From 698f8c2f01ea549d77d7dc3338a12e04c11057b9 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 17 Apr 2024 14:02:58 +0200 Subject: Adding upstream version 1.64.0+dfsg1. Signed-off-by: Daniel Baumann --- library/std/src/io/buffered/bufreader.rs | 496 +++++++++++++++++++++++++++++++ 1 file changed, 496 insertions(+) create mode 100644 library/std/src/io/buffered/bufreader.rs (limited to 'library/std/src/io/buffered/bufreader.rs') diff --git a/library/std/src/io/buffered/bufreader.rs b/library/std/src/io/buffered/bufreader.rs new file mode 100644 index 000000000..f7fbaa9c2 --- /dev/null +++ b/library/std/src/io/buffered/bufreader.rs @@ -0,0 +1,496 @@ +mod buffer; + +use crate::fmt; +use crate::io::{ + self, BufRead, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, SizeHint, DEFAULT_BUF_SIZE, +}; +use buffer::Buffer; + +/// The `BufReader` struct adds buffering to any reader. +/// +/// It can be excessively inefficient to work directly with a [`Read`] instance. +/// For example, every call to [`read`][`TcpStream::read`] on [`TcpStream`] +/// results in a system call. A `BufReader` performs large, infrequent reads on +/// the underlying [`Read`] and maintains an in-memory buffer of the results. +/// +/// `BufReader` can improve the speed of programs that make *small* and +/// *repeated* read calls to the same file or network socket. It does not +/// help when reading very large amounts at once, or reading just one or a few +/// times. It also provides no advantage when reading from a source that is +/// already in memory, like a [Vec]\. +/// +/// When the `BufReader` is dropped, the contents of its buffer will be +/// discarded. Creating multiple instances of a `BufReader` on the same +/// stream can cause data loss. Reading from the underlying reader after +/// unwrapping the `BufReader` with [`BufReader::into_inner`] can also cause +/// data loss. +/// +// HACK(#78696): can't use `crate` for associated items +/// [`TcpStream::read`]: super::super::super::net::TcpStream::read +/// [`TcpStream`]: crate::net::TcpStream +/// +/// # Examples +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::io::BufReader; +/// use std::fs::File; +/// +/// fn main() -> std::io::Result<()> { +/// let f = File::open("log.txt")?; +/// let mut reader = BufReader::new(f); +/// +/// let mut line = String::new(); +/// let len = reader.read_line(&mut line)?; +/// println!("First line is {len} bytes long"); +/// Ok(()) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub struct BufReader { + inner: R, + buf: Buffer, +} + +impl BufReader { + /// Creates a new `BufReader` with a default buffer capacity. The default is currently 8 KB, + /// but may change in the future. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let reader = BufReader::new(f); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn new(inner: R) -> BufReader { + BufReader::with_capacity(DEFAULT_BUF_SIZE, inner) + } + + /// Creates a new `BufReader` with the specified buffer capacity. + /// + /// # Examples + /// + /// Creating a buffer with ten bytes of capacity: + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let reader = BufReader::with_capacity(10, f); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn with_capacity(capacity: usize, inner: R) -> BufReader { + BufReader { inner, buf: Buffer::with_capacity(capacity) } + } +} + +impl BufReader { + /// Gets a reference to the underlying reader. + /// + /// It is inadvisable to directly read from the underlying reader. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f1 = File::open("log.txt")?; + /// let reader = BufReader::new(f1); + /// + /// let f2 = reader.get_ref(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_ref(&self) -> &R { + &self.inner + } + + /// Gets a mutable reference to the underlying reader. + /// + /// It is inadvisable to directly read from the underlying reader. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f1 = File::open("log.txt")?; + /// let mut reader = BufReader::new(f1); + /// + /// let f2 = reader.get_mut(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_mut(&mut self) -> &mut R { + &mut self.inner + } + + /// Returns a reference to the internally buffered data. + /// + /// Unlike [`fill_buf`], this will not attempt to fill the buffer if it is empty. + /// + /// [`fill_buf`]: BufRead::fill_buf + /// + /// # Examples + /// + /// ```no_run + /// use std::io::{BufReader, BufRead}; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let mut reader = BufReader::new(f); + /// assert!(reader.buffer().is_empty()); + /// + /// if reader.fill_buf()?.len() > 0 { + /// assert!(!reader.buffer().is_empty()); + /// } + /// Ok(()) + /// } + /// ``` + #[stable(feature = "bufreader_buffer", since = "1.37.0")] + pub fn buffer(&self) -> &[u8] { + self.buf.buffer() + } + + /// Returns the number of bytes the internal buffer can hold at once. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::{BufReader, BufRead}; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let mut reader = BufReader::new(f); + /// + /// let capacity = reader.capacity(); + /// let buffer = reader.fill_buf()?; + /// assert!