diff options
Diffstat (limited to 'library/std/src/io')
25 files changed, 12037 insertions, 0 deletions
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<R>` 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<R>` performs large, infrequent reads on +/// the underlying [`Read`] and maintains an in-memory buffer of the results. +/// +/// `BufReader<R>` 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 <code>[Vec]\<u8></code>. +/// +/// When the `BufReader<R>` is dropped, the contents of its buffer will be +/// discarded. Creating multiple instances of a `BufReader<R>` on the same +/// stream can cause data loss. Reading from the underlying reader after +/// unwrapping the `BufReader<R>` 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<R> { + inner: R, + buf: Buffer, +} + +impl<R: Read> BufReader<R> { + /// Creates a new `BufReader<R>` 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<R> { + BufReader::with_capacity(DEFAULT_BUF_SIZE, inner) + } + + /// Creates a new `BufReader<R>` 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<R> { + BufReader { inner, buf: Buffer::with_capacity(capacity) } + } +} + +impl<R> BufReader<R> { + /// 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<R>`, 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<R: Seek> BufReader<R> { + /// 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<R: Read> Read for BufReader<R> { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + // 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<usize> { + let total_len = bufs.iter().map(|b| b.len()).sum::<usize>(); + 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<u8>) -> io::Result<usize> { + 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<usize> { + // 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<R: Read> BufRead for BufReader<R> { + 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<R> fmt::Debug for BufReader<R> +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<R: Seek> Seek for BufReader<R> { + /// Seek to an offset, in bytes, in the underlying reader. + /// + /// The position used for seeking with <code>[SeekFrom::Current]\(_)</code> is the + /// position the underlying reader would be at if the `BufReader<R>` 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 <code>[SeekFrom::Current]\(n)</code> + /// 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 <code>[SeekFrom::Current]\(0)</code>. + /// + /// [`std::io::Seek`]: Seek + fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> { + 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<u64> { + 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<T> SizeHint for BufReader<T> { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(self.get_ref()) + self.buffer().len() + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + SizeHint::upper_bound(self.get_ref()).and_then(|up| self.buffer().len().checked_add(up)) + } +} diff --git a/library/std/src/io/buffered/bufreader/buffer.rs b/library/std/src/io/buffered/bufreader/buffer.rs new file mode 100644 index 000000000..8ae01f3b0 --- /dev/null +++ b/library/std/src/io/buffered/bufreader/buffer.rs @@ -0,0 +1,105 @@ +///! An encapsulation of `BufReader`'s buffer management logic. +/// +/// This module factors out the basic functionality of `BufReader` in order to protect two core +/// invariants: +/// * `filled` bytes of `buf` are always initialized +/// * `pos` is always <= `filled` +/// Since this module encapsulates the buffer management logic, we can ensure that the range +/// `pos..filled` is always a valid index into the initialized region of the buffer. This means +/// that user code which wants to do reads from a `BufReader` via `buffer` + `consume` can do so +/// without encountering any runtime bounds checks. +use crate::cmp; +use crate::io::{self, Read, ReadBuf}; +use crate::mem::MaybeUninit; + +pub struct Buffer { + // The buffer. + buf: Box<[MaybeUninit<u8>]>, + // The current seek offset into `buf`, must always be <= `filled`. + pos: usize, + // Each call to `fill_buf` sets `filled` to indicate how many bytes at the start of `buf` are + // initialized with bytes from a read. + filled: usize, +} + +impl Buffer { + #[inline] + pub fn with_capacity(capacity: usize) -> Self { + let buf = Box::new_uninit_slice(capacity); + Self { buf, pos: 0, filled: 0 } + } + + #[inline] + pub fn buffer(&self) -> &[u8] { + // SAFETY: self.pos and self.cap are valid, and self.cap => self.pos, and + // that region is initialized because those are all invariants of this type. + unsafe { MaybeUninit::slice_assume_init_ref(self.buf.get_unchecked(self.pos..self.filled)) } + } + + #[inline] + pub fn capacity(&self) -> usize { + self.buf.len() + } + + #[inline] + pub fn filled(&self) -> usize { + self.filled + } + + #[inline] + pub fn pos(&self) -> usize { + self.pos + } + + #[inline] + pub fn discard_buffer(&mut self) { + self.pos = 0; + self.filled = 0; + } + + #[inline] + pub fn consume(&mut self, amt: usize) { + self.pos = cmp::min(self.pos + amt, self.filled); + } + + /// If there are `amt` bytes available in the buffer, pass a slice containing those bytes to + /// `visitor` and return true. If there are not enough bytes available, return false. + #[inline] + pub fn consume_with<V>(&mut self, amt: usize, mut visitor: V) -> bool + where + V: FnMut(&[u8]), + { + if let Some(claimed) = self.buffer().get(..amt) { + visitor(claimed); + // If the indexing into self.buffer() succeeds, amt must be a valid increment. + self.pos += amt; + true + } else { + false + } + } + + #[inline] + pub fn unconsume(&mut self, amt: usize) { + self.pos = self.pos.saturating_sub(amt); + } + + #[inline] + pub fn fill_buf(&mut self, mut reader: impl Read) -> io::Result<&[u8]> { + // If we've reached the end of our internal buffer then we need to fetch + // some more data from the reader. + // Branch using `>=` instead of the more correct `==` + // to tell the compiler that the pos..cap slice is always valid. + if self.pos >= self.filled { + debug_assert!(self.pos == self.filled); + + let mut readbuf = ReadBuf::uninit(&mut self.buf); + + reader.read_buf(&mut readbuf)?; + + self.filled = readbuf.filled_len(); + self.pos = 0; + } + Ok(self.buffer()) + } +} diff --git a/library/std/src/io/buffered/bufwriter.rs b/library/std/src/io/buffered/bufwriter.rs new file mode 100644 index 000000000..6acb937e7 --- /dev/null +++ b/library/std/src/io/buffered/bufwriter.rs @@ -0,0 +1,674 @@ +use crate::error; +use crate::fmt; +use crate::io::{ + self, ErrorKind, IntoInnerError, IoSlice, Seek, SeekFrom, Write, DEFAULT_BUF_SIZE, +}; +use crate::mem; +use crate::ptr; + +/// Wraps a writer and buffers its output. +/// +/// It can be excessively inefficient to work directly with something that +/// implements [`Write`]. For example, every call to +/// [`write`][`TcpStream::write`] on [`TcpStream`] results in a system call. A +/// `BufWriter<W>` keeps an in-memory buffer of data and writes it to an underlying +/// writer in large, infrequent batches. +/// +/// `BufWriter<W>` can improve the speed of programs that make *small* and +/// *repeated* write calls to the same file or network socket. It does not +/// help when writing very large amounts at once, or writing just one or a few +/// times. It also provides no advantage when writing to a destination that is +/// in memory, like a <code>[Vec]\<u8></code>. +/// +/// It is critical to call [`flush`] before `BufWriter<W>` is dropped. Though +/// dropping will attempt to flush the contents of the buffer, any errors +/// that happen in the process of dropping will be ignored. Calling [`flush`] +/// ensures that the buffer is empty and thus dropping will not even attempt +/// file operations. +/// +/// # Examples +/// +/// Let's write the numbers one through ten to a [`TcpStream`]: +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::net::TcpStream; +/// +/// let mut stream = TcpStream::connect("127.0.0.1:34254").unwrap(); +/// +/// for i in 0..10 { +/// stream.write(&[i+1]).unwrap(); +/// } +/// ``` +/// +/// Because we're not buffering, we write each one in turn, incurring the +/// overhead of a system call per byte written. We can fix this with a +/// `BufWriter<W>`: +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::io::BufWriter; +/// use std::net::TcpStream; +/// +/// let mut stream = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); +/// +/// for i in 0..10 { +/// stream.write(&[i+1]).unwrap(); +/// } +/// stream.flush().unwrap(); +/// ``` +/// +/// By wrapping the stream with a `BufWriter<W>`, these ten writes are all grouped +/// together by the buffer and will all be written out in one system call when +/// the `stream` is flushed. +/// +// HACK(#78696): can't use `crate` for associated items +/// [`TcpStream::write`]: super::super::super::net::TcpStream::write +/// [`TcpStream`]: crate::net::TcpStream +/// [`flush`]: BufWriter::flush +#[stable(feature = "rust1", since = "1.0.0")] +pub struct BufWriter<W: Write> { + inner: W, + // The buffer. Avoid using this like a normal `Vec` in common code paths. + // That is, don't use `buf.push`, `buf.extend_from_slice`, or any other + // methods that require bounds checking or the like. This makes an enormous + // difference to performance (we may want to stop using a `Vec` entirely). + buf: Vec<u8>, + // #30888: If the inner writer panics in a call to write, we don't want to + // write the buffered data a second time in BufWriter's destructor. This + // flag tells the Drop impl if it should skip the flush. + panicked: bool, +} + +impl<W: Write> BufWriter<W> { + /// Creates a new `BufWriter<W>` with a default buffer capacity. The default is currently 8 KB, + /// but may change in the future. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut buffer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn new(inner: W) -> BufWriter<W> { + BufWriter::with_capacity(DEFAULT_BUF_SIZE, inner) + } + + /// Creates a new `BufWriter<W>` with at least the specified buffer capacity. + /// + /// # Examples + /// + /// Creating a buffer with a buffer of at least a hundred bytes. + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let stream = TcpStream::connect("127.0.0.1:34254").unwrap(); + /// let mut buffer = BufWriter::with_capacity(100, stream); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn with_capacity(capacity: usize, inner: W) -> BufWriter<W> { + BufWriter { inner, buf: Vec::with_capacity(capacity), panicked: false } + } + + /// Send data in our local buffer into the inner writer, looping as + /// necessary until either it's all been sent or an error occurs. + /// + /// Because all the data in the buffer has been reported to our owner as + /// "successfully written" (by returning nonzero success values from + /// `write`), any 0-length writes from `inner` must be reported as i/o + /// errors from this method. + pub(in crate::io) fn flush_buf(&mut self) -> io::Result<()> { + /// Helper struct to ensure the buffer is updated after all the writes + /// are complete. It tracks the number of written bytes and drains them + /// all from the front of the buffer when dropped. + struct BufGuard<'a> { + buffer: &'a mut Vec<u8>, + written: usize, + } + + impl<'a> BufGuard<'a> { + fn new(buffer: &'a mut Vec<u8>) -> Self { + Self { buffer, written: 0 } + } + + /// The unwritten part of the buffer + fn remaining(&self) -> &[u8] { + &self.buffer[self.written..] + } + + /// Flag some bytes as removed from the front of the buffer + fn consume(&mut self, amt: usize) { + self.written += amt; + } + + /// true if all of the bytes have been written + fn done(&self) -> bool { + self.written >= self.buffer.len() + } + } + + impl Drop for BufGuard<'_> { + fn drop(&mut self) { + if self.written > 0 { + self.buffer.drain(..self.written); + } + } + } + + let mut guard = BufGuard::new(&mut self.buf); + while !guard.done() { + self.panicked = true; + let r = self.inner.write(guard.remaining()); + self.panicked = false; + + match r { + Ok(0) => { + return Err(io::const_io_error!( + ErrorKind::WriteZero, + "failed to write the buffered data", + )); + } + Ok(n) => guard.consume(n), + Err(ref e) if e.kind() == io::ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } + Ok(()) + } + + /// Buffer some data without flushing it, regardless of the size of the + /// data. Writes as much as possible without exceeding capacity. Returns + /// the number of bytes written. + pub(super) fn write_to_buf(&mut self, buf: &[u8]) -> usize { + let available = self.spare_capacity(); + let amt_to_buffer = available.min(buf.len()); + + // SAFETY: `amt_to_buffer` is <= buffer's spare capacity by construction. + unsafe { + self.write_to_buffer_unchecked(&buf[..amt_to_buffer]); + } + + amt_to_buffer + } + + /// Gets a reference to the underlying writer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut buffer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // we can use reference just like buffer + /// let reference = buffer.get_ref(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_ref(&self) -> &W { + &self.inner + } + + /// Gets a mutable reference to the underlying writer. + /// + /// It is inadvisable to directly write to the underlying writer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut buffer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // we can use reference just like buffer + /// let reference = buffer.get_mut(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_mut(&mut self) -> &mut W { + &mut self.inner + } + + /// Returns a reference to the internally buffered data. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let buf_writer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // See how many bytes are currently buffered + /// let bytes_buffered = buf_writer.buffer().len(); + /// ``` + #[stable(feature = "bufreader_buffer", since = "1.37.0")] + pub fn buffer(&self) -> &[u8] { + &self.buf + } + + /// Returns a mutable reference to the internal buffer. + /// + /// This can be used to write data directly into the buffer without triggering writers + /// to the underlying writer. + /// + /// That the buffer is a `Vec` is an implementation detail. + /// Callers should not modify the capacity as there currently is no public API to do so + /// and thus any capacity changes would be unexpected by the user. + pub(in crate::io) fn buffer_mut(&mut self) -> &mut Vec<u8> { + &mut self.buf + } + + /// Returns the number of bytes the internal buffer can hold without flushing. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let buf_writer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // Check the capacity of the inner buffer + /// let capacity = buf_writer.capacity(); + /// // Calculate how many bytes can be written without flushing + /// let without_flush = capacity - buf_writer.buffer().len(); + /// ``` + #[stable(feature = "buffered_io_capacity", since = "1.46.0")] + pub fn capacity(&self) -> usize { + self.buf.capacity() + } + + /// Unwraps this `BufWriter<W>`, returning the underlying writer. + /// + /// The buffer is written out before returning the writer. + /// + /// # Errors + /// + /// An [`Err`] will be returned if an error occurs while flushing the buffer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut buffer = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // unwrap the TcpStream and flush the buffer + /// let stream = buffer.into_inner().unwrap(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn into_inner(mut self) -> Result<W, IntoInnerError<BufWriter<W>>> { + match self.flush_buf() { + Err(e) => Err(IntoInnerError::new(self, e)), + Ok(()) => Ok(self.into_parts().0), + } + } + + /// Disassembles this `BufWriter<W>`, returning the underlying writer, and any buffered but + /// unwritten data. + /// + /// If the underlying writer panicked, it is not known what portion of the data was written. + /// In this case, we return `WriterPanicked` for the buffered data (from which the buffer + /// contents can still be recovered). + /// + /// `into_parts` makes no attempt to flush data and cannot fail. + /// + /// # Examples + /// + /// ``` + /// use std::io::{BufWriter, Write}; + /// + /// let mut buffer = [0u8; 10]; + /// let mut stream = BufWriter::new(buffer.as_mut()); + /// write!(stream, "too much data").unwrap(); + /// stream.flush().expect_err("it doesn't fit"); + /// let (recovered_writer, buffered_data) = stream.into_parts(); + /// assert_eq!(recovered_writer.len(), 0); + /// assert_eq!(&buffered_data.unwrap(), b"ata"); + /// ``` + #[stable(feature = "bufwriter_into_parts", since = "1.56.0")] + pub fn into_parts(mut self) -> (W, Result<Vec<u8>, WriterPanicked>) { + let buf = mem::take(&mut self.buf); + let buf = if !self.panicked { Ok(buf) } else { Err(WriterPanicked { buf }) }; + + // SAFETY: forget(self) prevents double dropping inner + let inner = unsafe { ptr::read(&mut self.inner) }; + mem::forget(self); + + (inner, buf) + } + + // Ensure this function does not get inlined into `write`, so that it + // remains inlineable and its common path remains as short as possible. + // If this function ends up being called frequently relative to `write`, + // it's likely a sign that the client is using an improperly sized buffer + // or their write patterns are somewhat pathological. + #[cold] + #[inline(never)] + fn write_cold(&mut self, buf: &[u8]) -> io::Result<usize> { + if buf.len() > self.spare_capacity() { + self.flush_buf()?; + } + + // Why not len > capacity? To avoid a needless trip through the buffer when the input + // exactly fills it. We'd just need to flush it to the underlying writer anyway. + if buf.len() >= self.buf.capacity() { + self.panicked = true; + let r = self.get_mut().write(buf); + self.panicked = false; + r + } else { + // Write to the buffer. In this case, we write to the buffer even if it fills it + // exactly. Doing otherwise would mean flushing the buffer, then writing this + // input to the inner writer, which in many cases would be a worse strategy. + + // SAFETY: There was either enough spare capacity already, or there wasn't and we + // flushed the buffer to ensure that there is. In the latter case, we know that there + // is because flushing ensured that our entire buffer is spare capacity, and we entered + // this block because the input buffer length is less than that capacity. In either + // case, it's safe to write the input buffer to our buffer. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + Ok(buf.len()) + } + } + + // Ensure this function does not get inlined into `write_all`, so that it + // remains inlineable and its common path remains as short as possible. + // If this function ends up being called frequently relative to `write_all`, + // it's likely a sign that the client is using an improperly sized buffer + // or their write patterns are somewhat pathological. + #[cold] + #[inline(never)] + fn write_all_cold(&mut self, buf: &[u8]) -> io::Result<()> { + // Normally, `write_all` just calls `write` in a loop. We can do better + // by calling `self.get_mut().write_all()` directly, which avoids + // round trips through the buffer in the event of a series of partial + // writes in some circumstances. + + if buf.len() > self.spare_capacity() { + self.flush_buf()?; + } + + // Why not len > capacity? To avoid a needless trip through the buffer when the input + // exactly fills it. We'd just need to flush it to the underlying writer anyway. + if buf.len() >= self.buf.capacity() { + self.panicked = true; + let r = self.get_mut().write_all(buf); + self.panicked = false; + r + } else { + // Write to the buffer. In this case, we write to the buffer even if it fills it + // exactly. Doing otherwise would mean flushing the buffer, then writing this + // input to the inner writer, which in many cases would be a worse strategy. + + // SAFETY: There was either enough spare capacity already, or there wasn't and we + // flushed the buffer to ensure that there is. In the latter case, we know that there + // is because flushing ensured that our entire buffer is spare capacity, and we entered + // this block because the input buffer length is less than that capacity. In either + // case, it's safe to write the input buffer to our buffer. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + Ok(()) + } + } + + // SAFETY: Requires `buf.len() <= self.buf.capacity() - self.buf.len()`, + // i.e., that input buffer length is less than or equal to spare capacity. + #[inline] + unsafe fn write_to_buffer_unchecked(&mut self, buf: &[u8]) { + debug_assert!(buf.len() <= self.spare_capacity()); + let old_len = self.buf.len(); + let buf_len = buf.len(); + let src = buf.as_ptr(); + let dst = self.buf.as_mut_ptr().add(old_len); + ptr::copy_nonoverlapping(src, dst, buf_len); + self.buf.set_len(old_len + buf_len); + } + + #[inline] + fn spare_capacity(&self) -> usize { + self.buf.capacity() - self.buf.len() + } +} + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +/// Error returned for the buffered data from `BufWriter::into_parts`, when the underlying +/// writer has previously panicked. Contains the (possibly partly written) buffered data. +/// +/// # Example +/// +/// ``` +/// use std::io::{self, BufWriter, Write}; +/// use std::panic::{catch_unwind, AssertUnwindSafe}; +/// +/// struct PanickingWriter; +/// impl Write for PanickingWriter { +/// fn write(&mut self, buf: &[u8]) -> io::Result<usize> { panic!() } +/// fn flush(&mut self) -> io::Result<()> { panic!() } +/// } +/// +/// let mut stream = BufWriter::new(PanickingWriter); +/// write!(stream, "some data").unwrap(); +/// let result = catch_unwind(AssertUnwindSafe(|| { +/// stream.flush().unwrap() +/// })); +/// assert!(result.is_err()); +/// let (recovered_writer, buffered_data) = stream.into_parts(); +/// assert!(matches!(recovered_writer, PanickingWriter)); +/// assert_eq!(buffered_data.unwrap_err().into_inner(), b"some data"); +/// ``` +pub struct WriterPanicked { + buf: Vec<u8>, +} + +impl WriterPanicked { + /// Returns the perhaps-unwritten data. Some of this data may have been written by the + /// panicking call(s) to the underlying writer, so simply writing it again is not a good idea. + #[must_use = "`self` will be dropped if the result is not used"] + #[stable(feature = "bufwriter_into_parts", since = "1.56.0")] + pub fn into_inner(self) -> Vec<u8> { + self.buf + } + + const DESCRIPTION: &'static str = + "BufWriter inner writer panicked, what data remains unwritten is not known"; +} + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +impl error::Error for WriterPanicked { + #[allow(deprecated, deprecated_in_future)] + fn description(&self) -> &str { + Self::DESCRIPTION + } +} + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +impl fmt::Display for WriterPanicked { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "{}", Self::DESCRIPTION) + } +} + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +impl fmt::Debug for WriterPanicked { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("WriterPanicked") + .field("buffer", &format_args!("{}/{}", self.buf.len(), self.buf.capacity())) + .finish() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write> Write for BufWriter<W> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + // Use < instead of <= to avoid a needless trip through the buffer in some cases. + // See `write_cold` for details. + if buf.len() < self.spare_capacity() { + // SAFETY: safe by above conditional. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + Ok(buf.len()) + } else { + self.write_cold(buf) + } + } + + #[inline] + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + // Use < instead of <= to avoid a needless trip through the buffer in some cases. + // See `write_all_cold` for details. + if buf.len() < self.spare_capacity() { + // SAFETY: safe by above conditional. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + Ok(()) + } else { + self.write_all_cold(buf) + } + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + // FIXME: Consider applying `#[inline]` / `#[inline(never)]` optimizations already applied + // to `write` and `write_all`. The performance benefits can be significant. See #79930. + if self.get_ref().is_write_vectored() { + // We have to handle the possibility that the total length of the buffers overflows + // `usize` (even though this can only happen if multiple `IoSlice`s reference the + // same underlying buffer, as otherwise the buffers wouldn't fit in memory). If the + // computation overflows, then surely the input cannot fit in our buffer, so we forward + // to the inner writer's `write_vectored` method to let it handle it appropriately. + let saturated_total_len = + bufs.iter().fold(0usize, |acc, b| acc.saturating_add(b.len())); + + if saturated_total_len > self.spare_capacity() { + // Flush if the total length of the input exceeds our buffer's spare capacity. + // If we would have overflowed, this condition also holds, and we need to flush. + self.flush_buf()?; + } + + if saturated_total_len >= self.buf.capacity() { + // Forward to our inner writer if the total length of the input is greater than or + // equal to our buffer capacity. If we would have overflowed, this condition also + // holds, and we punt to the inner writer. + self.panicked = true; + let r = self.get_mut().write_vectored(bufs); + self.panicked = false; + r + } else { + // `saturated_total_len < self.buf.capacity()` implies that we did not saturate. + + // SAFETY: We checked whether or not the spare capacity was large enough above. If + // it was, then we're safe already. If it wasn't, we flushed, making sufficient + // room for any input <= the buffer size, which includes this input. + unsafe { + bufs.iter().for_each(|b| self.write_to_buffer_unchecked(b)); + }; + + Ok(saturated_total_len) + } + } else { + let mut iter = bufs.iter(); + let mut total_written = if let Some(buf) = iter.by_ref().find(|&buf| !buf.is_empty()) { + // This is the first non-empty slice to write, so if it does + // not fit in the buffer, we still get to flush and proceed. + if buf.len() > self.spare_capacity() { + self.flush_buf()?; + } + if buf.len() >= self.buf.capacity() { + // The slice is at least as large as the buffering capacity, + // so it's better to write it directly, bypassing the buffer. + self.panicked = true; + let r = self.get_mut().write(buf); + self.panicked = false; + return r; + } else { + // SAFETY: We checked whether or not the spare capacity was large enough above. + // If it was, then we're safe already. If it wasn't, we flushed, making + // sufficient room for any input <= the buffer size, which includes this input. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + buf.len() + } + } else { + return Ok(0); + }; + debug_assert!(total_written != 0); + for buf in iter { + if buf.len() <= self.spare_capacity() { + // SAFETY: safe by above conditional. + unsafe { + self.write_to_buffer_unchecked(buf); + } + + // This cannot overflow `usize`. If we are here, we've written all of the bytes + // so far to our buffer, and we've ensured that we never exceed the buffer's + // capacity. Therefore, `total_written` <= `self.buf.capacity()` <= `usize::MAX`. + total_written += buf.len(); + } else { + break; + } + } + Ok(total_written) + } + } + + fn is_write_vectored(&self) -> bool { + true + } + + fn flush(&mut self) -> io::Result<()> { + self.flush_buf().and_then(|()| self.get_mut().flush()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write> fmt::Debug for BufWriter<W> +where + W: fmt::Debug, +{ + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.debug_struct("BufWriter") + .field("writer", &self.inner) + .field("buffer", &format_args!("{}/{}", self.buf.len(), self.buf.capacity())) + .finish() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write + Seek> Seek for BufWriter<W> { + /// Seek to the offset, in bytes, in the underlying writer. + /// + /// Seeking always writes out the internal buffer before seeking. + fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> { + self.flush_buf()?; + self.get_mut().seek(pos) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write> Drop for BufWriter<W> { + fn drop(&mut self) { + if !self.panicked { + // dtors should not panic, so we ignore a failed flush + let _r = self.flush_buf(); + } + } +} diff --git a/library/std/src/io/buffered/linewriter.rs b/library/std/src/io/buffered/linewriter.rs new file mode 100644 index 000000000..a26a4ab33 --- /dev/null +++ b/library/std/src/io/buffered/linewriter.rs @@ -0,0 +1,232 @@ +use crate::fmt; +use crate::io::{self, buffered::LineWriterShim, BufWriter, IntoInnerError, IoSlice, Write}; + +/// Wraps a writer and buffers output to it, flushing whenever a newline +/// (`0x0a`, `'\n'`) is detected. +/// +/// The [`BufWriter`] struct wraps a writer and buffers its output. +/// But it only does this batched write when it goes out of scope, or when the +/// internal buffer is full. Sometimes, you'd prefer to write each line as it's +/// completed, rather than the entire buffer at once. Enter `LineWriter`. It +/// does exactly that. +/// +/// Like [`BufWriter`], a `LineWriter`’s buffer will also be flushed when the +/// `LineWriter` goes out of scope or when its internal buffer is full. +/// +/// If there's still a partial line in the buffer when the `LineWriter` is +/// dropped, it will flush those contents. +/// +/// # Examples +/// +/// We can use `LineWriter` to write one line at a time, significantly +/// reducing the number of actual writes to the file. +/// +/// ```no_run +/// use std::fs::{self, File}; +/// use std::io::prelude::*; +/// use std::io::LineWriter; +/// +/// fn main() -> std::io::Result<()> { +/// let road_not_taken = b"I shall be telling this with a sigh +/// Somewhere ages and ages hence: +/// Two roads diverged in a wood, and I - +/// I took the one less traveled by, +/// And that has made all the difference."; +/// +/// let file = File::create("poem.txt")?; +/// let mut file = LineWriter::new(file); +/// +/// file.write_all(b"I shall be telling this with a sigh")?; +/// +/// // No bytes are written until a newline is encountered (or +/// // the internal buffer is filled). +/// assert_eq!(fs::read_to_string("poem.txt")?, ""); +/// file.write_all(b"\n")?; +/// assert_eq!( +/// fs::read_to_string("poem.txt")?, +/// "I shall be telling this with a sigh\n", +/// ); +/// +/// // Write the rest of the poem. +/// file.write_all(b"Somewhere ages and ages hence: +/// Two roads diverged in a wood, and I - +/// I took the one less traveled by, +/// And that has made all the difference.")?; +/// +/// // The last line of the poem doesn't end in a newline, so +/// // we have to flush or drop the `LineWriter` to finish +/// // writing. +/// file.flush()?; +/// +/// // Confirm the whole poem was written. +/// assert_eq!(fs::read("poem.txt")?, &road_not_taken[..]); +/// Ok(()) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub struct LineWriter<W: Write> { + inner: BufWriter<W>, +} + +impl<W: Write> LineWriter<W> { + /// Creates a new `LineWriter`. + /// + /// # Examples + /// + /// ```no_run + /// use std::fs::File; + /// use std::io::LineWriter; + /// + /// fn main() -> std::io::Result<()> { + /// let file = File::create("poem.txt")?; + /// let file = LineWriter::new(file); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn new(inner: W) -> LineWriter<W> { + // Lines typically aren't that long, don't use a giant buffer + LineWriter::with_capacity(1024, inner) + } + + /// Creates a new `LineWriter` with at least the specified capacity for the + /// internal buffer. + /// + /// # Examples + /// + /// ```no_run + /// use std::fs::File; + /// use std::io::LineWriter; + /// + /// fn main() -> std::io::Result<()> { + /// let file = File::create("poem.txt")?; + /// let file = LineWriter::with_capacity(100, file); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn with_capacity(capacity: usize, inner: W) -> LineWriter<W> { + LineWriter { inner: BufWriter::with_capacity(capacity, inner) } + } + + /// Gets a reference to the underlying writer. + /// + /// # Examples + /// + /// ```no_run + /// use std::fs::File; + /// use std::io::LineWriter; + /// + /// fn main() -> std::io::Result<()> { + /// let file = File::create("poem.txt")?; + /// let file = LineWriter::new(file); + /// + /// let reference = file.get_ref(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_ref(&self) -> &W { + self.inner.get_ref() + } + + /// Gets a mutable reference to the underlying writer. + /// + /// Caution must be taken when calling methods on the mutable reference + /// returned as extra writes could corrupt the output stream. + /// + /// # Examples + /// + /// ```no_run + /// use std::fs::File; + /// use std::io::LineWriter; + /// + /// fn main() -> std::io::Result<()> { + /// let file = File::create("poem.txt")?; + /// let mut file = LineWriter::new(file); + /// + /// // we can use reference just like file + /// let reference = file.get_mut(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_mut(&mut self) -> &mut W { + self.inner.get_mut() + } + + /// Unwraps this `LineWriter`, returning the underlying writer. + /// + /// The internal buffer is written out before returning the writer. + /// + /// # Errors + /// + /// An [`Err`] will be returned if an error occurs while flushing the buffer. + /// + /// # Examples + /// + /// ```no_run + /// use std::fs::File; + /// use std::io::LineWriter; + /// + /// fn main() -> std::io::Result<()> { + /// let file = File::create("poem.txt")?; + /// + /// let writer: LineWriter<File> = LineWriter::new(file); + /// + /// let file: File = writer.into_inner()?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn into_inner(self) -> Result<W, IntoInnerError<LineWriter<W>>> { + self.inner.into_inner().map_err(|err| err.new_wrapped(|inner| LineWriter { inner })) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write> Write for LineWriter<W> { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + LineWriterShim::new(&mut self.inner).write(buf) + } + + fn flush(&mut self) -> io::Result<()> { + self.inner.flush() + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + LineWriterShim::new(&mut self.inner).write_vectored(bufs) + } + + fn is_write_vectored(&self) -> bool { + self.inner.is_write_vectored() + } + + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + LineWriterShim::new(&mut self.inner).write_all(buf) + } + + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + LineWriterShim::new(&mut self.inner).write_all_vectored(bufs) + } + + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> io::Result<()> { + LineWriterShim::new(&mut self.inner).write_fmt(fmt) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write> fmt::Debug for LineWriter<W> +where + W: fmt::Debug, +{ + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.debug_struct("LineWriter") + .field("writer", &self.get_ref()) + .field( + "buffer", + &format_args!