bitreader/src/lib.rs

// Copyright 2015 Ilkka Rauta
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

//! BitReader is a helper type to extract strings of bits from a slice of bytes.
//!
//! Here is how you read first a single bit, then three bits and finally four bits from a byte
//! buffer:
//!
//! ```
//! use bitreader::BitReader;
//!
//! let slice_of_u8 = &[0b1000_1111];
//! let mut reader = BitReader::new(slice_of_u8);
//!
//! // You probably should use try! or some other error handling mechanism in real code if the
//! // length of the input is not known in advance.
//! let a_single_bit = reader.read_u8(1).unwrap();
//! assert_eq!(a_single_bit, 1);
//!
//! let more_bits = reader.read_u8(3).unwrap();
//! assert_eq!(more_bits, 0);
//!
//! let last_bits_of_byte = reader.read_u8(4).unwrap();
//! assert_eq!(last_bits_of_byte, 0b1111);
//! ```
//! You can naturally read bits from longer buffer of data than just a single byte.
//!
//! As you read bits, the internal cursor of BitReader moves on along the stream of bits. Big
//! endian format is assumed when reading the multi-byte values. BitReader supports reading maximum
//! of 64 bits at a time (with read_u64). Reading signed values directly is not supported at the
//! moment.
//!
//! The reads do not need to be aligned in any particular way.
//!
//! Reading zero bits is a no-op.
//!
//! You can also skip over a number of bits, in which case there is no arbitrary small limits like
//! when reading the values to a variable. However, you can not seek past the end of the slice,
//! either when reading or when skipping bits.
//!
//! Note that the code will likely not work correctly if the slice is longer than 2^61 bytes, but
//! exceeding that should be pretty unlikely. Let's get back to this when people read exabytes of
//! information one bit at a time.
#![no_std]
cfg_if::cfg_if!{
    if #[cfg(feature = "std")] {
        extern crate std;
        use std::cmp::min;
        use std::prelude::v1::*;
        use std::fmt;
        use std::error::Error;
        use std::result;
    } else {
        use core::result;
        use core::fmt;
        use core::cmp::min;
    }
}

#[cfg(test)]
mod tests;

/// BitReader reads data from a byte slice at the granularity of a single bit.
pub struct BitReader<'a> {
    bytes: &'a [u8],
    /// Position from the start of the slice, counted as bits instead of bytes
    position: u64,
    relative_offset: u64,

    /// Length this reader is allowed to read from the slice, counted as bits instead of bytes.
    length: u64,
}

impl<'a> BitReader<'a> {
    /// Construct a new BitReader from a byte slice. The returned reader lives at most as long as
    /// the slice given to is valid.
    #[inline]
    pub fn new(bytes: &'a [u8]) -> BitReader<'a> {
        BitReader {
            bytes: bytes,
            position: 0,
            relative_offset: 0,
            length: bytes.len() as u64 * 8,
        }
    }

    /// Returns a copy of current BitReader, with the difference that its position() returns
    /// positions relative to the position of the original BitReader at the construction time.
    /// After construction, both readers are otherwise completely independent, except of course
    /// for sharing the same source data.
    ///
    /// ```
    /// use bitreader::BitReader;
    ///
    /// let bytes = &[0b11110000, 0b00001111];
    /// let mut original = BitReader::new(bytes);
    /// assert_eq!(original.read_u8(4).unwrap(), 0b1111);
    /// assert_eq!(original.position(), 4);
    ///
    /// let mut relative = original.relative_reader();
    /// assert_eq!(relative.position(), 0);
    ///
    /// assert_eq!(original.read_u8(8).unwrap(), 0);
    /// assert_eq!(relative.read_u8(8).unwrap(), 0);
    ///
    /// assert_eq!(original.position(), 12);
    /// assert_eq!(relative.position(), 8);
    /// ```
    #[inline]
    pub fn relative_reader(&self) -> BitReader<'a> {
        BitReader {
            bytes: self.bytes,
            position: self.position,
            relative_offset: self.position,
            length: self.length - self.position(),
        }
    }