(buffer.len() <= capacity); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "buffered_io_capacity", since = "1.46.0")] + pub fn capacity(&self) -> usize { + self.buf.capacity() + } + + /// Unwraps this `BufReader`, returning the underlying reader. + /// + /// Note that any leftover data in the internal buffer is lost. Therefore, + /// a following read from the underlying reader may lead to data loss. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f1 = File::open("log.txt")?; + /// let reader = BufReader::new(f1); + /// + /// let f2 = reader.into_inner(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn into_inner(self) -> R { + self.inner + } + + /// Invalidates all data in the internal buffer. + #[inline] + fn discard_buffer(&mut self) { + self.buf.discard_buffer() + } +} + +impl BufReader { + /// Seeks relative to the current position. If the new position lies within the buffer, + /// the buffer will not be flushed, allowing for more efficient seeks. + /// This method does not return the location of the underlying reader, so the caller + /// must track this information themselves if it is required. + #[stable(feature = "bufreader_seek_relative", since = "1.53.0")] + pub fn seek_relative(&mut self, offset: i64) -> io::Result<()> { + let pos = self.buf.pos() as u64; + if offset < 0 { + if let Some(_) = pos.checked_sub((-offset) as u64) { + self.buf.unconsume((-offset) as usize); + return Ok(()); + } + } else if let Some(new_pos) = pos.checked_add(offset as u64) { + if new_pos <= self.buf.filled() as u64 { + self.buf.consume(offset as usize); + return Ok(()); + } + } + + self.seek(SeekFrom::Current(offset)).map(drop) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for BufReader { + fn read(&mut self, buf: &mut [u8]) -> io::Result { + // If we don't have any buffered data and we're doing a massive read + // (larger than our internal buffer), bypass our internal buffer + // entirely. + if self.buf.pos() == self.buf.filled() && buf.len() >= self.capacity() { + self.discard_buffer(); + return self.inner.read(buf); + } + let nread = { + let mut rem = self.fill_buf()?; + rem.read(buf)? + }; + self.consume(nread); + Ok(nread) + } + + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + // If we don't have any buffered data and we're doing a massive read + // (larger than our internal buffer), bypass our internal buffer + // entirely. + if self.buf.pos() == self.buf.filled() && buf.remaining() >= self.capacity() { + self.discard_buffer(); + return self.inner.read_buf(buf); + } + + let prev = buf.filled_len(); + + let mut rem = self.fill_buf()?; + rem.read_buf(buf)?; + + self.consume(buf.filled_len() - prev); //slice impl of read_buf known to never unfill buf + + Ok(()) + } + + // Small read_exacts from a BufReader are extremely common when used with a deserializer. + // The default implementation calls read in a loop, which results in surprisingly poor code + // generation for the common path where the buffer has enough bytes to fill the passed-in + // buffer. + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + if self.buf.consume_with(buf.len(), |claimed| buf.copy_from_slice(claimed)) { + return Ok(()); + } + + crate::io::default_read_exact(self, buf) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + let total_len = bufs.iter().map(|b| b.len()).sum::(); + if self.buf.pos() == self.buf.filled() && total_len >= self.capacity() { + self.discard_buffer(); + return self.inner.read_vectored(bufs); + } + let nread = { + let mut rem = self.fill_buf()?; + rem.read_vectored(bufs)? + }; + self.consume(nread); + Ok(nread) + } + + fn is_read_vectored(&self) -> bool { + self.inner.is_read_vectored() + } + + // The inner reader might have an optimized `read_to_end`. Drain our buffer and then + // delegate to the inner implementation. + fn read_to_end(&mut self, buf: &mut Vec) -> io::Result { + let inner_buf = self.buffer(); + buf.extend_from_slice(inner_buf); + let nread = inner_buf.len(); + self.discard_buffer(); + Ok(nread + self.inner.read_to_end(buf)?) + } + + // The inner reader might have an optimized `read_to_end`. Drain our buffer and then + // delegate to the inner implementation. + fn read_to_string(&mut self, buf: &mut String) -> io::Result { + // In the general `else` case below we must read bytes into a side buffer, check + // that they are valid UTF-8, and then append them to `buf`. This requires a + // potentially large memcpy. + // + // If `buf` is empty--the most common case--we can leverage `append_to_string` + // to read directly into `buf`'s internal byte buffer, saving an allocation and + // a memcpy. + if buf.is_empty() { + // `append_to_string`'s safety relies on the buffer only being appended to since + // it only checks the UTF-8 validity of new data. If there were existing content in + // `buf` then an untrustworthy reader (i.e. `self.inner`) could not only append + // bytes but also modify existing bytes and render them invalid. On the other hand, + // if `buf` is empty then by definition any writes must be appends and + // `append_to_string` will validate all of the new bytes. + unsafe { crate::io::append_to_string(buf, |b| self.read_to_end(b)) } + } else { + // We cannot append our byte buffer directly onto the `buf` String as there could + // be an incomplete UTF-8 sequence that has only been partially read. We must read + // everything into a side buffer first and then call `from_utf8` on the complete + // buffer. + let mut bytes = Vec::new(); + self.read_to_end(&mut bytes)?; + let string = crate::str::from_utf8(&bytes).map_err(|_| { + io::const_io_error!