("{}/{}", self.inner.buffer().len(), self.inner.capacity()), + ) + .finish_non_exhaustive() + } +} diff --git a/library/std/src/io/buffered/linewritershim.rs b/library/std/src/io/buffered/linewritershim.rs new file mode 100644 index 000000000..0175d2693 --- /dev/null +++ b/library/std/src/io/buffered/linewritershim.rs @@ -0,0 +1,276 @@ +use crate::io::{self, BufWriter, IoSlice, Write}; +use crate::sys_common::memchr; + +/// Private helper struct for implementing the line-buffered writing logic. +/// This shim temporarily wraps a BufWriter, and uses its internals to +/// implement a line-buffered writer (specifically by using the internal +/// methods like write_to_buf and flush_buf). In this way, a more +/// efficient abstraction can be created than one that only had access to +/// `write` and `flush`, without needlessly duplicating a lot of the +/// implementation details of BufWriter. This also allows existing +/// `BufWriters` to be temporarily given line-buffering logic; this is what +/// enables Stdout to be alternately in line-buffered or block-buffered mode. +#[derive(Debug)] +pub struct LineWriterShim<'a, W: Write> { + buffer: &'a mut BufWriter<W>, +} + +impl<'a, W: Write> LineWriterShim<'a, W> { + pub fn new(buffer: &'a mut BufWriter<W>) -> Self { + Self { buffer } + } + + /// Get a reference to the inner writer (that is, the writer + /// wrapped by the BufWriter). + fn inner(&self) -> &W { + self.buffer.get_ref() + } + + /// Get a mutable reference to the inner writer (that is, the writer + /// wrapped by the BufWriter). Be careful with this writer, as writes to + /// it will bypass the buffer. + fn inner_mut(&mut self) -> &mut W { + self.buffer.get_mut() + } + + /// Get the content currently buffered in self.buffer + fn buffered(&self) -> &[u8] { + self.buffer.buffer() + } + + /// Flush the buffer iff the last byte is a newline (indicating that an + /// earlier write only succeeded partially, and we want to retry flushing + /// the buffered line before continuing with a subsequent write) + fn flush_if_completed_line(&mut self) -> io::Result<()> { + match self.buffered().last().copied() { + Some(b'\n') => self.buffer.flush_buf(), + _ => Ok(()), + } + } +} + +impl<'a, W: Write> Write for LineWriterShim<'a, W> { + /// Write some data into this BufReader with line buffering. This means + /// that, if any newlines are present in the data, the data up to the last + /// newline is sent directly to the underlying writer, and data after it + /// is buffered. Returns the number of bytes written. + /// + /// This function operates on a "best effort basis"; in keeping with the + /// convention of `Write::write`, it makes at most one attempt to write + /// new data to the underlying writer. If that write only reports a partial + /// success, the remaining data will be buffered. + /// + /// Because this function attempts to send completed lines to the underlying + /// writer, it will also flush the existing buffer if it ends with a + /// newline, even if the incoming data does not contain any newlines. + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + let newline_idx = match memchr::memrchr(b'\n', buf) { + // If there are no new newlines (that is, if this write is less than + // one line), just do a regular buffered write (which may flush if + // we exceed the inner buffer's size) + None => { + self.flush_if_completed_line()?; + return self.buffer.write(buf); + } + // Otherwise, arrange for the lines to be written directly to the + // inner writer. + Some(newline_idx) => newline_idx + 1, + }; + + // Flush existing content to prepare for our write. We have to do this + // before attempting to write `buf` in order to maintain consistency; + // if we add `buf` to the buffer then try to flush it all at once, + // we're obligated to return Ok(), which would mean suppressing any + // errors that occur during flush. + self.buffer.flush_buf()?; + + // This is what we're going to try to write directly to the inner + // writer. The rest will be buffered, if nothing goes wrong. + let lines = &buf[..newline_idx]; + + // Write `lines` directly to the inner writer. In keeping with the + // `write` convention, make at most one attempt to add new (unbuffered) + // data. Because this write doesn't touch the BufWriter state directly, + // and the buffer is known to be empty, we don't need to worry about + // self.buffer.panicked here. + let flushed = self.inner_mut().write(lines)?; + + // If buffer returns Ok(0), propagate that to the caller without + // doing additional buffering; otherwise we're just guaranteeing + // an "ErrorKind::WriteZero" later. + if flushed == 0 { + return Ok(0); + } + + // Now that the write has succeeded, buffer the rest (or as much of + // the rest as possible). If there were any unwritten newlines, we + // only buffer out to the last unwritten newline that fits in the + // buffer; this helps prevent flushing partial lines on subsequent + // calls to LineWriterShim::write. + + // Handle the cases in order of most-common to least-common, under + // the presumption that most writes succeed in totality, and that most + // writes are smaller than the buffer. + // - Is this a partial line (ie, no newlines left in the unwritten tail) + // - If not, does the data out to the last unwritten newline fit in + // the buffer? + // - If not, scan for the last newline that *does* fit in the buffer + let tail = if flushed >= newline_idx { + &buf[flushed..] + } else if newline_idx - flushed <= self.buffer.capacity() { + &buf[flushed..newline_idx] + } else { + let scan_area = &buf[flushed..]; + let scan_area = &scan_area[..self.buffer.capacity()]; + match memchr::memrchr(b'\n', scan_area) { + Some(newline_idx) => &scan_area[..newline_idx + 1], + None => scan_area, + } + }; + + let buffered = self.buffer.write_to_buf(tail); + Ok(flushed + buffered) + } + + fn flush(&mut self) -> io::Result<()> { + self.buffer.flush() + } + + /// Write some vectored data into this BufReader with line buffering. This + /// means that, if any newlines are present in the data, the data up to + /// and including the buffer containing the last newline is sent directly + /// to the inner writer, and the data after it is buffered. Returns the + /// number of bytes written. + /// + /// This function operates on a "best effort basis"; in keeping with the + /// convention of `Write::write`, it makes at most one attempt to write + /// new data to the underlying writer. + /// + /// Because this function attempts to send completed lines to the underlying + /// writer, it will also flush the existing buffer if it contains any + /// newlines. + /// + /// Because sorting through an array of `IoSlice` can be a bit convoluted, + /// This method differs from write in the following ways: + /// + /// - It attempts to write the full content of all the buffers up to and + /// including the one containing the last newline. This means that it + /// may attempt to write a partial line, that buffer has data past the + /// newline. + /// - If the write only reports partial success, it does not attempt to + /// find the precise location of the written bytes and buffer the rest. + /// + /// If the underlying vector doesn't support vectored writing, we instead + /// simply write the first non-empty buffer with `write`. This way, we + /// get the benefits of more granular partial-line handling without losing + /// anything in efficiency + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + // If there's no specialized behavior for write_vectored, just use + // write. This has the benefit of more granular partial-line handling. + if !self.is_write_vectored() { + return match bufs.iter().find(|buf| !buf.is_empty()) { + Some(buf) => self.write(buf), + None => Ok(0), + }; + } + + // Find the buffer containing the last newline + let last_newline_buf_idx = bufs + .iter() + .enumerate() + .rev() + .find_map(|(i, buf)| memchr::memchr(b'\n', buf).map(|_| i)); + + // If there are no new newlines (that is, if this write is less than + // one line), just do a regular buffered write + let last_newline_buf_idx = match last_newline_buf_idx { + // No newlines; just do a normal buffered write + None => { + self.flush_if_completed_line()?; + return self.buffer.write_vectored(bufs); + } + Some(i) => i, + }; + + // Flush existing content to prepare for our write + self.buffer.flush_buf()?; + + // This is what we're going to try to write directly to the inner + // writer. The rest will be buffered, if nothing goes wrong. + let (lines, tail) = bufs.split_at(last_newline_buf_idx + 1); + + // Write `lines` directly to the inner writer. In keeping with the + // `write` convention, make at most one attempt to add new (unbuffered) + // data. Because this write doesn't touch the BufWriter state directly, + // and the buffer is known to be empty, we don't need to worry about + // self.panicked here. + let flushed = self.inner_mut().write_vectored(lines)?; + + // If inner returns Ok(0), propagate that to the caller without + // doing additional buffering; otherwise we're just guaranteeing + // an "ErrorKind::WriteZero" later. + if flushed == 0 { + return Ok(0); + } + + // Don't try to reconstruct the exact amount written; just bail + // in the event of a partial write + let lines_len = lines.iter().map(|buf| buf.len()).sum(); + if flushed < lines_len { + return Ok(flushed); + } + + // Now that the write has succeeded, buffer the rest (or as much of the + // rest as possible) + let buffered: usize = tail + .iter() + .filter(|buf| !buf.is_empty()) + .map(|buf| self.buffer.write_to_buf(buf)) + .take_while(|&n| n > 0) + .sum(); + + Ok(flushed + buffered) + } + + fn is_write_vectored(&self) -> bool { + self.inner().is_write_vectored() + } + + /// Write some data into this BufReader with line buffering. This means + /// that, if any newlines are present in the data, the data up to the last + /// newline is sent directly to the underlying writer, and data after it + /// is buffered. + /// + /// Because this function attempts to send completed lines to the underlying + /// writer, it will also flush the existing buffer if it contains any + /// newlines, even if the incoming data does not contain any newlines. + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + match memchr::memrchr(b'\n', buf) { + // If there are no new newlines (that is, if this write is less than + // one line), just do a regular buffered write (which may flush if + // we exceed the inner buffer's size) + None => { + self.flush_if_completed_line()?; + self.buffer.write_all(buf) + } + Some(newline_idx) => { + let (lines, tail) = buf.split_at(newline_idx + 1); + + if self.buffered().is_empty() { + self.inner_mut().write_all(lines)?; + } else { + // If there is any buffered data, we add the incoming lines + // to that buffer before flushing, which saves us at least + // one write call. We can't really do this with `write`, + // since we can't do this *and* not suppress errors *and* + // report a consistent state to the caller in a return + // value, but here in write_all it's fine. + self.buffer.write_all(lines)?; + self.buffer.flush_buf()?; + } + + self.buffer.write_all(tail) + } + } + } +} diff --git a/library/std/src/io/buffered/mod.rs b/library/std/src/io/buffered/mod.rs new file mode 100644 index 000000000..100dab1e2 --- /dev/null +++ b/library/std/src/io/buffered/mod.rs @@ -0,0 +1,196 @@ +//! Buffering wrappers for I/O traits + +mod bufreader; +mod bufwriter; +mod linewriter; +mod linewritershim; + +#[cfg(test)] +mod tests; + +use crate::error; +use crate::fmt; +use crate::io::Error; + +#[stable(feature = "rust1", since = "1.0.0")] +pub use self::{bufreader::BufReader, bufwriter::BufWriter, linewriter::LineWriter}; +use linewritershim::LineWriterShim; + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +pub use bufwriter::WriterPanicked; + +/// An error returned by [`BufWriter::into_inner`] which combines an error that +/// happened while writing out the buffer, and the buffered writer object +/// which may be used to recover from the condition. +/// +/// # Examples +/// +/// ```no_run +/// use std::io::BufWriter; +/// use std::net::TcpStream; +/// +/// let mut stream = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); +/// +/// // do stuff with the stream +/// +/// // we want to get our `TcpStream` back, so let's try: +/// +/// let stream = match stream.into_inner() { +/// Ok(s) => s, +/// Err(e) => { +/// // Here, e is an IntoInnerError +/// panic!("An error occurred"); +/// } +/// }; +/// ``` +#[derive(Debug)] +#[stable(feature = "rust1", since = "1.0.0")] +pub struct IntoInnerError<W>(W, Error); + +impl<W> IntoInnerError<W> { + /// Construct a new IntoInnerError + fn new(writer: W, error: Error) -> Self { + Self(writer, error) + } + + /// Helper to construct a new IntoInnerError; intended to help with + /// adapters that wrap other adapters + fn new_wrapped<W2>(self, f: impl FnOnce(W) -> W2) -> IntoInnerError<W2> { + let Self(writer, error) = self; + IntoInnerError::new(f(writer), error) + } + + /// Returns the error which caused the call to [`BufWriter::into_inner()`] + /// to fail. + /// + /// This error was returned when attempting to write the internal buffer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut stream = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // do stuff with the stream + /// + /// // we want to get our `TcpStream` back, so let's try: + /// + /// let stream = match stream.into_inner() { + /// Ok(s) => s, + /// Err(e) => { + /// // Here, e is an IntoInnerError, let's log the inner error. + /// // + /// // We'll just 'log' to stdout for this example. + /// println!("{}", e.error()); + /// + /// panic!("An unexpected error occurred."); + /// } + /// }; + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn error(&self) -> &Error { + &self.1 + } + + /// Returns the buffered writer instance which generated the error. + /// + /// The returned object can be used for error recovery, such as + /// re-inspecting the buffer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufWriter; + /// use std::net::TcpStream; + /// + /// let mut stream = BufWriter::new(TcpStream::connect("127.0.0.1:34254").unwrap()); + /// + /// // do stuff with the stream + /// + /// // we want to get our `TcpStream` back, so let's try: + /// + /// let stream = match stream.into_inner() { + /// Ok(s) => s, + /// Err(e) => { + /// // Here, e is an IntoInnerError, let's re-examine the buffer: + /// let buffer = e.into_inner(); + /// + /// // do stuff to try to recover + /// + /// // afterwards, let's just return the stream + /// buffer.into_inner().unwrap() + /// } + /// }; + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn into_inner(self) -> W { + self.0 + } + + /// Consumes the [`IntoInnerError`] and returns the error which caused the call to + /// [`BufWriter::into_inner()`] to fail. Unlike `error`, this can be used to + /// obtain ownership of the underlying error. + /// + /// # Example + /// ``` + /// use std::io::{BufWriter, ErrorKind, Write}; + /// + /// let mut not_enough_space = [0u8; 10]; + /// let mut stream = BufWriter::new(not_enough_space.as_mut()); + /// write!(stream, "this cannot be actually written").unwrap(); + /// let into_inner_err = stream.into_inner().expect_err("now we discover it's too small"); + /// let err = into_inner_err.into_error(); + /// assert_eq!(err.kind(), ErrorKind::WriteZero); + /// ``` + #[stable(feature = "io_into_inner_error_parts", since = "1.55.0")] + pub fn into_error(self) -> Error { + self.1 + } + + /// Consumes the [`IntoInnerError`] and returns the error which caused the call to + /// [`BufWriter::into_inner()`] to fail, and the underlying writer. + /// + /// This can be used to simply obtain ownership of the underlying error; it can also be used for + /// advanced error recovery. + /// + /// # Example + /// ``` + /// use std::io::{BufWriter, ErrorKind, Write}; + /// + /// let mut not_enough_space = [0u8; 10]; + /// let mut stream = BufWriter::new(not_enough_space.as_mut()); + /// write!(stream, "this cannot be actually written").unwrap(); + /// let into_inner_err = stream.into_inner().expect_err("now we discover it's too small"); + /// let (err, recovered_writer) = into_inner_err.into_parts(); + /// assert_eq!(err.kind(), ErrorKind::WriteZero); + /// assert_eq!(recovered_writer.buffer(), b"t be actually written"); + /// ``` + #[stable(feature = "io_into_inner_error_parts", since = "1.55.0")] + pub fn into_parts(self) -> (Error, W) { + (self.1, self.0) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W> From<IntoInnerError<W>> for Error { + fn from(iie: IntoInnerError<W>) -> Error { + iie.1 + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Send + fmt::Debug> error::Error for IntoInnerError<W> { + #[allow(deprecated, deprecated_in_future)] + fn description(&self) -> &str { + error::Error::description(self.error()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<W> fmt::Display for IntoInnerError<W> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.error().fmt(f) + } +} diff --git a/library/std/src/io/buffered/tests.rs b/library/std/src/io/buffered/tests.rs new file mode 100644 index 000000000..fe45b1326 --- /dev/null +++ b/library/std/src/io/buffered/tests.rs @@ -0,0 +1,1039 @@ +use crate::io::prelude::*; +use crate::io::{self, BufReader, BufWriter, ErrorKind, IoSlice, LineWriter, ReadBuf, SeekFrom}; +use crate::mem::MaybeUninit; +use crate::panic; +use crate::sync::atomic::{AtomicUsize, Ordering}; +use crate::thread; + +/// A dummy reader intended at testing short-reads propagation. +pub struct ShortReader { + lengths: Vec<usize>, +} + +// FIXME: rustfmt and tidy disagree about the correct formatting of this +// function. This leads to issues for users with editors configured to +// rustfmt-on-save. +impl Read for ShortReader { + fn read(&mut self, _: &mut [u8]) -> io::Result<usize> { + if self.lengths.is_empty() { Ok(0) } else { Ok(self.lengths.remove(0)) } + } +} + +#[test] +fn test_buffered_reader() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(2, inner); + + let mut buf = [0, 0, 0]; + let nread = reader.read(&mut buf); + assert_eq!(nread.unwrap(), 3); + assert_eq!(buf, [5, 6, 7]); + assert_eq!(reader.buffer(), []); + + let mut buf = [0, 0]; + let nread = reader.read(&mut buf); + assert_eq!(nread.unwrap(), 2); + assert_eq!(buf, [0, 1]); + assert_eq!(reader.buffer(), []); + + let mut buf = [0]; + let nread = reader.read(&mut buf); + assert_eq!(nread.unwrap(), 1); + assert_eq!(buf, [2]); + assert_eq!(reader.buffer(), [3]); + + let mut buf = [0, 0, 0]; + let nread = reader.read(&mut buf); + assert_eq!(nread.unwrap(), 1); + assert_eq!(buf, [3, 0, 0]); + assert_eq!(reader.buffer(), []); + + let nread = reader.read(&mut buf); + assert_eq!(nread.unwrap(), 1); + assert_eq!(buf, [4, 0, 0]); + assert_eq!(reader.buffer(), []); + + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_buffered_reader_read_buf() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(2, inner); + + let mut buf = [MaybeUninit::uninit(); 3]; + let mut buf = ReadBuf::uninit(&mut buf); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled(), [5, 6, 7]); + assert_eq!(reader.buffer(), []); + + let mut buf = [MaybeUninit::uninit(); 2]; + let mut buf = ReadBuf::uninit(&mut buf); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled(), [0, 1]); + assert_eq!(reader.buffer(), []); + + let mut buf = [MaybeUninit::uninit(); 1]; + let mut buf = ReadBuf::uninit(&mut buf); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled(), [2]); + assert_eq!(reader.buffer(), [3]); + + let mut buf = [MaybeUninit::uninit(); 3]; + let mut buf = ReadBuf::uninit(&mut buf); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled(), [3]); + assert_eq!(reader.buffer(), []); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled(), [3, 4]); + assert_eq!(reader.buffer(), []); + + buf.clear(); + + reader.read_buf(&mut buf).unwrap(); + + assert_eq!(buf.filled_len(), 0); +} + +#[test] +fn test_buffered_reader_seek() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(2, io::Cursor::new(inner)); + + assert_eq!(reader.seek(SeekFrom::Start(3)).ok(), Some(3)); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert_eq!(reader.seek(SeekFrom::Current(0)).ok(), Some(3)); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert_eq!(reader.seek(SeekFrom::Current(1)).ok(), Some(4)); + assert_eq!(reader.fill_buf().ok(), Some(&[1, 2][..])); + reader.consume(1); + assert_eq!(reader.seek(SeekFrom::Current(-2)).ok(), Some(3)); +} + +#[test] +fn test_buffered_reader_seek_relative() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(2, io::Cursor::new(inner)); + + assert!(reader.seek_relative(3).is_ok()); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert!(reader.seek_relative(0).is_ok()); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert!(reader.seek_relative(1).is_ok()); + assert_eq!(reader.fill_buf().ok(), Some(&[1][..])); + assert!(reader.seek_relative(-1).is_ok()); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert!(reader.seek_relative(2).is_ok()); + assert_eq!(reader.fill_buf().ok(), Some(&[2, 3][..])); +} + +#[test] +fn test_buffered_reader_stream_position() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(2, io::Cursor::new(inner)); + + assert_eq!(reader.stream_position().ok(), Some(0)); + assert_eq!(reader.seek(SeekFrom::Start(3)).ok(), Some(3)); + assert_eq!(reader.stream_position().ok(), Some(3)); + // relative seeking within the buffer and reading position should keep the buffer + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1][..])); + assert!(reader.seek_relative(0).is_ok()); + assert_eq!(reader.stream_position().ok(), Some(3)); + assert_eq!(reader.buffer(), &[0, 1][..]); + assert!(reader.seek_relative(1).is_ok()); + assert_eq!(reader.stream_position().ok(), Some(4)); + assert_eq!(reader.buffer(), &[1][..]); + assert!(reader.seek_relative(-1).is_ok()); + assert_eq!(reader.stream_position().ok(), Some(3)); + assert_eq!(reader.buffer(), &[0, 1][..]); + // relative seeking outside the buffer will discard it + assert!(reader.seek_relative(2).is_ok()); + assert_eq!(reader.stream_position().ok(), Some(5)); + assert_eq!(reader.buffer(), &[][..]); +} + +#[test] +fn test_buffered_reader_stream_position_panic() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(4, io::Cursor::new(inner)); + + // cause internal buffer to be filled but read only partially + let mut buffer = [0, 0]; + assert!(reader.read_exact(&mut buffer).is_ok()); + // rewinding the internal reader will cause buffer to loose sync + let inner = reader.get_mut(); + assert!(inner.seek(SeekFrom::Start(0)).is_ok()); + // overflow when subtracting the remaining buffer size from current position + let result = panic::catch_unwind(panic::AssertUnwindSafe(|| reader.stream_position().ok())); + assert!(result.is_err()); +} + +#[test] +fn test_buffered_reader_invalidated_after_read() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(3, io::Cursor::new(inner)); + + assert_eq!(reader.fill_buf().ok(), Some(&[5, 6, 7][..])); + reader.consume(3); + + let mut buffer = [0, 0, 0, 0, 0]; + assert_eq!(reader.read(&mut buffer).ok(), Some(5)); + assert_eq!(buffer, [0, 1, 2, 3, 4]); + + assert!(reader.seek_relative(-2).is_ok()); + let mut buffer = [0, 0]; + assert_eq!(reader.read(&mut buffer).ok(), Some(2)); + assert_eq!(buffer, [3, 4]); +} + +#[test] +fn test_buffered_reader_invalidated_after_seek() { + let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; + let mut reader = BufReader::with_capacity(3, io::Cursor::new(inner)); + + assert_eq!(reader.fill_buf().ok(), Some(&[5, 6, 7][..])); + reader.consume(3); + + assert!(reader.seek(SeekFrom::Current(5)).is_ok()); + + assert!(reader.seek_relative(-2).is_ok()); + let mut buffer = [0, 0]; + assert_eq!(reader.read(&mut buffer).ok(), Some(2)); + assert_eq!(buffer, [3, 4]); +} + +#[test] +fn test_buffered_reader_seek_underflow() { + // gimmick reader that yields its position modulo 256 for each byte + struct PositionReader { + pos: u64, + } + impl Read for PositionReader { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + let len = buf.len(); + for x in buf { + *x = self.pos as u8; + self.pos = self.pos.wrapping_add(1); + } + Ok(len) + } + } + impl Seek for PositionReader { + fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> { + match pos { + SeekFrom::Start(n) => { + self.pos = n; + } + SeekFrom::Current(n) => { + self.pos = self.pos.wrapping_add(n as u64); + } + SeekFrom::End(n) => { + self.pos = u64::MAX.wrapping_add(n as u64); + } + } + Ok(self.pos) + } + } + + let mut reader = BufReader::with_capacity(5, PositionReader { pos: 0 }); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1, 2, 3, 4][..])); + assert_eq!(reader.seek(SeekFrom::End(-5)).ok(), Some(u64::MAX - 5)); + assert_eq!(reader.fill_buf().ok().map(|s| s.len()), Some(5)); + // the following seek will require two underlying seeks + let expected = 9223372036854775802; + assert_eq!(reader.seek(SeekFrom::Current(i64::MIN)).ok(), Some(expected)); + assert_eq!(reader.fill_buf().ok().map(|s| s.len()), Some(5)); + // seeking to 0 should empty the buffer. + assert_eq!(reader.seek(SeekFrom::Current(0)).ok(), Some(expected)); + assert_eq!(reader.get_ref().pos, expected); +} + +#[test] +fn test_buffered_reader_seek_underflow_discard_buffer_between_seeks() { + // gimmick reader that returns Err after first seek + struct ErrAfterFirstSeekReader { + first_seek: bool, + } + impl Read for ErrAfterFirstSeekReader { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + for x in &mut *buf { + *x = 0; + } + Ok(buf.len()) + } + } + impl Seek for ErrAfterFirstSeekReader { + fn seek(&mut self, _: SeekFrom) -> io::Result<u64> { + if self.first_seek { + self.first_seek = false; + Ok(0) + } else { + Err(io::Error::new(io::ErrorKind::Other, "oh no!")) + } + } + } + + let mut reader = BufReader::with_capacity(5, ErrAfterFirstSeekReader { first_seek: true }); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 0, 0, 0, 0][..])); + + // The following seek will require two underlying seeks. The first will + // succeed but the second will fail. This should still invalidate the + // buffer. + assert!(reader.seek(SeekFrom::Current(i64::MIN)).is_err()); + assert_eq!(reader.buffer().len(), 0); +} + +#[test] +fn test_buffered_reader_read_to_end_consumes_buffer() { + let data: &[u8] = &[0, 1, 2, 3, 4, 5, 6, 7]; + let mut reader = BufReader::with_capacity(3, data); + let mut buf = Vec::new(); + assert_eq!(reader.fill_buf().ok(), Some(&[0, 1, 2][..])); + assert_eq!(reader.read_to_end(&mut buf).ok(), Some(8)); + assert_eq!(&buf, &[0, 1, 2, 3, 4, 5, 6, 7]); + assert!(reader.buffer().is_empty()); +} + +#[test] +fn test_buffered_reader_read_to_string_consumes_buffer() { + let data: &[u8] = "deadbeef".as_bytes(); + let mut reader = BufReader::with_capacity(3, data); + let mut buf = String::new(); + assert_eq!(reader.fill_buf().ok(), Some("dea".as_bytes())); + assert_eq!(reader.read_to_string(&mut buf).ok(), Some(8)); + assert_eq!(&buf, "deadbeef"); + assert!(reader.buffer().is_empty()); +} + +#[test] +fn test_buffered_writer() { + let inner = Vec::new(); + let mut writer = BufWriter::with_capacity(2, inner); + + writer.write(&[0, 1]).unwrap(); + assert_eq!(writer.buffer(), []); + assert_eq!(*writer.get_ref(), [0, 1]); + + writer.write(&[2]).unwrap(); + assert_eq!(writer.buffer(), [2]); + assert_eq!(*writer.get_ref(), [0, 1]); + + writer.write(&[3]).unwrap(); + assert_eq!(writer.buffer(), [2, 3]); + assert_eq!(*writer.get_ref(), [0, 1]); + + writer.flush().unwrap(); + assert_eq!(writer.buffer(), []); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3]); + + writer.write(&[4]).unwrap(); + writer.write(&[5]).unwrap(); + assert_eq!(writer.buffer(), [4, 5]); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3]); + + writer.write(&[6]).unwrap(); + assert_eq!(writer.buffer(), [6]); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3, 4, 5]); + + writer.write(&[7, 8]).unwrap(); + assert_eq!(writer.buffer(), []); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3, 4, 5, 6, 7, 8]); + + writer.write(&[9, 10, 11]).unwrap(); + assert_eq!(writer.buffer(), []); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]); + + writer.flush().unwrap(); + assert_eq!(writer.buffer(), []); + assert_eq!(*writer.get_ref(), [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]); +} + +#[test] +fn test_buffered_writer_inner_flushes() { + let mut w = BufWriter::with_capacity(3, Vec::new()); + w.write(&[0, 1]).unwrap(); + assert_eq!(*w.get_ref(), []); + let w = w.into_inner().unwrap(); + assert_eq!(w, [0, 1]); +} + +#[test] +fn test_buffered_writer_seek() { + let mut w = BufWriter::with_capacity(3, io::Cursor::new(Vec::new())); + w.write_all(&[0, 1, 2, 3, 4, 5]).unwrap(); + w.write_all(&[6, 7]).unwrap(); + assert_eq!(w.seek(SeekFrom::Current(0)).ok(), Some(8)); + assert_eq!(&w.get_ref().get_ref()[..], &[0, 1, 2, 3, 4, 5, 6, 7][..]); + assert_eq!(w.seek(SeekFrom::Start(2)).ok(), Some(2)); + w.write_all(&[8, 9]).unwrap(); + assert_eq!(&w.into_inner().unwrap().into_inner()[..], &[0, 1, 8, 9, 4, 5, 6, 7]); +} + +#[test] +fn test_read_until() { + let inner: &[u8] = &[0, 1, 2, 1, 0]; + let mut reader = BufReader::with_capacity(2, inner); + let mut v = Vec::new(); + reader.read_until(0, &mut v).unwrap(); + assert_eq!(v, [0]); + v.truncate(0); + reader.read_until(2, &mut v).unwrap(); + assert_eq!(v, [1, 2]); + v.truncate(0); + reader.read_until(1, &mut v).unwrap(); + assert_eq!(v, [1]); + v.truncate(0); + reader.read_until(8, &mut v).unwrap(); + assert_eq!(v, [0]); + v.truncate(0); + reader.read_until(9, &mut v).unwrap(); + assert_eq!(v, []); +} + +#[test] +fn test_line_buffer() { + let mut writer = LineWriter::new(Vec::new()); + writer.write(&[0]).unwrap(); + assert_eq!(*writer.get_ref(), []); + writer.write(&[1]).unwrap(); + assert_eq!(*writer.get_ref(), []); + writer.flush().unwrap(); + assert_eq!(*writer.get_ref(), [0, 1]); + writer.write(&[0, b'\n', 1, b'\n', 2]).unwrap(); + assert_eq!(*writer.get_ref(), [0, 1, 0, b'\n', 1, b'\n']); + writer.flush().unwrap(); + assert_eq!(*writer.get_ref(), [0, 1, 0, b'\n', 1, b'\n', 2]); + writer.write(&[3, b'\n']).unwrap(); + assert_eq!(*writer.get_ref(), [0, 1, 0, b'\n', 1, b'\n', 2, 3, b'\n']); +} + +#[test] +fn test_read_line() { + let in_buf: &[u8] = b"a\nb\nc"; + let mut reader = BufReader::with_capacity(2, in_buf); + let mut s = String::new(); + reader.read_line(&mut s).unwrap(); + assert_eq!(s, "a\n"); + s.truncate(0); + reader.read_line(&mut s).unwrap(); + assert_eq!(s, "b\n"); + s.truncate(0); + reader.read_line(&mut s).unwrap(); + assert_eq!(s, "c"); + s.truncate(0); + reader.read_line(&mut s).unwrap(); + assert_eq!(s, ""); +} + +#[test] +fn test_lines() { + let in_buf: &[u8] = b"a\nb\nc"; + let reader = BufReader::with_capacity(2, in_buf); + let mut it = reader.lines(); + assert_eq!(it.next().unwrap().unwrap(), "a".to_string()); + assert_eq!(it.next().unwrap().unwrap(), "b".to_string()); + assert_eq!(it.next().unwrap().unwrap(), "c".to_string()); + assert!(it.next().is_none()); +} + +#[test] +fn test_short_reads() { + let inner = ShortReader { lengths: vec![0, 1, 2, 0, 1, 0] }; + let mut reader = BufReader::new(inner); + let mut buf = [0, 0]; + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.read(&mut buf).unwrap(), 2); + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +#[should_panic] +fn dont_panic_in_drop_on_panicked_flush() { + struct FailFlushWriter; + + impl Write for FailFlushWriter { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + Ok(buf.len()) + } + fn flush(&mut self) -> io::Result<()> { + Err(io::Error::last_os_error()) + } + } + + let writer = FailFlushWriter; + let _writer = BufWriter::new(writer); + + // If writer panics *again* due to the flush error then the process will + // abort. + panic!(); +} + +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn panic_in_write_doesnt_flush_in_drop() { + static WRITES: AtomicUsize = AtomicUsize::new(0); + + struct PanicWriter; + + impl Write for PanicWriter { + fn write(&mut self, _: &[u8]) -> io::Result<usize> { + WRITES.fetch_add(1, Ordering::SeqCst); + panic!(); + } + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } + } + + thread::spawn(|| { + let mut writer = BufWriter::new(PanicWriter); + let _ = writer.write(b"hello world"); + let _ = writer.flush(); + }) + .join() + .unwrap_err(); + + assert_eq!(WRITES.load(Ordering::SeqCst), 1); +} + +#[bench] +fn bench_buffered_reader(b: &mut test::Bencher) { + b.iter(|| BufReader::new(io::empty())); +} + +#[bench] +fn bench_buffered_reader_small_reads(b: &mut test::Bencher) { + let data = (0..u8::MAX).cycle().take(1024 * 4).collect::<Vec<_>>(); + b.iter(|| { + let mut reader = BufReader::new(&data[..]); + let mut buf = [0u8; 4]; + for _ in 0..1024 { + reader.read_exact(&mut buf).unwrap(); + core::hint::black_box(&buf); + } + }); +} + +#[bench] +fn bench_buffered_writer(b: &mut test::Bencher) { + b.iter(|| BufWriter::new(io::sink())); +} + +/// A simple `Write` target, designed to be wrapped by `LineWriter` / +/// `BufWriter` / etc, that can have its `write` & `flush` behavior +/// configured +#[derive(Default, Clone)] +struct ProgrammableSink { + // Writes append to this slice + pub buffer: Vec<u8>, + + // If true, writes will always be an error + pub always_write_error: bool, + + // If true, flushes will always be an error + pub always_flush_error: bool, + + // If set, only up to this number of bytes will be written in a single + // call to `write` + pub accept_prefix: Option<usize>, + + // If set, counts down with each write, and writes return an error + // when it hits 0 + pub max_writes: Option<usize>, + + // If set, attempting to write when max_writes == Some(0) will be an + // error; otherwise, it will return Ok(0). + pub error_after_max_writes: bool, +} + +impl Write for ProgrammableSink { + fn write(&mut self, data: &[u8]) -> io::Result<usize> { + if self.always_write_error { + return Err(io::Error::new(io::ErrorKind::Other, "test - always_write_error")); + } + + match self.max_writes { + Some(0) if self.error_after_max_writes => { + return Err(io::Error::new(io::ErrorKind::Other, "test - max_writes")); + } + Some(0) => return Ok(0), + Some(ref mut count) => *count -= 1, + None => {} + } + + let len = match self.accept_prefix { + None => data.len(), + Some(prefix) => data.len().min(prefix), + }; + + let data = &data[..len]; + self.buffer.extend_from_slice(data); + + Ok(len) + } + + fn flush(&mut self) -> io::Result<()> { + if self.always_flush_error { + Err(io::Error::new(io::ErrorKind::Other, "test - always_flush_error")) + } else { + Ok(()) + } + } +} + +/// Previously the `LineWriter` could successfully write some bytes but +/// then fail to report that it has done so. Additionally, an erroneous +/// flush after a successful write was permanently ignored. +/// +/// Test that a line writer correctly reports the number of written bytes, +/// and that it attempts to flush buffered lines from previous writes +/// before processing new data +/// +/// Regression test for #37807 +#[test] +fn erroneous_flush_retried() { + let writer = ProgrammableSink { + // Only write up to 4 bytes at a time + accept_prefix: Some(4), + + // Accept the first two writes, then error the others + max_writes: Some(2), + error_after_max_writes: true, + + ..Default::default() + }; + + // This should write the first 4 bytes. The rest will be buffered, out + // to the last newline. + let mut writer = LineWriter::new(writer); + assert_eq!(writer.write(b"a\nb\nc\nd\ne").unwrap(), 8); + + // This write should attempt to flush "c\nd\n", then buffer "e". No + // errors should happen here because no further writes should be + // attempted against `writer`. + assert_eq!(writer.write(b"e").unwrap(), 1); + assert_eq!(&writer.get_ref().buffer, b"a\nb\nc\nd\n"); +} + +#[test] +fn line_vectored() { + let mut a = LineWriter::new(Vec::new()); + assert_eq!( + a.write_vectored(&[ + IoSlice::new(&[]), + IoSlice::new(b"\n"), + IoSlice::new(&[]), + IoSlice::new(b"a"), + ]) + .unwrap(), + 2, + ); + assert_eq!(a.get_ref(), b"\n"); + + assert_eq!( + a.write_vectored(&[ + IoSlice::new(&[]), + IoSlice::new(b"b"), + IoSlice::new(&[]), + IoSlice::new(b"a"), + IoSlice::new(&[]), + IoSlice::new(b"c"), + ]) + .unwrap(), + 3, + ); + assert_eq!(a.get_ref(), b"\n"); + a.flush().unwrap(); + assert_eq!(a.get_ref(), b"\nabac"); + assert_eq!(a.write_vectored(&[]).unwrap(), 0); + assert_eq!( + a.write_vectored(&[ + IoSlice::new(&[]), + IoSlice::new(&[]), + IoSlice::new(&[]), + IoSlice::new(&[]), + ]) + .unwrap(), + 0, + ); + assert_eq!(a.write_vectored(&[IoSlice::new(b"a\nb"),]).unwrap(), 3); + assert_eq!(a.get_ref(), b"\nabaca\nb"); +} + +#[test] +fn line_vectored_partial_and_errors() { + use crate::collections::VecDeque; + + enum Call { + Write { inputs: Vec<&'static [u8]>, output: io::Result<usize> }, + Flush { output: io::Result<()> }, + } + + #[derive(Default)] + struct Writer { + calls: VecDeque<Call>, + } + + impl Write for Writer { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.write_vectored(&[IoSlice::new(buf)]) + } + + fn write_vectored(&mut self, buf: &[IoSlice<'_>]) -> io::Result<usize> { + match self.calls.pop_front().expect("unexpected call to write") { + Call::Write { inputs, output } => { + assert_eq!(inputs, buf.iter().map(|b| &**b).collect::<Vec<_>>()); + output + } + Call::Flush { .. } => panic!("unexpected call to write; expected a flush"), + } + } + + fn is_write_vectored(&self) -> bool { + true + } + + fn flush(&mut self) -> io::Result<()> { + match self.calls.pop_front().expect("Unexpected call to flush") { + Call::Flush { output } => output, + Call::Write { .. } => panic!("unexpected call to flush; expected a write"), + } + } + } + + impl Drop for Writer { + fn drop(&mut self) { + if !thread::panicking() { + assert_eq!(self.calls.len(), 0); + } + } + } + + // partial writes keep going + let mut a = LineWriter::new(Writer::default()); + a.write_vectored(&[IoSlice::new(&[]), IoSlice::new(b"abc")]).unwrap(); + + a.get_mut().calls.push_back(Call::Write { inputs: vec![b"abc"], output: Ok(1) }); + a.get_mut().calls.push_back(Call::Write { inputs: vec![b"bc"], output: Ok(2) }); + a.get_mut().calls.push_back(Call::Write { inputs: vec![b"x", b"\n"], output: Ok(2) }); + + a.write_vectored(&[IoSlice::new(b"x"), IoSlice::new(b"\n")]).unwrap(); + + a.get_mut().calls.push_back(Call::Flush { output: Ok(()) }); + a.flush().unwrap(); + + // erroneous writes stop and don't write more + a.get_mut().calls.push_back(Call::Write { inputs: vec![b"x", b"\na"], output: Err(err()) }); + a.get_mut().calls.push_back(Call::Flush { output: Ok(()) }); + assert!(a.write_vectored(&[IoSlice::new(b"x"), IoSlice::new(b"\na")]).is_err()); + a.flush().unwrap(); + + fn err() -> io::Error { + io::Error::new(io::ErrorKind::Other, "x") + } +} + +/// Test that, in cases where vectored writing is not enabled, the +/// LineWriter uses the normal `write` call, which more-correctly handles +/// partial lines +#[test] +fn line_vectored_ignored() { + let writer = ProgrammableSink::default(); + let mut writer = LineWriter::new(writer); + + let content = [ + IoSlice::new(&[]), + IoSlice::new(b"Line 1\nLine"), + IoSlice::new(b" 2\nLine 3\nL"), + IoSlice::new(&[]), + IoSlice::new(&[]), + IoSlice::new(b"ine 4"), + IoSlice::new(b"\nLine 5\n"), + ]; + + let count = writer.write_vectored(&content).unwrap(); + assert_eq!(count, 11); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n"); + + let count = writer.write_vectored(&content[2..]).unwrap(); + assert_eq!(count, 11); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\nLine 3\n"); + + let count = writer.write_vectored(&content[5..]).unwrap(); + assert_eq!(count, 5); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\nLine 3\n"); + + let count = writer.write_vectored(&content[6..]).unwrap(); + assert_eq!(count, 8); + assert_eq!( + writer.get_ref().buffer.as_slice(), + b"Line 1\nLine 2\nLine 3\nLine 4\nLine 5\n".as_ref() + ); +} + +/// Test that, given this input: +/// +/// Line 1\n +/// Line 2\n +/// Line 3\n +/// Line 4 +/// +/// And given a result that only writes to midway through Line 2 +/// +/// That only up to the end of Line 3 is buffered +/// +/// This behavior is desirable because it prevents flushing partial lines +#[test] +fn partial_write_buffers_line() { + let writer = ProgrammableSink { accept_prefix: Some(13), ..Default::default() }; + let mut writer = LineWriter::new(writer); + + assert_eq!(writer.write(b"Line 1\nLine 2\nLine 3\nLine4").unwrap(), 21); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2"); + + assert_eq!(writer.write(b"Line 4").unwrap(), 6); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\nLine 3\n"); +} + +/// Test that, given this input: +/// +/// Line 1\n +/// Line 2\n +/// Line 3 +/// +/// And given that the full write of lines 1 and 2 was successful +/// That data up to Line 3 is buffered +#[test] +fn partial_line_buffered_after_line_write() { + let writer = ProgrammableSink::default(); + let mut writer = LineWriter::new(writer); + + assert_eq!(writer.write(b"Line 1\nLine 2\nLine 3").unwrap(), 20); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\n"); + + assert!(writer.flush().is_ok()); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\nLine 3"); +} + +/// Test that, given a partial line that exceeds the length of +/// LineBuffer's buffer (that is, without a trailing newline), that that +/// line is written to the inner writer +#[test] +fn long_line_flushed() { + let writer = ProgrammableSink::default(); + let mut writer = LineWriter::with_capacity(5, writer); + + assert_eq!(writer.write(b"0123456789").unwrap(), 10); + assert_eq!(&writer.get_ref().buffer, b"0123456789"); +} + +/// Test that, given a very long partial line *after* successfully +/// flushing a complete line, that that line is buffered unconditionally, +/// and no additional writes take place. This assures the property that +/// `write` should make at-most-one attempt to write new data. +#[test] +fn line_long_tail_not_flushed() { + let writer = ProgrammableSink::default(); + let mut writer = LineWriter::with_capacity(5, writer); + + // Assert that Line 1\n is flushed, and 01234 is buffered + assert_eq!