    /// Returns a copy of current BitReader, with the difference that its position() returns
    /// positions relative to the position of the original BitReader at the construction time, and
    /// will not allow reading more than len bits. After construction, both readers are otherwise
    // completely independent, except of course for sharing the same source data.
    ///
    /// ```
    /// use bitreader::BitReader;
    /// use bitreader::BitReaderError;
    ///
    /// let bytes = &[0b11110000, 0b00001111];
    /// let mut original = BitReader::new(bytes);
    /// assert_eq!(original.read_u8(4).unwrap(), 0b1111);
    /// assert_eq!(original.position(), 4);
    ///
    /// let mut relative = original.relative_reader_atmost(8);
    /// assert_eq!(relative.position(), 0);
    ///
    /// assert_eq!(original.read_u8(8).unwrap(), 0);
    /// assert_eq!(relative.read_u8(8).unwrap(), 0);
    ///
    /// assert_eq!(original.position(), 12);
    /// assert_eq!(relative.position(), 8);
    ///
    /// assert_eq!(relative.read_u8(8).unwrap_err(), BitReaderError::NotEnoughData{
    ///    position: 8,
    ///    length: 8,
    ///    requested: 8
    /// });
    /// ```
    #[inline]
    pub fn relative_reader_atmost(&self, len: u64) -> BitReader<'a> {
        BitReader {
            bytes: self.bytes,
            position: self.position,
            relative_offset: self.position,
            length: min(self.length - self.position(), len),
        }
    }

    /// Read at most 8 bits into a u8.
    #[inline]
    pub fn read_u8(&mut self, bit_count: u8) -> Result<u8> {
        let value = self.read_value(bit_count, 8)?;
        Ok((value & 0xff) as u8)
    }

    /// Read at most 8 bits into a u8, but without moving the cursor forward.
    #[inline]
    pub fn peek_u8(&self, bit_count: u8) -> Result<u8> {
        self.relative_reader().read_u8(bit_count)
    }

    /// Fills the entire `output_bytes` slice. If there aren't enough bits remaining
    /// after the internal cursor's current position, the cursor won't be moved forward
    /// and the contents of `output_bytes` won't be modified.
    pub fn read_u8_slice(&mut self, output_bytes: &mut [u8]) -> Result<()> {
        let requested = output_bytes.len() as u64 * 8;
        if requested > self.remaining() {
            Err(BitReaderError::NotEnoughData {
                position: self.position(),
                length: self.length,
                requested,
            })
        } else {
            for byte in output_bytes.iter_mut() {
                *byte = self.read_u8(8)?;
            }
            Ok(())
        }
    }

    /// Read at most 16 bits into a u16.
    #[inline]
    pub fn read_u16(&mut self, bit_count: u8) -> Result<u16> {
        let value = self.read_value(bit_count, 16)?;
        Ok((value & 0xffff) as u16)
    }

    /// Read at most 16 bits into a u16, but without moving the cursor forward.
    #[inline]
    pub fn peek_u16(&self, bit_count: u8) -> Result<u16> {
        self.relative_reader().read_u16(bit_count)
    }

    /// Read at most 32 bits into a u32.
    #[inline]
    pub fn read_u32(&mut self, bit_count: u8) -> Result<u32> {
        let value = self.read_value(bit_count, 32)?;
        Ok((value & 0xffffffff) as u32)
    }

    /// Read at most 32 bits into a u32, but without moving the cursor forward.
    #[inline]
    pub fn peek_u32(&self, bit_count: u8) -> Result<u32> {
        self.relative_reader().read_u32(bit_count)
    }

    /// Read at most 64 bits into a u64.
    #[inline]
    pub fn read_u64(&mut self, bit_count: u8) -> Result<u64> {
        let value = self.read_value(bit_count, 64)?;
        Ok(value)
    }

    /// Read at most 64 bits into a u64, but without moving the cursor forward.
    #[inline]
    pub fn peek_u64(&self, bit_count: u8) -> Result<u64> {
        self.relative_reader().read_u64(bit_count)
    }

    /// Read at most 8 bits into a i8.
    /// Assumes the bits are stored in two's complement format.
    #[inline]
    pub fn read_i8(&mut self, bit_count: u8) -> Result<i8> {
        let value = self.read_signed_value(bit_count, 8)?;
        Ok((value & 0xff) as i8)
    }

    /// Read at most 16 bits into a i16.
    /// Assumes the bits are stored in two's complement format.
    #[inline]
    pub fn read_i16(&mut self, bit_count: u8) -> Result<i16> {
        let value = self.read_signed_value(bit_count, 16)?;
        Ok((value & 0xffff) as i16)
    }

    /// Read at most 32 bits into a i32.
    /// Assumes the bits are stored in two's complement format.
    #[inline]
    pub fn read_i32(&mut self, bit_count: u8) -> Result<i32> {
        let value = self.read_signed_value(bit_count, 32)?;
        Ok((value & 0xffffffff) as i32)
    }

