme URI may follow, but we don't use it right now. + skip_box_remain(src)?; + Ok(SchemeTypeBox { + scheme_type, + scheme_version, + }) +} + +/// Parse a metadata box inside a moov, trak, or mdia box. +/// See ISOBMFF (ISO 14496-12:2020) § 8.10.1. +fn read_udta<T: Read>(src: &mut BMFFBox<T>) -> Result<UserdataBox> { + let mut iter = src.box_iter(); + let mut udta = UserdataBox { meta: None }; + + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MetadataBox => { + let meta = read_meta(&mut b)?; + udta.meta = Some(meta); + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(udta) +} + +/// Parse the meta box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.1 +fn read_meta<T: Read>(src: &mut BMFFBox<T>) -> Result<MetadataBox> { + let (_, _) = read_fullbox_extra(src)?; + let mut iter = src.box_iter(); + let mut meta = MetadataBox::default(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MetadataItemListEntry => read_ilst(&mut b, &mut meta)?, + #[cfg(feature = "meta-xml")] + BoxType::MetadataXMLBox => read_xml_(&mut b, &mut meta)?, + #[cfg(feature = "meta-xml")] + BoxType::MetadataBXMLBox => read_bxml(&mut b, &mut meta)?, + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(meta) +} + +/// Parse a XML box inside a meta box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.2 +#[cfg(feature = "meta-xml")] +fn read_xml_<T: Read>(src: &mut BMFFBox<T>, meta: &mut MetadataBox) -> Result<()> { + if read_fullbox_version_no_flags(src)? != 0 { + return Err(Error::Unsupported("unsupported XmlBox version")); + } + meta.xml = Some(XmlBox::StringXmlBox(src.read_into_try_vec()?)); + Ok(()) +} + +/// Parse a Binary XML box inside a meta box +/// See ISOBMFF (ISO 14496-12:2020) § 8.11.2 +#[cfg(feature = "meta-xml")] +fn read_bxml<T: Read>(src: &mut BMFFBox<T>, meta: &mut MetadataBox) -> Result<()> { + if read_fullbox_version_no_flags(src)? != 0 { + return Err(Error::Unsupported("unsupported XmlBox version")); + } + meta.xml = Some(XmlBox::BinaryXmlBox(src.read_into_try_vec()?)); + Ok(()) +} + +/// Parse a metadata box inside a udta box +fn read_ilst<T: Read>(src: &mut BMFFBox<T>, meta: &mut MetadataBox) -> Result<()> { + let mut iter = src.box_iter(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::AlbumEntry => meta.album = read_ilst_string_data(&mut b)?, + BoxType::ArtistEntry | BoxType::ArtistLowercaseEntry => { + meta.artist = read_ilst_string_data(&mut b)? + } + BoxType::AlbumArtistEntry => meta.album_artist = read_ilst_string_data(&mut b)?, + BoxType::CommentEntry => meta.comment = read_ilst_string_data(&mut b)?, + BoxType::DateEntry => meta.year = read_ilst_string_data(&mut b)?, + BoxType::TitleEntry => meta.title = read_ilst_string_data(&mut b)?, + BoxType::CustomGenreEntry => { + meta.genre = read_ilst_string_data(&mut b)?.map(Genre::CustomGenre) + } + BoxType::StandardGenreEntry => { + meta.genre = read_ilst_u8_data(&mut b)? + .and_then(|gnre| Some(Genre::StandardGenre(gnre.get(1).copied()?))) + } + BoxType::ComposerEntry => meta.composer = read_ilst_string_data(&mut b)?, + BoxType::EncoderEntry => meta.encoder = read_ilst_string_data(&mut b)?, + BoxType::EncodedByEntry => meta.encoded_by = read_ilst_string_data(&mut b)?, + BoxType::CopyrightEntry => meta.copyright = read_ilst_string_data(&mut b)?, + BoxType::GroupingEntry => meta.grouping = read_ilst_string_data(&mut b)?, + BoxType::CategoryEntry => meta.category = read_ilst_string_data(&mut b)?, + BoxType::KeywordEntry => meta.keyword = read_ilst_string_data(&mut b)?, + BoxType::PodcastUrlEntry => meta.podcast_url = read_ilst_string_data(&mut b)?, + BoxType::PodcastGuidEntry => meta.podcast_guid = read_ilst_string_data(&mut b)?, + BoxType::DescriptionEntry => meta.description = read_ilst_string_data(&mut b)?, + BoxType::LongDescriptionEntry => meta.long_description = read_ilst_string_data(&mut b)?, + BoxType::LyricsEntry => meta.lyrics = read_ilst_string_data(&mut b)?, + BoxType::TVNetworkNameEntry => meta.tv_network_name = read_ilst_string_data(&mut b)?, + BoxType::TVEpisodeNameEntry => meta.tv_episode_name = read_ilst_string_data(&mut b)?, + BoxType::TVShowNameEntry => meta.tv_show_name = read_ilst_string_data(&mut b)?, + BoxType::PurchaseDateEntry => meta.purchase_date = read_ilst_string_data(&mut b)?, + BoxType::RatingEntry => meta.rating = read_ilst_string_data(&mut b)?, + BoxType::OwnerEntry => meta.owner = read_ilst_string_data(&mut b)?, + BoxType::HDVideoEntry => meta.hd_video = read_ilst_bool_data(&mut b)?, + BoxType::SortNameEntry => meta.sort_name = read_ilst_string_data(&mut b)?, + BoxType::SortArtistEntry => meta.sort_artist = read_ilst_string_data(&mut