Revision control
Copy as Markdown
Other Tools
#![allow(clippy::len_without_is_empty)]
use core::convert::TryInto;
use core::ops::Range;
use core::{mem, result};
use crate::pod::{from_bytes, slice_from_bytes, Pod};
type Result<T> = result::Result<T, ()>;
/// A trait for reading references to [`Pod`] types from a block of data.
///
/// This allows parsers to handle both of these cases:
/// - the block of data exists in memory, and it is desirable
/// to use references to this block instead of copying it,
/// - the block of data exists in storage, and it is desirable
/// to read on demand to minimize I/O and memory usage.
///
/// A block of data typically exists in memory as a result of using a memory
/// mapped file, and the crate was written with this use case in mind.
/// Reading the entire file into a `Vec` is also possible, but it often uses
/// more I/O and memory.
/// Both of these are handled by the `ReadRef` implementation for `&[u8]`.
///
/// For the second use case, the `ReadRef` trait is implemented for
/// [`&ReadCache`](super::ReadCache). This is useful for environments where
/// memory mapped files are not available or not suitable, such as WebAssembly.
/// This differs from reading into a `Vec` in that it only reads the portions
/// of the file that are needed for parsing.
///
/// The methods accept `self` by value because `Self` is expected to behave
/// similar to a reference: it may be a reference with a lifetime of `'a`,
/// or it may be a wrapper of a reference.
///
/// The `Clone` and `Copy` bounds are for convenience, and since `Self` is
/// expected to be similar to a reference, these are easily satisfied.
///
/// Object file parsers typically use offsets to locate the structures
/// in the block, and will most commonly use the `*_at` methods to
/// read a structure at a known offset.
///
/// Occasionally file parsers will need to treat the block as a stream,
/// and so convenience methods are provided that update an offset with
/// the size that was read.
//
// An alternative would be for methods to accept `&mut self` and use a
// `seek` method instead of the `offset` parameters, but this is less
// convenient for implementers.
pub trait ReadRef<'a>: Clone + Copy {
/// The total size of the block of data.
fn len(self) -> Result<u64>;
/// Get a reference to a `u8` slice at the given offset.
///
/// Returns an error if offset or size are out of bounds.
fn read_bytes_at(self, offset: u64, size: u64) -> Result<&'a [u8]>;
/// Get a reference to a delimited `u8` slice which starts at range.start.
///
/// Does not include the delimiter.
///
/// Returns an error if the range is out of bounds or the delimiter is
/// not found in the range.
fn read_bytes_at_until(self, range: Range<u64>, delimiter: u8) -> Result<&'a [u8]>;
/// Get a reference to a `u8` slice at the given offset, and update the offset.
///
/// Returns an error if offset or size are out of bounds.
fn read_bytes(self, offset: &mut u64, size: u64) -> Result<&'a [u8]> {
let bytes = self.read_bytes_at(*offset, size)?;
*offset = offset.wrapping_add(size);
Ok(bytes)
}
/// Get a reference to a `Pod` type at the given offset, and update the offset.
///
/// Returns an error if offset or size are out of bounds.
///
/// The default implementation uses `read_bytes`, and returns an error if
/// `read_bytes` does not return bytes with the correct alignment for `T`.
/// Implementors may want to provide their own implementation that ensures
/// the alignment can be satisfied. Alternatively, only use this method with
/// types that do not need alignment (see the `unaligned` feature of this crate).
fn read<T: Pod>(self, offset: &mut u64) -> Result<&'a T> {
let size = mem::size_of::<T>().try_into().map_err(|_| ())?;
let bytes = self.read_bytes(offset, size)?;
let (t, _) = from_bytes(bytes)?;
Ok(t)
}
/// Get a reference to a `Pod` type at the given offset.
///
/// Returns an error if offset or size are out of bounds.
///
/// Also see the `read` method for information regarding alignment of `T`.
fn read_at<T: Pod>(self, mut offset: u64) -> Result<&'a T> {
self.read(&mut offset)
}
/// Get a reference to a slice of a `Pod` type at the given offset, and update the offset.
///
/// Returns an error if offset or size are out of bounds.
///
/// Also see the `read` method for information regarding alignment of `T`.
fn read_slice<T: Pod>(self, offset: &mut u64, count: usize) -> Result<&'a [T]> {
let size = count
.checked_mul(mem::size_of::<T>())
.ok_or(())?
.try_into()
.map_err(|_| ())?;
let bytes = self.read_bytes(offset, size)?;
let (t, _) = slice_from_bytes(bytes, count)?;
Ok(t)
}
/// Get a reference to a slice of a `Pod` type at the given offset.
///
/// Returns an error if offset or size are out of bounds.
///
/// Also see the `read` method for information regarding alignment of `T`.
fn read_slice_at<T: Pod>(self, mut offset: u64, count: usize) -> Result<&'a [T]> {
self.read_slice(&mut offset, count)
}
}
impl<'a> ReadRef<'a> for &'a [u8] {
fn len(self) -> Result<u64> {
self.len().try_into().map_err(|_| ())
}
fn read_bytes_at(self, offset: u64, size: u64) -> Result<&'a [u8]> {
let offset: usize = offset.try_into().map_err(|_| ())?;
let size: usize = size.try_into().map_err(|_| ())?;
self.get(offset..).ok_or(())?.get(..size).ok_or(())
}
fn read_bytes_at_until(self, range: Range<u64>, delimiter: u8) -> Result<&'a [u8]> {
let start: usize = range.start.try_into().map_err(|_| ())?;
let end: usize = range.end.try_into().map_err(|_| ())?;
let bytes = self.get(start..end).ok_or(())?;
match memchr::memchr(delimiter, bytes) {
Some(len) => {
// This will never fail.
bytes.get(..len).ok_or(())
}
None => Err(()),
}
}
}