(writer.write(b"Line 1\n0123456789").unwrap(), 12); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n"); + + // Because the buffer is full, this subsequent write will flush it + assert_eq!(writer.write(b"5").unwrap(), 1); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n01234"); +} + +/// Test that, if an attempt to pre-flush buffered data returns Ok(0), +/// this is propagated as an error. +#[test] +fn line_buffer_write0_error() { + let writer = ProgrammableSink { + // Accept one write, then return Ok(0) on subsequent ones + max_writes: Some(1), + + ..Default::default() + }; + let mut writer = LineWriter::new(writer); + + // This should write "Line 1\n" and buffer "Partial" + assert_eq!(writer.write(b"Line 1\nPartial").unwrap(), 14); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n"); + + // This will attempt to flush "partial", which will return Ok(0), which + // needs to be an error, because we've already informed the client + // that we accepted the write. + let err = writer.write(b" Line End\n").unwrap_err(); + assert_eq!(err.kind(), ErrorKind::WriteZero); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n"); +} + +/// Test that, if a write returns Ok(0) after a successful pre-flush, this +/// is propagated as Ok(0) +#[test] +fn line_buffer_write0_normal() { + let writer = ProgrammableSink { + // Accept two writes, then return Ok(0) on subsequent ones + max_writes: Some(2), + + ..Default::default() + }; + let mut writer = LineWriter::new(writer); + + // This should write "Line 1\n" and buffer "Partial" + assert_eq!(writer.write(b"Line 1\nPartial").unwrap(), 14); + assert_eq!(&writer.get_ref().buffer, b"Line 1\n"); + + // This will flush partial, which will succeed, but then return Ok(0) + // when flushing " Line End\n" + assert_eq!(writer.write(b" Line End\n").unwrap(), 0); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nPartial"); +} + +/// LineWriter has a custom `write_all`; make sure it works correctly +#[test] +fn line_write_all() { + let writer = ProgrammableSink { + // Only write 5 bytes at a time + accept_prefix: Some(5), + ..Default::default() + }; + let mut writer = LineWriter::new(writer); + + writer.write_all(b"Line 1\nLine 2\nLine 3\nLine 4\nPartial").unwrap(); + assert_eq!(&writer.get_ref().buffer, b"Line 1\nLine 2\nLine 3\nLine 4\n"); + writer.write_all(b" Line 5\n").unwrap(); + assert_eq!( + writer.get_ref().buffer.as_slice(), + b"Line 1\nLine 2\nLine 3\nLine 4\nPartial Line 5\n".as_ref(), + ); +} + +#[test] +fn line_write_all_error() { + let writer = ProgrammableSink { + // Only accept up to 3 writes of up to 5 bytes each + accept_prefix: Some(5), + max_writes: Some(3), + ..Default::default() + }; + + let mut writer = LineWriter::new(writer); + let res = writer.write_all(b"Line 1\nLine 2\nLine 3\nLine 4\nPartial"); + assert!(res.is_err()); + // An error from write_all leaves everything in an indeterminate state, + // so there's nothing else to test here +} + +/// Under certain circumstances, the old implementation of LineWriter +/// would try to buffer "to the last newline" but be forced to buffer +/// less than that, leading to inappropriate partial line writes. +/// Regression test for that issue. +#[test] +fn partial_multiline_buffering() { + let writer = ProgrammableSink { + // Write only up to 5 bytes at a time + accept_prefix: Some(5), + ..Default::default() + }; + + let mut writer = LineWriter::with_capacity(10, writer); + + let content = b"AAAAABBBBB\nCCCCDDDDDD\nEEE"; + + // When content is written, LineWriter will try to write blocks A, B, + // C, and D. Only block A will succeed. Under the old behavior, LineWriter + // would then try to buffer B, C and D, but because its capacity is 10, + // it will only be able to buffer B and C. We don't want to buffer + // partial lines concurrent with whole lines, so the correct behavior + // is to buffer only block B (out to the newline) + assert_eq!(writer.write(content).unwrap(), 11); + assert_eq!(writer.get_ref().buffer, *b"AAAAA"); + + writer.flush().unwrap(); + assert_eq!(writer.get_ref().buffer, *b"AAAAABBBBB\n"); +} + +/// Same as test_partial_multiline_buffering, but in the event NO full lines +/// fit in the buffer, just buffer as much as possible +#[test] +fn partial_multiline_buffering_without_full_line() { + let writer = ProgrammableSink { + // Write only up to 5 bytes at a time + accept_prefix: Some(5), + ..Default::default() + }; + + let mut writer = LineWriter::with_capacity(5, writer); + + let content = b"AAAAABBBBBBBBBB\nCCCCC\nDDDDD"; + + // When content is written, LineWriter will try to write blocks A, B, + // and C. Only block A will succeed. Under the old behavior, LineWriter + // would then try to buffer B and C, but because its capacity is 5, + // it will only be able to buffer part of B. Because it's not possible + // for it to buffer any complete lines, it should buffer as much of B as + // possible + assert_eq!(writer.write(content).unwrap(), 10); + assert_eq!(writer.get_ref().buffer, *b"AAAAA"); + + writer.flush().unwrap(); + assert_eq!(writer.get_ref().buffer, *b"AAAAABBBBB"); +} + +#[derive(Debug, Clone, PartialEq, Eq)] +enum RecordedEvent { + Write(String), + Flush, +} + +#[derive(Debug, Clone, Default)] +struct WriteRecorder { + pub events: Vec<RecordedEvent>, +} + +impl Write for WriteRecorder { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + use crate::str::from_utf8; + + self.events.push(RecordedEvent::Write(from_utf8(buf).unwrap().to_string())); + Ok(buf.len()) + } + + fn flush(&mut self) -> io::Result<()> { + self.events.push(RecordedEvent::Flush); + Ok(()) + } +} + +/// Test that a normal, formatted writeln only results in a single write +/// call to the underlying writer. A naive implementation of +/// LineWriter::write_all results in two writes: one of the buffered data, +/// and another of the final substring in the formatted set +#[test] +fn single_formatted_write() { + let writer = WriteRecorder::default(); + let mut writer = LineWriter::new(writer); + + // Under a naive implementation of LineWriter, this will result in two + // writes: "hello, world" and "!\n", because write() has to flush the + // buffer before attempting to write the last "!\n". write_all shouldn't + // have this limitation. + writeln!(&mut writer, "{}, {}!", "hello", "world").unwrap(); + assert_eq!(writer.get_ref().events, [RecordedEvent::Write("hello, world!\n".to_string())]); +} diff --git a/library/std/src/io/copy.rs b/library/std/src/io/copy.rs new file mode 100644 index 000000000..1a10245e4 --- /dev/null +++ b/library/std/src/io/copy.rs @@ -0,0 +1,161 @@ +use super::{BufWriter, ErrorKind, Read, ReadBuf, Result, Write, DEFAULT_BUF_SIZE}; +use crate::mem::MaybeUninit; + +/// Copies the entire contents of a reader into a writer. +/// +/// This function will continuously read data from `reader` and then +/// write it into `writer` in a streaming fashion until `reader` +/// returns EOF. +/// +/// On success, the total number of bytes that were copied from +/// `reader` to `writer` is returned. +/// +/// If you’re wanting to copy the contents of one file to another and you’re +/// working with filesystem paths, see the [`fs::copy`] function. +/// +/// [`fs::copy`]: crate::fs::copy +/// +/// # Errors +/// +/// This function will return an error immediately if any call to [`read`] or +/// [`write`] returns an error. All instances of [`ErrorKind::Interrupted`] are +/// handled by this function and the underlying operation is retried. +/// +/// [`read`]: Read::read +/// [`write`]: Write::write +/// +/// # Examples +/// +/// ``` +/// use std::io; +/// +/// fn main() -> io::Result<()> { +/// let mut reader: &[u8] = b"hello"; +/// let mut writer: Vec<u8> = vec![]; +/// +/// io::copy(&mut reader, &mut writer)?; +/// +/// assert_eq!(&b"hello"[..], &writer[..]); +/// Ok(()) +/// } +/// ``` +/// +/// # Platform-specific behavior +/// +/// On Linux (including Android), this function uses `copy_file_range(2)`, +/// `sendfile(2)` or `splice(2)` syscalls to move data directly between file +/// descriptors if possible. +/// +/// Note that platform-specific behavior [may change in the future][changes]. +/// +/// [changes]: crate::io#platform-specific-behavior +#[stable(feature = "rust1", since = "1.0.0")] +pub fn copy<R: ?Sized, W: ?Sized>(reader: &mut R, writer: &mut W) -> Result<u64> +where + R: Read, + W: Write, +{ + cfg_if::cfg_if! { + if #[cfg(any(target_os = "linux", target_os = "android"))] { + crate::sys::kernel_copy::copy_spec(reader, writer) + } else { + generic_copy(reader, writer) + } + } +} + +/// The userspace read-write-loop implementation of `io::copy` that is used when +/// OS-specific specializations for copy offloading are not available or not applicable. +pub(crate) fn generic_copy<R: ?Sized, W: ?Sized>(reader: &mut R, writer: &mut W) -> Result<u64> +where + R: Read, + W: Write, +{ + BufferedCopySpec::copy_to(reader, writer) +} + +/// Specialization of the read-write loop that either uses a stack buffer +/// or reuses the internal buffer of a BufWriter +trait BufferedCopySpec: Write { + fn copy_to<R: Read + ?Sized>(reader: &mut R, writer: &mut Self) -> Result<u64>; +} + +impl<W: Write + ?Sized> BufferedCopySpec for W { + default fn copy_to<R: Read + ?Sized>(reader: &mut R, writer: &mut Self) -> Result<u64> { + stack_buffer_copy(reader, writer) + } +} + +impl<I: Write> BufferedCopySpec for BufWriter<I> { + fn copy_to<R: Read + ?Sized>(reader: &mut R, writer: &mut Self) -> Result<u64> { + if writer.capacity() < DEFAULT_BUF_SIZE { + return stack_buffer_copy(reader, writer); + } + + let mut len = 0; + let mut init = 0; + + loop { + let buf = writer.buffer_mut(); + let mut read_buf = ReadBuf::uninit(buf.spare_capacity_mut()); + + // SAFETY: init is either 0 or the initialized_len of the previous iteration + unsafe { + read_buf.assume_init(init); + } + + if read_buf.capacity() >= DEFAULT_BUF_SIZE { + match reader.read_buf(&mut read_buf) { + Ok(()) => { + let bytes_read = read_buf.filled_len(); + + if bytes_read == 0 { + return Ok(len); + } + + init = read_buf.initialized_len() - bytes_read; + + // SAFETY: ReadBuf guarantees all of its filled bytes are init + unsafe { buf.set_len(buf.len() + bytes_read) }; + len += bytes_read as u64; + // Read again if the buffer still has enough capacity, as BufWriter itself would do + // This will occur if the reader returns short reads + continue; + } + Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + } + } + + writer.flush_buf()?; + } + } +} + +fn stack_buffer_copy<R: Read + ?Sized, W: Write + ?Sized>( + reader: &mut R, + writer: &mut W, +) -> Result<u64> { + let mut buf = [MaybeUninit::uninit(); DEFAULT_BUF_SIZE]; + let mut buf = ReadBuf::uninit(&mut buf); + + let mut len = 0; + + loop { + match reader.read_buf(&mut buf) { + Ok(()) => {} + Err(e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + }; + + if buf.filled().is_empty() { + break; + } + + len += buf.filled().len() as u64; + writer.write_all(buf.filled())?; + buf.clear(); + } + + Ok(len) +} diff --git a/library/std/src/io/cursor.rs b/library/std/src/io/cursor.rs new file mode 100644 index 000000000..f3fbfc447 --- /dev/null +++ b/library/std/src/io/cursor.rs @@ -0,0 +1,640 @@ +#[cfg(test)] +mod tests; + +use crate::io::prelude::*; + +use crate::alloc::Allocator; +use crate::cmp; +use crate::io::{self, ErrorKind, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; + +/// A `Cursor` wraps an in-memory buffer and provides it with a +/// [`Seek`] implementation. +/// +/// `Cursor`s are used with in-memory buffers, anything implementing +/// <code>[AsRef]<\[u8]></code>, to allow them to implement [`Read`] and/or [`Write`], +/// allowing these buffers to be used anywhere you might use a reader or writer +/// that does actual I/O. +/// +/// The standard library implements some I/O traits on various types which +/// are commonly used as a buffer, like <code>Cursor<[Vec]\<u8>></code> and +/// <code>Cursor<[&\[u8\]][bytes]></code>. +/// +/// # Examples +/// +/// We may want to write bytes to a [`File`] in our production +/// code, but use an in-memory buffer in our tests. We can do this with +/// `Cursor`: +/// +/// [bytes]: crate::slice "slice" +/// [`File`]: crate::fs::File +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::io::{self, SeekFrom}; +/// use std::fs::File; +/// +/// // a library function we've written +/// fn write_ten_bytes_at_end<W: Write + Seek>(writer: &mut W) -> io::Result<()> { +/// writer.seek(SeekFrom::End(-10))?; +/// +/// for i in 0..10 { +/// writer.write(&[i])?; +/// } +/// +/// // all went well +/// Ok(()) +/// } +/// +/// # fn foo() -> io::Result<()> { +/// // Here's some code that uses this library function. +/// // +/// // We might want to use a BufReader here for efficiency, but let's +/// // keep this example focused. +/// let mut file = File::create("foo.txt")?; +/// +/// write_ten_bytes_at_end(&mut file)?; +/// # Ok(()) +/// # } +/// +/// // now let's write a test +/// #[test] +/// fn test_writes_bytes() { +/// // setting up a real File is much slower than an in-memory buffer, +/// // let's use a cursor instead +/// use std::io::Cursor; +/// let mut buff = Cursor::new(vec![0; 15]); +/// +/// write_ten_bytes_at_end(&mut buff).unwrap(); +/// +/// assert_eq!(&buff.get_ref()[5..15], &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]); +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug, Default, Eq, PartialEq)] +pub struct Cursor<T> { + inner: T, + pos: u64, +} + +impl<T> Cursor<T> { + /// Creates a new cursor wrapping the provided underlying in-memory buffer. + /// + /// Cursor initial position is `0` even if underlying buffer (e.g., [`Vec`]) + /// is not empty. So writing to cursor starts with overwriting [`Vec`] + /// content, not with appending to it. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// + /// let buff = Cursor::new(Vec::new()); + /// # fn force_inference(_: &Cursor<Vec<u8>>) {} + /// # force_inference(&buff); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] + pub const fn new(inner: T) -> Cursor<T> { + Cursor { pos: 0, inner } + } + + /// Consumes this cursor, returning the underlying value. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// + /// let buff = Cursor::new(Vec::new()); + /// # fn force_inference(_: &Cursor<Vec<u8>>) {} + /// # force_inference(&buff); + /// + /// let vec = buff.into_inner(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn into_inner(self) -> T { + self.inner + } + + /// Gets a reference to the underlying value in this cursor. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// + /// let buff = Cursor::new(Vec::new()); + /// # fn force_inference(_: &Cursor<Vec<u8>>) {} + /// # force_inference(&buff); + /// + /// let reference = buff.get_ref(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] + pub const fn get_ref(&self) -> &T { + &self.inner + } + + /// Gets a mutable reference to the underlying value in this cursor. + /// + /// Care should be taken to avoid modifying the internal I/O state of the + /// underlying value as it may corrupt this cursor's position. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// + /// let mut buff = Cursor::new(Vec::new()); + /// # fn force_inference(_: &Cursor<Vec<u8>>) {} + /// # force_inference(&buff); + /// + /// let reference = buff.get_mut(); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn get_mut(&mut self) -> &mut T { + &mut self.inner + } + + /// Returns the current position of this cursor. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// use std::io::prelude::*; + /// use std::io::SeekFrom; + /// + /// let mut buff = Cursor::new(vec![1, 2, 3, 4, 5]); + /// + /// assert_eq!(buff.position(), 0); + /// + /// buff.seek(SeekFrom::Current(2)).unwrap(); + /// assert_eq!(buff.position(), 2); + /// + /// buff.seek(SeekFrom::Current(-1)).unwrap(); + /// assert_eq!(buff.position(), 1); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] + pub const fn position(&self) -> u64 { + self.pos + } + + /// Sets the position of this cursor. + /// + /// # Examples + /// + /// ``` + /// use std::io::Cursor; + /// + /// let mut buff = Cursor::new(vec![1, 2, 3, 4, 5]); + /// + /// assert_eq!(buff.position(), 0); + /// + /// buff.set_position(2); + /// assert_eq!(buff.position(), 2); + /// + /// buff.set_position(4); + /// assert_eq!(buff.position(), 4); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn set_position(&mut self, pos: u64) { + self.pos = pos; + } +} + +impl<T> Cursor<T> +where + T: AsRef<[u8]>, +{ + /// Returns the remaining slice. + /// + /// # Examples + /// + /// ``` + /// #![feature(cursor_remaining)] + /// use std::io::Cursor; + /// + /// let mut buff = Cursor::new(vec![1, 2, 3, 4, 5]); + /// + /// assert_eq!(buff.remaining_slice(), &[1, 2, 3, 4, 5]); + /// + /// buff.set_position(2); + /// assert_eq!(buff.remaining_slice(), &[3, 4, 5]); + /// + /// buff.set_position(4); + /// assert_eq!(buff.remaining_slice(), &[5]); + /// + /// buff.set_position(6); + /// assert_eq!(buff.remaining_slice(), &[]); + /// ``` + #[unstable(feature = "cursor_remaining", issue = "86369")] + pub fn remaining_slice(&self) -> &[u8] { + let len = self.pos.min(self.inner.as_ref().len() as u64); + &self.inner.as_ref()[(len as usize)..] + } + + /// Returns `true` if the remaining slice is empty. + /// + /// # Examples + /// + /// ``` + /// #![feature(cursor_remaining)] + /// use std::io::Cursor; + /// + /// let mut buff = Cursor::new(vec![1, 2, 3, 4, 5]); + /// + /// buff.set_position(2); + /// assert!(!buff.is_empty()); + /// + /// buff.set_position(5); + /// assert!(buff.is_empty()); + /// + /// buff.set_position(10); + /// assert!(buff.is_empty()); + /// ``` + #[unstable(feature = "cursor_remaining", issue = "86369")] + pub fn is_empty(&self) -> bool { + self.pos >= self.inner.as_ref().len() as u64 + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T> Clone for Cursor<T> +where + T: Clone, +{ + #[inline] + fn clone(&self) -> Self { + Cursor { inner: self.inner.clone(), pos: self.pos } + } + + #[inline] + fn clone_from(&mut self, other: &Self) { + self.inner.clone_from(&other.inner); + self.pos = other.pos; + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T> io::Seek for Cursor<T> +where + T: AsRef<[u8]>, +{ + fn seek(&mut self, style: SeekFrom) -> io::Result<u64> { + let (base_pos, offset) = match style { + SeekFrom::Start(n) => { + self.pos = n; + return Ok(n); + } + SeekFrom::End(n) => (self.inner.as_ref().len() as u64, n), + SeekFrom::Current(n) => (self.pos, n), + }; + match base_pos.checked_add_signed(offset) { + Some(n) => { + self.pos = n; + Ok(self.pos) + } + None => Err(io::const_io_error!( + ErrorKind::InvalidInput, + "invalid seek to a negative or overflowing position", + )), + } + } + + fn stream_len(&mut self) -> io::Result<u64> { + Ok(self.inner.as_ref().len() as u64) + } + + fn stream_position(&mut self) -> io::Result<u64> { + Ok(self.pos) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T> Read for Cursor<T> +where + T: AsRef<[u8]>, +{ + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + let n = Read::read(&mut self.remaining_slice(), buf)?; + self.pos += n as u64; + Ok(n) + } + + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + let prev_filled = buf.filled_len(); + + Read::read_buf(&mut self.fill_buf()?, buf)?; + + self.pos += (buf.filled_len() - prev_filled) as u64; + + Ok(()) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + let mut nread = 0; + for buf in bufs { + let n = self.read(buf)?; + nread += n; + if n < buf.len() { + break; + } + } + Ok(nread) + } + + fn is_read_vectored(&self) -> bool { + true + } + + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + let n = buf.len(); + Read::read_exact(&mut self.remaining_slice(), buf)?; + self.pos += n as u64; + Ok(()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T> BufRead for Cursor<T> +where + T: AsRef<[u8]>, +{ + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Ok(self.remaining_slice()) + } + fn consume(&mut self, amt: usize) { + self.pos += amt as u64; + } +} + +// Non-resizing write implementation +#[inline] +fn slice_write(pos_mut: &mut u64, slice: &mut [u8], buf: &[u8]) -> io::Result<usize> { + let pos = cmp::min(*pos_mut, slice.len() as u64); + let amt = (&mut slice[(pos as usize)..]).write(buf)?; + *pos_mut += amt as u64; + Ok(amt) +} + +#[inline] +fn slice_write_vectored( + pos_mut: &mut u64, + slice: &mut [u8], + bufs: &[IoSlice<'_>], +) -> io::Result<usize> { + let mut nwritten = 0; + for buf in bufs { + let n = slice_write(pos_mut, slice, buf)?; + nwritten += n; + if n < buf.len() { + break; + } + } + Ok(nwritten) +} + +/// Reserves the required space, and pads the vec with 0s if necessary. +fn reserve_and_pad<A: Allocator>( + pos_mut: &mut u64, + vec: &mut Vec<u8, A>, + buf_len: usize, +) -> io::Result<usize> { + let pos: usize = (*pos_mut).try_into().map_err(|_| { + io::const_io_error!( + ErrorKind::InvalidInput, + "cursor position exceeds maximum possible vector length", + ) + })?; + + // For safety reasons, we don't want these numbers to overflow + // otherwise our allocation won't be enough + let desired_cap = pos.saturating_add(buf_len); + if desired_cap > vec.capacity() { + // We want our vec's total capacity + // to have room for (pos+buf_len) bytes. Reserve allocates + // based on additional elements from the length, so we need to + // reserve the difference + vec.reserve(desired_cap - vec.len()); + } + // Pad if pos is above the current len. + if pos > vec.len() { + let diff = pos - vec.len(); + // Unfortunately, `resize()` would suffice but the optimiser does not + // realise the `reserve` it does can be eliminated. So we do it manually + // to eliminate that extra branch + let spare = vec.spare_capacity_mut(); + debug_assert!(spare.len() >= diff); + // Safety: we have allocated enough capacity for this. + // And we are only writing, not reading + unsafe { + spare.get_unchecked_mut(..diff).fill(core::mem::MaybeUninit::new(0)); + vec.set_len(pos); + } + } + + Ok(pos) +} + +/// Writes the slice to the vec without allocating +/// # Safety: vec must have buf.len() spare capacity +unsafe fn vec_write_unchecked<A>(pos: usize, vec: &mut Vec<u8, A>, buf: &[u8]) -> usize +where + A: Allocator, +{ + debug_assert!(vec.capacity() >= pos + buf.len()); + vec.as_mut_ptr().add(pos).copy_from(buf.as_ptr(), buf.len()); + pos + buf.len() +} + +/// Resizing write implementation for [`Cursor`] +/// +/// Cursor is allowed to have a pre-allocated and initialised +/// vector body, but with a position of 0. This means the [`Write`] +/// will overwrite the contents of the vec. +/// +/// This also allows for the vec body to be empty, but with a position of N. +/// This means that [`Write`] will pad the vec with 0 initially, +/// before writing anything from that point +fn vec_write<A>(pos_mut: &mut u64, vec: &mut Vec<u8, A>, buf: &[u8]) -> io::Result<usize> +where + A: Allocator, +{ + let buf_len = buf.len(); + let mut pos = reserve_and_pad(pos_mut, vec, buf_len)?; + + // Write the buf then progress the vec forward if necessary + // Safety: we have ensured that the capacity is available + // and that all bytes get written up to pos + unsafe { + pos = vec_write_unchecked(pos, vec, buf); + if pos > vec.len() { + vec.set_len(pos); + } + }; + + // Bump us forward + *pos_mut += buf_len as u64; + Ok(buf_len) +} + +/// Resizing write_vectored implementation for [`Cursor`] +/// +/// Cursor is allowed to have a pre-allocated and initialised +/// vector body, but with a position of 0. This means the [`Write`] +/// will overwrite the contents of the vec. +/// +/// This also allows for the vec body to be empty, but with a position of N. +/// This means that [`Write`] will pad the vec with 0 initially, +/// before writing anything from that point +fn vec_write_vectored<A>( + pos_mut: &mut u64, + vec: &mut Vec<u8, A>, + bufs: &[IoSlice<'_>], +) -> io::Result<usize> +where + A: Allocator, +{ + // For safety reasons, we don't want this sum to overflow ever. + // If this saturates, the reserve should panic to avoid any unsound writing. + let buf_len = bufs.iter().fold(0usize, |a, b| a.saturating_add(b.len())); + let mut pos = reserve_and_pad(pos_mut, vec, buf_len)?; + + // Write the buf then progress the vec forward if necessary + // Safety: we have ensured that the capacity is available + // and that all bytes get written up to the last pos + unsafe { + for buf in bufs { + pos = vec_write_unchecked(pos, vec, buf); + } + if pos > vec.len() { + vec.set_len(pos); + } + } + + // Bump us forward + *pos_mut += buf_len as u64; + Ok(buf_len) +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for Cursor<&mut [u8]> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + slice_write(&mut self.pos, self.inner, buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + slice_write_vectored(&mut self.pos, self.inner, bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "cursor_mut_vec", since = "1.25.0")] +impl<A> Write for Cursor<&mut Vec<u8, A>> +where + A: Allocator, +{ + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + vec_write(&mut self.pos, self.inner, buf) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + vec_write_vectored(&mut self.pos, self.inner, bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<A> Write for Cursor<Vec<u8, A>> +where + A: Allocator, +{ + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + vec_write(&mut self.pos, &mut self.inner, buf) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + vec_write_vectored(&mut self.pos, &mut self.inner, bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "cursor_box_slice", since = "1.5.0")] +impl<A> Write for Cursor<Box<[u8], A>> +where + A: Allocator, +{ + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + slice_write(&mut self.pos, &mut self.inner, buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + slice_write_vectored(&mut self.pos, &mut self.inner, bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "cursor_array", since = "1.61.0")] +impl<const N: usize> Write for Cursor<[u8; N]> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + slice_write(&mut self.pos, &mut self.inner, buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + slice_write_vectored(&mut self.pos, &mut self.inner, bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} diff --git a/library/std/src/io/cursor/tests.rs b/library/std/src/io/cursor/tests.rs new file mode 100644 index 000000000..d7c203c29 --- /dev/null +++ b/library/std/src/io/cursor/tests.rs @@ -0,0 +1,567 @@ +use crate::io::prelude::*; +use crate::io::{Cursor, IoSlice, IoSliceMut, SeekFrom}; + +#[test] +fn test_vec_writer() { + let mut writer = Vec::new(); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + assert_eq!( + writer + .write_vectored(&[IoSlice::new(&[]), IoSlice::new(&[8, 9]), IoSlice::new(&[10])],) + .unwrap(), + 3 + ); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; + assert_eq!(writer, b); +} + +#[test] +fn test_mem_writer() { + let mut writer = Cursor::new(Vec::new()); + writer.set_position(10); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + assert_eq!( + writer + .write_vectored(&[IoSlice::new(&[]), IoSlice::new(&[8, 9]), IoSlice::new(&[10])],) + .unwrap(), + 3 + ); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; + assert_eq!(&writer.get_ref()[..10], &[0; 10]); + assert_eq!(&writer.get_ref()[10..], b); +} + +#[test] +fn test_mem_writer_preallocated() { + let mut writer = Cursor::new(vec![0, 0, 0, 0, 0, 0, 0, 0, 8, 9, 10]); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; + assert_eq!(&writer.get_ref()[..], b); +} + +#[test] +fn test_mem_mut_writer() { + let mut vec = Vec::new(); + let mut writer = Cursor::new(&mut vec); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + assert_eq!( + writer + .write_vectored(&[IoSlice::new(&[]), IoSlice::new(&[8, 9]), IoSlice::new(&[10])],) + .unwrap(), + 3 + ); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; + assert_eq!(&writer.get_ref()[..], b); +} + +fn test_slice_writer<T>(writer: &mut Cursor<T>) +where + T: AsRef<[u8]>, + Cursor<T>: Write, +{ + assert_eq!(writer.position(), 0); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.position(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + assert_eq!(writer.position(), 8); + assert_eq!(writer.write(&[]).unwrap(), 0); + assert_eq!(writer.position(), 8); + + assert_eq!(writer.write(&[8, 9]).unwrap(), 1); + assert_eq!(writer.write(&[10]).unwrap(), 0); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8]; + assert_eq!(writer.get_ref().as_ref(), b); +} + +fn test_slice_writer_vectored<T>(writer: &mut Cursor<T>) +where + T: AsRef<[u8]>, + Cursor<T>: Write, +{ + assert_eq!(writer.position(), 0); + assert_eq!(writer.write_vectored(&[IoSlice::new(&[0])]).unwrap(), 1); + assert_eq!(writer.position(), 1); + assert_eq!( + writer.write_vectored(&[IoSlice::new(&[1, 2, 3]), IoSlice::new(&[4, 5, 6, 7]),]).unwrap(), + 7, + ); + assert_eq!(writer.position(), 8); + assert_eq!(writer.write_vectored(&[]).unwrap(), 0); + assert_eq!(writer.position(), 8); + + assert_eq!(writer.write_vectored(&[IoSlice::new(&[8, 9])]).unwrap(), 1); + assert_eq!(writer.write_vectored(&[IoSlice::new(&[10])]).unwrap(), 0); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7, 8]; + assert_eq!(writer.get_ref().as_ref(), b); +} + +#[test] +fn test_box_slice_writer() { + let mut writer = Cursor::new(vec![0u8; 9].into_boxed_slice()); + test_slice_writer(&mut writer); +} + +#[test] +fn test_box_slice_writer_vectored() { + let mut writer = Cursor::new(vec![0u8; 9].into_boxed_slice()); + test_slice_writer_vectored(&mut writer); +} + +#[test] +fn test_array_writer() { + let mut writer = Cursor::new([0u8; 9]); + test_slice_writer(&mut writer); +} + +#[test] +fn test_array_writer_vectored() { + let mut writer = Cursor::new([0u8; 9]); + test_slice_writer_vectored(&mut writer); +} + +#[test] +fn test_buf_writer() { + let mut buf = [0 as u8; 9]; + let mut writer = Cursor::new(&mut buf[..]); + test_slice_writer(&mut writer); +} + +#[test] +fn test_buf_writer_vectored() { + let mut buf = [0 as u8; 9]; + let mut writer = Cursor::new(&mut buf[..]); + test_slice_writer_vectored(&mut writer); +} + +#[test] +fn test_buf_writer_seek() { + let mut buf = [0 as u8; 8]; + { + let mut writer = Cursor::new(&mut buf[..]); + assert_eq!(writer.position(), 0); + assert_eq!(writer.write(&[1]).unwrap(), 1); + assert_eq!(writer.position(), 1); + + assert_eq!(writer.seek(SeekFrom::Start(2)).unwrap(), 2); + assert_eq!(writer.position(), 2); + assert_eq!(writer.write(&[2]).unwrap(), 1); + assert_eq!(writer.position(), 3); + + assert_eq!(writer.seek(SeekFrom::Current(-2)).unwrap(), 1); + assert_eq!(writer.position(), 1); + assert_eq!(writer.write(&[3]).unwrap(), 1); + assert_eq!(writer.position(), 2); + + assert_eq!(writer.seek(SeekFrom::End(-1)).unwrap(), 7); + assert_eq!(writer.position(), 7); + assert_eq!(writer.write(&[4]).unwrap(), 1); + assert_eq!(writer.position(), 8); + } + let b: &[_] = &[1, 3, 2, 0, 0, 0, 0, 4]; + assert_eq!(buf, b); +} + +#[test] +fn test_buf_writer_error() { + let mut buf = [0 as u8; 2]; + let mut writer = Cursor::new(&mut buf[..]); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.write(&[0, 0]).unwrap(), 1); + assert_eq!(writer.write(&[0, 0]).unwrap(), 0); +} + +#[test] +fn test_mem_reader() { + let mut reader = Cursor::new(vec![0, 1, 2, 3, 4, 5, 6, 7]); + let mut buf = []; + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.position(), 0); + let mut buf = [0]; + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.position(), 1); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf = [0; 4]; + assert_eq!(reader.read(&mut buf).unwrap(), 4); + assert_eq!(reader.position(), 5); + let b: &[_] = &[1, 2, 3, 4]; + assert_eq!(buf, b); + assert_eq!(reader.read(&mut buf).unwrap(), 3); + let b: &[_] = &[5, 6, 7]; + assert_eq!(&buf[..3], b); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_mem_reader_vectored() { + let mut reader = Cursor::new(vec![0, 1, 2, 3, 4, 5, 6, 7]); + let mut buf = []; + assert_eq!(reader.read_vectored(&mut [IoSliceMut::new(&mut buf)]).unwrap(), 0); + assert_eq!(reader.position(), 0); + let mut buf = [0]; + assert_eq!( + reader.read_vectored(&mut [IoSliceMut::new(&mut []), IoSliceMut::new(&mut buf),]).unwrap(), + 1, + ); + assert_eq!(reader.position(), 1); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf1 = [0; 4]; + let mut buf2 = [0; 4]; + assert_eq!( + reader + .read_vectored(&mut [IoSliceMut::new(&mut buf1), IoSliceMut::new(&mut buf2),]) + .unwrap(), + 7, + ); + let b1: &[_] = &[1, 2, 3, 4]; + let b2: &[_] = &[5, 6, 7]; + assert_eq!(buf1, b1); + assert_eq!(&buf2[..3], b2); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_boxed_slice_reader() { + let mut reader = Cursor::new(vec![0, 1, 2, 3, 4, 5, 6, 7].into_boxed_slice()); + let mut buf = []; + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.position(), 0); + let mut buf = [0]; + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.position(), 1); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf = [0; 4]; + assert_eq!(reader.read(&mut buf).unwrap(), 4); + assert_eq!(reader.position(), 5); + let b: &[_] = &[1, 2, 3, 4]; + assert_eq!(buf, b); + assert_eq!(reader.read(&mut buf).unwrap(), 3); + let b: &[_] = &[5, 6, 7]; + assert_eq!(&buf[..3], b); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_boxed_slice_reader_vectored() { + let mut reader = Cursor::new(vec![0, 1, 2, 3, 4, 5, 6, 7].into_boxed_slice()); + let mut buf = []; + assert_eq!(reader.read_vectored(&mut [IoSliceMut::new(&mut buf)]).unwrap(), 0); + assert_eq!(reader.position(), 0); + let mut buf = [0]; + assert_eq!( + reader.read_vectored(&mut [IoSliceMut::new(&mut []), IoSliceMut::new(&mut buf),]).unwrap(), + 1, + ); + assert_eq!(reader.position(), 1); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf1 = [0; 4]; + let mut buf2 = [0; 4]; + assert_eq!( + reader + .read_vectored(&mut [IoSliceMut::new(&mut buf1), IoSliceMut::new(&mut buf2)],) + .unwrap(), + 7, + ); + let b1: &[_] = &[1, 2, 3, 4]; + let b2: &[_] = &[5, 6, 7]; + assert_eq!(buf1, b1); + assert_eq!(&buf2[..3], b2); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn read_to_end() { + let mut reader = Cursor::new(vec![0, 1, 2, 3, 4, 5, 6, 7]); + let mut v = Vec::new(); + reader.read_to_end(&mut v).unwrap(); + assert_eq!(v, [0, 1, 2, 3, 4, 5, 6, 7]); +} + +#[test] +fn test_slice_reader() { + let in_buf = vec![0, 1, 2, 3, 4, 5, 6, 7]; + let reader = &mut &in_buf[..]; + let mut buf = []; + assert_eq!(reader.read(&mut buf).unwrap(), 0); + let mut buf = [0]; + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.len(), 7); + let b: &[_] = &[0]; + assert_eq!(&buf[..], b); + let mut buf = [0; 4]; + assert_eq!(reader.read(&mut buf).unwrap(), 4); + assert_eq!(reader.len(), 3); + let b: &[_] = &[1, 2, 3, 4]; + assert_eq!(&buf[..], b); + assert_eq!(reader.read(&mut buf).unwrap(), 3); + let b: &[_] = &[5, 6, 7]; + assert_eq!(&buf[..3], b); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_slice_reader_vectored() { + let in_buf = vec![0, 1, 2, 3, 4, 5, 6, 7]; + let reader = &mut &in_buf[..]; + let mut buf = []; + assert_eq!(reader.read_vectored(&mut [IoSliceMut::new(&mut buf)]).unwrap(), 0); + let mut buf = [0]; + assert_eq!( + reader.read_vectored(&mut [IoSliceMut::new(&mut []), IoSliceMut::new(&mut buf),]).unwrap(), + 1, + ); + assert_eq!(reader.len(), 7); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf1 = [0; 4]; + let mut buf2 = [0; 4]; + assert_eq!( + reader + .read_vectored(&mut [IoSliceMut::new(&mut buf1), IoSliceMut::new(&mut buf2)],) + .unwrap(), + 7, + ); + let b1: &[_] = &[1, 2, 3, 4]; + let b2: &[_] = &[5, 6, 7]; + assert_eq!(buf1, b1); + assert_eq!(&buf2[..3], b2); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn test_read_exact() { + let in_buf = vec![0, 1, 2, 3, 4, 5, 6, 7]; + let reader = &mut &in_buf[..]; + let mut buf = []; + assert!(reader.read_exact(&mut buf).is_ok()); + let mut buf = [8]; + assert!(reader.read_exact(&mut buf).is_ok()); + assert_eq!(buf[0], 0); + assert_eq!(reader.len(), 7); + let mut buf = [0, 0, 0, 0, 0, 0, 0]; + assert!(reader.read_exact(&mut buf).is_ok()); + assert_eq!(buf, [1, 2, 3, 4, 5, 6, 7]); + assert_eq!(reader.len(), 0); + let mut buf = [0]; + assert!(reader.read_exact(&mut buf).is_err()); +} + +#[test] +fn test_buf_reader() { + let in_buf = vec![0, 1, 2, 3, 4, 5, 6, 7]; + let mut reader = Cursor::new(&in_buf[..]); + let mut buf = []; + assert_eq!(reader.read(&mut buf).unwrap(), 0); + assert_eq!(reader.position(), 0); + let mut buf = [0]; + assert_eq!(reader.read(&mut buf).unwrap(), 1); + assert_eq!(reader.position(), 1); + let b: &[_] = &[0]; + assert_eq!(buf, b); + let mut buf = [0; 4]; + assert_eq!(reader.read(&mut buf).unwrap(), 4); + assert_eq!(reader.position(), 5); + let b: &[_] = &[1, 2, 3, 4]; + assert_eq!(buf, b); + assert_eq!(reader.read(&mut buf).unwrap(), 3); + let b: &[_] = &[5, 6, 7]; + assert_eq!(&buf[..3], b); + assert_eq!(reader.read(&mut buf).unwrap(), 0); +} + +#[test] +fn seek_past_end() { + let buf = [0xff]; + let mut r = Cursor::new(&buf[..]); + assert_eq!(r.seek(SeekFrom::Start(10)).unwrap(), 10); + assert_eq!(r.read(&mut [0]).unwrap(), 0); + + let mut r = Cursor::new(vec![10]); + assert_eq!(r.seek(SeekFrom::Start(10)).unwrap(), 10); + assert_eq!(r.read(&mut [0]).unwrap(), 0); + + let mut buf = [0]; + let mut r = Cursor::new(&mut buf[..]); + assert_eq!(r.seek(SeekFrom::Start(10)).unwrap(), 10); + assert_eq!(r.write(&[3]).unwrap(), 0); + + let mut r = Cursor::new(vec![10].into_boxed_slice()); + assert_eq!(r.seek(SeekFrom::Start(10)).unwrap(), 10); + assert_eq!(r.write(&[3]).unwrap(), 0); +} + +#[test] +fn seek_past_i64() { + let buf = [0xff]; + let mut r = Cursor::new(&buf[..]); + assert_eq!(r.seek(SeekFrom::Start(6)).unwrap(), 6); + assert_eq!(r.seek(SeekFrom::Current(0x7ffffffffffffff0)).unwrap(), 0x7ffffffffffffff6); + assert_eq!(r.seek(SeekFrom::Current(0x10)).unwrap(), 0x8000000000000006); + assert_eq!(r.seek(SeekFrom::Current(0)).unwrap(), 0x8000000000000006); + assert!(r.seek(SeekFrom::Current(0x7ffffffffffffffd)).is_err()); + assert_eq!(r.seek(SeekFrom::Current(-0x8000000000000000)).unwrap(), 6); + + let mut r = Cursor::new(vec![10]); + assert_eq!(r.seek(SeekFrom::Start(6)).unwrap(), 6); + assert_eq!(r.seek(SeekFrom::Current(0x7ffffffffffffff0)).unwrap(), 0x7ffffffffffffff6); + assert_eq!(r.seek(SeekFrom::Current(0x10)).unwrap(), 0x8000000000000006); + assert_eq!(r.seek(SeekFrom::Current(0)).unwrap(), 0x8000000000000006); + assert!(r.seek(SeekFrom::Current(0x7ffffffffffffffd)).is_err()); + assert_eq!