b)?, + BoxType::SortAlbumEntry => meta.sort_album = read_ilst_string_data(&mut b)?, + BoxType::SortAlbumArtistEntry => { + meta.sort_album_artist = read_ilst_string_data(&mut b)? + } + BoxType::SortComposerEntry => meta.sort_composer = read_ilst_string_data(&mut b)?, + BoxType::TrackNumberEntry => { + if let Some(trkn) = read_ilst_u8_data(&mut b)? { + meta.track_number = trkn.get(3).copied(); + meta.total_tracks = trkn.get(5).copied(); + }; + } + BoxType::DiskNumberEntry => { + if let Some(disk) = read_ilst_u8_data(&mut b)? { + meta.disc_number = disk.get(3).copied(); + meta.total_discs = disk.get(5).copied(); + }; + } + BoxType::TempoEntry => { + meta.beats_per_minute = + read_ilst_u8_data(&mut b)?.and_then(|tmpo| tmpo.get(1).copied()) + } + BoxType::CompilationEntry => meta.compilation = read_ilst_bool_data(&mut b)?, + BoxType::AdvisoryEntry => { + meta.advisory = read_ilst_u8_data(&mut b)?.and_then(|rtng| { + Some(match rtng.get(0)? { + 2 => AdvisoryRating::Clean, + 0 => AdvisoryRating::Inoffensive, + r => AdvisoryRating::Explicit(*r), + }) + }) + } + BoxType::MediaTypeEntry => { + meta.media_type = read_ilst_u8_data(&mut b)?.and_then(|stik| { + Some(match stik.get(0)? { + 0 => MediaType::Movie, + 1 => MediaType::Normal, + 2 => MediaType::AudioBook, + 5 => MediaType::WhackedBookmark, + 6 => MediaType::MusicVideo, + 9 => MediaType::ShortFilm, + 10 => MediaType::TVShow, + 11 => MediaType::Booklet, + s => MediaType::Unknown(*s), + }) + }) + } + BoxType::PodcastEntry => meta.podcast = read_ilst_bool_data(&mut b)?, + BoxType::TVSeasonNumberEntry => { + meta.tv_season = read_ilst_u8_data(&mut b)?.and_then(|tvsn| tvsn.get(3).copied()) + } + BoxType::TVEpisodeNumberEntry => { + meta.tv_episode_number = + read_ilst_u8_data(&mut b)?.and_then(|tves| tves.get(3).copied()) + } + BoxType::GaplessPlaybackEntry => meta.gapless_playback = read_ilst_bool_data(&mut b)?, + BoxType::CoverArtEntry => meta.cover_art = read_ilst_multiple_u8_data(&mut b).ok(), + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(()) +} + +fn read_ilst_bool_data<T: Read>(src: &mut BMFFBox<T>) -> Result<Option<bool>> { + Ok(read_ilst_u8_data(src)?.and_then(|d| Some(d.get(0)? == &1))) +} + +fn read_ilst_string_data<T: Read>(src: &mut BMFFBox<T>) -> Result<Option<TryString>> { + read_ilst_u8_data(src) +} + +fn read_ilst_u8_data<T: Read>(src: &mut BMFFBox<T>) -> Result<Option<TryVec<u8>>> { + // For all non-covr atoms, there must only be one data atom. + Ok(read_ilst_multiple_u8_data(src)?.pop()) +} + +fn read_ilst_multiple_u8_data<T: Read>(src: &mut BMFFBox<T>) -> Result<TryVec<TryVec<u8>>> { + let mut iter = src.box_iter(); + let mut data = TryVec::new(); + while let Some(mut b) = iter.next_box()? { + match b.head.name { + BoxType::MetadataItemDataEntry => { + data.push(read_ilst_data(&mut b)?)?; + } + _ => skip_box_content(&mut b)?, + }; + check_parser_state!(b.content); + } + Ok(data) +} + +fn read_ilst_data<T: Read>(src: &mut BMFFBox<T>) -> Result<TryVec<u8>> { + // Skip past the padding bytes + skip(&mut src.content, src.head.offset)?; + let size = src.content.limit(); + read_buf(&mut src.content, size) +} + +/// Skip a number of bytes that we don't care to parse. +fn skip<T: Read>(src: &mut T, bytes: u64) -> Result<()> { + std::io::copy(&mut src.take(bytes), &mut std::io::sink())?; + Ok(()) +} + +/// Read size bytes into a Vector or return error. +fn read_buf<T: Read>(src: &mut T, size: u64) -> Result<TryVec<u8>> { + let buf = src.take(size).read_into_try_vec()?; + if buf.len().to_u64() != size { + return Err(Error::InvalidData("failed buffer read")); + } + + Ok(buf) +} + +fn be_i16<T: ReadBytesExt>(src: &mut T) -> Result<i16> { + src.read_i16::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_i32<T: ReadBytesExt>(src: &mut T) -> Result<i32> { + src.read_i32::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_i64<T: ReadBytesExt>(src: &mut T) -> Result<i64> { + src.read_i64::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_u16<T: ReadBytesExt>(src: &mut T) -> Result<u16> { + src.read_u16::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_u24<T: ReadBytesExt>(src: &mut T) -> Result<u32> { + src.read_u24::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_u32<T: ReadBytesExt>(src: &mut T) -> Result<u32> { + src.read_u32::<byteorder::BigEndian>().map_err(From::from) +} + +fn be_u64<T: ReadBytesExt>(src: &mut T) -> Result<u64> { + src.read_u64::<byteorder::BigEndian>().map_err(From::from) +} + +fn write_be_u32<T: WriteBytesExt>(des: &mut T, num: u32) -> Result<()> { + des.write_u32::<byteorder::BigEndian>(num) + .map_err(From::from) +} |