( + io::ErrorKind::InvalidData, + "stream did not contain valid UTF-8", + ) + })?; + *buf += string; + Ok(string.len()) + } + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl BufRead for BufReader { + fn fill_buf(&mut self) -> io::Result<&[u8]> { + self.buf.fill_buf(&mut self.inner) + } + + fn consume(&mut self, amt: usize) { + self.buf.consume(amt) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl fmt::Debug for BufReader +where + R: fmt::Debug, +{ + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.debug_struct("BufReader") + .field("reader", &self.inner) + .field( + "buffer", + &format_args!("{}/{}", self.buf.filled() - self.buf.pos(), self.capacity()), + ) + .finish() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Seek for BufReader { + /// Seek to an offset, in bytes, in the underlying reader. + /// + /// The position used for seeking with [SeekFrom::Current]\(_) is the + /// position the underlying reader would be at if the `BufReader` had no + /// internal buffer. + /// + /// Seeking always discards the internal buffer, even if the seek position + /// would otherwise fall within it. This guarantees that calling + /// [`BufReader::into_inner()`] immediately after a seek yields the underlying reader + /// at the same position. + /// + /// To seek without discarding the internal buffer, use [`BufReader::seek_relative`]. + /// + /// See [`std::io::Seek`] for more details. + /// + /// Note: In the edge case where you're seeking with [SeekFrom::Current]\(n) + /// where `n` minus the internal buffer length overflows an `i64`, two + /// seeks will be performed instead of one. If the second seek returns + /// [`Err`], the underlying reader will be left at the same position it would + /// have if you called `seek` with [SeekFrom::Current]\(0). + /// + /// [`std::io::Seek`]: Seek + fn seek(&mut self, pos: SeekFrom) -> io::Result { + let result: u64; + if let SeekFrom::Current(n) = pos { + let remainder = (self.buf.filled() - self.buf.pos()) as i64; + // it should be safe to assume that remainder fits within an i64 as the alternative + // means we managed to allocate 8 exbibytes and that's absurd. + // But it's not out of the realm of possibility for some weird underlying reader to + // support seeking by i64::MIN so we need to handle underflow when subtracting + // remainder. + if let Some(offset) = n.checked_sub(remainder) { + result = self.inner.seek(SeekFrom::Current(offset))?; + } else { + // seek backwards by our remainder, and then by the offset + self.inner.seek(SeekFrom::Current(-remainder))?; + self.discard_buffer(); + result = self.inner.seek(SeekFrom::Current(n))?; + } + } else { + // Seeking with Start/End doesn't care about our buffer length. + result = self.inner.seek(pos)?; + } + self.discard_buffer(); + Ok(result) + } + + /// Returns the current seek position from the start of the stream. + /// + /// The value returned is equivalent to `self.seek(SeekFrom::Current(0))` + /// but does not flush the internal buffer. Due to this optimization the + /// function does not guarantee that calling `.into_inner()` immediately + /// afterwards will yield the underlying reader at the same position. Use + /// [`BufReader::seek`] instead if you require that guarantee. + /// + /// # Panics + /// + /// This function will panic if the position of the inner reader is smaller + /// than the amount of buffered data. That can happen if the inner reader + /// has an incorrect implementation of [`Seek::stream_position`], or if the + /// position has gone out of sync due to calling [`Seek::seek`] directly on + /// the underlying reader. + /// + /// # Example + /// + /// ```no_run + /// use std::{ + /// io::{self, BufRead, BufReader, Seek}, + /// fs::File, + /// }; + /// + /// fn main() -> io::Result<()> { + /// let mut f = BufReader::new(File::open("foo.txt")?); + /// + /// let before = f.stream_position()?; + /// f.read_line(&mut String::new())?; + /// let after = f.stream_position()?; + /// + /// println!("The first line was {} bytes long", after - before); + /// Ok(()) + /// } + /// ``` + fn stream_position(&mut self) -> io::Result { + let remainder = (self.buf.filled() - self.buf.pos()) as u64; + self.inner.stream_position().map(|pos| { + pos.checked_sub(remainder).expect( + "overflow when subtracting remaining buffer size from inner stream position", + ) + }) + } +} + +impl SizeHint for BufReader { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(self.get_ref()) + self.buffer().len() + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(self.get_ref()).and_then(|up| self.buffer().len().checked_add(up)) + } +} -- cgit v1.2.3