    /// Read at most 64 bits into a i64.
    /// Assumes the bits are stored in two's complement format.
    #[inline]
    pub fn read_i64(&mut self, bit_count: u8) -> Result<i64> {
        let value = self.read_signed_value(bit_count, 64)?;
        Ok(value)
    }

    /// Read a single bit as a boolean value.
    /// Interprets 1 as true and 0 as false.
    #[inline]
    pub fn read_bool(&mut self) -> Result<bool> {
        match self.read_value(1, 1)? {
            0 => Ok(false),
            _ => Ok(true),
        }
    }

    /// Read a single bit as a boolean value, but without moving the cursor forward.
    /// Interprets 1 as true and 0 as false.
    #[inline]
    pub fn peek_bool(&self) -> Result<bool> {
        self.relative_reader().read_bool()
    }

    /// Skip arbitrary number of bits. However, you can skip at most to the end of the byte slice.
    pub fn skip(&mut self, bit_count: u64) -> Result<()> {
        let end_position = self.position + bit_count;
        if end_position > (self.relative_offset + self.length) {
            return Err(BitReaderError::NotEnoughData {
                position: self.position(),
                length: self.length,
                requested: bit_count,
            });
        }
        self.position = end_position;
        Ok(())
    }

    /// Returns the position of the cursor, or how many bits have been read so far.
    #[inline]
    pub fn position(&self) -> u64 {
        self.position - self.relative_offset
    }

    /// Returns the number of bits not yet read from the underlying slice.
    #[inline]
    pub fn remaining(&self) -> u64 {
        self.length - self.position()
    }

    /// Helper to make sure the "bit cursor" is exactly at the beginning of a byte, or at specific
    /// multi-byte alignment position.
    ///
    /// For example `reader.is_aligned(1)` returns true if exactly n bytes, or n * 8 bits, has been
    /// read. Similarly, `reader.is_aligned(4)` returns true if exactly n * 32 bits, or n 4-byte
    /// sequences has been read.
    ///
    /// This function can be used to validate the data is being read properly, for example by
    /// adding invocations wrapped into `debug_assert!()` to places where it is known the data
    /// should be n-byte aligned.
    #[inline]
    pub fn is_aligned(&self, alignment_bytes: u32) -> bool {
        self.position % (alignment_bytes as u64 * 8) == 0
    }

    /// Helper to move the "bit cursor" to exactly the beginning of a byte, or to a specific
    /// multi-byte alignment position.
    ///
    /// That is, `reader.align(n)` moves the cursor to the next position that
    /// is a multiple of n * 8 bits, if it's not correctly aligned already.
    pub fn align(&mut self, alignment_bytes: u32) -> Result<()> {
        let alignment_bits = alignment_bytes as u64 * 8;
        let cur_alignment = self.position % alignment_bits;
        let bits_to_skip = (alignment_bits - cur_alignment) % alignment_bits;
        self.skip(bits_to_skip)
    }

    #[inline]
    fn read_signed_value(&mut self, bit_count: u8, maximum_count: u8) -> Result<i64> {
        if bit_count == 0 {
            return Ok(0);
        }
        let unsigned = self.read_value(bit_count, maximum_count)?;
        // Fill the bits above the requested bits with all ones or all zeros,
        // depending on the sign bit.
        let sign_bit = unsigned >> (bit_count - 1) & 1;
        let high_bits = if sign_bit == 1 { -1 } else { 0 };
        if bit_count == 64 {
            // Avoid left-shift-with-overflow exception
            return Ok(unsigned as i64);
        }
        Ok(high_bits << bit_count | unsigned as i64)
    }

    #[cold]
    fn too_many_bits_for_type_error(&self, bit_count: u8, maximum_count: u8) -> Result<u64> {
        Err(BitReaderError::TooManyBitsForType {
            position: self.position,
            requested: bit_count,
            allowed: maximum_count,
        })
    }

