use crate::cmp; use crate::fmt; use crate::io::{ self, BufRead, Initializer, IoSliceMut, Read, Seek, SeekFrom, SizeHint, DEFAULT_BUF_SIZE, }; /// 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 {} bytes long", len); /// Ok(()) /// } /// ``` #[stable(feature = "rust1", since = "1.0.0")] pub struct BufReader { inner: R, buf: Box<[u8]>, pos: usize, cap: usize, } 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 { unsafe { let mut buf = Box::new_uninit_slice(capacity).assume_init(); inner.initializer().initialize(&mut buf); BufReader { inner, buf, pos: 0, cap: 0 } } } } 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[self.pos..self.cap] } /// 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.len() } /// 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.pos = 0; self.cap = 0; } } 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.pos as u64; if offset < 0 { if let Some(new_pos) = pos.checked_sub((-offset) as u64) { self.pos = new_pos as usize; return Ok(()); } } else if let Some(new_pos) = pos.checked_add(offset as u64) { if new_pos <= self.cap as u64 { self.pos = new_pos 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.pos == self.cap && buf.len() >= self.buf.len() { self.discard_buffer(); return self.inner.read(buf); } let nread = { let mut rem = self.fill_buf()?; rem.read(buf)? }; self.consume(nread); Ok(nread) } // 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.buffer().len() >= buf.len() { buf.copy_from_slice(&self.buffer()[..buf.len()]); self.consume(buf.len()); 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.pos == self.cap && total_len >= self.buf.len() { 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() } // we can't skip unconditionally because of the large buffer case in read. unsafe fn initializer(&self) -> Initializer { self.inner.initializer() } // 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 nread = self.cap - self.pos; buf.extend_from_slice(&self.buf[self.pos..self.cap]); 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::Error::new_const( 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]> { // If we've reached the end of our internal buffer then we need to fetch // some more data from the underlying reader. // Branch using `>=` instead of the more correct `==` // to tell the compiler that the pos..cap slice is always valid. if self.pos >= self.cap { debug_assert!(self.pos == self.cap); self.cap = self.inner.read(&mut self.buf)?; self.pos = 0; } Ok(&self.buf[self.pos..self.cap]) } fn consume(&mut self, amt: usize) { self.pos = cmp::min(self.pos + amt, self.cap); } } #[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.cap - self.pos, self.buf.len())) .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.cap - self.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.cap - self.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)) } }