//! a pack data fileuse crate::MMap;use std::path::Path; /// The offset to an entry into the pack data file, relative to its beginning.pub type Offset = u64; /// An identifier to uniquely identify all packs loaded within a known context or namespace.pub type Id = u32; /// An representing an full- or delta-object within a pack#[derive(PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone)]#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]pub struct Entry { /// The entry's header pub header: entry::Header, /// The decompressed size of the entry in bytes. /// /// Note that for non-delta entries this will be the size of the object itself. pub decompressed_size: u64, /// absolute offset to compressed object data in the pack, just behind the entry's header pub data_offset: Offset, /// The amount of bytes used to encode the entry header in the pack. /// /// Git accepts non-canonical size encodings in pack entry headers. These are encodings that /// continue the variable-length size field even though the remaining size bits are zero, for /// example `b3 00` for a blob of size `3` instead of the canonical single-byte `33`. /// /// Git does not usually write such headers itself, but they can appear when pack data is reused /// verbatim, for example when a server constructs fetch or clone responses from already stored /// pack entries. Readers have to preserve the byte length that was actually consumed: recomputing /// the canonical length from the decoded size would shift [`Entry::pack_offset()`] backwards and /// make ofs-delta base-offset calculations refer to the wrong entry. /// /// A value of `0` means the actual length is unknown, causing [`Entry::header_size()`] to /// recompute the canonical length from the decoded size. This is useful for deserializing data /// produced by older versions. #[cfg_attr(feature = "serde", serde(default))] pub encoded_header_size: u16,} mod file;pub use file::{Header, decode, verify};///pub mod header; ///pub mod init { pub use super::header::decode::Error;} ///pub mod entry; ///#[cfg(feature = "streaming-input")]pub mod input; /// Utilities to encode pack data entries and write them to a `Write` implementation to resemble a pack data file.#[cfg(feature = "generate")]pub mod output; /// A slice into a pack file denoting a pack entry.////// An entry can be decoded into an object.pub type EntryRange = std::ops::Range<Offset>; /// Supported versions of a pack data file#[derive(Default, PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone, Copy)]#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]pub enum Version { /// The default pack data version. /// /// This is the version generated by Git and by `gix-pack` writers. #[default] V2, /// A pack data version accepted by Git and recognized by `gix-pack` readers. /// /// Git does not generate this version, and `gix-pack` writers currently reject it. /// Entries are decoded with the same layout as [`V2`](Version::V2); the difference /// visible to this crate is the version number stored in the pack header. V3,} /// A pack data filepub struct File<T = MMap> { data: T, path: std::path::PathBuf, /// A value to represent this pack uniquely when used with cache lookup, or a way to identify this pack by its location on disk. /// The same location on disk should yield the same id. /// /// These must be unique per pack and must be stable, that is they don't change if the pack doesn't change. /// If the same id is assigned (or reassigned) to different packs, pack creation or cache access will fail in hard-to-debug ways. /// /// This value is controlled by the owning object store, which can use it in whichever way it wants as long as the above constraints are met. pub id: Id, version: Version, num_objects: u32, /// The kind of hash contained within. This is entirely determined by the caller, and repositories have to know which hash to use /// based on their configuration. object_hash: gix_hash::Kind, /// The maximum size of a single allocation caused by user-controlled on-disk pack data. /// /// If `None`, no additional limit is enforced. pub alloc_limit_bytes: Option<usize>,} /// Information about the pack data file itselfimpl<T> File<T>where T: crate::FileData,{ /// The pack data version of this file pub fn version(&self) -> Version { self.version } /// The number of objects stored in this pack data file pub fn num_objects(&self) -> u32 { self.num_objects } /// The length of all mapped data, including the pack header and the pack trailer pub fn data_len(&self) -> usize { self.data.len() } /// The kind of hash we use internally. pub fn object_hash(&self) -> gix_hash::Kind { self.object_hash } /// The position of the byte one past the last pack entry, or in other terms, the first byte of the trailing hash. pub fn pack_end(&self) -> usize { self.data.len() - self.object_hash.len_in_bytes() } /// The path to the pack data file on disk pub fn path(&self) -> &Path { &self.path } /// Returns the pack data at the given slice if its range is contained in the mapped pack data pub fn entry_slice(&self, slice: EntryRange) -> Option<&[u8]> { let entry_end: usize = slice.end.try_into().expect("end of pack fits into usize"); let entry_start = slice.start as usize; self.data.get(entry_start..entry_end) } /// Returns the CRC32 of the pack data indicated by `pack_offset` and the `size` of the mapped data. /// /// _Note:_ finding the right size is only possible by decompressing /// the pack entry beforehand, or by using the (to be sorted) offsets stored in an index file. /// /// # Panics /// /// If `pack_offset` or `size` are pointing to a range outside of the mapped pack data. pub fn entry_crc32(&self, pack_offset: Offset, size: usize) -> u32 { let pack_offset: usize = pack_offset.try_into().expect("pack_size fits into usize"); gix_features::hash::crc32(&self.data[pack_offset..pack_offset + size]) }} ///pub mod delta;
//! a pack data file
use crate::MMap;
use std::path::Path;
/// The offset to an entry into the pack data file, relative to its beginning.
pub type Offset = u64;
/// An identifier to uniquely identify all packs loaded within a known context or namespace.
pub type Id = u32;
/// An representing an full- or delta-object within a pack
#[derive(PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone)]
#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
pub struct Entry {
/// The entry's header
pub header: entry::Header,
/// The decompressed size of the entry in bytes.