    #[inline]
    // Inlining allows the size checks to be eliminated for constant values
    fn read_value(&mut self, bit_count: u8, maximum_count: u8) -> Result<u64> {
        if bit_count == 0 {
            return Ok(0);
        }
        if bit_count > maximum_count {
            return self.too_many_bits_for_type_error(bit_count, maximum_count);
        }
        self.read_bits(bit_count)
    }

    fn read_bits(&mut self, bit_count: u8) -> Result<u64> {
        let start_position = self.position;
        let end_position = self.position + bit_count as u64;
        if end_position > (self.relative_offset + self.length) {
            return Err(BitReaderError::NotEnoughData {
                position: self.position(),
                length: self.length,
                requested: bit_count as u64,
            });
        }

        let mut value: u64 = 0;

        for i in start_position..end_position {
            let byte_index = (i / 8) as usize;
            // This will never fail, but prevents Rust from generating panic code
            let byte = if let Some(byte) = self.bytes.get(byte_index).copied() {
                byte
            } else {
                break;
            };
            let shift = 7 - (i % 8);
            let bit = (byte >> shift) as u64 & 1;
            value = (value << 1) | bit;
        }

        self.position = end_position;
        Ok(value)
    }
}

/// Result type for those BitReader operations that can fail.
pub type Result<T> = result::Result<T, BitReaderError>;

/// Error enumeration of BitReader errors.
#[derive(Debug,PartialEq,Copy,Clone)]
pub enum BitReaderError {
    /// Requested more bits than there are left in the byte slice at the current position.
    NotEnoughData {
        /// Current posititon in bits relative to the beginning of the reader.
        position: u64,

        /// Total readable length in bits of the underlaying slice.
        length: u64,

        /// Bits requested to be read.
        requested: u64,
    },
    /// Requested more bits than the returned variable can hold, for example more than 8 bits when
    /// reading into a u8.
    TooManyBitsForType {
        position: u64,
        requested: u8,
        allowed: u8,
    }
}

#[cfg(feature = "std")]
impl Error for BitReaderError {
    fn description(&self) -> &str {
        match *self {
            BitReaderError::NotEnoughData {..} => "Requested more bits than the byte slice has left",
            BitReaderError::TooManyBitsForType {..} => "Requested more bits than the requested integer type can hold",
        }
    }
}

impl fmt::Display for BitReaderError {
    fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result {
        //self.description().fmt(fmt)
        match *self {
            BitReaderError::NotEnoughData { position, length, requested } => write!(fmt, "BitReader: Requested {} bits with only {}/{} bits left (position {})", requested, length - position, length, position),
            BitReaderError::TooManyBitsForType { position, requested, allowed } => write!(fmt, "BitReader: Requested {} bits while the type can only hold {} (position {})", requested, allowed, position),
        }
    }
}

/// Helper trait to allow reading bits into a variable without explicitly mentioning its type.
///
/// If you can't or want, for some reason, to use BitReader's read methods (`read_u8` etc.) but
/// want to rely on type inference instead, you can use the ReadInto trait. The trait is
/// implemented for all basic integer types (8/16/32/64 bits, signed/unsigned)
/// and the boolean type.
///
/// ```
/// use bitreader::{BitReader,ReadInto};
///
/// let slice_of_u8 = &[0b1110_0000];
/// let mut reader = BitReader::new(slice_of_u8);
///
/// struct Foo {
///     bar: u8,
///     valid: bool,
/// }
///
/// // No type mentioned here, instead the type of bits is inferred from the type of Foo::bar,
/// // and consequently the correct "overload" is used.
/// let bits = ReadInto::read(&mut reader, 2).unwrap();
/// let valid = ReadInto::read(&mut reader, 1).unwrap();
///
/// let foo = Foo { bar: bits, valid: valid };
/// assert_eq!(foo.bar, 3);
/// assert!(foo.valid);
/// ```
pub trait ReadInto
    where Self: Sized
{
    fn read(reader: &mut BitReader, bits: u8) -> Result<Self>;
}

// There's eight almost identical implementations, let's make this easier.
macro_rules! impl_read_into {
    ($T:ty, $method:ident) => (
        impl ReadInto for $T {
            fn read(reader: &mut BitReader, bits: u8) -> Result<Self> {
                reader.$method(bits)
            }
        }
    )
}

impl_read_into!(u8, read_u8);
impl_read_into!(u16, read_u16);
impl_read_into!(u32, read_u32);
impl_read_into!(u64, read_u64);

impl_read_into!(i8, read_i8);
impl_read_into!(i16, read_i16);
impl_read_into!(i32, read_i32);
impl_read_into!(i64, read_i64);

// We can't cast to bool, so this requires a separate method.
impl ReadInto for bool {
    fn read(reader: &mut BitReader, bits: u8) -> Result<Self> {
        match reader.read_u8(bits)? {
            0 => Ok(false),
            _ => Ok(true),
        }
    }
}