(r.seek(SeekFrom::Current(-0x8000000000000000)).unwrap(), 6); + + let mut buf = [0]; + let mut r = Cursor::new(&mut buf[..]); + assert_eq!(r.seek(SeekFrom::Start(6)).unwrap(), 6); + assert_eq!(r.seek(SeekFrom::Current(0x7ffffffffffffff0)).unwrap(), 0x7ffffffffffffff6); + assert_eq!(r.seek(SeekFrom::Current(0x10)).unwrap(), 0x8000000000000006); + assert_eq!(r.seek(SeekFrom::Current(0)).unwrap(), 0x8000000000000006); + assert!(r.seek(SeekFrom::Current(0x7ffffffffffffffd)).is_err()); + assert_eq!(r.seek(SeekFrom::Current(-0x8000000000000000)).unwrap(), 6); + + let mut r = Cursor::new(vec![10].into_boxed_slice()); + assert_eq!(r.seek(SeekFrom::Start(6)).unwrap(), 6); + assert_eq!(r.seek(SeekFrom::Current(0x7ffffffffffffff0)).unwrap(), 0x7ffffffffffffff6); + assert_eq!(r.seek(SeekFrom::Current(0x10)).unwrap(), 0x8000000000000006); + assert_eq!(r.seek(SeekFrom::Current(0)).unwrap(), 0x8000000000000006); + assert!(r.seek(SeekFrom::Current(0x7ffffffffffffffd)).is_err()); + assert_eq!(r.seek(SeekFrom::Current(-0x8000000000000000)).unwrap(), 6); +} + +#[test] +fn seek_before_0() { + let buf = [0xff]; + let mut r = Cursor::new(&buf[..]); + assert!(r.seek(SeekFrom::End(-2)).is_err()); + + let mut r = Cursor::new(vec![10]); + assert!(r.seek(SeekFrom::End(-2)).is_err()); + + let mut buf = [0]; + let mut r = Cursor::new(&mut buf[..]); + assert!(r.seek(SeekFrom::End(-2)).is_err()); + + let mut r = Cursor::new(vec![10].into_boxed_slice()); + assert!(r.seek(SeekFrom::End(-2)).is_err()); +} + +#[test] +fn test_seekable_mem_writer() { + let mut writer = Cursor::new(Vec::<u8>::new()); + assert_eq!(writer.position(), 0); + assert_eq!(writer.write(&[0]).unwrap(), 1); + assert_eq!(writer.position(), 1); + assert_eq!(writer.write(&[1, 2, 3]).unwrap(), 3); + assert_eq!(writer.write(&[4, 5, 6, 7]).unwrap(), 4); + assert_eq!(writer.position(), 8); + let b: &[_] = &[0, 1, 2, 3, 4, 5, 6, 7]; + assert_eq!(&writer.get_ref()[..], b); + + assert_eq!(writer.seek(SeekFrom::Start(0)).unwrap(), 0); + assert_eq!(writer.position(), 0); + assert_eq!(writer.write(&[3, 4]).unwrap(), 2); + let b: &[_] = &[3, 4, 2, 3, 4, 5, 6, 7]; + assert_eq!(&writer.get_ref()[..], b); + + assert_eq!(writer.seek(SeekFrom::Current(1)).unwrap(), 3); + assert_eq!(writer.write(&[0, 1]).unwrap(), 2); + let b: &[_] = &[3, 4, 2, 0, 1, 5, 6, 7]; + assert_eq!(&writer.get_ref()[..], b); + + assert_eq!(writer.seek(SeekFrom::End(-1)).unwrap(), 7); + assert_eq!(writer.write(&[1, 2]).unwrap(), 2); + let b: &[_] = &[3, 4, 2, 0, 1, 5, 6, 1, 2]; + assert_eq!(&writer.get_ref()[..], b); + + assert_eq!(writer.seek(SeekFrom::End(1)).unwrap(), 10); + assert_eq!(writer.write(&[1]).unwrap(), 1); + let b: &[_] = &[3, 4, 2, 0, 1, 5, 6, 1, 2, 0, 1]; + assert_eq!(&writer.get_ref()[..], b); +} + +#[test] +fn vec_seek_past_end() { + let mut r = Cursor::new(Vec::new()); + assert_eq!(r.seek(SeekFrom::Start(10)).unwrap(), 10); + assert_eq!(r.write(&[3]).unwrap(), 1); +} + +#[test] +fn vec_seek_before_0() { + let mut r = Cursor::new(Vec::new()); + assert!(r.seek(SeekFrom::End(-2)).is_err()); +} + +#[test] +#[cfg(target_pointer_width = "32")] +fn vec_seek_and_write_past_usize_max() { + let mut c = Cursor::new(Vec::new()); + c.set_position(usize::MAX as u64 + 1); + assert!(c.write_all(&[1, 2, 3]).is_err()); +} + +#[test] +fn test_partial_eq() { + assert_eq!(Cursor::new(Vec::<u8>::new()), Cursor::new(Vec::<u8>::new())); +} + +#[test] +fn test_eq() { + struct AssertEq<T: Eq>(pub T); + + let _: AssertEq<Cursor<Vec<u8>>> = AssertEq(Cursor::new(Vec::new())); +} + +#[allow(dead_code)] +fn const_cursor() { + const CURSOR: Cursor<&[u8]> = Cursor::new(&[0]); + const _: &&[u8] = CURSOR.get_ref(); + const _: u64 = CURSOR.position(); +} + +#[bench] +fn bench_write_vec(b: &mut test::Bencher) { + let slice = &[1; 128]; + + b.iter(|| { + let mut buf = b"some random data to overwrite".to_vec(); + let mut cursor = Cursor::new(&mut buf); + + let _ = cursor.write_all(slice); + test::black_box(&cursor); + }) +} + +#[bench] +fn bench_write_vec_vectored(b: &mut test::Bencher) { + let slices = [ + IoSlice::new(&[1; 128]), + IoSlice::new(&[2; 256]), + IoSlice::new(&[3; 512]), + IoSlice::new(&[4; 1024]), + IoSlice::new(&[5; 2048]), + IoSlice::new(&[6; 4096]), + IoSlice::new(&[7; 8192]), + IoSlice::new(&[8; 8192 * 2]), + ]; + + b.iter(|| { + let mut buf = b"some random data to overwrite".to_vec(); + let mut cursor = Cursor::new(&mut buf); + + let mut slices = slices; + let _ = cursor.write_all_vectored(&mut slices); + test::black_box(&cursor); + }) +} diff --git a/library/std/src/io/error.rs b/library/std/src/io/error.rs new file mode 100644 index 000000000..ff7fdcae1 --- /dev/null +++ b/library/std/src/io/error.rs @@ -0,0 +1,960 @@ +#[cfg(test)] +mod tests; + +#[cfg(target_pointer_width = "64")] +mod repr_bitpacked; +#[cfg(target_pointer_width = "64")] +use repr_bitpacked::Repr; + +#[cfg(not(target_pointer_width = "64"))] +mod repr_unpacked; +#[cfg(not(target_pointer_width = "64"))] +use repr_unpacked::Repr; + +use crate::convert::From; +use crate::error; +use crate::fmt; +use crate::result; +use crate::sys; + +/// A specialized [`Result`] type for I/O operations. +/// +/// This type is broadly used across [`std::io`] for any operation which may +/// produce an error. +/// +/// This typedef is generally used to avoid writing out [`io::Error`] directly and +/// is otherwise a direct mapping to [`Result`]. +/// +/// While usual Rust style is to import types directly, aliases of [`Result`] +/// often are not, to make it easier to distinguish between them. [`Result`] is +/// generally assumed to be [`std::result::Result`][`Result`], and so users of this alias +/// will generally use `io::Result` instead of shadowing the [prelude]'s import +/// of [`std::result::Result`][`Result`]. +/// +/// [`std::io`]: crate::io +/// [`io::Error`]: Error +/// [`Result`]: crate::result::Result +/// [prelude]: crate::prelude +/// +/// # Examples +/// +/// A convenience function that bubbles an `io::Result` to its caller: +/// +/// ``` +/// use std::io; +/// +/// fn get_string() -> io::Result<String> { +/// let mut buffer = String::new(); +/// +/// io::stdin().read_line(&mut buffer)?; +/// +/// Ok(buffer) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub type Result<T> = result::Result<T, Error>; + +/// The error type for I/O operations of the [`Read`], [`Write`], [`Seek`], and +/// associated traits. +/// +/// Errors mostly originate from the underlying OS, but custom instances of +/// `Error` can be created with crafted error messages and a particular value of +/// [`ErrorKind`]. +/// +/// [`Read`]: crate::io::Read +/// [`Write`]: crate::io::Write +/// [`Seek`]: crate::io::Seek +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Error { + repr: Repr, +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl fmt::Debug for Error { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt::Debug::fmt(&self.repr, f) + } +} + +// Only derive debug in tests, to make sure it +// doesn't accidentally get printed. +#[cfg_attr(test, derive(Debug))] +enum ErrorData<C> { + Os(i32), + Simple(ErrorKind), + SimpleMessage(&'static SimpleMessage), + Custom(C), +} + +// `#[repr(align(4))]` is probably redundant, it should have that value or +// higher already. We include it just because repr_bitpacked.rs's encoding +// requires an alignment >= 4 (note that `#[repr(align)]` will not reduce the +// alignment required by the struct, only increase it). +// +// If we add more variants to ErrorData, this can be increased to 8, but it +// should probably be behind `#[cfg_attr(target_pointer_width = "64", ...)]` or +// whatever cfg we're using to enable the `repr_bitpacked` code, since only the +// that version needs the alignment, and 8 is higher than the alignment we'll +// have on 32 bit platforms. +// +// (For the sake of being explicit: the alignment requirement here only matters +// if `error/repr_bitpacked.rs` is in use — for the unpacked repr it doesn't +// matter at all) +#[repr(align(4))] +#[derive(Debug)] +pub(crate) struct SimpleMessage { + kind: ErrorKind, + message: &'static str, +} + +impl SimpleMessage { + pub(crate) const fn new(kind: ErrorKind, message: &'static str) -> Self { + Self { kind, message } + } +} + +/// Create and return an `io::Error` for a given `ErrorKind` and constant +/// message. This doesn't allocate. +pub(crate) macro const_io_error($kind:expr, $message:expr $(,)?) { + $crate::io::error::Error::from_static_message({ + const MESSAGE_DATA: $crate::io::error::SimpleMessage = + $crate::io::error::SimpleMessage::new($kind, $message); + &MESSAGE_DATA + }) +} + +// As with `SimpleMessage`: `#[repr(align(4))]` here is just because +// repr_bitpacked's encoding requires it. In practice it almost certainly be +// already be this high or higher. +#[derive(Debug)] +#[repr(align(4))] +struct Custom { + kind: ErrorKind, + error: Box<dyn error::Error + Send + Sync>, +} + +/// A list specifying general categories of I/O error. +/// +/// This list is intended to grow over time and it is not recommended to +/// exhaustively match against it. +/// +/// It is used with the [`io::Error`] type. +/// +/// [`io::Error`]: Error +/// +/// # Handling errors and matching on `ErrorKind` +/// +/// In application code, use `match` for the `ErrorKind` values you are +/// expecting; use `_` to match "all other errors". +/// +/// In comprehensive and thorough tests that want to verify that a test doesn't +/// return any known incorrect error kind, you may want to cut-and-paste the +/// current full list of errors from here into your test code, and then match +/// `_` as the correct case. This seems counterintuitive, but it will make your +/// tests more robust. In particular, if you want to verify that your code does +/// produce an unrecognized error kind, the robust solution is to check for all +/// the recognized error kinds and fail in those cases. +#[derive(Clone, Copy, Debug, Eq, Hash, Ord, PartialEq, PartialOrd)] +#[stable(feature = "rust1", since = "1.0.0")] +#[allow(deprecated)] +#[non_exhaustive] +pub enum ErrorKind { + /// An entity was not found, often a file. + #[stable(feature = "rust1", since = "1.0.0")] + NotFound, + /// The operation lacked the necessary privileges to complete. + #[stable(feature = "rust1", since = "1.0.0")] + PermissionDenied, + /// The connection was refused by the remote server. + #[stable(feature = "rust1", since = "1.0.0")] + ConnectionRefused, + /// The connection was reset by the remote server. + #[stable(feature = "rust1", since = "1.0.0")] + ConnectionReset, + /// The remote host is not reachable. + #[unstable(feature = "io_error_more", issue = "86442")] + HostUnreachable, + /// The network containing the remote host is not reachable. + #[unstable(feature = "io_error_more", issue = "86442")] + NetworkUnreachable, + /// The connection was aborted (terminated) by the remote server. + #[stable(feature = "rust1", since = "1.0.0")] + ConnectionAborted, + /// The network operation failed because it was not connected yet. + #[stable(feature = "rust1", since = "1.0.0")] + NotConnected, + /// A socket address could not be bound because the address is already in + /// use elsewhere. + #[stable(feature = "rust1", since = "1.0.0")] + AddrInUse, + /// A nonexistent interface was requested or the requested address was not + /// local. + #[stable(feature = "rust1", since = "1.0.0")] + AddrNotAvailable, + /// The system's networking is down. + #[unstable(feature = "io_error_more", issue = "86442")] + NetworkDown, + /// The operation failed because a pipe was closed. + #[stable(feature = "rust1", since = "1.0.0")] + BrokenPipe, + /// An entity already exists, often a file. + #[stable(feature = "rust1", since = "1.0.0")] + AlreadyExists, + /// The operation needs to block to complete, but the blocking operation was + /// requested to not occur. + #[stable(feature = "rust1", since = "1.0.0")] + WouldBlock, + /// A filesystem object is, unexpectedly, not a directory. + /// + /// For example, a filesystem path was specified where one of the intermediate directory + /// components was, in fact, a plain file. + #[unstable(feature = "io_error_more", issue = "86442")] + NotADirectory, + /// The filesystem object is, unexpectedly, a directory. + /// + /// A directory was specified when a non-directory was expected. + #[unstable(feature = "io_error_more", issue = "86442")] + IsADirectory, + /// A non-empty directory was specified where an empty directory was expected. + #[unstable(feature = "io_error_more", issue = "86442")] + DirectoryNotEmpty, + /// The filesystem or storage medium is read-only, but a write operation was attempted. + #[unstable(feature = "io_error_more", issue = "86442")] + ReadOnlyFilesystem, + /// Loop in the filesystem or IO subsystem; often, too many levels of symbolic links. + /// + /// There was a loop (or excessively long chain) resolving a filesystem object + /// or file IO object. + /// + /// On Unix this is usually the result of a symbolic link loop; or, of exceeding the + /// system-specific limit on the depth of symlink traversal. + #[unstable(feature = "io_error_more", issue = "86442")] + FilesystemLoop, + /// Stale network file handle. + /// + /// With some network filesystems, notably NFS, an open file (or directory) can be invalidated + /// by problems with the network or server. + #[unstable(feature = "io_error_more", issue = "86442")] + StaleNetworkFileHandle, + /// A parameter was incorrect. + #[stable(feature = "rust1", since = "1.0.0")] + InvalidInput, + /// Data not valid for the operation were encountered. + /// + /// Unlike [`InvalidInput`], this typically means that the operation + /// parameters were valid, however the error was caused by malformed + /// input data. + /// + /// For example, a function that reads a file into a string will error with + /// `InvalidData` if the file's contents are not valid UTF-8. + /// + /// [`InvalidInput`]: ErrorKind::InvalidInput + #[stable(feature = "io_invalid_data", since = "1.2.0")] + InvalidData, + /// The I/O operation's timeout expired, causing it to be canceled. + #[stable(feature = "rust1", since = "1.0.0")] + TimedOut, + /// An error returned when an operation could not be completed because a + /// call to [`write`] returned [`Ok(0)`]. + /// + /// This typically means that an operation could only succeed if it wrote a + /// particular number of bytes but only a smaller number of bytes could be + /// written. + /// + /// [`write`]: crate::io::Write::write + /// [`Ok(0)`]: Ok + #[stable(feature = "rust1", since = "1.0.0")] + WriteZero, + /// The underlying storage (typically, a filesystem) is full. + /// + /// This does not include out of quota errors. + #[unstable(feature = "io_error_more", issue = "86442")] + StorageFull, + /// Seek on unseekable file. + /// + /// Seeking was attempted on an open file handle which is not suitable for seeking - for + /// example, on Unix, a named pipe opened with `File::open`. + #[unstable(feature = "io_error_more", issue = "86442")] + NotSeekable, + /// Filesystem quota was exceeded. + #[unstable(feature = "io_error_more", issue = "86442")] + FilesystemQuotaExceeded, + /// File larger than allowed or supported. + /// + /// This might arise from a hard limit of the underlying filesystem or file access API, or from + /// an administratively imposed resource limitation. Simple disk full, and out of quota, have + /// their own errors. + #[unstable(feature = "io_error_more", issue = "86442")] + FileTooLarge, + /// Resource is busy. + #[unstable(feature = "io_error_more", issue = "86442")] + ResourceBusy, + /// Executable file is busy. + /// + /// An attempt was made to write to a file which is also in use as a running program. (Not all + /// operating systems detect this situation.) + #[unstable(feature = "io_error_more", issue = "86442")] + ExecutableFileBusy, + /// Deadlock (avoided). + /// + /// A file locking operation would result in deadlock. This situation is typically detected, if + /// at all, on a best-effort basis. + #[unstable(feature = "io_error_more", issue = "86442")] + Deadlock, + /// Cross-device or cross-filesystem (hard) link or rename. + #[unstable(feature = "io_error_more", issue = "86442")] + CrossesDevices, + /// Too many (hard) links to the same filesystem object. + /// + /// The filesystem does not support making so many hardlinks to the same file. + #[unstable(feature = "io_error_more", issue = "86442")] + TooManyLinks, + /// A filename was invalid. + /// + /// This error can also cause if it exceeded the filename length limit. + #[unstable(feature = "io_error_more", issue = "86442")] + InvalidFilename, + /// Program argument list too long. + /// + /// When trying to run an external program, a system or process limit on the size of the + /// arguments would have been exceeded. + #[unstable(feature = "io_error_more", issue = "86442")] + ArgumentListTooLong, + /// This operation was interrupted. + /// + /// Interrupted operations can typically be retried. + #[stable(feature = "rust1", since = "1.0.0")] + Interrupted, + + /// This operation is unsupported on this platform. + /// + /// This means that the operation can never succeed. + #[stable(feature = "unsupported_error", since = "1.53.0")] + Unsupported, + + // ErrorKinds which are primarily categorisations for OS error + // codes should be added above. + // + /// An error returned when an operation could not be completed because an + /// "end of file" was reached prematurely. + /// + /// This typically means that an operation could only succeed if it read a + /// particular number of bytes but only a smaller number of bytes could be + /// read. + #[stable(feature = "read_exact", since = "1.6.0")] + UnexpectedEof, + + /// An operation could not be completed, because it failed + /// to allocate enough memory. + #[stable(feature = "out_of_memory_error", since = "1.54.0")] + OutOfMemory, + + // "Unusual" error kinds which do not correspond simply to (sets + // of) OS error codes, should be added just above this comment. + // `Other` and `Uncategorised` should remain at the end: + // + /// A custom error that does not fall under any other I/O error kind. + /// + /// This can be used to construct your own [`Error`]s that do not match any + /// [`ErrorKind`]. + /// + /// This [`ErrorKind`] is not used by the standard library. + /// + /// Errors from the standard library that do not fall under any of the I/O + /// error kinds cannot be `match`ed on, and will only match a wildcard (`_`) pattern. + /// New [`ErrorKind`]s might be added in the future for some of those. + #[stable(feature = "rust1", since = "1.0.0")] + Other, + + /// Any I/O error from the standard library that's not part of this list. + /// + /// Errors that are `Uncategorized` now may move to a different or a new + /// [`ErrorKind`] variant in the future. It is not recommended to match + /// an error against `Uncategorized`; use a wildcard match (`_`) instead. + #[unstable(feature = "io_error_uncategorized", issue = "none")] + #[doc(hidden)] + Uncategorized, +} + +impl ErrorKind { + pub(crate) fn as_str(&self) -> &'static str { + use ErrorKind::*; + // Strictly alphabetical, please. (Sadly rustfmt cannot do this yet.) + match *self { + AddrInUse => "address in use", + AddrNotAvailable => "address not available", + AlreadyExists => "entity already exists", + ArgumentListTooLong => "argument list too long", + BrokenPipe => "broken pipe", + ConnectionAborted => "connection aborted", + ConnectionRefused => "connection refused", + ConnectionReset => "connection reset", + CrossesDevices => "cross-device link or rename", + Deadlock => "deadlock", + DirectoryNotEmpty => "directory not empty", + ExecutableFileBusy => "executable file busy", + FileTooLarge => "file too large", + FilesystemLoop => "filesystem loop or indirection limit (e.g. symlink loop)", + FilesystemQuotaExceeded => "filesystem quota exceeded", + HostUnreachable => "host unreachable", + Interrupted => "operation interrupted", + InvalidData => "invalid data", + InvalidFilename => "invalid filename", + InvalidInput => "invalid input parameter", + IsADirectory => "is a directory", + NetworkDown => "network down", + NetworkUnreachable => "network unreachable", + NotADirectory => "not a directory", + NotConnected => "not connected", + NotFound => "entity not found", + NotSeekable => "seek on unseekable file", + Other => "other error", + OutOfMemory => "out of memory", + PermissionDenied => "permission denied", + ReadOnlyFilesystem => "read-only filesystem or storage medium", + ResourceBusy => "resource busy", + StaleNetworkFileHandle => "stale network file handle", + StorageFull => "no storage space", + TimedOut => "timed out", + TooManyLinks => "too many links", + Uncategorized => "uncategorized error", + UnexpectedEof => "unexpected end of file", + Unsupported => "unsupported", + WouldBlock => "operation would block", + WriteZero => "write zero", + } + } +} + +#[stable(feature = "io_errorkind_display", since = "1.60.0")] +impl fmt::Display for ErrorKind { + /// Shows a human-readable description of the `ErrorKind`. + /// + /// This is similar to `impl Display for Error`, but doesn't require first converting to Error. + /// + /// # Examples + /// ``` + /// use std::io::ErrorKind; + /// assert_eq!("entity not found", ErrorKind::NotFound.to_string()); + /// ``` + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.write_str(self.as_str()) + } +} + +/// Intended for use for errors not exposed to the user, where allocating onto +/// the heap (for normal construction via Error::new) is too costly. +#[stable(feature = "io_error_from_errorkind", since = "1.14.0")] +impl From<ErrorKind> for Error { + /// Converts an [`ErrorKind`] into an [`Error`]. + /// + /// This conversion creates a new error with a simple representation of error kind. + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// let not_found = ErrorKind::NotFound; + /// let error = Error::from(not_found); + /// assert_eq!("entity not found", format!("{error}")); + /// ``` + #[inline] + fn from(kind: ErrorKind) -> Error { + Error { repr: Repr::new_simple(kind) } + } +} + +impl Error { + /// Creates a new I/O error from a known kind of error as well as an + /// arbitrary error payload. + /// + /// This function is used to generically create I/O errors which do not + /// originate from the OS itself. The `error` argument is an arbitrary + /// payload which will be contained in this [`Error`]. + /// + /// If no extra payload is required, use the `From` conversion from + /// `ErrorKind`. + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// // errors can be created from strings + /// let custom_error = Error::new(ErrorKind::Other, "oh no!"); + /// + /// // errors can also be created from other errors + /// let custom_error2 = Error::new(ErrorKind::Interrupted, custom_error); + /// + /// // creating an error without payload + /// let eof_error = Error::from(ErrorKind::UnexpectedEof); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn new<E>(kind: ErrorKind, error: E) -> Error + where + E: Into<Box<dyn error::Error + Send + Sync>>, + { + Self::_new(kind, error.into()) + } + + /// Creates a new I/O error from an arbitrary error payload. + /// + /// This function is used to generically create I/O errors which do not + /// originate from the OS itself. It is a shortcut for [`Error::new`] + /// with [`ErrorKind::Other`]. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_error_other)] + /// + /// use std::io::Error; + /// + /// // errors can be created from strings + /// let custom_error = Error::other("oh no!"); + /// + /// // errors can also be created from other errors + /// let custom_error2 = Error::other(custom_error); + /// ``` + #[unstable(feature = "io_error_other", issue = "91946")] + pub fn other<E>(error: E) -> Error + where + E: Into<Box<dyn error::Error + Send + Sync>>, + { + Self::_new(ErrorKind::Other, error.into()) + } + + fn _new(kind: ErrorKind, error: Box<dyn error::Error + Send + Sync>) -> Error { + Error { repr: Repr::new_custom(Box::new(Custom { kind, error })) } + } + + /// Creates a new I/O error from a known kind of error as well as a constant + /// message. + /// + /// This function does not allocate. + /// + /// You should not use this directly, and instead use the `const_io_error!` + /// macro: `io::const_io_error!(ErrorKind::Something, "some_message")`. + /// + /// This function should maybe change to `from_static_message<const MSG: &'static + /// str>(kind: ErrorKind)` in the future, when const generics allow that. + #[inline] + pub(crate) const fn from_static_message(msg: &'static SimpleMessage) -> Error { + Self { repr: Repr::new_simple_message(msg) } + } + + /// Returns an error representing the last OS error which occurred. + /// + /// This function reads the value of `errno` for the target platform (e.g. + /// `GetLastError` on Windows) and will return a corresponding instance of + /// [`Error`] for the error code. + /// + /// This should be called immediately after a call to a platform function, + /// otherwise the state of the error value is indeterminate. In particular, + /// other standard library functions may call platform functions that may + /// (or may not) reset the error value even if they succeed. + /// + /// # Examples + /// + /// ``` + /// use std::io::Error; + /// + /// let os_error = Error::last_os_error(); + /// println!("last OS error: {os_error:?}"); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + #[inline] + pub fn last_os_error() -> Error { + Error::from_raw_os_error(sys::os::errno() as i32) + } + + /// Creates a new instance of an [`Error`] from a particular OS error code. + /// + /// # Examples + /// + /// On Linux: + /// + /// ``` + /// # if cfg!(target_os = "linux") { + /// use std::io; + /// + /// let error = io::Error::from_raw_os_error(22); + /// assert_eq!(error.kind(), io::ErrorKind::InvalidInput); + /// # } + /// ``` + /// + /// On Windows: + /// + /// ``` + /// # if cfg!(windows) { + /// use std::io; + /// + /// let error = io::Error::from_raw_os_error(10022); + /// assert_eq!(error.kind(), io::ErrorKind::InvalidInput); + /// # } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + #[inline] + pub fn from_raw_os_error(code: i32) -> Error { + Error { repr: Repr::new_os(code) } + } + + /// Returns the OS error that this error represents (if any). + /// + /// If this [`Error`] was constructed via [`last_os_error`] or + /// [`from_raw_os_error`], then this function will return [`Some`], otherwise + /// it will return [`None`]. + /// + /// [`last_os_error`]: Error::last_os_error + /// [`from_raw_os_error`]: Error::from_raw_os_error + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// fn print_os_error(err: &Error) { + /// if let Some(raw_os_err) = err.raw_os_error() { + /// println!("raw OS error: {raw_os_err:?}"); + /// } else { + /// println!("Not an OS error"); + /// } + /// } + /// + /// fn main() { + /// // Will print "raw OS error: ...". + /// print_os_error(&Error::last_os_error()); + /// // Will print "Not an OS error". + /// print_os_error(&Error::new(ErrorKind::Other, "oh no!")); + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + #[inline] + pub fn raw_os_error(&self) -> Option<i32> { + match self.repr.data() { + ErrorData::Os(i) => Some(i), + ErrorData::Custom(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + } + } + + /// Returns a reference to the inner error wrapped by this error (if any). + /// + /// If this [`Error`] was constructed via [`new`] then this function will + /// return [`Some`], otherwise it will return [`None`]. + /// + /// [`new`]: Error::new + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// fn print_error(err: &Error) { + /// if let Some(inner_err) = err.get_ref() { + /// println!("Inner error: {inner_err:?}"); + /// } else { + /// println!("No inner error"); + /// } + /// } + /// + /// fn main() { + /// // Will print "No inner error". + /// print_error(&Error::last_os_error()); + /// // Will print "Inner error: ...". + /// print_error(&Error::new(ErrorKind::Other, "oh no!")); + /// } + /// ``` + #[stable(feature = "io_error_inner", since = "1.3.0")] + #[must_use] + #[inline] + pub fn get_ref(&self) -> Option<&(dyn error::Error + Send + Sync + 'static)> { + match self.repr.data() { + ErrorData::Os(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + ErrorData::Custom(c) => Some(&*c.error), + } + } + + /// Returns a mutable reference to the inner error wrapped by this error + /// (if any). + /// + /// If this [`Error`] was constructed via [`new`] then this function will + /// return [`Some`], otherwise it will return [`None`]. + /// + /// [`new`]: Error::new + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// use std::{error, fmt}; + /// use std::fmt::Display; + /// + /// #[derive(Debug)] + /// struct MyError { + /// v: String, + /// } + /// + /// impl MyError { + /// fn new() -> MyError { + /// MyError { + /// v: "oh no!".to_string() + /// } + /// } + /// + /// fn change_message(&mut self, new_message: &str) { + /// self.v = new_message.to_string(); + /// } + /// } + /// + /// impl error::Error for MyError {} + /// + /// impl Display for MyError { + /// fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + /// write!(f, "MyError: {}", &self.v) + /// } + /// } + /// + /// fn change_error(mut err: Error) -> Error { + /// if let Some(inner_err) = err.get_mut() { + /// inner_err.downcast_mut::<MyError>().unwrap().change_message("I've been changed!"); + /// } + /// err + /// } + /// + /// fn print_error(err: &Error) { + /// if let Some(inner_err) = err.get_ref() { + /// println!("Inner error: {inner_err}"); + /// } else { + /// println!("No inner error"); + /// } + /// } + /// + /// fn main() { + /// // Will print "No inner error". + /// print_error(&change_error(Error::last_os_error())); + /// // Will print "Inner error: ...". + /// print_error(&change_error(Error::new(ErrorKind::Other, MyError::new()))); + /// } + /// ``` + #[stable(feature = "io_error_inner", since = "1.3.0")] + #[must_use] + #[inline] + pub fn get_mut(&mut self) -> Option<&mut (dyn error::Error + Send + Sync + 'static)> { + match self.repr.data_mut() { + ErrorData::Os(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + ErrorData::Custom(c) => Some(&mut *c.error), + } + } + + /// Consumes the `Error`, returning its inner error (if any). + /// + /// If this [`Error`] was constructed via [`new`] then this function will + /// return [`Some`], otherwise it will return [`None`]. + /// + /// [`new`]: Error::new + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// fn print_error(err: Error) { + /// if let Some(inner_err) = err.into_inner() { + /// println!("Inner error: {inner_err}"); + /// } else { + /// println!("No inner error"); + /// } + /// } + /// + /// fn main() { + /// // Will print "No inner error". + /// print_error(Error::last_os_error()); + /// // Will print "Inner error: ...". + /// print_error(Error::new(ErrorKind::Other, "oh no!")); + /// } + /// ``` + #[stable(feature = "io_error_inner", since = "1.3.0")] + #[must_use = "`self` will be dropped if the result is not used"] + #[inline] + pub fn into_inner(self) -> Option<Box<dyn error::Error + Send + Sync>> { + match self.repr.into_data() { + ErrorData::Os(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + ErrorData::Custom(c) => Some(c.error), + } + } + + /// Attempt to downgrade the inner error to `E` if any. + /// + /// If this [`Error`] was constructed via [`new`] then this function will + /// attempt to perform downgrade on it, otherwise it will return [`Err`]. + /// + /// If downgrade succeeds, it will return [`Ok`], otherwise it will also + /// return [`Err`]. + /// + /// [`new`]: Error::new + /// + /// # Examples + /// + /// ``` + /// #![feature(io_error_downcast)] + /// + /// use std::fmt; + /// use std::io; + /// use std::error::Error; + /// + /// #[derive(Debug)] + /// enum E { + /// Io(io::Error), + /// SomeOtherVariant, + /// } + /// + /// impl fmt::Display for E { + /// // ... + /// # fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + /// # todo!() + /// # } + /// } + /// impl Error for E {} + /// + /// impl From<io::Error> for E { + /// fn from(err: io::Error) -> E { + /// err.downcast::<E>() + /// .map(|b| *b) + /// .unwrap_or_else(E::Io) + /// } + /// } + /// ``` + #[unstable(feature = "io_error_downcast", issue = "99262")] + pub fn downcast<E>(self) -> result::Result<Box<E>, Self> + where + E: error::Error + Send + Sync + 'static, + { + match self.repr.into_data() { + ErrorData::Custom(b) if b.error.is::<E>() => { + let res = (*b).error.downcast::<E>(); + + // downcast is a really trivial and is marked as inline, so + // it's likely be inlined here. + // + // And the compiler should be able to eliminate the branch + // that produces `Err` here since b.error.is::<E>() + // returns true. + Ok(res.unwrap()) + } + repr_data => Err(Self { repr: Repr::new(repr_data) }), + } + } + + /// Returns the corresponding [`ErrorKind`] for this error. + /// + /// # Examples + /// + /// ``` + /// use std::io::{Error, ErrorKind}; + /// + /// fn print_error(err: Error) { + /// println!("{:?}", err.kind()); + /// } + /// + /// fn main() { + /// // Will print "Uncategorized". + /// print_error(Error::last_os_error()); + /// // Will print "AddrInUse". + /// print_error(Error::new(ErrorKind::AddrInUse, "oh no!")); + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + #[inline] + pub fn kind(&self) -> ErrorKind { + match self.repr.data() { + ErrorData::Os(code) => sys::decode_error_kind(code), + ErrorData::Custom(c) => c.kind, + ErrorData::Simple(kind) => kind, + ErrorData::SimpleMessage(m) => m.kind, + } + } +} + +impl fmt::Debug for Repr { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + match self.data() { + ErrorData::Os(code) => fmt + .debug_struct("Os") + .field("code", &code) + .field("kind", &sys::decode_error_kind(code)) + .field("message", &sys::os::error_string(code)) + .finish(), + ErrorData::Custom(c) => fmt::Debug::fmt(&c, fmt), + ErrorData::Simple(kind) => fmt.debug_tuple("Kind").field(&kind).finish(), + ErrorData::SimpleMessage(msg) => fmt + .debug_struct("Error") + .field("kind", &msg.kind) + .field("message", &msg.message) + .finish(), + } + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl fmt::Display for Error { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + match self.repr.data() { + ErrorData::Os(code) => { + let detail = sys::os::error_string(code); + write!(fmt, "{detail} (os error {code})") + } + ErrorData::Custom(ref c) => c.error.fmt(fmt), + ErrorData::Simple(kind) => write!(fmt, "{}", kind.as_str()), + ErrorData::SimpleMessage(msg) => msg.message.fmt(fmt), + } + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl error::Error for Error { + #[allow(deprecated, deprecated_in_future)] + fn description(&self) -> &str { + match self.repr.data() { + ErrorData::Os(..) | ErrorData::Simple(..) => self.kind().as_str(), + ErrorData::SimpleMessage(msg) => msg.message, + ErrorData::Custom(c) => c.error.description(), + } + } + + #[allow(deprecated)] + fn cause(&self) -> Option<&dyn error::Error> { + match self.repr.data() { + ErrorData::Os(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + ErrorData::Custom(c) => c.error.cause(), + } + } + + fn source(&self) -> Option<&(dyn error::Error + 'static)> { + match self.repr.data() { + ErrorData::Os(..) => None, + ErrorData::Simple(..) => None, + ErrorData::SimpleMessage(..) => None, + ErrorData::Custom(c) => c.error.source(), + } + } +} + +fn _assert_error_is_sync_send() { + fn _is_sync_send<T: Sync + Send>() {} + _is_sync_send::<Error>(); +} diff --git a/library/std/src/io/error/repr_bitpacked.rs b/library/std/src/io/error/repr_bitpacked.rs new file mode 100644 index 000000000..292bf4826 --- /dev/null +++ b/library/std/src/io/error/repr_bitpacked.rs @@ -0,0 +1,409 @@ +//! This is a densely packed error representation which is used on targets with +//! 64-bit pointers. +//! +//! (Note that `bitpacked` vs `unpacked` here has no relationship to +//! `#[repr(packed)]`, it just refers to attempting to use any available bits in +//! a more clever manner than `rustc`'s default layout algorithm would). +//! +//! Conceptually, it stores the same data as the "unpacked" equivalent we use on +//! other targets. Specifically, you can imagine it as an optimized version of +//! the following enum (which is roughly equivalent to what's stored by +//! `repr_unpacked::Repr`, e.g. `super::ErrorData<Box<Custom>>`): +//! +//! ```ignore (exposition-only) +//! enum ErrorData { +//! Os(i32), +//! Simple(ErrorKind), +//! SimpleMessage(&'static SimpleMessage), +//! Custom(Box<Custom>), +//! } +//! ``` +//! +//! However, it packs this data into a 64bit non-zero value. +//! +//! This optimization not only allows `io::Error` to occupy a single pointer, +//! but improves `io::Result` as well, especially for situations like +//! `io::Result<()>` (which is now 64 bits) or `io::Result<u64>` (which is now +//! 128 bits), which are quite common. +//! +//! # Layout +//! Tagged values are 64 bits, with the 2 least significant bits used for the +//! tag. This means there are there are 4 "variants": +//! +//! - **Tag 0b00**: The first variant is equivalent to +//! `ErrorData::SimpleMessage`, and holds a `&'static SimpleMessage` directly. +//! +//! `SimpleMessage` has an alignment >= 4 (which is requested with +//! `#[repr(align)]` and checked statically at the bottom of this file), which +//! means every `&'static SimpleMessage` should have the both tag bits as 0, +//! meaning its tagged and untagged representation are equivalent. +//! +//! This means we can skip tagging it, which is necessary as this variant can +//! be constructed from a `const fn`, which probably cannot tag pointers (or +//! at least it would be difficult). +//! +//! - **Tag 0b01**: The other pointer variant holds the data for +//! `ErrorData::Custom` and the remaining 62 bits are used to store a +//! `Box<Custom>`. `Custom` also has alignment >= 4, so the bottom two bits +//! are free to use for the tag. +//! +//! The only important thing to note is that `ptr::wrapping_add` and +//! `ptr::wrapping_sub` are used to tag the pointer, rather than bitwise +//! operations. This should preserve the pointer's provenance, which would +//! otherwise be lost. +//! +//! - **Tag 0b10**: Holds the data for `ErrorData::Os(i32)`. We store the `i32` +//! in the pointer's most significant 32 bits, and don't use the bits `2..32` +//! for anything. Using the top 32 bits is just to let us easily recover the +//! `i32` code with the correct sign. +//! +//! - **Tag 0b11**: Holds the data for `ErrorData::Simple(ErrorKind)`. This +//! stores the `ErrorKind` in the top 32 bits as well, although it doesn't +//! occupy nearly that many. Most of the bits are unused here, but it's not +//! like we need them for anything else yet. +//! +//! # Use of `NonNull<()>` +//! +//! Everything is stored in a `NonNull<()>`, which is odd, but actually serves a +//! purpose. +//! +//! Conceptually you might think of this more like: +//! +//! ```ignore (exposition-only) +//! union Repr { +//! // holds integer (Simple/Os) variants, and +//! // provides access to the tag bits. +//! bits: NonZeroU64, +//! // Tag is 0, so this is stored untagged. +//! msg: &'static SimpleMessage, +//! // Tagged (offset) `Box<Custom>` pointer. +//! tagged_custom: NonNull<()>, +//! } +//! ``` +//! +//! But there are a few problems with this: +//! +//! 