///
/// Note that for non-delta entries this will be the size of the object itself.
pub decompressed_size: u64,
/// absolute offset to compressed object data in the pack, just behind the entry's header
pub data_offset: Offset,
/// The amount of bytes used to encode the entry header in the pack.
///
/// Git accepts non-canonical size encodings in pack entry headers. These are encodings that
/// continue the variable-length size field even though the remaining size bits are zero, for
/// example `b3 00` for a blob of size `3` instead of the canonical single-byte `33`.
///
/// Git does not usually write such headers itself, but they can appear when pack data is reused
/// verbatim, for example when a server constructs fetch or clone responses from already stored
/// pack entries. Readers have to preserve the byte length that was actually consumed: recomputing
/// the canonical length from the decoded size would shift [`Entry::pack_offset()`] backwards and
/// make ofs-delta base-offset calculations refer to the wrong entry.
///
/// A value of `0` means the actual length is unknown, causing [`Entry::header_size()`] to
/// recompute the canonical length from the decoded size. This is useful for deserializing data
/// produced by older versions.
#[cfg_attr(feature = "serde", serde(default))]
pub encoded_header_size: u16,
}
mod file;
pub use file::{Header, decode, verify};
///
pub mod header;
///
pub mod init {
pub use super::header::decode::Error;
}
///
pub mod entry;
///
#[cfg(feature = "streaming-input")]
pub mod input;
/// Utilities to encode pack data entries and write them to a `Write` implementation to resemble a pack data file.
#[cfg(feature = "generate")]
pub mod output;
/// A slice into a pack file denoting a pack entry.
///
/// An entry can be decoded into an object.
pub type EntryRange = std::ops::Range<Offset>;
/// Supported versions of a pack data file
#[derive(Default, PartialEq, Eq, Debug, Hash, Ord, PartialOrd, Clone, Copy)]
#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
pub enum Version {
/// The default pack data version.
///
/// This is the version generated by Git and by `gix-pack` writers.
#[default]
V2,
/// A pack data version accepted by Git and recognized by `gix-pack` readers.
///
/// Git does not generate this version, and `gix-pack` writers currently reject it.
/// Entries are decoded with the same layout as [`V2`](Version::V2); the difference
/// visible to this crate is the version number stored in the pack header.
V3,
}
/// A pack data file
pub struct File<T = MMap> {
data: T,
path: std::path::PathBuf,
/// A value to represent this pack uniquely when used with cache lookup, or a way to identify this pack by its location on disk.
/// The same location on disk should yield the same id.
///
/// These must be unique per pack and must be stable, that is they don't change if the pack doesn't change.
/// If the same id is assigned (or reassigned) to different packs, pack creation or cache access will fail in hard-to-debug ways.
///
/// This value is controlled by the owning object store, which can use it in whichever way it wants as long as the above constraints are met.
pub id: Id,
version: Version,
num_objects: u32,
/// The kind of hash contained within. This is entirely determined by the caller, and repositories have to know which hash to use
/// based on their configuration.
object_hash: gix_hash::Kind,
/// The maximum size of a single allocation caused by user-controlled on-disk pack data.
///
/// If `None`, no additional limit is enforced.
pub alloc_limit_bytes: Option<usize>,
}
/// Information about the pack data file itself
impl<T> File<T>
where
T: crate::FileData,
{
/// The pack data version of this file
pub fn version(&self) -> Version {
self.version
}
/// The number of objects stored in this pack data file
pub fn num_objects(&self) -> u32 {
self.num_objects
}
/// The length of all mapped data, including the pack header and the pack trailer
pub fn data_len(&self) -> usize {
self.data.len()
}
/// The kind of hash we use internally.
pub fn object_hash(&self) -> gix_hash::Kind {
self.object_hash
}
/// The position of the byte one past the last pack entry, or in other terms, the first byte of the trailing hash.
pub fn pack_end(&self) -> usize {
self.data.len() - self.object_hash.len_in_bytes()
}
/// The path to the pack data file on disk
pub fn path(&self) -> &Path {
&self.path
}
/// Returns the pack data at the given slice if its range is contained in the mapped pack data
pub fn entry_slice(&self, slice: EntryRange) -> Option<&[u8]> {
let entry_end: usize = slice.end.try_into().expect("end of pack fits into usize");
let entry_start = slice.start as usize;
self.data.get(entry_start..entry_end)
}
/// Returns the CRC32 of the pack data indicated by `pack_offset` and the `size` of the mapped data.
///
/// _Note:_ finding the right size is only possible by decompressing
/// the pack entry beforehand, or by using the (to be sorted) offsets stored in an index file.
///
/// # Panics
///
/// If `pack_offset` or `size` are pointing to a range outside of the mapped pack data.
pub fn entry_crc32(&self, pack_offset: Offset, size: usize) -> u32 {
let pack_offset: usize = pack_offset.try_into().expect("pack_size fits into usize");
gix_features::hash::crc32(&self.data[pack_offset..pack_offset + size])
}
}
///
pub mod delta;