1. Union access is equivalent to a transmute, so this representation would +//! require we transmute between integers and pointers in at least one +//! direction, which may be UB (and even if not, it is likely harder for a +//! compiler to reason about than explicit ptr->int operations). +//! +//! 2. Even if all fields of a union have a niche, the union itself doesn't, +//! although this may change in the future. This would make things like +//! `io::Result<()>` and `io::Result<usize>` larger, which defeats part of +//! the motivation of this bitpacking. +//! +//! Storing everything in a `NonZeroUsize` (or some other integer) would be a +//! bit more traditional for pointer tagging, but it would lose provenance +//! information, couldn't be constructed from a `const fn`, and would probably +//! run into other issues as well. +//! +//! The `NonNull<()>` seems like the only alternative, even if it's fairly odd +//! to use a pointer type to store something that may hold an integer, some of +//! the time. + +use super::{Custom, ErrorData, ErrorKind, SimpleMessage}; +use alloc::boxed::Box; +use core::marker::PhantomData; +use core::mem::{align_of, size_of}; +use core::ptr::{self, NonNull}; + +// The 2 least-significant bits are used as tag. +const TAG_MASK: usize = 0b11; +const TAG_SIMPLE_MESSAGE: usize = 0b00; +const TAG_CUSTOM: usize = 0b01; +const TAG_OS: usize = 0b10; +const TAG_SIMPLE: usize = 0b11; + +/// The internal representation. +/// +/// See the module docs for more, this is just a way to hack in a check that we +/// indeed are not unwind-safe. +/// +/// ```compile_fail,E0277 +/// fn is_unwind_safe<T: core::panic::UnwindSafe>() {} +/// is_unwind_safe::<std::io::Error>(); +/// ``` +#[repr(transparent)] +pub(super) struct Repr(NonNull<()>, PhantomData<ErrorData<Box<Custom>>>); + +// All the types `Repr` stores internally are Send + Sync, and so is it. +unsafe impl Send for Repr {} +unsafe impl Sync for Repr {} + +impl Repr { + pub(super) fn new(dat: ErrorData<Box<Custom>>) -> Self { + match dat { + ErrorData::Os(code) => Self::new_os(code), + ErrorData::Simple(kind) => Self::new_simple(kind), + ErrorData::SimpleMessage(simple_message) => Self::new_simple_message(simple_message), + ErrorData::Custom(b) => Self::new_custom(b), + } + } + + pub(super) fn new_custom(b: Box<Custom>) -> Self { + let p = Box::into_raw(b).cast::<u8>(); + // Should only be possible if an allocator handed out a pointer with + // wrong alignment. + debug_assert_eq!(p.addr() & TAG_MASK, 0); + // Note: We know `TAG_CUSTOM <= size_of::<Custom>()` (static_assert at + // end of file), and both the start and end of the expression must be + // valid without address space wraparound due to `Box`'s semantics. + // + // This means it would be correct to implement this using `ptr::add` + // (rather than `ptr::wrapping_add`), but it's unclear this would give + // any benefit, so we just use `wrapping_add` instead. + let tagged = p.wrapping_add(TAG_CUSTOM).cast::<()>(); + // Safety: `TAG_CUSTOM + p` is the same as `TAG_CUSTOM | p`, + // because `p`'s alignment means it isn't allowed to have any of the + // `TAG_BITS` set (you can verify that addition and bitwise-or are the + // same when the operands have no bits in common using a truth table). + // + // Then, `TAG_CUSTOM | p` is not zero, as that would require + // `TAG_CUSTOM` and `p` both be zero, and neither is (as `p` came from a + // box, and `TAG_CUSTOM` just... isn't zero -- it's `0b01`). Therefore, + // `TAG_CUSTOM + p` isn't zero and so `tagged` can't be, and the + // `new_unchecked` is safe. + let res = Self(unsafe { NonNull::new_unchecked(tagged) }, PhantomData); + // quickly smoke-check we encoded the right thing (This generally will + // only run in libstd's tests, unless the user uses -Zbuild-std) + debug_assert!(matches!(res.data(), ErrorData::Custom(_)), "repr(custom) encoding failed"); + res + } + + #[inline] + pub(super) fn new_os(code: i32) -> Self { + let utagged = ((code as usize) << 32) | TAG_OS; + // Safety: `TAG_OS` is not zero, so the result of the `|` is not 0. + let res = Self(unsafe { NonNull::new_unchecked(ptr::invalid_mut(utagged)) }, PhantomData); + // quickly smoke-check we encoded the right thing (This generally will + // only run in libstd's tests, unless the user uses -Zbuild-std) + debug_assert!( + matches!(res.data(), ErrorData::Os(c) if c == code), + "repr(os) encoding failed for {code}" + ); + res + } + + #[inline] + pub(super) fn new_simple(kind: ErrorKind) -> Self { + let utagged = ((kind as usize) << 32) | TAG_SIMPLE; + // Safety: `TAG_SIMPLE` is not zero, so the result of the `|` is not 0. + let res = Self(unsafe { NonNull::new_unchecked(ptr::invalid_mut(utagged)) }, PhantomData); + // quickly smoke-check we encoded the right thing (This generally will + // only run in libstd's tests, unless the user uses -Zbuild-std) + debug_assert!( + matches!(res.data(), ErrorData::Simple(k) if k == kind), + "repr(simple) encoding failed {:?}", + kind, + ); + res + } + + #[inline] + pub(super) const fn new_simple_message(m: &'static SimpleMessage) -> Self { + // Safety: References are never null. + Self(unsafe { NonNull::new_unchecked(m as *const _ as *mut ()) }, PhantomData) + } + + #[inline] + pub(super) fn data(&self) -> ErrorData<&Custom> { + // Safety: We're a Repr, decode_repr is fine. + unsafe { decode_repr(self.0, |c| &*c) } + } + + #[inline] + pub(super) fn data_mut(&mut self) -> ErrorData<&mut Custom> { + // Safety: We're a Repr, decode_repr is fine. + unsafe { decode_repr(self.0, |c| &mut *c) } + } + + #[inline] + pub(super) fn into_data(self) -> ErrorData<Box<Custom>> { + let this = core::mem::ManuallyDrop::new(self); + // Safety: We're a Repr, decode_repr is fine. The `Box::from_raw` is + // safe because we prevent double-drop using `ManuallyDrop`. + unsafe { decode_repr(this.0, |p| Box::from_raw(p)) } + } +} + +impl Drop for Repr { + #[inline] + fn drop(&mut self) { + // Safety: We're a Repr, decode_repr is fine. The `Box::from_raw` is + // safe because we're being dropped. + unsafe { + let _ = decode_repr(self.0, |p| Box::<Custom>::from_raw(p)); + } + } +} + +// Shared helper to decode a `Repr`'s internal pointer into an ErrorData. +// +// Safety: `ptr`'s bits should be encoded as described in the document at the +// top (it should `some_repr.0`) +#[inline] +unsafe fn decode_repr<C, F>(ptr: NonNull<()>, make_custom: F) -> ErrorData<C> +where + F: FnOnce(*mut Custom) -> C, +{ + let bits = ptr.as_ptr().addr(); + match bits & TAG_MASK { + TAG_OS => { + let code = ((bits as i64) >> 32) as i32; + ErrorData::Os(code) + } + TAG_SIMPLE => { + let kind_bits = (bits >> 32) as u32; + let kind = kind_from_prim(kind_bits).unwrap_or_else(|| { + debug_assert!(false, "Invalid io::error::Repr bits: `Repr({:#018x})`", bits); + // This means the `ptr` passed in was not valid, which violates + // the unsafe contract of `decode_repr`. + // + // Using this rather than unwrap meaningfully improves the code + // for callers which only care about one variant (usually + // `Custom`) + core::hint::unreachable_unchecked(); + }); + ErrorData::Simple(kind) + } + TAG_SIMPLE_MESSAGE => ErrorData::SimpleMessage(&*ptr.cast::<SimpleMessage>().as_ptr()), + TAG_CUSTOM => { + // It would be correct for us to use `ptr::sub` here (see the + // comment above the `wrapping_add` call in `new_custom` for why), + // but it isn't clear that it makes a difference, so we don't. + let custom = ptr.as_ptr().cast::<u8>().wrapping_sub(TAG_CUSTOM).cast::<Custom>(); + ErrorData::Custom(make_custom(custom)) + } + _ => { + // Can't happen, and compiler can tell + unreachable!(); + } + } +} + +// This compiles to the same code as the check+transmute, but doesn't require +// unsafe, or to hard-code max ErrorKind or its size in a way the compiler +// couldn't verify. +#[inline] +fn kind_from_prim(ek: u32) -> Option<ErrorKind> { + macro_rules! from_prim { + ($prim:expr => $Enum:ident { $($Variant:ident),* $(,)? }) => {{ + // Force a compile error if the list gets out of date. + const _: fn(e: $Enum) = |e: $Enum| match e { + $($Enum::$Variant => ()),* + }; + match $prim { + $(v if v == ($Enum::$Variant as _) => Some($Enum::$Variant),)* + _ => None, + } + }} + } + from_prim!(ek => ErrorKind { + NotFound, + PermissionDenied, + ConnectionRefused, + ConnectionReset, + HostUnreachable, + NetworkUnreachable, + ConnectionAborted, + NotConnected, + AddrInUse, + AddrNotAvailable, + NetworkDown, + BrokenPipe, + AlreadyExists, + WouldBlock, + NotADirectory, + IsADirectory, + DirectoryNotEmpty, + ReadOnlyFilesystem, + FilesystemLoop, + StaleNetworkFileHandle, + InvalidInput, + InvalidData, + TimedOut, + WriteZero, + StorageFull, + NotSeekable, + FilesystemQuotaExceeded, + FileTooLarge, + ResourceBusy, + ExecutableFileBusy, + Deadlock, + CrossesDevices, + TooManyLinks, + InvalidFilename, + ArgumentListTooLong, + Interrupted, + Other, + UnexpectedEof, + Unsupported, + OutOfMemory, + Uncategorized, + }) +} + +// Some static checking to alert us if a change breaks any of the assumptions +// that our encoding relies on for correctness and soundness. (Some of these are +// a bit overly thorough/cautious, admittedly) +// +// If any of these are hit on a platform that libstd supports, we should likely +// just use `repr_unpacked.rs` there instead (unless the fix is easy). +macro_rules! static_assert { + ($condition:expr) => { + const _: () = assert!($condition); + }; + (@usize_eq: $lhs:expr, $rhs:expr) => { + const _: [(); $lhs] = [(); $rhs]; + }; +} + +// The bitpacking we use requires pointers be exactly 64 bits. +static_assert!(@usize_eq: size_of::<NonNull<()>>(), 8); + +// We also require pointers and usize be the same size. +static_assert!(@usize_eq: size_of::<NonNull<()>>(), size_of::<usize>()); + +// `Custom` and `SimpleMessage` need to be thin pointers. +static_assert!(@usize_eq: size_of::<&'static SimpleMessage>(), 8); +static_assert!(@usize_eq: size_of::<Box<Custom>>(), 8); + +static_assert!((TAG_MASK + 1).is_power_of_two()); +// And they must have sufficient alignment. +static_assert!(align_of::<SimpleMessage>() >= TAG_MASK + 1); +static_assert!(align_of::<Custom>() >= TAG_MASK + 1); + +static_assert!(@usize_eq: (TAG_MASK & TAG_SIMPLE_MESSAGE), TAG_SIMPLE_MESSAGE); +static_assert!(@usize_eq: (TAG_MASK & TAG_CUSTOM), TAG_CUSTOM); +static_assert!(@usize_eq: (TAG_MASK & TAG_OS), TAG_OS); +static_assert!(@usize_eq: (TAG_MASK & TAG_SIMPLE), TAG_SIMPLE); + +// This is obviously true (`TAG_CUSTOM` is `0b01`), but in `Repr::new_custom` we +// offset a pointer by this value, and expect it to both be within the same +// object, and to not wrap around the address space. See the comment in that +// function for further details. +// +// Actually, at the moment we use `ptr::wrapping_add`, not `ptr::add`, so this +// check isn't needed for that one, although the assertion that we don't +// actually wrap around in that wrapping_add does simplify the safety reasoning +// elsewhere considerably. +static_assert!(size_of::<Custom>() >= TAG_CUSTOM); + +// These two store a payload which is allowed to be zero, so they must be +// non-zero to preserve the `NonNull`'s range invariant. +static_assert!(TAG_OS != 0); +static_assert!(TAG_SIMPLE != 0); +// We can't tag `SimpleMessage`s, the tag must be 0. +static_assert!(@usize_eq: TAG_SIMPLE_MESSAGE, 0); + +// Check that the point of all of this still holds. +// +// We'd check against `io::Error`, but *technically* it's allowed to vary, +// as it's not `#[repr(transparent)]`/`#[repr(C)]`. We could add that, but +// the `#[repr()]` would show up in rustdoc, which might be seen as a stable +// commitment. +static_assert!(@usize_eq: size_of::<Repr>(), 8); +static_assert!(@usize_eq: size_of::<Option<Repr>>(), 8); +static_assert!(@usize_eq: size_of::<Result<(), Repr>>(), 8); +static_assert!(@usize_eq: size_of::<Result<usize, Repr>>(), 16); diff --git a/library/std/src/io/error/repr_unpacked.rs b/library/std/src/io/error/repr_unpacked.rs new file mode 100644 index 000000000..d6ad55b99 --- /dev/null +++ b/library/std/src/io/error/repr_unpacked.rs @@ -0,0 +1,54 @@ +//! This is a fairly simple unpacked error representation that's used on +//! non-64bit targets, where the packed 64 bit representation wouldn't work, and +//! would have no benefit. + +use super::{Custom, ErrorData, ErrorKind, SimpleMessage}; +use alloc::boxed::Box; + +type Inner = ErrorData<Box<Custom>>; + +pub(super) struct Repr(Inner); + +impl Repr { + #[inline] + pub(super) fn new(dat: ErrorData<Box<Custom>>) -> Self { + Self(dat) + } + pub(super) fn new_custom(b: Box<Custom>) -> Self { + Self(Inner::Custom(b)) + } + #[inline] + pub(super) fn new_os(code: i32) -> Self { + Self(Inner::Os(code)) + } + #[inline] + pub(super) fn new_simple(kind: ErrorKind) -> Self { + Self(Inner::Simple(kind)) + } + #[inline] + pub(super) const fn new_simple_message(m: &'static SimpleMessage) -> Self { + Self(Inner::SimpleMessage(m)) + } + #[inline] + pub(super) fn into_data(self) -> ErrorData<Box<Custom>> { + self.0 + } + #[inline] + pub(super) fn data(&self) -> ErrorData<&Custom> { + match &self.0 { + Inner::Os(c) => ErrorData::Os(*c), + Inner::Simple(k) => ErrorData::Simple(*k), + Inner::SimpleMessage(m) => ErrorData::SimpleMessage(*m), + Inner::Custom(m) => ErrorData::Custom(&*m), + } + } + #[inline] + pub(super) fn data_mut(&mut self) -> ErrorData<&mut Custom> { + match &mut self.0 { + Inner::Os(c) => ErrorData::Os(*c), + Inner::Simple(k) => ErrorData::Simple(*k), + Inner::SimpleMessage(m) => ErrorData::SimpleMessage(*m), + Inner::Custom(m) => ErrorData::Custom(&mut *m), + } + } +} diff --git a/library/std/src/io/error/tests.rs b/library/std/src/io/error/tests.rs new file mode 100644 index 000000000..c897a5e87 --- /dev/null +++ b/library/std/src/io/error/tests.rs @@ -0,0 +1,194 @@ +use super::{const_io_error, Custom, Error, ErrorData, ErrorKind, Repr, SimpleMessage}; +use crate::assert_matches::assert_matches; +use crate::error; +use crate::fmt; +use crate::mem::size_of; +use crate::sys::decode_error_kind; +use crate::sys::os::error_string; + +#[test] +fn test_size() { + assert!(size_of::<Error>() <= size_of::<[usize; 2]>()); +} + +#[test] +fn test_debug_error() { + let code = 6; + let msg = error_string(code); + let kind = decode_error_kind(code); + let err = Error { + repr: Repr::new_custom(Box::new(Custom { + kind: ErrorKind::InvalidInput, + error: Box::new(Error { repr: super::Repr::new_os(code) }), + })), + }; + let expected = format!( + "Custom {{ \ + kind: InvalidInput, \ + error: Os {{ \ + code: {:?}, \ + kind: {:?}, \ + message: {:?} \ + }} \ + }}", + code, kind, msg + ); + assert_eq!(format!("{err:?}"), expected); +} + +#[test] +fn test_downcasting() { + #[derive(Debug)] + struct TestError; + + impl fmt::Display for TestError { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.write_str("asdf") + } + } + + impl error::Error for TestError {} + + // we have to call all of these UFCS style right now since method + // resolution won't implicitly drop the Send+Sync bounds + let mut err = Error::new(ErrorKind::Other, TestError); + assert!(err.get_ref().unwrap().is::<TestError>()); + assert_eq!("asdf", err.get_ref().unwrap().to_string()); + assert!(err.get_mut().unwrap().is::<TestError>()); + let extracted = err.into_inner().unwrap(); + extracted.downcast::<TestError>().unwrap(); +} + +#[test] +fn test_const() { + const E: Error = const_io_error!(ErrorKind::NotFound, "hello"); + + assert_eq!(E.kind(), ErrorKind::NotFound); + assert_eq!(E.to_string(), "hello"); + assert!(format!("{E:?}").contains("\"hello\"")); + assert!(format!("{E:?}").contains("NotFound")); +} + +#[test] +fn test_os_packing() { + for code in -20i32..20i32 { + let e = Error::from_raw_os_error(code); + assert_eq!(e.raw_os_error(), Some(code)); + assert_matches!( + e.repr.data(), + ErrorData::Os(c) if c == code, + ); + } +} + +#[test] +fn test_errorkind_packing() { + assert_eq!(Error::from(ErrorKind::NotFound).kind(), ErrorKind::NotFound); + assert_eq!(Error::from(ErrorKind::PermissionDenied).kind(), ErrorKind::PermissionDenied); + assert_eq!(Error::from(ErrorKind::Uncategorized).kind(), ErrorKind::Uncategorized); + // Check that the innards look like like what we want. + assert_matches!( + Error::from(ErrorKind::OutOfMemory).repr.data(), + ErrorData::Simple(ErrorKind::OutOfMemory), + ); +} + +#[test] +fn test_simple_message_packing() { + use super::{ErrorKind::*, SimpleMessage}; + macro_rules! check_simple_msg { + ($err:expr, $kind:ident, $msg:literal) => {{ + let e = &$err; + // Check that the public api is right. + assert_eq!(e.kind(), $kind); + assert!(format!("{e:?}").contains($msg)); + // and we got what we expected + assert_matches!( + e.repr.data(), + ErrorData::SimpleMessage(SimpleMessage { kind: $kind, message: $msg }) + ); + }}; + } + + let not_static = const_io_error!(Uncategorized, "not a constant!"); + check_simple_msg!(not_static, Uncategorized, "not a constant!"); + + const CONST: Error = const_io_error!(NotFound, "definitely a constant!"); + check_simple_msg!(CONST, NotFound, "definitely a constant!"); + + static STATIC: Error = const_io_error!(BrokenPipe, "a constant, sort of!"); + check_simple_msg!(STATIC, BrokenPipe, "a constant, sort of!"); +} + +#[derive(Debug, PartialEq)] +struct Bojji(bool); +impl error::Error for Bojji {} +impl fmt::Display for Bojji { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "ah! {:?}", self) + } +} + +#[test] +fn test_custom_error_packing() { + use super::Custom; + let test = Error::new(ErrorKind::Uncategorized, Bojji(true)); + assert_matches!( + test.repr.data(), + ErrorData::Custom(Custom { + kind: ErrorKind::Uncategorized, + error, + }) if error.downcast_ref::<Bojji>().as_deref() == Some(&Bojji(true)), + ); +} + +#[derive(Debug)] +struct E; + +impl fmt::Display for E { + fn fmt(&self, _f: &mut fmt::Formatter<'_>) -> fmt::Result { + Ok(()) + } +} + +impl error::Error for E {} + +#[test] +fn test_std_io_error_downcast() { + // Case 1: custom error, downcast succeeds + let io_error = Error::new(ErrorKind::Other, Bojji(true)); + let e: Box<Bojji> = io_error.downcast().unwrap(); + assert!(e.0); + + // Case 2: custom error, downcast fails + let io_error = Error::new(ErrorKind::Other, Bojji(true)); + let io_error = io_error.downcast::<E>().unwrap_err(); + + // ensures that the custom error is intact + assert_eq!(ErrorKind::Other, io_error.kind()); + let e: Box<Bojji> = io_error.downcast().unwrap(); + assert!(e.0); + + // Case 3: os error + let errno = 20; + let io_error = Error::from_raw_os_error(errno); + let io_error = io_error.downcast::<E>().unwrap_err(); + + assert_eq!(errno, io_error.raw_os_error().unwrap()); + + // Case 4: simple + let kind = ErrorKind::OutOfMemory; + let io_error: Error = kind.into(); + let io_error = io_error.downcast::<E>().unwrap_err(); + + assert_eq!(kind, io_error.kind()); + + // Case 5: simple message + const SIMPLE_MESSAGE: SimpleMessage = + SimpleMessage { kind: ErrorKind::Other, message: "simple message error test" }; + let io_error = Error::from_static_message(&SIMPLE_MESSAGE); + let io_error = io_error.downcast::<E>().unwrap_err(); + + assert_eq!(SIMPLE_MESSAGE.kind, io_error.kind()); + assert_eq!(SIMPLE_MESSAGE.message, &*format!("{io_error}")); +} diff --git a/library/std/src/io/impls.rs b/library/std/src/io/impls.rs new file mode 100644 index 000000000..950725473 --- /dev/null +++ b/library/std/src/io/impls.rs @@ -0,0 +1,458 @@ +#[cfg(test)] +mod tests; + +use crate::alloc::Allocator; +use crate::cmp; +use crate::collections::VecDeque; +use crate::fmt; +use crate::io::{ + self, BufRead, ErrorKind, IoSlice, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, Write, +}; +use crate::mem; + +// ============================================================================= +// Forwarding implementations + +#[stable(feature = "rust1", since = "1.0.0")] +impl<R: Read + ?Sized> Read for &mut R { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + (**self).read(buf) + } + + #[inline] + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + (**self).read_buf(buf) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + (**self).read_vectored(bufs) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + (**self).is_read_vectored() + } + + #[inline] + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + (**self).read_to_end(buf) + } + + #[inline] + fn read_to_string(&mut self, buf: &mut String) -> io::Result<usize> { + (**self).read_to_string(buf) + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + (**self).read_exact(buf) + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write + ?Sized> Write for &mut W { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + (**self).write(buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + (**self).write_vectored(bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + (**self).is_write_vectored() + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + (**self).flush() + } + + #[inline] + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + (**self).write_all(buf) + } + + #[inline] + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> io::Result<()> { + (**self).write_fmt(fmt) + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<S: Seek + ?Sized> Seek for &mut S { + #[inline] + fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> { + (**self).seek(pos) + } + + #[inline] + fn stream_position(&mut self) -> io::Result<u64> { + (**self).stream_position() + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<B: BufRead + ?Sized> BufRead for &mut B { + #[inline] + fn fill_buf(&mut self) -> io::Result<&[u8]> { + (**self).fill_buf() + } + + #[inline] + fn consume(&mut self, amt: usize) { + (**self).consume(amt) + } + + #[inline] + fn read_until(&mut self, byte: u8, buf: &mut Vec<u8>) -> io::Result<usize> { + (**self).read_until(byte, buf) + } + + #[inline] + fn read_line(&mut self, buf: &mut String) -> io::Result<usize> { + (**self).read_line(buf) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<R: Read + ?Sized> Read for Box<R> { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + (**self).read(buf) + } + + #[inline] + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + (**self).read_buf(buf) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + (**self).read_vectored(bufs) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + (**self).is_read_vectored() + } + + #[inline] + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + (**self).read_to_end(buf) + } + + #[inline] + fn read_to_string(&mut self, buf: &mut String) -> io::Result<usize> { + (**self).read_to_string(buf) + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + (**self).read_exact(buf) + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<W: Write + ?Sized> Write for Box<W> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + (**self).write(buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + (**self).write_vectored(bufs) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + (**self).is_write_vectored() + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + (**self).flush() + } + + #[inline] + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + (**self).write_all(buf) + } + + #[inline] + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> io::Result<()> { + (**self).write_fmt(fmt) + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<S: Seek + ?Sized> Seek for Box<S> { + #[inline] + fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> { + (**self).seek(pos) + } + + #[inline] + fn stream_position(&mut self) -> io::Result<u64> { + (**self).stream_position() + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl<B: BufRead + ?Sized> BufRead for Box<B> { + #[inline] + fn fill_buf(&mut self) -> io::Result<&[u8]> { + (**self).fill_buf() + } + + #[inline] + fn consume(&mut self, amt: usize) { + (**self).consume(amt) + } + + #[inline] + fn read_until(&mut self, byte: u8, buf: &mut Vec<u8>) -> io::Result<usize> { + (**self).read_until(byte, buf) + } + + #[inline] + fn read_line(&mut self, buf: &mut String) -> io::Result<usize> { + (**self).read_line(buf) + } +} + +// ============================================================================= +// In-memory buffer implementations + +/// Read is implemented for `&[u8]` by copying from the slice. +/// +/// Note that reading updates the slice to point to the yet unread part. +/// The slice will be empty when EOF is reached. +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for &[u8] { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + let amt = cmp::min(buf.len(), self.len()); + let (a, b) = self.split_at(amt); + + // First check if the amount of bytes we want to read is small: + // `copy_from_slice` will generally expand to a call to `memcpy`, and + // for a single byte the overhead is significant. + if amt == 1 { + buf[0] = a[0]; + } else { + buf[..amt].copy_from_slice(a); + } + + *self = b; + Ok(amt) + } + + #[inline] + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + let amt = cmp::min(buf.remaining(), self.len()); + let (a, b) = self.split_at(amt); + + buf.append(a); + + *self = b; + Ok(()) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + let mut nread = 0; + for buf in bufs { + nread += self.read(buf)?; + if self.is_empty() { + break; + } + } + + Ok(nread) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + true + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + if buf.len() > self.len() { + return Err(io::const_io_error!( + ErrorKind::UnexpectedEof, + "failed to fill whole buffer" + )); + } + let (a, b) = self.split_at(buf.len()); + + // First check if the amount of bytes we want to read is small: + // `copy_from_slice` will generally expand to a call to `memcpy`, and + // for a single byte the overhead is significant. + if buf.len() == 1 { + buf[0] = a[0]; + } else { + buf.copy_from_slice(a); + } + + *self = b; + Ok(()) + } + + #[inline] + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + buf.extend_from_slice(*self); + let len = self.len(); + *self = &self[len..]; + Ok(len) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl BufRead for &[u8] { + #[inline] + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Ok(*self) + } + + #[inline] + fn consume(&mut self, amt: usize) { + *self = &self[amt..]; + } +} + +/// Write is implemented for `&mut [u8]` by copying into the slice, overwriting +/// its data. +/// +/// Note that writing updates the slice to point to the yet unwritten part. +/// The slice will be empty when it has been completely overwritten. +/// +/// If the number of bytes to be written exceeds the size of the slice, write operations will +/// return short writes: ultimately, `Ok(0)`; in this situation, `write_all` returns an error of +/// kind `ErrorKind::WriteZero`. +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for &mut [u8] { + #[inline] + fn write(&mut self, data: &[u8]) -> io::Result<usize> { + let amt = cmp::min(data.len(), self.len()); + let (a, b) = mem::replace(self, &mut []).split_at_mut(amt); + a.copy_from_slice(&data[..amt]); + *self = b; + Ok(amt) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let mut nwritten = 0; + for buf in bufs { + nwritten += self.write(buf)?; + if self.is_empty() { + break; + } + } + + Ok(nwritten) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn write_all(&mut self, data: &[u8]) -> io::Result<()> { + if self.write(data)? == data.len() { + Ok(()) + } else { + Err(io::const_io_error!(ErrorKind::WriteZero, "failed to write whole buffer")) + } + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +/// Write is implemented for `Vec<u8>` by appending to the vector. +/// The vector will grow as needed. +#[stable(feature = "rust1", since = "1.0.0")] +impl<A: Allocator> Write for Vec<u8, A> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.extend_from_slice(buf); + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let len = bufs.iter().map(|b| b.len()).sum(); + self.reserve(len); + for buf in bufs { + self.extend_from_slice(buf); + } + Ok(len) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.extend_from_slice(buf); + Ok(()) + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +/// Read is implemented for `VecDeque<u8>` by consuming bytes from the front of the `VecDeque`. +#[stable(feature = "vecdeque_read_write", since = "1.63.0")] +impl<A: Allocator> Read for VecDeque<u8, A> { + /// Fill `buf` with the contents of the "front" slice as returned by + /// [`as_slices`][`VecDeque::as_slices`]. If the contained byte slices of the `VecDeque` are + /// discontiguous, multiple calls to `read` will be needed to read the entire content. + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + let (ref mut front, _) = self.as_slices(); + let n = Read::read(front, buf)?; + self.drain(..n); + Ok(n) + } + + #[inline] + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + let (ref mut front, _) = self.as_slices(); + let n = cmp::min(buf.remaining(), front.len()); + Read::read_buf(front, buf)?; + self.drain(..n); + Ok(()) + } +} + +/// Write is implemented for `VecDeque<u8>` by appending to the `VecDeque`, growing it as needed. +#[stable(feature = "vecdeque_read_write", since = "1.63.0")] +impl<A: Allocator> Write for VecDeque<u8, A> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.extend(buf); + Ok(buf.len()) + } + + #[inline] + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.extend(buf); + Ok(()) + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} diff --git a/library/std/src/io/impls/tests.rs b/library/std/src/io/impls/tests.rs new file mode 100644 index 000000000..d1cd84a67 --- /dev/null +++ b/library/std/src/io/impls/tests.rs @@ -0,0 +1,57 @@ +use crate::io::prelude::*; + +#[bench] +fn bench_read_slice(b: &mut test::Bencher) { + let buf = [5; 1024]; + let mut dst = [0; 128]; + + b.iter(|| { + let mut rd = &buf[..]; + for _ in 0..8 { + let _ = rd.read(&mut dst); + test::black_box(&dst); + } + }) +} + +#[bench] +fn bench_write_slice(b: &mut test::Bencher) { + let mut buf = [0; 1024]; + let src = [5; 128]; + + b.iter(|| { + let mut wr = &mut buf[..]; + for _ in 0..8 { + let _ = wr.write_all(&src); + test::black_box(&wr); + } + }) +} + +#[bench] +fn bench_read_vec(b: &mut test::Bencher) { + let buf = vec![5; 1024]; + let mut dst = [0; 128]; + + b.iter(|| { + let mut rd = &buf[..]; + for _ in 0..8 { + let _ = rd.read(&mut dst); + test::black_box(&dst); + } + }) +} + +#[bench] +fn bench_write_vec(b: &mut test::Bencher) { + let mut buf = Vec::with_capacity(1024); + let src = [5; 128]; + + b.iter(|| { + let mut wr = &mut buf[..]; + for _ in 0..8 { + let _ = wr.write_all(&src); + test::black_box(&wr); + } + }) +} diff --git a/library/std/src/io/mod.rs b/library/std/src/io/mod.rs new file mode 100644 index 000000000..96addbd1a --- /dev/null +++ b/library/std/src/io/mod.rs @@ -0,0 +1,2827 @@ +//! Traits, helpers, and type definitions for core I/O functionality. +//! +//! The `std::io` module contains a number of common things you'll need +//! when doing input and output. The most core part of this module is +//! the [`Read`] and [`Write`] traits, which provide the +//! most general interface for reading and writing input and output. +//! +//! # Read and Write +//! +//! Because they are traits, [`Read`] and [`Write`] are implemented by a number +//! of other types, and you can implement them for your types too. As such, +//! you'll see a few different types of I/O throughout the documentation in +//! this module: [`File`]s, [`TcpStream`]s, and sometimes even [`Vec<T>`]s. For +//! example, [`Read`] adds a [`read`][`Read::read`] method, which we can use on +//! [`File`]s: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! use std::fs::File; +//! +//! fn main() -> io::Result<()> { +//! let mut f = File::open("foo.txt")?; +//! let mut buffer = [0; 10]; +//! +//! // read up to 10 bytes +//! let n = f.read(&mut buffer)?; +//! +//! println!("The bytes: {:?}", &buffer[..n]); +//! Ok(()) +//! } +//! ``` +//! +//! [`Read`] and [`Write`] are so important, implementors of the two traits have a +//! nickname: readers and writers. So you'll sometimes see 'a reader' instead +//! of 'a type that implements the [`Read`] trait'. Much easier! +//! +//! ## Seek and BufRead +//! +//! Beyond that, there are two important traits that are provided: [`Seek`] +//! and [`BufRead`]. Both of these build on top of a reader to control +//! how the reading happens. [`Seek`] lets you control where the next byte is +//! coming from: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! use std::io::SeekFrom; +//! use std::fs::File; +//! +//! fn main() -> io::Result<()> { +//! let mut f = File::open("foo.txt")?; +//! let mut buffer = [0; 10]; +//! +//! // skip to the last 10 bytes of the file +//! f.seek(SeekFrom::End(-10))?; +//! +//! // read up to 10 bytes +//! let n = f.read(&mut buffer)?; +//! +//! println!("The bytes: {:?}", &buffer[..n]); +//! Ok(()) +//! } +//! ``` +//! +//! [`BufRead`] uses an internal buffer to provide a number of other ways to read, but +//! to show it off, we'll need to talk about buffers in general. Keep reading! +//! +//! ## BufReader and BufWriter +//! +//! Byte-based interfaces are unwieldy and can be inefficient, as we'd need to be +//! making near-constant calls to the operating system. To help with this, +//! `std::io` comes with two structs, [`BufReader`] and [`BufWriter`], which wrap +//! readers and writers. The wrapper uses a buffer, reducing the number of +//! calls and providing nicer methods for accessing exactly what you want. +//! +//! For example, [`BufReader`] works with the [`BufRead`] trait to add extra +//! methods to any reader: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! use std::io::BufReader; +//! use std::fs::File; +//! +//! fn main() -> io::Result<()> { +//! let f = File::open("foo.txt")?; +//! let mut reader = BufReader::new(f); +//! let mut buffer = String::new(); +//! +//! // read a line into buffer +//! reader.read_line(&mut buffer)?; +//! +//! println!("{buffer}"); +//! Ok(()) +//! } +//! ``` +//! +//! [`BufWriter`] doesn't add any new ways of writing; it just buffers every call +//! to [`write`][`Write::write`]: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! use std::io::BufWriter; +//! use std::fs::File; +//! +//! fn main() -> io::Result<()> { +//! let f = File::create("foo.txt")?; +//! { +//! let mut writer = BufWriter::new(f); +//! +//! // write a byte to the buffer +//! writer.write(&[42])?; +//! +//! } // the buffer is flushed once writer goes out of scope +//! +//! Ok(()) +//! } +//! ``` +//! +//! ## Standard input and output +//! +//! A very common source of input is standard input: +//! +//! ```no_run +//! use std::io; +//! +//! fn main() -> io::Result<()> { +//! let mut input = String::new(); +//! +//! io::stdin().read_line(&mut input)?; +//! +//! println!("You typed: {}", input.trim()); +//! Ok(()) +//! } +//! ``` +//! +//! Note that you cannot use the [`?` operator] in functions that do not return +//! a [`Result<T, E>`][`Result`]. Instead, you can call [`.unwrap()`] +//! or `match` on the return value to catch any possible errors: +//! +//! ```no_run +//! use std::io; +//! +//! let mut input = String::new(); +//! +//! io::stdin().read_line(&mut input).unwrap(); +//! ``` +//! +//! And a very common source of output is standard output: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! +//! fn main() -> io::Result<()> { +//! io::stdout().write(&[42])?; +//! Ok(()) +//! } +//! ``` +//! +//! Of course, using [`io::stdout`] directly is less common than something like +//! [`println!`]. +//! +//! ## Iterator types +//! +//! A large number of the structures provided by `std::io` are for various +//! ways of iterating over I/O. For example, [`Lines`] is used to split over +//! lines: +//! +//! ```no_run +//! use std::io; +//! use std::io::prelude::*; +//! use std::io::BufReader; +//! use std::fs::File; +//! +//! fn main() -> io::Result<()> { +//! let f = File::open("foo.txt")?; +//! let reader = BufReader::new(f); +//! +//! for line in reader.lines() { +//! println!("{}", line?); +//! } +//! Ok(()) +//! } +//! ``` +//! +//! ## Functions +//! +//! There are a number of [functions][functions-list] that offer access to various +//! features. For example, we can use three of these functions to copy everything +//! from standard input to standard output: +//! +//! ```no_run +//! use std::io; +//! +//! fn main() -> io::Result<()> { +//! io::copy(&mut io::stdin(), &mut io::stdout())?; +//! Ok(()) +//! } +//! ``` +//! +//! [functions-list]: #functions-1 +//! +//! ## io::Result +//! +//! Last, but certainly not least, is [`io::Result`]. This type is used +//! as the return type of many `std::io` functions that can cause an error, and +//! can be returned from your own functions as well. Many of the examples in this +//! module use the [`?` operator]: +//! +//! ``` +//! use std::io; +//! +//! fn read_input() -> io::Result<()> { +//! let mut input = String::new(); +//! +//! io::stdin().read_line(&mut input)?; +//! +//! println!("You typed: {}", input.trim()); +//! +//! Ok(()) +//! } +//! ``` +//! +//! The return type of `read_input()`, [`io::Result<()>`][`io::Result`], is a very +//! common type for functions which don't have a 'real' return value, but do want to +//! return errors if they happen. In this case, the only purpose of this function is +//! to read the line and print it, so we use `()`. +//! +//! ## Platform-specific behavior +//! +//! Many I/O functions throughout the standard library are documented to indicate +//! what various library or syscalls they are delegated to. This is done to help +//! applications both understand what's happening under the hood as well as investigate +//! any possibly unclear semantics. Note, however, that this is informative, not a binding +//! contract. The implementation of many of these functions are subject to change over +//! time and may call fewer or more syscalls/library functions. +//! +//! [`File`]: crate::fs::File +//! [`TcpStream`]: crate::net::TcpStream +//! [`io::stdout`]: stdout +//! [`io::Result`]: self::Result +//! [`?` operator]: ../../book/appendix-02-operators.html +//! [`Result`]: crate::result::Result +//! [`.unwrap()`]: crate::result::Result::unwrap + +#![stable(feature = "rust1", since = "1.0.0")] + +#[cfg(test)] +mod tests; + +use crate::cmp; +use crate::fmt; +use crate::mem::replace; +use crate::ops::{Deref, DerefMut}; +use crate::slice; +use crate::str; +use crate::sys; +use crate::sys_common::memchr; + +#[stable(feature = "bufwriter_into_parts", since = "1.56.0")] +pub use self::buffered::WriterPanicked; +#[unstable(feature = "internal_output_capture", issue = "none")] +#[doc(no_inline, hidden)] +pub use self::stdio::set_output_capture; +#[unstable(feature = "print_internals", issue = "none")] +pub use self::stdio::{_eprint, _print}; +#[stable(feature = "rust1", since = "1.0.0")] +pub use self::{ + buffered::{BufReader, BufWriter, IntoInnerError, LineWriter}, + copy::copy, + cursor::Cursor, + error::{Error, ErrorKind, Result}, + stdio::{stderr, stdin, stdout, Stderr, StderrLock, Stdin, StdinLock, Stdout, StdoutLock}, + util::{empty, repeat, sink, Empty, Repeat, Sink}, +}; + +#[unstable(feature = "read_buf", issue = "78485")] +pub use self::readbuf::ReadBuf; +pub(crate) use error::const_io_error; + +mod buffered; +pub(crate) mod copy; +mod cursor; +mod error; +mod impls; +pub mod prelude; +mod readbuf; +mod stdio; +mod util; + +const DEFAULT_BUF_SIZE: usize = crate::sys_common::io::DEFAULT_BUF_SIZE; + +pub(crate) use stdio::cleanup; + +struct Guard<'a> { + buf: &'a mut Vec<u8>, + len: usize, +} + +impl Drop for Guard<'_> { + fn drop(&mut self) { + unsafe { + self.buf.set_len(self.len); + } + } +} + +// Several `read_to_string` and `read_line` methods in the standard library will +// append data into a `String` buffer, but we need to be pretty careful when +// doing this. The implementation will just call `.as_mut_vec()` and then +// delegate to a byte-oriented reading method, but we must ensure that when +// returning we never leave `buf` in a state such that it contains invalid UTF-8 +// in its bounds. +// +// To this end, we use an RAII guard (to protect against panics) which updates +// the length of the string when it is dropped. This guard initially truncates +// the string to the prior length and only after we've validated that the +// new contents are valid UTF-8 do we allow it to set a longer length. +// +// The unsafety in this function is twofold: +// +// 1. We're looking at the raw bytes of `buf`, so we take on the burden of UTF-8 +// checks. +// 2. We're passing a raw buffer to the function `f`, and it is expected that +// the function only *appends* bytes to the buffer. We'll get undefined +// behavior if existing bytes are overwritten to have non-UTF-8 data. +pub(crate) unsafe fn append_to_string<F>(buf: &mut String, f: F) -> Result<usize> +where + F: FnOnce(&mut Vec<u8>) -> Result<usize>, +{ + let mut g = Guard { len: buf.len(), buf: buf.as_mut_vec() }; + let ret = f(g.buf); + if str::from_utf8(&g.buf[g.len..]).is_err() { + ret.and_then(|_| { + Err(error::const_io_error!( + ErrorKind::InvalidData, + "stream did not contain valid UTF-8" + )) + }) + } else { + g.len = g.buf.len(); + ret + } +} + +// This uses an adaptive system to extend the vector when it fills. We want to +// avoid paying to allocate and zero a huge chunk of memory if the reader only +// has 4 bytes while still making large reads if the reader does have a ton +// of data to return. Simply tacking on an extra DEFAULT_BUF_SIZE space every +// time is 4,500 times (!) slower than a default reservation size of 32 if the +// reader has a very small amount of data to return. +pub(crate) fn default_read_to_end<R: Read + ?Sized>(r: &mut R, buf: &mut Vec<u8>) -> Result<usize> { + let start_len = buf.len(); + let start_cap = buf.capacity(); + + let mut initialized = 0; // Extra initialized bytes from previous loop iteration + loop { + if buf.len() == buf.capacity() { + buf.reserve(32); // buf is full, need more space + } + + let mut read_buf = ReadBuf::uninit(buf.spare_capacity_mut()); + + // SAFETY: These bytes were initialized but not filled in the previous loop + unsafe { + read_buf.assume_init(initialized); + } + + match r.read_buf(&mut read_buf) { + Ok(()) => {} + Err(e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + } + + if read_buf.filled_len() == 0 { + return Ok(buf.len() - start_len); + } + + // store how much was initialized but not filled + initialized = read_buf.initialized_len() - read_buf.filled_len(); + let new_len = read_buf.filled_len() + buf.len(); + + // SAFETY: ReadBuf's invariants mean this much memory is init + unsafe { + buf.set_len(new_len); + } + + if buf.len() == buf.capacity() && buf.capacity() == start_cap { + // The buffer might be an exact fit. Let's read into a probe buffer + // and see if it returns `Ok(0)`. If so, we've avoided an + // unnecessary doubling of the capacity. But if not, append the + // probe buffer to the primary buffer and let its capacity grow. + let mut probe = [0u8; 32]; + + loop { + match r.read(&mut probe) { + Ok(0) => return Ok(buf.len() - start_len), + Ok(n) => { + buf.extend_from_slice(&probe[..n]); + break; + } + Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + } + } + } + } +} + +pub(crate) fn default_read_to_string<R: Read + ?Sized>( + r: &mut R, + buf: &mut String, +) -> Result<usize> { + // Note that we do *not* call `r.read_to_end()` here. We are passing + // `&mut Vec<u8>` (the raw contents of `buf`) into the `read_to_end` + // method to fill it up. An arbitrary implementation could overwrite the + // entire contents of the vector, not just append to it (which is what + // we are expecting). + // + // To prevent extraneously checking the UTF-8-ness of the entire buffer + // we pass it to our hardcoded `default_read_to_end` implementation which + // we know is guaranteed to only read data into the end of the buffer. + unsafe { append_to_string(buf, |b| default_read_to_end(r, b)) } +} + +pub(crate) fn default_read_vectored<F>(read: F, bufs: &mut [IoSliceMut<'_>]) -> Result<usize> +where + F: FnOnce(&mut [u8]) -> Result<usize>, +{ + let buf = bufs.iter_mut().find(|b| !b.is_empty()).map_or(&mut [][..], |b| &mut **b); + read(buf) +} + +pub(crate) fn default_write_vectored<F>(write: F, bufs: &[IoSlice<'_>]) -> Result<usize> +where + F: FnOnce(&[u8]) -> Result<usize>, +{ + let buf = bufs.iter().find(|b| !b.is_empty()).map_or(&[][..], |b| &**b); + write(buf) +} + +pub(crate) fn default_read_exact<R: Read + ?Sized>(this: &mut R, mut buf: &mut [u8]) -> Result<()> { + while !buf.is_empty() { + match this.read(buf) { + Ok(0) => break, + Ok(n) => { + let tmp = buf; + buf = &mut tmp[n..]; + } + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } + if !buf.is_empty() { + Err(error::const_io_error!(ErrorKind::UnexpectedEof, "failed to fill whole buffer")) + } else { + Ok(()) + } +} + +pub(crate) fn default_read_buf<F>(read: F, buf: &mut ReadBuf<'_>) -> Result<()> +where + F: FnOnce(&mut [u8]) -> Result<usize>, +{ + let n = read(buf.initialize_unfilled())?; + buf.add_filled(n); + Ok(()) +} + +/// The `Read` trait allows for reading bytes from a source. +/// +/// Implementors of the `Read` trait are called 'readers'. +/// +/// Readers are defined by one required method, [`read()`]. Each call to [`read()`] +/// will attempt to pull bytes from this source into a provided buffer. A +/// number of other methods are implemented in terms of [`read()`], giving +/// implementors a number of ways to read bytes while only needing to implement +/// a single method. +/// +/// Readers are intended to be composable with one another. Many implementors +/// throughout [`std::io`] take and provide types which implement the `Read` +/// trait. +/// +/// Please note that each call to [`read()`] may involve a system call, and +/// therefore, using something that implements [`BufRead`], such as +/// [`BufReader`], will be more efficient. +/// +/// # Examples +/// +/// [`File`]s implement `Read`: +/// +/// ```no_run +/// use std::io; +/// use std::io::prelude::*; +/// use std::fs::File; +/// +/// fn main() -> io::Result<()> { +/// let mut f = File::open("foo.txt")?; +/// let mut buffer = [0; 10]; +/// +/// // read up to 10 bytes +/// f.read(&mut buffer)?; +/// +/// let mut buffer = Vec::new(); +/// // read the whole file +/// f.read_to_end(&mut buffer)?; +/// +/// // read into a String, so that you don't need to do the conversion. +/// let mut buffer = String::new(); +/// f.read_to_string(&mut buffer)?; +/// +/// // and more! See the other methods for more details. +/// Ok(()) +/// } +/// ``` +/// +/// Read from [`&str`] because [`&[u8]`][prim@slice] implements `Read`: +/// +/// ```no_run +/// # use std::io; +/// use std::io::prelude::*; +/// +/// fn main() -> io::Result<()> { +/// let mut b = "This string will be read".as_bytes(); +/// let mut buffer = [0; 10]; +/// +/// // read up to 10 bytes +/// b.read(&mut buffer)?; +/// +/// // etc... it works exactly as a File does! +/// Ok(()) +/// } +/// ``` +/// +/// [`read()`]: Read::read +/// [`&str`]: prim@str +/// [`std::io`]: self +/// [`File`]: crate::fs::File +#[stable(feature = "rust1", since = "1.0.0")] +#[doc(notable_trait)] +#[cfg_attr(not(test), rustc_diagnostic_item = "IoRead")] +pub trait Read { + /// Pull some bytes from this source into the specified buffer, returning + /// how many bytes were read. + /// + /// This function does not provide any guarantees about whether it blocks + /// waiting for data, but if an object needs to block for a read and cannot, + /// it will typically signal this via an [`Err`] return value. + /// + /// If the return value of this method is [`Ok(n)`], then implementations must + /// guarantee that `0 <= n <= buf.len()`. A nonzero `n` value indicates + /// that the buffer `buf` has been filled in with `n` bytes of data from this + /// source. If `n` is `0`, then it can indicate one of two scenarios: + /// + /// 1. This reader has reached its "end of file" and will likely no longer + /// be able to produce bytes. Note that this does not mean that the + /// reader will *always* no longer be able to produce bytes. As an example, + /// on Linux, this method will call the `recv` syscall for a [`TcpStream`], + /// where returning zero indicates the connection was shut down correctly. While + /// for [`File`], it is possible to reach the end of file and get zero as result, + /// but if more data is appended to the file, future calls to `read` will return + /// more data. + /// 2. The buffer specified was 0 bytes in length. + /// + /// It is not an error if the returned value `n` is smaller than the buffer size, + /// even when the reader is not at the end of the stream yet. + /// This may happen for example because fewer bytes are actually available right now + /// (e. g. being close to end-of-file) or because read() was interrupted by a signal. + /// + /// As this trait is safe to implement, callers cannot rely on `n <= buf.len()` for safety. + /// Extra care needs to be taken when `unsafe` functions are used to access the read bytes. + /// Callers have to ensure that no unchecked out-of-bounds accesses are possible even if + /// `n > buf.len()`. + /// + /// No guarantees are provided about the contents of `buf` when this + /// function is called, implementations cannot rely on any property of the + /// contents of `buf` being true. It is recommended that *implementations* + /// only write data to `buf` instead of reading its contents. + /// + /// Correspondingly, however, *callers* of this method must not assume any guarantees + /// about how the implementation uses `buf`. The trait is safe to implement, + /// so it is possible that the code that's supposed to write to the buffer might also read + /// from it. It is your responsibility to make sure that `buf` is initialized + /// before calling `read`. Calling `read` with an uninitialized `buf` (of the kind one + /// obtains via [`MaybeUninit<T>`]) is not safe, and can lead to undefined behavior. + /// + /// [`MaybeUninit<T>`]: crate::mem::MaybeUninit + /// + /// # Errors + /// + /// If this function encounters any form of I/O or other error, an error + /// variant will be returned. If an error is returned then it must be + /// guaranteed that no bytes were read. + /// + /// An error of the [`ErrorKind::Interrupted`] kind is non-fatal and the read + /// operation should be retried if there is nothing else to do. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`Ok(n)`]: Ok + /// [`File`]: crate::fs::File + /// [`TcpStream`]: crate::net::TcpStream + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// let mut buffer = [0; 10]; + /// + /// // read up to 10 bytes + /// let n = f.read(&mut buffer[..])?; + /// + /// println!("The bytes: {:?}", &buffer[..n]); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn read(&mut self, buf: &mut [u8]) -> Result<usize>; + + /// Like `read`, except that it reads into a slice of buffers. + /// + /// Data is copied to fill each buffer in order, with the final buffer + /// written to possibly being only partially filled. This method must + /// behave equivalently to a single call to `read` with concatenated + /// buffers. + /// + /// The default implementation calls `read` with either the first nonempty + /// buffer provided, or an empty one if none exists. + #[stable(feature = "iovec", since = "1.36.0")] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> Result<usize> { + default_read_vectored(|b| self.read(b), bufs) + } + + /// Determines if this `Read`er has an efficient `read_vectored` + /// implementation. + /// + /// If a `Read`er does not override the default `read_vectored` + /// implementation, code using it may want to avoid the method all together + /// and coalesce writes into a single buffer for higher performance. + /// + /// The default implementation returns `false`. + #[unstable(feature = "can_vector", issue = "69941")] + fn is_read_vectored(&self) -> bool { + false + } + + /// Read all bytes until EOF in this source, placing them into `buf`. + /// + /// All bytes read from this source will be appended to the specified buffer + /// `buf`. This function will continuously call [`read()`] to append more data to + /// `buf` until [`read()`] returns either [`Ok(0)`] or an error of + /// non-[`ErrorKind::Interrupted`] kind. + /// + /// If successful, this function will return the total number of bytes read. + /// + /// # Errors + /// + /// If this function encounters an error of the kind + /// [`ErrorKind::Interrupted`] then the error is ignored and the operation + /// will continue. + /// + /// If any other read error is encountered then this function immediately + /// returns. Any bytes which have already been read will be appended to + /// `buf`. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`read()`]: Read::read + /// [`Ok(0)`]: Ok + /// [`File`]: crate::fs::File + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// let mut buffer = Vec::new(); + /// + /// // read the whole file + /// f.read_to_end(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + /// + /// (See also the [`std::fs::read`] convenience function for reading from a + /// file.) + /// + /// [`std::fs::read`]: crate::fs::read + #[stable(feature = "rust1", since = "1.0.0")] + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> Result<usize> { + default_read_to_end(self, buf) + } + + /// Read all bytes until EOF in this source, appending them to `buf`. + /// + /// If successful, this function returns the number of bytes which were read + /// and appended to `buf`. + /// + /// # Errors + /// + /// If the data in this stream is *not* valid UTF-8 then an error is + /// returned and `buf` is unchanged. + /// + /// See [`read_to_end`] for other error semantics. + /// + /// [`read_to_end`]: Read::read_to_end + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`File`]: crate::fs::File + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// let mut buffer = String::new(); + /// + /// f.read_to_string(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + /// + /// (See also the [`std::fs::read_to_string`] convenience function for + /// reading from a file.) + /// + /// [`std::fs::read_to_string`]: crate::fs::read_to_string + #[stable(feature = "rust1", since = "1.0.0")] + fn read_to_string(&mut self, buf: &mut String) -> Result<usize> { + default_read_to_string(self, buf) + } + + /// Read the exact number of bytes required to fill `buf`. + /// + /// This function reads as many bytes as necessary to completely fill the + /// specified buffer `buf`. + /// + /// No guarantees are provided about the contents of `buf` when this + /// function is called, implementations cannot rely on any property of the + /// contents of `buf` being true. It is recommended that implementations + /// only write data to `buf` instead of reading its contents. The + /// documentation on [`read`] has a more detailed explanation on this + /// subject. + /// + /// # Errors + /// + /// If this function encounters an error of the kind + /// [`ErrorKind::Interrupted`] then the error is ignored and the operation + /// will continue. + /// + /// If this function encounters an "end of file" before completely filling + /// the buffer, it returns an error of the kind [`ErrorKind::UnexpectedEof`]. + /// The contents of `buf` are unspecified in this case. + /// + /// If any other read error is encountered then this function immediately + /// returns. The contents of `buf` are unspecified in this case. + /// + /// If this function returns an error, it is unspecified how many bytes it + /// has read, but it will never read more than would be necessary to + /// completely fill the buffer. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`read`]: Read::read + /// [`File`]: crate::fs::File + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// let mut buffer = [0; 10]; + /// + /// // read exactly 10 bytes + /// f.read_exact(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "read_exact", since = "1.6.0")] + fn read_exact(&mut self, buf: &mut [u8]) -> Result<()> { + default_read_exact(self, buf) + } + + /// Pull some bytes from this source into the specified buffer. + /// + /// This is equivalent to the [`read`](Read::read) method, except that it is passed a [`ReadBuf`] rather than `[u8]` to allow use + /// with uninitialized buffers. The new data will be appended to any existing contents of `buf`. + /// + /// The default implementation delegates to `read`. + #[unstable(feature = "read_buf", issue = "78485")] + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { + default_read_buf(|b| self.read(b), buf) + } + + /// Read the exact number of bytes required to fill `buf`. + /// + /// This is equivalent to the [`read_exact`](Read::read_exact) method, except that it is passed a [`ReadBuf`] rather than `[u8]` to + /// allow use with uninitialized buffers. + #[unstable(feature = "read_buf", issue = "78485")] + fn read_buf_exact(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { + while buf.remaining() > 0 { + let prev_filled = buf.filled().len(); + match self.read_buf(buf) { + Ok(()) => {} + Err(e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + } + + if buf.filled().len() == prev_filled { + return Err(Error::new(ErrorKind::UnexpectedEof, "failed to fill buffer")); + } + } + + Ok(()) + } + + /// Creates a "by reference" adaptor for this instance of `Read`. + /// + /// The returned adapter also implements `Read` and will simply borrow this + /// current reader. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`File`]: crate::fs::File + /// + /// ```no_run + /// use std::io; + /// use std::io::Read; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// let mut buffer = Vec::new(); + /// let mut other_buffer = Vec::new(); + /// + /// { + /// let reference = f.by_ref(); + /// + /// // read at most 5 bytes + /// reference.take(5).read_to_end(&mut buffer)?; + /// + /// } // drop our &mut reference so we can use f again + /// + /// // original file still usable, read the rest + /// f.read_to_end(&mut other_buffer)?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn by_ref(&mut self) -> &mut Self + where + Self: Sized, + { + self + } + + /// Transforms this `Read` instance to an [`Iterator`] over its bytes. + /// + /// The returned type implements [`Iterator`] where the [`Item`] is + /// <code>[Result]<[u8], [io::Error]></code>. + /// The yielded item is [`Ok`] if a byte was successfully read and [`Err`] + /// otherwise. EOF is mapped to returning [`None`] from this iterator. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`Item`]: Iterator::Item + /// [`File`]: crate::fs::File "fs::File" + /// [Result]: crate::result::Result "Result" + /// [io::Error]: self::Error "io::Error" + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let f = File::open("foo.txt")?; + /// + /// for byte in f.bytes() { + /// println!("{}", byte.unwrap()); + /// } + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn bytes(self) -> Bytes<Self> + where + Self: Sized, + { + Bytes { inner: self } + } + + /// Creates an adapter which will chain this stream with another. + /// + /// The returned `Read` instance will first read all bytes from this object + /// until EOF is encountered. Afterwards the output is equivalent to the + /// output of `next`. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`File`]: crate::fs::File + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let f1 = File::open("foo.txt")?; + /// let f2 = File::open("bar.txt")?; + /// + /// let mut handle = f1.chain(f2); + /// let mut buffer = String::new(); + /// + /// // read the value into a String. We could use any Read method here, + /// // this is just one example. + /// handle.read_to_string(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn chain<R: Read>(self, next: R) -> Chain<Self, R> + where + Self: Sized, + { + Chain { first: self, second: next, done_first: false } + } + + /// Creates an adapter which will read at most `limit` bytes from it. + /// + /// This function returns a new instance of `Read` which will read at most + /// `limit` bytes, after which it will always return EOF ([`Ok(0)`]). Any + /// read errors will not count towards the number of bytes read and future + /// calls to [`read()`] may succeed. + /// + /// # Examples + /// + /// [`File`]s implement `Read`: + /// + /// [`File`]: crate::fs::File + /// [`Ok(0)`]: Ok + /// [`read()`]: Read::read + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let f = File::open("foo.txt")?; + /// let mut buffer = [0; 5]; + /// + /// // read at most five bytes + /// let mut handle = f.take(5); + /// + /// handle.read(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn take(self, limit: u64) -> Take<Self> + where + Self: Sized, + { + Take { inner: self, limit } + } +} + +/// Read all bytes from a [reader][Read] into a new [`String`]. +/// +/// This is a convenience function for [`Read::read_to_string`]. Using this +/// function avoids having to create a variable first and provides more type +/// safety since you can only get the buffer out if there were no errors. (If you +/// use [`Read::read_to_string`] you have to remember to check whether the read +/// succeeded because otherwise your buffer will be empty or only partially full.) +/// +/// # Performance +/// +/// The downside of this function's increased ease of use and type safety is +/// that it gives you less control over performance. For example, you can't +/// pre-allocate memory like you can using [`String::with_capacity`] and +/// [`Read::read_to_string`]. Also, you can't re-use the buffer if an error +/// occurs while reading. +/// +/// In many cases, this function's performance will be adequate and the ease of use +/// and type safety tradeoffs will be worth it. However, there are cases where you +/// need more control over performance, and in those cases you should definitely use +/// [`Read::read_to_string`] directly. +/// +/// Note that in some special cases, such as when reading files, this function will +/// pre-allocate memory based on the size of the input it is reading. In those +/// cases, the performance should be as good as if you had used +/// [`Read::read_to_string`] with a manually pre-allocated buffer. +/// +/// # Errors +/// +/// This function forces you to handle errors because the output (the `String`) +/// is wrapped in a [`Result`]. See [`Read::read_to_string`] for the errors +/// that can occur. If any error occurs, you will get an [`Err`], so you +/// don't have to worry about your buffer being empty or partially full. +/// +/// # Examples +/// +/// ```no_run +/// #![feature(io_read_to_string)] +/// +/// # use std::io; +/// fn main() -> io::Result<()> { +/// let stdin = io::read_to_string(io::stdin())?; +/// println!("Stdin was:"); +/// println!("{stdin}"); +/// Ok(()) +/// } +/// ``` +#[unstable(feature = "io_read_to_string", issue = "80218")] +pub fn read_to_string<R: Read>(mut reader: R) -> Result<String> { + let mut buf = String::new(); + reader.read_to_string(&mut buf)?; + Ok(buf) +} + +/// A buffer type used with `Read::read_vectored`. +/// +/// It is semantically a wrapper around an `&mut [u8]`, but is guaranteed to be +/// ABI compatible with the `iovec` type on Unix platforms and `WSABUF` on +/// Windows. +#[stable(feature = "iovec", since = "1.36.0")] +#[repr(transparent)] +pub struct IoSliceMut<'a>(sys::io::IoSliceMut<'a>); + +#[stable(feature = "iovec-send-sync", since = "1.44.0")] +unsafe impl<'a> Send for IoSliceMut<'a> {} + +#[stable(feature = "iovec-send-sync", since = "1.44.0")] +unsafe impl<'a> Sync for IoSliceMut<'a> {} + +#[stable(feature = "iovec", since = "1.36.0")] +impl<'a> fmt::Debug for IoSliceMut<'a> { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt::Debug::fmt(self.0.as_slice(), fmt) + } +} + +impl<'a> IoSliceMut<'a> { + /// Creates a new `IoSliceMut` wrapping a byte slice. + /// + /// # Panics + /// + /// Panics on Windows if the slice is larger than 4GB. + #[stable(feature = "iovec", since = "1.36.0")] + #[inline] + pub fn new(buf: &'a mut [u8]) -> IoSliceMut<'a> { + IoSliceMut(sys::io::IoSliceMut::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSliceMut::advance_slices`] to advance the cursors of + /// multiple buffers. + /// + /// # Panics + /// + /// Panics when trying to advance beyond the end of the slice. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSliceMut; + /// use std::ops::Deref; + /// + /// let mut data = [1; 8]; + /// let mut buf = IoSliceMut::new(&mut data); + /// + /// // Mark 3 bytes as read. + /// buf.advance(3); + /// assert_eq!(buf.deref(), [1; 5].as_ref()); + /// ``` + #[unstable(feature = "io_slice_advance", issue = "62726")] + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance a slice of slices. + /// + /// Shrinks the slice to remove any `IoSliceMut`s that are fully advanced over. + /// If the cursor ends up in the middle of an `IoSliceMut`, it is modified + /// to start at that cursor. + /// + /// For example, if we have a slice of two 8-byte `IoSliceMut`s, and we advance by 10 bytes, + /// the result will only include the second `IoSliceMut`, advanced by 2 bytes. + /// + /// # Panics + /// + /// Panics when trying to advance beyond the end of the slices. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSliceMut; + /// use std::ops::Deref; + /// + /// let mut buf1 = [1; 8]; + /// let mut buf2 = [2; 16]; + /// let mut buf3 = [3; 8]; + /// let mut bufs = &mut [ + /// IoSliceMut::new(&mut buf1), + /// IoSliceMut::new(&mut buf2), + /// IoSliceMut::new(&mut buf3), + /// ][..]; + /// + /// // Mark 10 bytes as read. + /// IoSliceMut::advance_slices(&mut bufs, 10); + /// assert_eq!(bufs[0].deref(), [2; 14].as_ref()); + /// assert_eq!(bufs[1].deref(), [3; 8].as_ref()); + /// ``` + #[unstable(feature = "io_slice_advance", issue = "62726")] + #[inline] + pub fn advance_slices(bufs: &mut &mut [IoSliceMut<'a>], n: usize) { + // Number of buffers to remove. + let mut remove = 0; + // Total length of all the to be removed buffers. + let mut accumulated_len = 0; + for buf in bufs.iter() { + if accumulated_len + buf.len() > n { + break; + } else { + accumulated_len += buf.len(); + remove += 1; + } + } + + *bufs = &mut replace(bufs, &mut [])[remove..]; + if bufs.is_empty() { + assert!(n == accumulated_len, "advancing io slices beyond their length"); + } else { + bufs[0].advance(n - accumulated_len) + } + } +} + +#[stable(feature = "iovec", since = "1.36.0")] +impl<'a> Deref for IoSliceMut<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +#[stable(feature = "iovec", since = "1.36.0")] +impl<'a> DerefMut for IoSliceMut<'a> { + #[inline] + fn deref_mut(&mut self) -> &mut [u8] { + self.0.as_mut_slice() + } +} + +/// A buffer type used with `Write::write_vectored`. +/// +/// It is semantically a wrapper around a `&[u8]`, but is guaranteed to be +/// ABI compatible with the `iovec` type on Unix platforms and `WSABUF` on +/// Windows. +#[stable(feature = "iovec", since = "1.36.0")] +#[derive(Copy, Clone)] +#[repr(transparent)] +pub struct IoSlice<'a>(sys::io::IoSlice<'a>); + +#[stable(feature = "iovec-send-sync", since = "1.44.0")] +unsafe impl<'a> Send for IoSlice<'a> {} + +#[stable(feature = "iovec-send-sync", since = "1.44.0")] +unsafe impl<'a> Sync for IoSlice<'a> {} + +#[stable(feature = "iovec", since = "1.36.0")] +impl<'a> fmt::Debug for IoSlice<'a> { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt::Debug::fmt(self.0.as_slice(), fmt) + } +} + +impl<'a> IoSlice<'a> { + /// Creates a new `IoSlice` wrapping a byte slice. + /// + /// # Panics + /// + /// Panics on Windows if the slice is larger than 4GB. + #[stable(feature = "iovec", since = "1.36.0")] + #[must_use] + #[inline] + pub fn new(buf: &'a [u8]) -> IoSlice<'a> { + IoSlice(sys::io::IoSlice::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSlice::advance_slices`] to advance the cursors of multiple + /// buffers. + /// + /// # Panics + /// + /// Panics when trying to advance beyond the end of the slice. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSlice; + /// use std::ops::Deref; + /// + /// let data = [1; 8]; + /// let mut buf = IoSlice::new(&data); + /// + /// // Mark 3 bytes as read. + /// buf.advance(3); + /// assert_eq!(buf.deref(), [1; 5].as_ref()); + /// ``` + #[unstable(feature = "io_slice_advance", issue = "62726")] + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance a slice of slices. + /// + /// Shrinks the slice to remove any `IoSlice`s that are fully advanced over. + /// If the cursor ends up in the middle of an `IoSlice`, it is modified + /// to start at that cursor. + /// + /// For example, if we have a slice of two 8-byte `IoSlice`s, and we advance by 10 bytes, + /// the result will only include the second `IoSlice`, advanced by 2 bytes. + /// + /// # Panics + /// + /// Panics when trying to advance beyond the end of the slices. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSlice; + /// use std::ops::Deref; + /// + /// let buf1 = [1; 8]; + /// let buf2 = [2; 16]; + /// let buf3 = [3; 8]; + /// let mut bufs = &mut [ + /// IoSlice::new(&buf1), + /// IoSlice::new(&buf2), + /// IoSlice::new(&buf3), + /// ][..]; + /// + /// // Mark 10 bytes as written. + /// IoSlice::advance_slices(&mut bufs, 10); + /// assert_eq!(bufs[0].deref(), [2; 14].as_ref()); + /// assert_eq!(bufs[1].deref(), [3; 8].as_ref()); + #[unstable(feature = "io_slice_advance", issue = "62726")] + #[inline] + pub fn advance_slices(bufs: &mut &mut [IoSlice<'a>], n: usize) { + // Number of buffers to remove. + let mut remove = 0; + // Total length of all the to be removed buffers. + let mut accumulated_len = 0; + for buf in bufs.iter() { + if accumulated_len + buf.len() > n { + break; + } else { + accumulated_len += buf.len(); + remove += 1; + } + } + + *bufs = &mut replace(bufs, &mut [])[remove..]; + if bufs.is_empty() { + assert!(n == accumulated_len, "advancing io slices beyond their length"); + } else { + bufs[0].advance(n - accumulated_len) + } + } +} + +#[stable(feature = "iovec", since = "1.36.0")] +impl<'a> Deref for IoSlice<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +/// A trait for objects which are byte-oriented sinks. +/// +/// Implementors of the `Write` trait are sometimes called 'writers'. +/// +/// Writers are defined by two required methods, [`write`] and [`flush`]: +/// +/// * The [`write`] method will attempt to write some data into the object, +/// returning how many bytes were successfully written. +/// +/// * The [`flush`] method is useful for adapters and explicit buffers +/// themselves for ensuring that all buffered data has been pushed out to the +/// 'true sink'. +/// +/// Writers are intended to be composable with one another. Many implementors +/// throughout [`std::io`] take and provide types which implement the `Write` +/// trait. +/// +/// [`write`]: Write::write +/// [`flush`]: Write::flush +/// [`std::io`]: self +/// +/// # Examples +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::fs::File; +/// +/// fn main() -> std::io::Result<()> { +/// let data = b"some bytes"; +/// +/// let mut pos = 0; +/// let mut buffer = File::create("foo.txt")?; +/// +/// while pos < data.len() { +/// let bytes_written = buffer.write(&data[pos..])?; +/// pos += bytes_written; +/// } +/// Ok(()) +/// } +/// ``` +/// +/// The trait also provides convenience methods like [`write_all`], which calls +/// `write` in a loop until its entire input has been written. +/// +/// [`write_all`]: Write::write_all +#[stable(feature = "rust1", since = "1.0.0")] +#[doc(notable_trait)] +#[cfg_attr(not(test), rustc_diagnostic_item = "IoWrite")] +pub trait Write { + /// Write a buffer into this writer, returning how many bytes were written. + /// + /// This function will attempt to write the entire contents of `buf`, but + /// the entire write might not succeed, or the write may also generate an + /// error. A call to `write` represents *at most one* attempt to write to + /// any wrapped object. + /// + /// Calls to `write` are not guaranteed to block waiting for data to be + /// written, and a write which would otherwise block can be indicated through + /// an [`Err`] variant. + /// + /// If the return value is [`Ok(n)`] then it must be guaranteed that + /// `n <= buf.len()`. A return value of `0` typically means that the + /// underlying object is no longer able to accept bytes and will likely not + /// be able to in the future as well, or that the buffer provided is empty. + /// + /// # Errors + /// + /// Each call to `write` may generate an I/O error indicating that the + /// operation could not be completed. If an error is returned then no bytes + /// in the buffer were written to this writer. + /// + /// It is **not** considered an error if the entire buffer could not be + /// written to this writer. + /// + /// An error of the [`ErrorKind::Interrupted`] kind is non-fatal and the + /// write operation should be retried if there is nothing else to do. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut buffer = File::create("foo.txt")?; + /// + /// // Writes some prefix of the byte string, not necessarily all of it. + /// buffer.write(b"some bytes")?; + /// Ok(()) + /// } + /// ``` + /// + /// [`Ok(n)`]: Ok + #[stable(feature = "rust1", since = "1.0.0")] + fn write(&mut self, buf: &[u8]) -> Result<usize>; + + /// Like [`write`], except that it writes from a slice of buffers. + /// + /// Data is copied from each buffer in order, with the final buffer + /// read from possibly being only partially consumed. This method must + /// behave as a call to [`write`] with the buffers concatenated would. + /// + /// The default implementation calls [`write`] with either the first nonempty + /// buffer provided, or an empty one if none exists. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::IoSlice; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let data1 = [1; 8]; + /// let data2 = [15; 8]; + /// let io_slice1 = IoSlice::new(&data1); + /// let io_slice2 = IoSlice::new(&data2); + /// + /// let mut buffer = File::create("foo.txt")?; + /// + /// // Writes some prefix of the byte string, not necessarily all of it. + /// buffer.write_vectored(&[io_slice1, io_slice2])?; + /// Ok(()) + /// } + /// ``` + /// + /// [`write`]: Write::write + #[stable(feature = "iovec", since = "1.36.0")] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> Result<usize> { + default_write_vectored(|b| self.write(b), bufs) + } + + /// Determines if this `Write`r has an efficient [`write_vectored`] + /// implementation. + /// + /// If a `Write`r does not override the default [`write_vectored`] + /// implementation, code using it may want to avoid the method all together + /// and coalesce writes into a single buffer for higher performance. + /// + /// The default implementation returns `false`. + /// + /// [`write_vectored`]: Write::write_vectored + #[unstable(feature = "can_vector", issue = "69941")] + fn is_write_vectored(&self) -> bool { + false + } + + /// Flush this output stream, ensuring that all intermediately buffered + /// contents reach their destination. + /// + /// # Errors + /// + /// It is considered an error if not all bytes could be written due to + /// I/O errors or EOF being reached. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::prelude::*; + /// use std::io::BufWriter; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut buffer = BufWriter::new(File::create("foo.txt")?); + /// + /// buffer.write_all(b"some bytes")?; + /// buffer.flush()?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn flush(&mut self) -> Result<()>; + + /// Attempts to write an entire buffer into this writer. + /// + /// This method will continuously call [`write`] until there is no more data + /// to be written or an error of non-[`ErrorKind::Interrupted`] kind is + /// returned. This method will not return until the entire buffer has been + /// successfully written or such an error occurs. The first error that is + /// not of [`ErrorKind::Interrupted`] kind generated from this method will be + /// returned. + /// + /// If the buffer contains no data, this will never call [`write`]. + /// + /// # Errors + /// + /// This function will return the first error of + /// non-[`ErrorKind::Interrupted`] kind that [`write`] returns. + /// + /// [`write`]: Write::write + /// + /// # Examples + /// + /// ```no_run + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut buffer = File::create("foo.txt")?; + /// + /// buffer.write_all(b"some bytes")?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn write_all(&mut self, mut buf: &[u8]) -> Result<()> { + while !buf.is_empty() { + match self.write(buf) { + Ok(0) => { + return Err(error::const_io_error!( + ErrorKind::WriteZero, + "failed to write whole buffer", + )); + } + Ok(n) => buf = &buf[n..], + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } + Ok(()) + } + + /// Attempts to write multiple buffers into this writer. + /// + /// This method will continuously call [`write_vectored`] until there is no + /// more data to be written or an error of non-[`ErrorKind::Interrupted`] + /// kind is returned. This method will not return until all buffers have + /// been successfully written or such an error occurs. The first error that + /// is not of [`ErrorKind::Interrupted`] kind generated from this method + /// will be returned. + /// + /// If the buffer contains no data, this will never call [`write_vectored`]. + /// + /// # Notes + /// + /// Unlike [`write_vectored`], this takes a *mutable* reference to + /// a slice of [`IoSlice`]s, not an immutable one. That's because we need to + /// modify the slice to keep track of the bytes already written. + /// + /// Once this function returns, the contents of `bufs` are unspecified, as + /// this depends on how many calls to [`write_vectored`] were necessary. It is + /// best to understand this function as taking ownership of `bufs` and to + /// not use `bufs` afterwards. The underlying buffers, to which the + /// [`IoSlice`]s point (but not the [`IoSlice`]s themselves), are unchanged and + /// can be reused. + /// + /// [`write_vectored`]: Write::write_vectored + /// + /// # Examples + /// + /// ``` + /// #![feature(write_all_vectored)] + /// # fn main() -> std::io::Result<()> { + /// + /// use std::io::{Write, IoSlice}; + /// + /// let mut writer = Vec::new(); + /// let bufs = &mut [ + /// IoSlice::new(&[1]), + /// IoSlice::new(&[2, 3]), + /// IoSlice::new(&[4, 5, 6]), + /// ]; + /// + /// writer.write_all_vectored(bufs)?; + /// // Note: the contents of `bufs` is now undefined, see the Notes section. + /// + /// assert_eq!(writer, &[1, 2, 3, 4, 5, 6]); + /// # Ok(()) } + /// ``` + #[unstable(feature = "write_all_vectored", issue = "70436")] + fn write_all_vectored(&mut self, mut bufs: &mut [IoSlice<'_>]) -> Result<()> { + // Guarantee that bufs is empty if it contains no data, + // to avoid calling write_vectored if there is no data to be written. + IoSlice::advance_slices(&mut bufs, 0); + while !bufs.is_empty() { + match self.write_vectored(bufs) { + Ok(0) => { + return Err(error::const_io_error!( + ErrorKind::WriteZero, + "failed to write whole buffer", + )); + } + Ok(n) => IoSlice::advance_slices(&mut bufs, n), + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } + Ok(()) + } + + /// Writes a formatted string into this writer, returning any error + /// encountered. + /// + /// This method is primarily used to interface with the + /// [`format_args!()`] macro, and it is rare that this should + /// explicitly be called. The [`write!()`] macro should be favored to + /// invoke this method instead. + /// + /// This function internally uses the [`write_all`] method on + /// this trait and hence will continuously write data so long as no errors + /// are received. This also means that partial writes are not indicated in + /// this signature. + /// + /// [`write_all`]: Write::write_all + /// + /// # Errors + /// + /// This function will return any I/O error reported while formatting. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut buffer = File::create("foo.txt")?; + /// + /// // this call + /// write!(buffer, "{:.*}", 2, 1.234567)?; + /// // turns into this: + /// buffer.write_fmt(format_args!("{:.*}", 2, 1.234567))?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> Result<()> { + // Create a shim which translates a Write to a fmt::Write and saves + // off I/O errors. instead of discarding them + struct Adapter<'a, T: ?Sized + 'a> { + inner: &'a mut T, + error: Result<()>, + } + + impl<T: Write + ?Sized> fmt::Write for Adapter<'_, T> { + fn write_str(&mut self, s: &str) -> fmt::Result { + match self.inner.write_all(s.as_bytes()) { + Ok(()) => Ok(()), + Err(e) => { + self.error = Err(e); + Err(fmt::Error) + } + } + } + } + + let mut output = Adapter { inner: self, error: Ok(()) }; + match fmt::write(&mut output, fmt) { + Ok(()) => Ok(()), + Err(..) => { + // check if the error came from the underlying `Write` or not + if output.error.is_err() { + output.error + } else { + Err(error::const_io_error!(ErrorKind::Uncategorized, "formatter error")) + } + } + } + } + + /// Creates a "by reference" adapter for this instance of `Write`. + /// + /// The returned adapter also implements `Write` and will simply borrow this + /// current writer. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::Write; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut buffer = File::create("foo.txt")?; + /// + /// let reference = buffer.by_ref(); + /// + /// // we can use reference just like our original buffer + /// reference.write_all(b"some bytes")?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn by_ref(&mut self) -> &mut Self + where + Self: Sized, + { + self + } +} + +/// The `Seek` trait provides a cursor which can be moved within a stream of +/// bytes. +/// +/// The stream typically has a fixed size, allowing seeking relative to either +/// end or the current offset. +/// +/// # Examples +/// +/// [`File`]s implement `Seek`: +/// +/// [`File`]: crate::fs::File +/// +/// ```no_run +/// use std::io; +/// use std::io::prelude::*; +/// use std::fs::File; +/// use std::io::SeekFrom; +/// +/// fn main() -> io::Result<()> { +/// let mut f = File::open("foo.txt")?; +/// +/// // move the cursor 42 bytes from the start of the file +/// f.seek(SeekFrom::Start(42))?; +/// Ok(()) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub trait Seek { + /// Seek to an offset, in bytes, in a stream. + /// + /// A seek beyond the end of a stream is allowed, but behavior is defined + /// by the implementation. + /// + /// If the seek operation completed successfully, + /// this method returns the new position from the start of the stream. + /// That position can be used later with [`SeekFrom::Start`]. + /// + /// # Errors + /// + /// Seeking can fail, for example because it might involve flushing a buffer. + /// + /// Seeking to a negative offset is considered an error. + #[stable(feature = "rust1", since = "1.0.0")] + fn seek(&mut self, pos: SeekFrom) -> Result<u64>; + + /// Rewind to the beginning of a stream. + /// + /// This is a convenience method, equivalent to `seek(SeekFrom::Start(0))`. + /// + /// # Errors + /// + /// Rewinding can fail, for example because it might involve flushing a buffer. + /// + /// # Example + /// + /// ```no_run + /// use std::io::{Read, Seek, Write}; + /// use std::fs::OpenOptions; + /// + /// let mut f = OpenOptions::new() + /// .write(true) + /// .read(true) + /// .create(true) + /// .open("foo.txt").unwrap(); + /// + /// let hello = "Hello!\n"; + /// write!(f, "{hello}").unwrap(); + /// f.rewind().unwrap(); + /// + /// let mut buf = String::new(); + /// f.read_to_string(&mut buf).unwrap(); + /// assert_eq!(&buf, hello); + /// ``` + #[stable(feature = "seek_rewind", since = "1.55.0")] + fn rewind(&mut self) -> Result<()> { + self.seek(SeekFrom::Start(0))?; + Ok(()) + } + + /// Returns the length of this stream (in bytes). + /// + /// This method is implemented using up to three seek operations. If this + /// method returns successfully, the seek position is unchanged (i.e. the + /// position before calling this method is the same as afterwards). + /// However, if this method returns an error, the seek position is + /// unspecified. + /// + /// If you need to obtain the length of *many* streams and you don't care + /// about the seek position afterwards, you can reduce the number of seek + /// operations by simply calling `seek(SeekFrom::End(0))` and using its + /// return value (it is also the stream length). + /// + /// Note that length of a stream can change over time (for example, when + /// data is appended to a file). So calling this method multiple times does + /// not necessarily return the same length each time. + /// + /// # Example + /// + /// ```no_run + /// #![feature(seek_stream_len)] + /// use std::{ + /// io::{self, Seek}, + /// fs::File, + /// }; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// + /// let len = f.stream_len()?; + /// println!("The file is currently {len} bytes long"); + /// Ok(()) + /// } + /// ``` + #[unstable(feature = "seek_stream_len", issue = "59359")] + fn stream_len(&mut self) -> Result<u64> { + let old_pos = self.stream_position()?; + let len = self.seek(SeekFrom::End(0))?; + + // Avoid seeking a third time when we were already at the end of the + // stream. The branch is usually way cheaper than a seek operation. + if old_pos != len { + self.seek(SeekFrom::Start(old_pos))?; + } + + Ok(len) + } + + /// Returns the current seek position from the start of the stream. + /// + /// This is equivalent to `self.seek(SeekFrom::Current(0))`. + /// + /// # 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(()) + /// } + /// ``` + #[stable(feature = "seek_convenience", since = "1.51.0")] + fn stream_position(&mut self) -> Result<u64> { + self.seek(SeekFrom::Current(0)) + } +} + +/// Enumeration of possible methods to seek within an I/O object. +/// +/// It is used by the [`Seek`] trait. +#[derive(Copy, PartialEq, Eq, Clone, Debug)] +#[stable(feature = "rust1", since = "1.0.0")] +pub enum SeekFrom { + /// Sets the offset to the provided number of bytes. + #[stable(feature = "rust1", since = "1.0.0")] + Start(#[stable(feature = "rust1", since = "1.0.0")] u64), + + /// Sets the offset to the size of this object plus the specified number of + /// bytes. + /// + /// It is possible to seek beyond the end of an object, but it's an error to + /// seek before byte 0. + #[stable(feature = "rust1", since = "1.0.0")] + End(#[stable(feature = "rust1", since = "1.0.0")] i64), + + /// Sets the offset to the current position plus the specified number of + /// bytes. + /// + /// It is possible to seek beyond the end of an object, but it's an error to + /// seek before byte 0. + #[stable(feature = "rust1", since = "1.0.0")] + Current(#[stable(feature = "rust1", since = "1.0.0")] i64), +} + +fn read_until<R: BufRead + ?Sized>(r: &mut R, delim: u8, buf: &mut Vec<u8>) -> Result<usize> { + let mut read = 0; + loop { + let (done, used) = { + let available = match r.fill_buf() { + Ok(n) => n, + Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + }; + match memchr::memchr(delim, available) { + Some(i) => { + buf.extend_from_slice(&available[..=i]); + (true, i + 1) + } + None => { + buf.extend_from_slice(available); + (false, available.len()) + } + } + }; + r.consume(used); + read += used; + if done || used == 0 { + return Ok(read); + } + } +} + +/// A `BufRead` is a type of `Read`er which has an internal buffer, allowing it +/// to perform extra ways of reading. +/// +/// For example, reading line-by-line is inefficient without using a buffer, so +/// if you want to read by line, you'll need `BufRead`, which includes a +/// [`read_line`] method as well as a [`lines`] iterator. +/// +/// # Examples +/// +/// A locked standard input implements `BufRead`: +/// +/// ```no_run +/// use std::io; +/// use std::io::prelude::*; +/// +/// let stdin = io::stdin(); +/// for line in stdin.lock().lines() { +/// println!("{}", line.unwrap()); +/// } +/// ``` +/// +/// If you have something that implements [`Read`], you can use the [`BufReader` +/// type][`BufReader`] to turn it into a `BufRead`. +/// +/// For example, [`File`] implements [`Read`], but not `BufRead`. +/// [`BufReader`] to the rescue! +/// +/// [`File`]: crate::fs::File +/// [`read_line`]: BufRead::read_line +/// [`lines`]: BufRead::lines +/// +/// ```no_run +/// use std::io::{self, BufReader}; +/// use std::io::prelude::*; +/// use std::fs::File; +/// +/// fn main() -> io::Result<()> { +/// let f = File::open("foo.txt")?; +/// let f = BufReader::new(f); +/// +/// for line in f.lines() { +/// println!("{}", line.unwrap()); +/// } +/// +/// Ok(()) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub trait BufRead: Read { + /// Returns the contents of the internal buffer, filling it with more data + /// from the inner reader if it is empty. + /// + /// This function is a lower-level call. It needs to be paired with the + /// [`consume`] method to function properly. When calling this + /// method, none of the contents will be "read" in the sense that later + /// calling `read` may return the same contents. As such, [`consume`] must + /// be called with the number of bytes that are consumed from this buffer to + /// ensure that the bytes are never returned twice. + /// + /// [`consume`]: BufRead::consume + /// + /// An empty buffer returned indicates that the stream has reached EOF. + /// + /// # Errors + /// + /// This function will return an I/O error if the underlying reader was + /// read, but returned an error. + /// + /// # Examples + /// + /// A locked standard input implements `BufRead`: + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// + /// let stdin = io::stdin(); + /// let mut stdin = stdin.lock(); + /// + /// let buffer = stdin.fill_buf().unwrap(); + /// + /// // work with buffer + /// println!("{buffer:?}"); + /// + /// // ensure the bytes we worked with aren't returned again later + /// let length = buffer.len(); + /// stdin.consume(length); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn fill_buf(&mut self) -> Result<&[u8]>; + + /// Tells this buffer that `amt` bytes have been consumed from the buffer, + /// so they should no longer be returned in calls to `read`. + /// + /// This function is a lower-level call. It needs to be paired with the + /// [`fill_buf`] method to function properly. This function does + /// not perform any I/O, it simply informs this object that some amount of + /// its buffer, returned from [`fill_buf`], has been consumed and should + /// no longer be returned. As such, this function may do odd things if + /// [`fill_buf`] isn't called before calling it. + /// + /// The `amt` must be `<=` the number of bytes in the buffer returned by + /// [`fill_buf`]. + /// + /// # Examples + /// + /// Since `consume()` is meant to be used with [`fill_buf`], + /// that method's example includes an example of `consume()`. + /// + /// [`fill_buf`]: BufRead::fill_buf + #[stable(feature = "rust1", since = "1.0.0")] + fn consume(&mut self, amt: usize); + + /// Check if the underlying `Read` has any data left to be read. + /// + /// This function may fill the buffer to check for data, + /// so this functions returns `Result<bool>`, not `bool`. + /// + /// Default implementation calls `fill_buf` and checks that + /// returned slice is empty (which means that there is no data left, + /// since EOF is reached). + /// + /// Examples + /// + /// ``` + /// #![feature(buf_read_has_data_left)] + /// use std::io; + /// use std::io::prelude::*; + /// + /// let stdin = io::stdin(); + /// let mut stdin = stdin.lock(); + /// + /// while stdin.has_data_left().unwrap() { + /// let mut line = String::new(); + /// stdin.read_line(&mut line).unwrap(); + /// // work with line + /// println!("{line:?}"); + /// } + /// ``` + #[unstable(feature = "buf_read_has_data_left", reason = "recently added", issue = "86423")] + fn has_data_left(&mut self) -> Result<bool> { + self.fill_buf().map(|b| !b.is_empty()) + } + + /// Read all bytes into `buf` until the delimiter `byte` or EOF is reached. + /// + /// This function will read bytes from the underlying stream until the + /// delimiter or EOF is found. Once found, all bytes up to, and including, + /// the delimiter (if found) will be appended to `buf`. + /// + /// If successful, this function will return the total number of bytes read. + /// + /// This function is blocking and should be used carefully: it is possible for + /// an attacker to continuously send bytes without ever sending the delimiter + /// or EOF. + /// + /// # Errors + /// + /// This function will ignore all instances of [`ErrorKind::Interrupted`] and + /// will otherwise return any errors returned by [`fill_buf`]. + /// + /// If an I/O error is encountered then all bytes read so far will be + /// present in `buf` and its length will have been adjusted appropriately. + /// + /// [`fill_buf`]: BufRead::fill_buf + /// + /// # Examples + /// + /// [`std::io::Cursor`][`Cursor`] is a type that implements `BufRead`. In + /// this example, we use [`Cursor`] to read all the bytes in a byte slice + /// in hyphen delimited segments: + /// + /// ``` + /// use std::io::{self, BufRead}; + /// + /// let mut cursor = io::Cursor::new(b"lorem-ipsum"); + /// let mut buf = vec![]; + /// + /// // cursor is at 'l' + /// let num_bytes = cursor.read_until(b'-', &mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 6); + /// assert_eq!(buf, b"lorem-"); + /// buf.clear(); + /// + /// // cursor is at 'i' + /// let num_bytes = cursor.read_until(b'-', &mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 5); + /// assert_eq!(buf, b"ipsum"); + /// buf.clear(); + /// + /// // cursor is at EOF + /// let num_bytes = cursor.read_until(b'-', &mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 0); + /// assert_eq!(buf, b""); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn read_until(&mut self, byte: u8, buf: &mut Vec<u8>) -> Result<usize> { + read_until(self, byte, buf) + } + + /// Read all bytes until a newline (the `0xA` byte) is reached, and append + /// them to the provided buffer. You do not need to clear the buffer before + /// appending. + /// + /// This function will read bytes from the underlying stream until the + /// newline delimiter (the `0xA` byte) or EOF is found. Once found, all bytes + /// up to, and including, the delimiter (if found) will be appended to + /// `buf`. + /// + /// If successful, this function will return the total number of bytes read. + /// + /// If this function returns [`Ok(0)`], the stream has reached EOF. + /// + /// This function is blocking and should be used carefully: it is possible for + /// an attacker to continuously send bytes without ever sending a newline + /// or EOF. + /// + /// [`Ok(0)`]: Ok + /// + /// # Errors + /// + /// This function has the same error semantics as [`read_until`] and will + /// also return an error if the read bytes are not valid UTF-8. If an I/O + /// error is encountered then `buf` may contain some bytes already read in + /// the event that all data read so far was valid UTF-8. + /// + /// [`read_until`]: BufRead::read_until + /// + /// # Examples + /// + /// [`std::io::Cursor`][`Cursor`] is a type that implements `BufRead`. In + /// this example, we use [`Cursor`] to read all the lines in a byte slice: + /// + /// ``` + /// use std::io::{self, BufRead}; + /// + /// let mut cursor = io::Cursor::new(b"foo\nbar"); + /// let mut buf = String::new(); + /// + /// // cursor is at 'f' + /// let num_bytes = cursor.read_line(&mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 4); + /// assert_eq!(buf, "foo\n"); + /// buf.clear(); + /// + /// // cursor is at 'b' + /// let num_bytes = cursor.read_line(&mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 3); + /// assert_eq!(buf, "bar"); + /// buf.clear(); + /// + /// // cursor is at EOF + /// let num_bytes = cursor.read_line(&mut buf) + /// .expect("reading from cursor won't fail"); + /// assert_eq!(num_bytes, 0); + /// assert_eq!(buf, ""); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn read_line(&mut self, buf: &mut String) -> Result<usize> { + // Note that we are not calling the `.read_until` method here, but + // rather our hardcoded implementation. For more details as to why, see + // the comments in `read_to_end`. + unsafe { append_to_string(buf, |b| read_until(self, b'\n', b)) } + } + + /// Returns an iterator over the contents of this reader split on the byte + /// `byte`. + /// + /// The iterator returned from this function will return instances of + /// <code>[io::Result]<[Vec]\<u8>></code>. Each vector returned will *not* have + /// the delimiter byte at the end. + /// + /// This function will yield errors whenever [`read_until`] would have + /// also yielded an error. + /// + /// [io::Result]: self::Result "io::Result" + /// [`read_until`]: BufRead::read_until + /// + /// # Examples + /// + /// [`std::io::Cursor`][`Cursor`] is a type that implements `BufRead`. In + /// this example, we use [`Cursor`] to iterate over all hyphen delimited + /// segments in a byte slice + /// + /// ``` + /// use std::io::{self, BufRead}; + /// + /// let cursor = io::Cursor::new(b"lorem-ipsum-dolor"); + /// + /// let mut split_iter = cursor.split(b'-').map(|l| l.unwrap()); + /// assert_eq!(split_iter.next(), Some(b"lorem".to_vec())); + /// assert_eq!(split_iter.next(), Some(b"ipsum".to_vec())); + /// assert_eq!(split_iter.next(), Some(b"dolor".to_vec())); + /// assert_eq!(split_iter.next(), None); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + fn split(self, byte: u8) -> Split<Self> + where + Self: Sized, + { + Split { buf: self, delim: byte } + } + + /// Returns an iterator over the lines of this reader. + /// + /// The iterator returned from this function will yield instances of + /// <code>[io::Result]<[String]></code>. Each string returned will *not* have a newline + /// byte (the `0xA` byte) or `CRLF` (`0xD`, `0xA` bytes) at the end. + /// + /// [io::Result]: self::Result "io::Result" + /// + /// # Examples + /// + /// [`std::io::Cursor`][`Cursor`] is a type that implements `BufRead`. In + /// this example, we use [`Cursor`] to iterate over all the lines in a byte + /// slice. + /// + /// ``` + /// use std::io::{self, BufRead}; + /// + /// let cursor = io::Cursor::new(b"lorem\nipsum\r\ndolor"); + /// + /// let mut lines_iter = cursor.lines().map(|l| l.unwrap()); + /// assert_eq!(lines_iter.next(), Some(String::from("lorem"))); + /// assert_eq!(lines_iter.next(), Some(String::from("ipsum"))); + /// assert_eq!(lines_iter.next(), Some(String::from("dolor"))); + /// assert_eq!(lines_iter.next(), None); + /// ``` + /// + /// # Errors + /// + /// Each line of the iterator has the same error semantics as [`BufRead::read_line`]. + #[stable(feature = "rust1", since = "1.0.0")] + fn lines(self) -> Lines<Self> + where + Self: Sized, + { + Lines { buf: self } + } +} + +/// Adapter to chain together two readers. +/// +/// This struct is generally created by calling [`chain`] on a reader. +/// Please see the documentation of [`chain`] for more details. +/// +/// [`chain`]: Read::chain +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug)] +pub struct Chain<T, U> { + first: T, + second: U, + done_first: bool, +} + +impl<T, U> Chain<T, U> { + /// Consumes the `Chain`, returning the wrapped readers. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut foo_file = File::open("foo.txt")?; + /// let mut bar_file = File::open("bar.txt")?; + /// + /// let chain = foo_file.chain(bar_file); + /// let (foo_file, bar_file) = chain.into_inner(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "more_io_inner_methods", since = "1.20.0")] + pub fn into_inner(self) -> (T, U) { + (self.first, self.second) + } + + /// Gets references to the underlying readers in this `Chain`. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut foo_file = File::open("foo.txt")?; + /// let mut bar_file = File::open("bar.txt")?; + /// + /// let chain = foo_file.chain(bar_file); + /// let (foo_file, bar_file) = chain.get_ref(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "more_io_inner_methods", since = "1.20.0")] + pub fn get_ref(&self) -> (&T, &U) { + (&self.first, &self.second) + } + + /// Gets mutable references to the underlying readers in this `Chain`. + /// + /// Care should be taken to avoid modifying the internal I/O state of the + /// underlying readers as doing so may corrupt the internal state of this + /// `Chain`. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut foo_file = File::open("foo.txt")?; + /// let mut bar_file = File::open("bar.txt")?; + /// + /// let mut chain = foo_file.chain(bar_file); + /// let (foo_file, bar_file) = chain.get_mut(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "more_io_inner_methods", since = "1.20.0")] + pub fn get_mut(&mut self) -> (&mut T, &mut U) { + (&mut self.first, &mut self.second) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T: Read, U: Read> Read for Chain<T, U> { + fn read(&mut self, buf: &mut [u8]) -> Result<usize> { + if !self.done_first { + match self.first.read(buf)? { + 0 if !buf.is_empty() => self.done_first = true, + n => return Ok(n), + } + } + self.second.read(buf) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> Result<usize> { + if !self.done_first { + match self.first.read_vectored(bufs)? { + 0 if bufs.iter().any(|b| !b.is_empty()) => self.done_first = true, + n => return Ok(n), + } + } + self.second.read_vectored(bufs) + } +} + +#[stable(feature = "chain_bufread", since = "1.9.0")] +impl<T: BufRead, U: BufRead> BufRead for Chain<T, U> { + fn fill_buf(&mut self) -> Result<&[u8]> { + if !self.done_first { + match self.first.fill_buf()? { + buf if buf.is_empty() => { + self.done_first = true; + } + buf => return Ok(buf), + } + } + self.second.fill_buf() + } + + fn consume(&mut self, amt: usize) { + if !self.done_first { self.first.consume(amt) } else { self.second.consume(amt) } + } +} + +impl<T, U> SizeHint for Chain<T, U> { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&self.first) + SizeHint::lower_bound(&self.second) + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + match (SizeHint::upper_bound(&self.first), SizeHint::upper_bound(&self.second)) { + (Some(first), Some(second)) => first.checked_add(second), + _ => None, + } + } +} + +/// Reader adapter which limits the bytes read from an underlying reader. +/// +/// This struct is generally created by calling [`take`] on a reader. +/// Please see the documentation of [`take`] for more details. +/// +/// [`take`]: Read::take +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug)] +pub struct Take<T> { + inner: T, + limit: u64, +} + +impl<T> Take<T> { + /// Returns the number of bytes that can be read before this instance will + /// return EOF. + /// + /// # Note + /// + /// This instance may reach `EOF` after reading fewer bytes than indicated by + /// this method if the underlying [`Read`] instance reaches EOF. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let f = File::open("foo.txt")?; + /// + /// // read at most five bytes + /// let handle = f.take(5); + /// + /// println!("limit: {}", handle.limit()); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn limit(&self) -> u64 { + self.limit + } + + /// Sets the number of bytes that can be read before this instance will + /// return EOF. This is the same as constructing a new `Take` instance, so + /// the amount of bytes read and the previous limit value don't matter when + /// calling this method. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let f = File::open("foo.txt")?; + /// + /// // read at most five bytes + /// let mut handle = f.take(5); + /// handle.set_limit(10); + /// + /// assert_eq!(handle.limit(), 10); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "take_set_limit", since = "1.27.0")] + pub fn set_limit(&mut self, limit: u64) { + self.limit = limit; + } + + /// Consumes the `Take`, returning the wrapped reader. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut file = File::open("foo.txt")?; + /// + /// let mut buffer = [0; 5]; + /// let mut handle = file.take(5); + /// handle.read(&mut buffer)?; + /// + /// let file = handle.into_inner(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "io_take_into_inner", since = "1.15.0")] + pub fn into_inner(self) -> T { + self.inner + } + + /// Gets a reference to the underlying reader. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut file = File::open("foo.txt")?; + /// + /// let mut buffer = [0; 5]; + /// let mut handle = file.take(5); + /// handle.read(&mut buffer)?; + /// + /// let file = handle.get_ref(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "more_io_inner_methods", since = "1.20.0")] + pub fn get_ref(&self) -> &T { + &self.inner + } + + /// Gets a mutable reference to the underlying reader. + /// + /// Care should be taken to avoid modifying the internal I/O state of the + /// underlying reader as doing so may corrupt the internal limit of this + /// `Take`. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut file = File::open("foo.txt")?; + /// + /// let mut buffer = [0; 5]; + /// let mut handle = file.take(5); + /// handle.read(&mut buffer)?; + /// + /// let file = handle.get_mut(); + /// Ok(()) + /// } + /// ``` + #[stable(feature = "more_io_inner_methods", since = "1.20.0")] + pub fn get_mut(&mut self) -> &mut T { + &mut self.inner + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T: Read> Read for Take<T> { + fn read(&mut self, buf: &mut [u8]) -> Result<usize> { + // Don't call into inner reader at all at EOF because it may still block + if self.limit == 0 { + return Ok(0); + } + + let max = cmp::min(buf.len() as u64, self.limit) as usize; + let n = self.inner.read(&mut buf[..max])?; + assert!(n as u64 <= self.limit, "number of read bytes exceeds limit"); + self.limit -= n as u64; + Ok(n) + } + + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { + // Don't call into inner reader at all at EOF because it may still block + if self.limit == 0 { + return Ok(()); + } + + let prev_filled = buf.filled_len(); + + if self.limit <= buf.remaining() as u64 { + // if we just use an as cast to convert, limit may wrap around on a 32 bit target + let limit = cmp::min(self.limit, usize::MAX as u64) as usize; + + let extra_init = cmp::min(limit as usize, buf.initialized_len() - buf.filled_len()); + + // SAFETY: no uninit data is written to ibuf + let ibuf = unsafe { &mut buf.unfilled_mut()[..limit] }; + + let mut sliced_buf = ReadBuf::uninit(ibuf); + + // SAFETY: extra_init bytes of ibuf are known to be initialized + unsafe { + sliced_buf.assume_init(extra_init); + } + + self.inner.read_buf(&mut sliced_buf)?; + + let new_init = sliced_buf.initialized_len(); + let filled = sliced_buf.filled_len(); + + // sliced_buf / ibuf must drop here + + // SAFETY: new_init bytes of buf's unfilled buffer have been initialized + unsafe { + buf.assume_init(new_init); + } + + buf.add_filled(filled); + + self.limit -= filled as u64; + } else { + self.inner.read_buf(buf)?; + + //inner may unfill + self.limit -= buf.filled_len().saturating_sub(prev_filled) as u64; + } + + Ok(()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<T: BufRead> BufRead for Take<T> { + fn fill_buf(&mut self) -> Result<&[u8]> { + // Don't call into inner reader at all at EOF because it may still block + if self.limit == 0 { + return Ok(&[]); + } + + let buf = self.inner.fill_buf()?; + let cap = cmp::min(buf.len() as u64, self.limit) as usize; + Ok(&buf[..cap]) + } + + fn consume(&mut self, amt: usize) { + // Don't let callers reset the limit by passing an overlarge value + let amt = cmp::min(amt as u64, self.limit) as usize; + self.limit -= amt as u64; + self.inner.consume(amt); + } +} + +impl<T> SizeHint for Take<T> { + #[inline] + fn lower_bound(&self) -> usize { + cmp::min(SizeHint::lower_bound(&self.inner) as u64, self.limit) as usize + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + match SizeHint::upper_bound(&self.inner) { + Some(upper_bound) => Some(cmp::min(upper_bound as u64, self.limit) as usize), + None => self.limit.try_into().ok(), + } + } +} + +/// An iterator over `u8` values of a reader. +/// +/// This struct is generally created by calling [`bytes`] on a reader. +/// Please see the documentation of [`bytes`] for more details. +/// +/// [`bytes`]: Read::bytes +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug)] +pub struct Bytes<R> { + inner: R, +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<R: Read> Iterator for Bytes<R> { + type Item = Result<u8>; + + fn next(&mut self) -> Option<Result<u8>> { + let mut byte = 0; + loop { + return match self.inner.read(slice::from_mut(&mut byte)) { + Ok(0) => None, + Ok(..) => Some(Ok(byte)), + Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => Some(Err(e)), + }; + } + } + + fn size_hint(&self) -> (usize, Option<usize>) { + SizeHint::size_hint(&self.inner) + } +} + +trait SizeHint { + fn lower_bound(&self) -> usize; + + fn upper_bound(&self) -> Option<usize>; + + fn size_hint(&self) -> (usize, Option<usize>) { + (self.lower_bound(), self.upper_bound()) + } +} + +impl<T> SizeHint for T { + #[inline] + default fn lower_bound(&self) -> usize { + 0 + } + + #[inline] + default fn upper_bound(&self) -> Option<usize> { + None + } +} + +impl<T> SizeHint for &mut T { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(*self) + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + SizeHint::upper_bound(*self) + } +} + +impl<T> SizeHint for Box<T> { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&**self) + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + SizeHint::upper_bound(&**self) + } +} + +impl SizeHint for &[u8] { + #[inline] + fn lower_bound(&self) -> usize { + self.len() + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + Some(self.len()) + } +} + +/// An iterator over the contents of an instance of `BufRead` split on a +/// particular byte. +/// +/// This struct is generally created by calling [`split`] on a `BufRead`. +/// Please see the documentation of [`split`] for more details. +/// +/// [`split`]: BufRead::split +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug)] +pub struct Split<B> { + buf: B, + delim: u8, +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<B: BufRead> Iterator for Split<B> { + type Item = Result<Vec<u8>>; + + fn next(&mut self) -> Option<Result<Vec<u8>>> { + let mut buf = Vec::new(); + match self.buf.read_until(self.delim, &mut buf) { + Ok(0) => None, + Ok(_n) => { + if buf[buf.len() - 1] == self.delim { + buf.pop(); + } + Some(Ok(buf)) + } + Err(e) => Some(Err(e)), + } + } +} + +/// An iterator over the lines of an instance of `BufRead`. +/// +/// This struct is generally created by calling [`lines`] on a `BufRead`. +/// Please see the documentation of [`lines`] for more details. +/// +/// [`lines`]: BufRead::lines +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Debug)] +pub struct Lines<B> { + buf: B, +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<B: BufRead> Iterator for Lines<B> { + type Item = Result<String>; + + fn next(&mut self) -> Option<Result<String>> { + let mut buf = String::new(); + match self.buf.read_line(&mut buf) { + Ok(0) => None, + Ok(_n) => { + if buf.ends_with('\n') { + buf.pop(); + if buf.ends_with('\r') { + buf.pop(); + } + } + Some(Ok(buf)) + } + Err(e) => Some(Err(e)), + } + } +} diff --git a/library/std/src/io/prelude.rs b/library/std/src/io/prelude.rs new file mode 100644 index 000000000..d80643101 --- /dev/null +++ b/library/std/src/io/prelude.rs @@ -0,0 +1,14 @@ +//! The I/O Prelude. +//! +//! The purpose of this module is to alleviate imports of many common I/O traits +//! by adding a glob import to the top of I/O heavy modules: +//! +//! ``` +//! # #![allow(unused_imports)] +//! use std::io::prelude::*; +//! ``` + +#![stable(feature = "rust1", since = "1.0.0")] + +#[stable(feature = "rust1", since = "1.0.0")] +pub use super::{BufRead, Read, Seek, Write}; diff --git a/library/std/src/io/readbuf.rs b/library/std/src/io/readbuf.rs new file mode 100644 index 000000000..78d1113f8 --- /dev/null +++ b/library/std/src/io/readbuf.rs @@ -0,0 +1,249 @@ +#![unstable(feature = "read_buf", issue = "78485")] + +#[cfg(test)] +mod tests; + +use crate::cmp; +use crate::fmt::{self, Debug, Formatter}; +use crate::mem::MaybeUninit; + +/// A wrapper around a byte buffer that is incrementally filled and initialized. +/// +/// This type is a sort of "double cursor". It tracks three regions in the buffer: a region at the beginning of the +/// buffer that has been logically filled with data, a region that has been initialized at some point but not yet +/// logically filled, and a region at the end that is fully uninitialized. The filled region is guaranteed to be a +/// subset of the initialized region. +/// +/// In summary, the contents of the buffer can be visualized as: +/// ```not_rust +/// [ capacity ] +/// [ filled | unfilled ] +/// [ initialized | uninitialized ] +/// ``` +pub struct ReadBuf<'a> { + buf: &'a mut [MaybeUninit<u8>], + filled: usize, + initialized: usize, +} + +impl Debug for ReadBuf<'_> { + fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result { + f.debug_struct("ReadBuf") + .field("init", &self.initialized()) + .field("filled", &self.filled) + .field("capacity", &self.capacity()) + .finish() + } +} + +impl<'a> ReadBuf<'a> { + /// Creates a new `ReadBuf` from a fully initialized buffer. + #[inline] + pub fn new(buf: &'a mut [u8]) -> ReadBuf<'a> { + let len = buf.len(); + + ReadBuf { + //SAFETY: initialized data never becoming uninitialized is an invariant of ReadBuf + buf: unsafe { (buf as *mut [u8]).as_uninit_slice_mut().unwrap() }, + filled: 0, + initialized: len, + } + } + + /// Creates a new `ReadBuf` from a fully uninitialized buffer. + /// + /// Use `assume_init` if part of the buffer is known to be already initialized. + #[inline] + pub fn uninit(buf: &'a mut [MaybeUninit<u8>]) -> ReadBuf<'a> { + ReadBuf { buf, filled: 0, initialized: 0 } + } + + /// Returns the total capacity of the buffer. + #[inline] + pub fn capacity(&self) -> usize { + self.buf.len() + } + + /// Returns a shared reference to the filled portion of the buffer. + #[inline] + pub fn filled(&self) -> &[u8] { + //SAFETY: We only slice the filled part of the buffer, which is always valid + unsafe { MaybeUninit::slice_assume_init_ref(&self.buf[0..self.filled]) } + } + + /// Returns a mutable reference to the filled portion of the buffer. + #[inline] + pub fn filled_mut(&mut self) -> &mut [u8] { + //SAFETY: We only slice the filled part of the buffer, which is always valid + unsafe { MaybeUninit::slice_assume_init_mut(&mut self.buf[0..self.filled]) } + } + + /// Returns a shared reference to the initialized portion of the buffer. + /// + /// This includes the filled portion. + #[inline] + pub fn initialized(&self) -> &[u8] { + //SAFETY: We only slice the initialized part of the buffer, which is always valid + unsafe { MaybeUninit::slice_assume_init_ref(&self.buf[0..self.initialized]) } + } + + /// Returns a mutable reference to the initialized portion of the buffer. + /// + /// This includes the filled portion. + #[inline] + pub fn initialized_mut(&mut self) -> &mut [u8] { + //SAFETY: We only slice the initialized part of the buffer, which is always valid + unsafe { MaybeUninit::slice_assume_init_mut(&mut self.buf[0..self.initialized]) } + } + + /// Returns a mutable reference to the unfilled part of the buffer without ensuring that it has been fully + /// initialized. + /// + /// # Safety + /// + /// The caller must not de-initialize portions of the buffer that have already been initialized. + #[inline] + pub unsafe fn unfilled_mut(&mut self) -> &mut [MaybeUninit<u8>] { + &mut self.buf[self.filled..] + } + + /// Returns a mutable reference to the uninitialized part of the buffer. + /// + /// It is safe to uninitialize any of these bytes. + #[inline] + pub fn uninitialized_mut(&mut self) -> &mut [MaybeUninit<u8>] { + &mut self.buf[self.initialized..] + } + + /// Returns a mutable reference to the unfilled part of the buffer, ensuring it is fully initialized. + /// + /// Since `ReadBuf` tracks the region of the buffer that has been initialized, this is effectively "free" after + /// the first use. + #[inline] + pub fn initialize_unfilled(&mut self) -> &mut [u8] { + // should optimize out the assertion + self.initialize_unfilled_to(self.remaining()) + } + + /// Returns a mutable reference to the first `n` bytes of the unfilled part of the buffer, ensuring it is + /// fully initialized. + /// + /// # Panics + /// + /// Panics if `self.remaining()` is less than `n`. + #[inline] + pub fn initialize_unfilled_to(&mut self, n: usize) -> &mut [u8] { + assert!(self.remaining() >= n); + + let extra_init = self.initialized - self.filled; + // If we don't have enough initialized, do zeroing + if n > extra_init { + let uninit = n - extra_init; + let unfilled = &mut self.uninitialized_mut()[0..uninit]; + + for byte in unfilled.iter_mut() { + byte.write(0); + } + + // SAFETY: we just initialized uninit bytes, and the previous bytes were already init + unsafe { + self.assume_init(n); + } + } + + let filled = self.filled; + + &mut self.initialized_mut()[filled..filled + n] + } + + /// Returns the number of bytes at the end of the slice that have not yet been filled. + #[inline] + pub fn remaining(&self) -> usize { + self.capacity() - self.filled + } + + /// Clears the buffer, resetting the filled region to empty. + /// + /// The number of initialized bytes is not changed, and the contents of the buffer are not modified. + #[inline] + pub fn clear(&mut self) -> &mut Self { + self.set_filled(0) // The assertion in `set_filled` is optimized out + } + + /// Increases the size of the filled region of the buffer. + /// + /// The number of initialized bytes is not changed. + /// + /// # Panics + /// + /// Panics if the filled region of the buffer would become larger than the initialized region. + #[inline] + pub fn add_filled(&mut self, n: usize) -> &mut Self { + self.set_filled(self.filled + n) + } + + /// Sets the size of the filled region of the buffer. + /// + /// The number of initialized bytes is not changed. + /// + /// Note that this can be used to *shrink* the filled region of the buffer in addition to growing it (for + /// example, by a `Read` implementation that compresses data in-place). + /// + /// # Panics + /// + /// Panics if the filled region of the buffer would become larger than the initialized region. + #[inline] + pub fn set_filled(&mut self, n: usize) -> &mut Self { + assert!(n <= self.initialized); + + self.filled = n; + self + } + + /// Asserts that the first `n` unfilled bytes of the buffer are initialized. + /// + /// `ReadBuf` assumes that bytes are never de-initialized, so this method does nothing when called with fewer + /// bytes than are already known to be initialized. + /// + /// # Safety + /// + /// The caller must ensure that the first `n` unfilled bytes of the buffer have already been initialized. + #[inline] + pub unsafe fn assume_init(&mut self, n: usize) -> &mut Self { + self.initialized = cmp::max(self.initialized, self.filled + n); + self + } + + /// Appends data to the buffer, advancing the written position and possibly also the initialized position. + /// + /// # Panics + /// + /// Panics if `self.remaining()` is less than `buf.len()`. + #[inline] + pub fn append(&mut self, buf: &[u8]) { + assert!(self.remaining() >= buf.len()); + + // SAFETY: we do not de-initialize any of the elements of the slice + unsafe { + MaybeUninit::write_slice(&mut self.unfilled_mut()[..buf.len()], buf); + } + + // SAFETY: We just added the entire contents of buf to the filled section. + unsafe { + self.assume_init(buf.len()); + } + self.add_filled(buf.len()); + } + + /// Returns the amount of bytes that have been filled. + #[inline] + pub fn filled_len(&self) -> usize { + self.filled + } + + /// Returns the amount of bytes that have been initialized. + #[inline] + pub fn initialized_len(&self) -> usize { + self.initialized + } +} diff --git a/library/std/src/io/readbuf/tests.rs b/library/std/src/io/readbuf/tests.rs new file mode 100644 index 000000000..3b7a5a56d --- /dev/null +++ b/library/std/src/io/readbuf/tests.rs @@ -0,0 +1,181 @@ +use super::ReadBuf; +use crate::mem::MaybeUninit; + +/// Test that ReadBuf has the correct numbers when created with new +#[test] +fn new() { + let mut buf = [0; 16]; + let rbuf = ReadBuf::new(&mut buf); + + assert_eq!(rbuf.filled_len(), 0); + assert_eq!(rbuf.initialized_len(), 16); + assert_eq!(rbuf.capacity(), 16); + assert_eq!(rbuf.remaining(), 16); +} + +/// Test that ReadBuf has the correct numbers when created with uninit +#[test] +fn uninit() { + let mut buf = [MaybeUninit::uninit(); 16]; + let rbuf = ReadBuf::uninit(&mut buf); + + assert_eq!(rbuf.filled_len(), 0); + assert_eq!(rbuf.initialized_len(), 0); + assert_eq!(rbuf.capacity(), 16); + assert_eq!(rbuf.remaining(), 16); +} + +#[test] +fn initialize_unfilled() { + let mut buf = [MaybeUninit::uninit(); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + rbuf.initialize_unfilled(); + + assert_eq!(rbuf.initialized_len(), 16); +} + +#[test] +fn initialize_unfilled_to() { + let mut buf = [MaybeUninit::uninit(); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + rbuf.initialize_unfilled_to(8); + + assert_eq!(rbuf.initialized_len(), 8); + + rbuf.initialize_unfilled_to(4); + + assert_eq!(rbuf.initialized_len(), 8); + + rbuf.set_filled(8); + + rbuf.initialize_unfilled_to(6); + + assert_eq!(rbuf.initialized_len(), 14); + + rbuf.initialize_unfilled_to(8); + + assert_eq!(rbuf.initialized_len(), 16); +} + +#[test] +fn add_filled() { + let mut buf = [0; 16]; + let mut rbuf = ReadBuf::new(&mut buf); + + rbuf.add_filled(1); + + assert_eq!(rbuf.filled_len(), 1); + assert_eq!(rbuf.remaining(), 15); +} + +#[test] +#[should_panic] +fn add_filled_panic() { + let mut buf = [MaybeUninit::uninit(); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + rbuf.add_filled(1); +} + +#[test] +fn set_filled() { + let mut buf = [0; 16]; + let mut rbuf = ReadBuf::new(&mut buf); + + rbuf.set_filled(16); + + assert_eq!(rbuf.filled_len(), 16); + assert_eq!(rbuf.remaining(), 0); + + rbuf.set_filled(6); + + assert_eq!(rbuf.filled_len(), 6); + assert_eq!(rbuf.remaining(), 10); +} + +#[test] +#[should_panic] +fn set_filled_panic() { + let mut buf = [MaybeUninit::uninit(); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + rbuf.set_filled(16); +} + +#[test] +fn clear() { + let mut buf = [255; 16]; + let mut rbuf = ReadBuf::new(&mut buf); + + rbuf.set_filled(16); + + assert_eq!(rbuf.filled_len(), 16); + assert_eq!(rbuf.remaining(), 0); + + rbuf.clear(); + + assert_eq!(rbuf.filled_len(), 0); + assert_eq!(rbuf.remaining(), 16); + + assert_eq!(rbuf.initialized(), [255; 16]); +} + +#[test] +fn assume_init() { + let mut buf = [MaybeUninit::uninit(); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + unsafe { + rbuf.assume_init(8); + } + + assert_eq!(rbuf.initialized_len(), 8); + + rbuf.add_filled(4); + + unsafe { + rbuf.assume_init(2); + } + + assert_eq!(rbuf.initialized_len(), 8); + + unsafe { + rbuf.assume_init(8); + } + + assert_eq!(rbuf.initialized_len(), 12); +} + +#[test] +fn append() { + let mut buf = [MaybeUninit::new(255); 16]; + let mut rbuf = ReadBuf::uninit(&mut buf); + + rbuf.append(&[0; 8]); + + assert_eq!(rbuf.initialized_len(), 8); + assert_eq!(rbuf.filled_len(), 8); + assert_eq!(rbuf.filled(), [0; 8]); + + rbuf.clear(); + + rbuf.append(&[1; 16]); + + assert_eq!(rbuf.initialized_len(), 16); + assert_eq!(rbuf.filled_len(), 16); + assert_eq!(rbuf.filled(), [1; 16]); +} + +#[test] +fn filled_mut() { + let mut buf = [0; 16]; + let mut rbuf = ReadBuf::new(&mut buf); + + rbuf.add_filled(8); + + let filled = rbuf.filled().to_vec(); + + assert_eq!(&*filled, &*rbuf.filled_mut()); +} diff --git a/library/std/src/io/stdio.rs b/library/std/src/io/stdio.rs new file mode 100644 index 000000000..4d3736f79 --- /dev/null +++ b/library/std/src/io/stdio.rs @@ -0,0 +1,1042 @@ +#![cfg_attr(test, allow(unused))] + +#[cfg(test)] +mod tests; + +use crate::io::prelude::*; + +use crate::cell::{Cell, RefCell}; +use crate::fmt; +use crate::io::{self, BufReader, IoSlice, IoSliceMut, LineWriter, Lines}; +use crate::pin::Pin; +use crate::sync::atomic::{AtomicBool, Ordering}; +use crate::sync::{Arc, Mutex, MutexGuard, OnceLock}; +use crate::sys::stdio; +use crate::sys_common::remutex::{ReentrantMutex, ReentrantMutexGuard}; + +type LocalStream = Arc<Mutex<Vec<u8>>>; + +thread_local! { + /// Used by the test crate to capture the output of the print macros and panics. + static OUTPUT_CAPTURE: Cell<Option<LocalStream>> = { + Cell::new(None) + } +} + +/// Flag to indicate OUTPUT_CAPTURE is used. +/// +/// If it is None and was never set on any thread, this flag is set to false, +/// and OUTPUT_CAPTURE can be safely ignored on all threads, saving some time +/// and memory registering an unused thread local. +/// +/// Note about memory ordering: This contains information about whether a +/// thread local variable might be in use. Although this is a global flag, the +/// memory ordering between threads does not matter: we only want this flag to +/// have a consistent order between set_output_capture and print_to *within +/// the same thread*. Within the same thread, things always have a perfectly +/// consistent order. So Ordering::Relaxed is fine. +static OUTPUT_CAPTURE_USED: AtomicBool = AtomicBool::new(false); + +/// A handle to a raw instance of the standard input stream of this process. +/// +/// This handle is not synchronized or buffered in any fashion. Constructed via +/// the `std::io::stdio::stdin_raw` function. +struct StdinRaw(stdio::Stdin); + +/// A handle to a raw instance of the standard output stream of this process. +/// +/// This handle is not synchronized or buffered in any fashion. Constructed via +/// the `std::io::stdio::stdout_raw` function. +struct StdoutRaw(stdio::Stdout); + +/// A handle to a raw instance of the standard output stream of this process. +/// +/// This handle is not synchronized or buffered in any fashion. Constructed via +/// the `std::io::stdio::stderr_raw` function. +struct StderrRaw(stdio::Stderr); + +/// Constructs a new raw handle to the standard input of this process. +/// +/// The returned handle does not interact with any other handles created nor +/// handles returned by `std::io::stdin`. Data buffered by the `std::io::stdin` +/// handles is **not** available to raw handles returned from this function. +/// +/// The returned handle has no external synchronization or buffering. +#[unstable(feature = "libstd_sys_internals", issue = "none")] +const fn stdin_raw() -> StdinRaw { + StdinRaw(stdio::Stdin::new()) +} + +/// Constructs a new raw handle to the standard output stream of this process. +/// +/// The returned handle does not interact with any other handles created nor +/// handles returned by `std::io::stdout`. Note that data is buffered by the +/// `std::io::stdout` handles so writes which happen via this raw handle may +/// appear before previous writes. +/// +/// The returned handle has no external synchronization or buffering layered on +/// top. +#[unstable(feature = "libstd_sys_internals", issue = "none")] +const fn stdout_raw() -> StdoutRaw { + StdoutRaw(stdio::Stdout::new()) +} + +/// Constructs a new raw handle to the standard error stream of this process. +/// +/// The returned handle does not interact with any other handles created nor +/// handles returned by `std::io::stderr`. +/// +/// The returned handle has no external synchronization or buffering layered on +/// top. +#[unstable(feature = "libstd_sys_internals", issue = "none")] +const fn stderr_raw() -> StderrRaw { + StderrRaw(stdio::Stderr::new()) +} + +impl Read for StdinRaw { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + handle_ebadf(self.0.read(buf), 0) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + handle_ebadf(self.0.read_vectored(bufs), 0) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + self.0.is_read_vectored() + } + + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + handle_ebadf(self.0.read_to_end(buf), 0) + } + + fn read_to_string(&mut self, buf: &mut String) -> io::Result<usize> { + handle_ebadf(self.0.read_to_string(buf), 0) + } +} + +impl Write for StdoutRaw { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + handle_ebadf(self.0.write(buf), buf.len()) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let total = bufs.iter().map(|b| b.len()).sum(); + handle_ebadf(self.0.write_vectored(bufs), total) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + self.0.is_write_vectored() + } + + fn flush(&mut self) -> io::Result<()> { + handle_ebadf(self.0.flush(), ()) + } + + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + handle_ebadf(self.0.write_all(buf), ()) + } + + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + handle_ebadf(self.0.write_all_vectored(bufs), ()) + } + + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> io::Result<()> { + handle_ebadf(self.0.write_fmt(fmt), ()) + } +} + +impl Write for StderrRaw { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + handle_ebadf(self.0.write(buf), buf.len()) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let total = bufs.iter().map(|b| b.len()).sum(); + handle_ebadf(self.0.write_vectored(bufs), total) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + self.0.is_write_vectored() + } + + fn flush(&mut self) -> io::Result<()> { + handle_ebadf(self.0.flush(), ()) + } + + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + handle_ebadf(self.0.write_all(buf), ()) + } + + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + handle_ebadf(self.0.write_all_vectored(bufs), ()) + } + + fn write_fmt(&mut self, fmt: fmt::Arguments<'_>) -> io::Result<()> { + handle_ebadf(self.0.write_fmt(fmt), ()) + } +} + +fn handle_ebadf<T>(r: io::Result<T>, default: T) -> io::Result<T> { + match r { + Err(ref e) if stdio::is_ebadf(e) => Ok(default), + r => r, + } +} + +/// A handle to the standard input stream of a process. +/// +/// Each handle is a shared reference to a global buffer of input data to this +/// process. A handle can be `lock`'d to gain full access to [`BufRead`] methods +/// (e.g., `.lines()`). Reads to this handle are otherwise locked with respect +/// to other reads. +/// +/// This handle implements the `Read` trait, but beware that concurrent reads +/// of `Stdin` must be executed with care. +/// +/// Created by the [`io::stdin`] method. +/// +/// [`io::stdin`]: stdin +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to read bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// # Examples +/// +/// ```no_run +/// use std::io; +/// +/// fn main() -> io::Result<()> { +/// let mut buffer = String::new(); +/// let stdin = io::stdin(); // We get `Stdin` here. +/// stdin.read_line(&mut buffer)?; +/// Ok(()) +/// } +/// ``` +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Stdin { + inner: &'static Mutex<BufReader<StdinRaw>>, +} + +/// A locked reference to the [`Stdin`] handle. +/// +/// This handle implements both the [`Read`] and [`BufRead`] traits, and +/// is constructed via the [`Stdin::lock`] method. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to read bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// # Examples +/// +/// ```no_run +/// use std::io::{self, BufRead}; +/// +/// fn main() -> io::Result<()> { +/// let mut buffer = String::new(); +/// let stdin = io::stdin(); // We get `Stdin` here. +/// { +/// let mut handle = stdin.lock(); // We get `StdinLock` here. +/// handle.read_line(&mut buffer)?; +/// } // `StdinLock` is dropped here. +/// Ok(()) +/// } +/// ``` +#[must_use = "if unused stdin will immediately unlock"] +#[stable(feature = "rust1", since = "1.0.0")] +pub struct StdinLock<'a> { + inner: MutexGuard<'a, BufReader<StdinRaw>>, +} + +/// Constructs a new handle to the standard input of the current process. +/// +/// Each handle returned is a reference to a shared global buffer whose access +/// is synchronized via a mutex. If you need more explicit control over +/// locking, see the [`Stdin::lock`] method. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to read bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// # Examples +/// +/// Using implicit synchronization: +/// +/// ```no_run +/// use std::io; +/// +/// fn main() -> io::Result<()> { +/// let mut buffer = String::new(); +/// io::stdin().read_line(&mut buffer)?; +/// Ok(()) +/// } +/// ``` +/// +/// Using explicit synchronization: +/// +/// ```no_run +/// use std::io::{self, BufRead}; +/// +/// fn main() -> io::Result<()> { +/// let mut buffer = String::new(); +/// let stdin = io::stdin(); +/// let mut handle = stdin.lock(); +/// +/// handle.read_line(&mut buffer)?; +/// Ok(()) +/// } +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +pub fn stdin() -> Stdin { + static INSTANCE: OnceLock<Mutex<BufReader<StdinRaw>>> = OnceLock::new(); + Stdin { + inner: INSTANCE.get_or_init(|| { + Mutex::new(BufReader::with_capacity(stdio::STDIN_BUF_SIZE, stdin_raw())) + }), + } +} + +impl Stdin { + /// Locks this handle to the standard input stream, returning a readable + /// guard. + /// + /// The lock is released when the returned lock goes out of scope. The + /// returned guard also implements the [`Read`] and [`BufRead`] traits for + /// accessing the underlying data. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::{self, BufRead}; + /// + /// fn main() -> io::Result<()> { + /// let mut buffer = String::new(); + /// let stdin = io::stdin(); + /// let mut handle = stdin.lock(); + /// + /// handle.read_line(&mut buffer)?; + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn lock(&self) -> StdinLock<'static> { + // Locks this handle with 'static lifetime. This depends on the + // implementation detail that the underlying `Mutex` is static. + StdinLock { inner: self.inner.lock().unwrap_or_else(|e| e.into_inner()) } + } + + /// Locks this handle and reads a line of input, appending it to the specified buffer. + /// + /// For detailed semantics of this method, see the documentation on + /// [`BufRead::read_line`]. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// + /// let mut input = String::new(); + /// match io::stdin().read_line(&mut input) { + /// Ok(n) => { + /// println!("{n} bytes read"); + /// println!("{input}"); + /// } + /// Err(error) => println!("error: {error}"), + /// } + /// ``` + /// + /// You can run the example one of two ways: + /// + /// - Pipe some text to it, e.g., `printf foo | path/to/executable` + /// - Give it text interactively by running the executable directly, + /// in which case it will wait for the Enter key to be pressed before + /// continuing + #[stable(feature = "rust1", since = "1.0.0")] + pub fn read_line(&self, buf: &mut String) -> io::Result<usize> { + self.lock().read_line(buf) + } + + /// Consumes this handle and returns an iterator over input lines. + /// + /// For detailed semantics of this method, see the documentation on + /// [`BufRead::lines`]. + /// + /// # Examples + /// + /// ```no_run + /// use std::io; + /// + /// let lines = io::stdin().lines(); + /// for line in lines { + /// println!("got a line: {}", line.unwrap()); + /// } + /// ``` + #[must_use = "`self` will be dropped if the result is not used"] + #[stable(feature = "stdin_forwarders", since = "1.62.0")] + pub fn lines(self) -> Lines<StdinLock<'static>> { + self.lock().lines() + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Stdin { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Stdin").finish_non_exhaustive() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for Stdin { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + self.lock().read(buf) + } + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + self.lock().read_vectored(bufs) + } + #[inline] + fn is_read_vectored(&self) -> bool { + self.lock().is_read_vectored() + } + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + self.lock().read_to_end(buf) + } + fn read_to_string(&mut self, buf: &mut String) -> io::Result<usize> { + self.lock().read_to_string(buf) + } + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + self.lock().read_exact(buf) + } +} + +// only used by platform-dependent io::copy specializations, i.e. unused on some platforms +#[cfg(any(target_os = "linux", target_os = "android"))] +impl StdinLock<'_> { + pub(crate) fn as_mut_buf(&mut self) -> &mut BufReader<impl Read> { + &mut self.inner + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for StdinLock<'_> { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + self.inner.read(buf) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + self.inner.read_vectored(bufs) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + self.inner.is_read_vectored() + } + + fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> { + self.inner.read_to_end(buf) + } + + fn read_to_string(&mut self, buf: &mut String) -> io::Result<usize> { + self.inner.read_to_string(buf) + } + + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + self.inner.read_exact(buf) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl BufRead for StdinLock<'_> { + fn fill_buf(&mut self) -> io::Result<&[u8]> { + self.inner.fill_buf() + } + + fn consume(&mut self, n: usize) { + self.inner.consume(n) + } + + fn read_until(&mut self, byte: u8, buf: &mut Vec<u8>) -> io::Result<usize> { + self.inner.read_until(byte, buf) + } + + fn read_line(&mut self, buf: &mut String) -> io::Result<usize> { + self.inner.read_line(buf) + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for StdinLock<'_> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("StdinLock").finish_non_exhaustive() + } +} + +/// A handle to the global standard output stream of the current process. +/// +/// Each handle shares a global buffer of data to be written to the standard +/// output stream. Access is also synchronized via a lock and explicit control +/// over locking is available via the [`lock`] method. +/// +/// Created by the [`io::stdout`] method. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// [`lock`]: Stdout::lock +/// [`io::stdout`]: stdout +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Stdout { + // FIXME: this should be LineWriter or BufWriter depending on the state of + // stdout (tty or not). Note that if this is not line buffered it + // should also flush-on-panic or some form of flush-on-abort. + inner: Pin<&'static ReentrantMutex<RefCell<LineWriter<StdoutRaw>>>>, +} + +/// A locked reference to the [`Stdout`] handle. +/// +/// This handle implements the [`Write`] trait, and is constructed via +/// the [`Stdout::lock`] method. See its documentation for more. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +#[must_use = "if unused stdout will immediately unlock"] +#[stable(feature = "rust1", since = "1.0.0")] +pub struct StdoutLock<'a> { + inner: ReentrantMutexGuard<'a, RefCell<LineWriter<StdoutRaw>>>, +} + +static STDOUT: OnceLock<ReentrantMutex<RefCell<LineWriter<StdoutRaw>>>> = OnceLock::new(); + +/// Constructs a new handle to the standard output of the current process. +/// +/// Each handle returned is a reference to a shared global buffer whose access +/// is synchronized via a mutex. If you need more explicit control over +/// locking, see the [`Stdout::lock`] method. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// # Examples +/// +/// Using implicit synchronization: +/// +/// ```no_run +/// use std::io::{self, Write}; +/// +/// fn main() -> io::Result<()> { +/// io::stdout().write_all(b"hello world")?; +/// +/// Ok(()) +/// } +/// ``` +/// +/// Using explicit synchronization: +/// +/// ```no_run +/// use std::io::{self, Write}; +/// +/// fn main() -> io::Result<()> { +/// let stdout = io::stdout(); +/// let mut handle = stdout.lock(); +/// +/// handle.write_all(b"hello world")?; +/// +/// Ok(()) +/// } +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +pub fn stdout() -> Stdout { + Stdout { + inner: Pin::static_ref(&STDOUT).get_or_init_pin( + || unsafe { ReentrantMutex::new(RefCell::new(LineWriter::new(stdout_raw()))) }, + |mutex| unsafe { mutex.init() }, + ), + } +} + +pub fn cleanup() { + if let Some(instance) = STDOUT.get() { + // Flush the data and disable buffering during shutdown + // by replacing the line writer by one with zero + // buffering capacity. + // We use try_lock() instead of lock(), because someone + // might have leaked a StdoutLock, which would + // otherwise cause a deadlock here. + if let Some(lock) = Pin::static_ref(instance).try_lock() { + *lock.borrow_mut() = LineWriter::with_capacity(0, stdout_raw()); + } + } +} + +impl Stdout { + /// Locks this handle to the standard output stream, returning a writable + /// guard. + /// + /// The lock is released when the returned lock goes out of scope. The + /// returned guard also implements the `Write` trait for writing data. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::{self, Write}; + /// + /// fn main() -> io::Result<()> { + /// let mut stdout = io::stdout().lock(); + /// + /// stdout.write_all(b"hello world")?; + /// + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn lock(&self) -> StdoutLock<'static> { + // Locks this handle with 'static lifetime. This depends on the + // implementation detail that the underlying `ReentrantMutex` is + // static. + StdoutLock { inner: self.inner.lock() } + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Stdout { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Stdout").finish_non_exhaustive() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for Stdout { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + (&*self).write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + (&*self).write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + io::Write::is_write_vectored(&&*self) + } + fn flush(&mut self) -> io::Result<()> { + (&*self).flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + (&*self).write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + (&*self).write_all_vectored(bufs) + } + fn write_fmt(&mut self, args: fmt::Arguments<'_>) -> io::Result<()> { + (&*self).write_fmt(args) + } +} + +#[stable(feature = "write_mt", since = "1.48.0")] +impl Write for &Stdout { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.lock().write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + self.lock().write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + self.lock().is_write_vectored() + } + fn flush(&mut self) -> io::Result<()> { + self.lock().flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.lock().write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + self.lock().write_all_vectored(bufs) + } + fn write_fmt(&mut self, args: fmt::Arguments<'_>) -> io::Result<()> { + self.lock().write_fmt(args) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for StdoutLock<'_> { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.inner.borrow_mut().write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + self.inner.borrow_mut().write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + self.inner.borrow_mut().is_write_vectored() + } + fn flush(&mut self) -> io::Result<()> { + self.inner.borrow_mut().flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.inner.borrow_mut().write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + self.inner.borrow_mut().write_all_vectored(bufs) + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for StdoutLock<'_> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("StdoutLock").finish_non_exhaustive() + } +} + +/// A handle to the standard error stream of a process. +/// +/// For more information, see the [`io::stderr`] method. +/// +/// [`io::stderr`]: stderr +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Stderr { + inner: Pin<&'static ReentrantMutex<RefCell<StderrRaw>>>, +} + +/// A locked reference to the [`Stderr`] handle. +/// +/// This handle implements the [`Write`] trait and is constructed via +/// the [`Stderr::lock`] method. See its documentation for more. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +#[must_use = "if unused stderr will immediately unlock"] +#[stable(feature = "rust1", since = "1.0.0")] +pub struct StderrLock<'a> { + inner: ReentrantMutexGuard<'a, RefCell<StderrRaw>>, +} + +/// Constructs a new handle to the standard error of the current process. +/// +/// This handle is not buffered. +/// +/// ### Note: Windows Portability Considerations +/// +/// When operating in a console, the Windows implementation of this stream does not support +/// non-UTF-8 byte sequences. Attempting to write bytes that are not valid UTF-8 will return +/// an error. +/// +/// In a process with a detached console, such as one using +/// `#![windows_subsystem = "windows"]`, or in a child process spawned from such a process, +/// the contained handle will be null. In such cases, the standard library's `Read` and +/// `Write` will do nothing and silently succeed. All other I/O operations, via the +/// standard library or via raw Windows API calls, will fail. +/// +/// # Examples +/// +/// Using implicit synchronization: +/// +/// ```no_run +/// use std::io::{self, Write}; +/// +/// fn main() -> io::Result<()> { +/// io::stderr().write_all(b"hello world")?; +/// +/// Ok(()) +/// } +/// ``` +/// +/// Using explicit synchronization: +/// +/// ```no_run +/// use std::io::{self, Write}; +/// +/// fn main() -> io::Result<()> { +/// let stderr = io::stderr(); +/// let mut handle = stderr.lock(); +/// +/// handle.write_all(b"hello world")?; +/// +/// Ok(()) +/// } +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +pub fn stderr() -> Stderr { + // Note that unlike `stdout()` we don't use `at_exit` here to register a + // destructor. Stderr is not buffered , so there's no need to run a + // destructor for flushing the buffer + static INSTANCE: OnceLock<ReentrantMutex<RefCell<StderrRaw>>> = OnceLock::new(); + + Stderr { + inner: Pin::static_ref(&INSTANCE).get_or_init_pin( + || unsafe { ReentrantMutex::new(RefCell::new(stderr_raw())) }, + |mutex| unsafe { mutex.init() }, + ), + } +} + +impl Stderr { + /// Locks this handle to the standard error stream, returning a writable + /// guard. + /// + /// The lock is released when the returned lock goes out of scope. The + /// returned guard also implements the [`Write`] trait for writing data. + /// + /// # Examples + /// + /// ``` + /// use std::io::{self, Write}; + /// + /// fn foo() -> io::Result<()> { + /// let stderr = io::stderr(); + /// let mut handle = stderr.lock(); + /// + /// handle.write_all(b"hello world")?; + /// + /// Ok(()) + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn lock(&self) -> StderrLock<'static> { + // Locks this handle with 'static lifetime. This depends on the + // implementation detail that the underlying `ReentrantMutex` is + // static. + StderrLock { inner: self.inner.lock() } + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Stderr { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Stderr").finish_non_exhaustive() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for Stderr { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + (&*self).write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + (&*self).write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + io::Write::is_write_vectored(&&*self) + } + fn flush(&mut self) -> io::Result<()> { + (&*self).flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + (&*self).write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + (&*self).write_all_vectored(bufs) + } + fn write_fmt(&mut self, args: fmt::Arguments<'_>) -> io::Result<()> { + (&*self).write_fmt(args) + } +} + +#[stable(feature = "write_mt", since = "1.48.0")] +impl Write for &Stderr { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.lock().write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + self.lock().write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + self.lock().is_write_vectored() + } + fn flush(&mut self) -> io::Result<()> { + self.lock().flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.lock().write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + self.lock().write_all_vectored(bufs) + } + fn write_fmt(&mut self, args: fmt::Arguments<'_>) -> io::Result<()> { + self.lock().write_fmt(args) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for StderrLock<'_> { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.inner.borrow_mut().write(buf) + } + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + self.inner.borrow_mut().write_vectored(bufs) + } + #[inline] + fn is_write_vectored(&self) -> bool { + self.inner.borrow_mut().is_write_vectored() + } + fn flush(&mut self) -> io::Result<()> { + self.inner.borrow_mut().flush() + } + fn write_all(&mut self, buf: &[u8]) -> io::Result<()> { + self.inner.borrow_mut().write_all(buf) + } + fn write_all_vectored(&mut self, bufs: &mut [IoSlice<'_>]) -> io::Result<()> { + self.inner.borrow_mut().write_all_vectored(bufs) + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for StderrLock<'_> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("StderrLock").finish_non_exhaustive() + } +} + +/// Sets the thread-local output capture buffer and returns the old one. +#[unstable( + feature = "internal_output_capture", + reason = "this function is meant for use in the test crate \ + and may disappear in the future", + issue = "none" +)] +#[doc(hidden)] +pub fn set_output_capture(sink: Option<LocalStream>) -> Option<LocalStream> { + if sink.is_none() && !OUTPUT_CAPTURE_USED.load(Ordering::Relaxed) { + // OUTPUT_CAPTURE is definitely None since OUTPUT_CAPTURE_USED is false. + return None; + } + OUTPUT_CAPTURE_USED.store(true, Ordering::Relaxed); + OUTPUT_CAPTURE.with(move |slot| slot.replace(sink)) +} + +/// Write `args` to the capture buffer if enabled and possible, or `global_s` +/// otherwise. `label` identifies the stream in a panic message. +/// +/// This function is used to print error messages, so it takes extra +/// care to avoid causing a panic when `local_s` is unusable. +/// For instance, if the TLS key for the local stream is +/// already destroyed, or if the local stream is locked by another +/// thread, it will just fall back to the global stream. +/// +/// However, if the actual I/O causes an error, this function does panic. +fn print_to<T>(args: fmt::Arguments<'_>, global_s: fn() -> T, label: &str) +where + T: Write, +{ + if OUTPUT_CAPTURE_USED.load(Ordering::Relaxed) + && OUTPUT_CAPTURE.try_with(|s| { + // Note that we completely remove a local sink to write to in case + // our printing recursively panics/prints, so the recursive + // panic/print goes to the global sink instead of our local sink. + s.take().map(|w| { + let _ = w.lock().unwrap_or_else(|e| e.into_inner()).write_fmt(args); + s.set(Some(w)); + }) + }) == Ok(Some(())) + { + // Successfully wrote to capture buffer. + return; + } + + if let Err(e) = global_s().write_fmt(args) { + panic!("failed printing to {label}: {e}"); + } +} + +#[unstable( + feature = "print_internals", + reason = "implementation detail which may disappear or be replaced at any time", + issue = "none" +)] +#[doc(hidden)] +#[cfg(not(test))] +pub fn _print(args: fmt::Arguments<'_>) { + print_to(args, stdout, "stdout"); +} + +#[unstable( + feature = "print_internals", + reason = "implementation detail which may disappear or be replaced at any time", + issue = "none" +)] +#[doc(hidden)] +#[cfg(not(test))] +pub fn _eprint(args: fmt::Arguments<'_>) { + print_to(args, stderr, "stderr"); +} + +#[cfg(test)] +pub use realstd::io::{_eprint, _print}; diff --git a/library/std/src/io/stdio/tests.rs b/library/std/src/io/stdio/tests.rs new file mode 100644 index 000000000..f89fd27ce --- /dev/null +++ b/library/std/src/io/stdio/tests.rs @@ -0,0 +1,166 @@ +use super::*; +use crate::panic::{RefUnwindSafe, UnwindSafe}; +use crate::sync::mpsc::sync_channel; +use crate::thread; + +#[test] +fn stdout_unwind_safe() { + assert_unwind_safe::<Stdout>(); +} +#[test] +fn stdoutlock_unwind_safe() { + assert_unwind_safe::<StdoutLock<'_>>(); + assert_unwind_safe::<StdoutLock<'static>>(); +} +#[test] +fn stderr_unwind_safe() { + assert_unwind_safe::<Stderr>(); +} +#[test] +fn stderrlock_unwind_safe() { + assert_unwind_safe::<StderrLock<'_>>(); + assert_unwind_safe::<StderrLock<'static>>(); +} + +fn assert_unwind_safe<T: UnwindSafe + RefUnwindSafe>() {} + +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn panic_doesnt_poison() { + thread::spawn(|| { + let _a = stdin(); + let _a = _a.lock(); + let _a = stdout(); + let _a = _a.lock(); + let _a = stderr(); + let _a = _a.lock(); + panic!(); + }) + .join() + .unwrap_err(); + + let _a = stdin(); + let _a = _a.lock(); + let _a = stdout(); + let _a = _a.lock(); + let _a = stderr(); + let _a = _a.lock(); +} + +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn test_lock_stderr() { + test_lock(stderr, || stderr().lock()); +} +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn test_lock_stdin() { + test_lock(stdin, || stdin().lock()); +} +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn test_lock_stdout() { + test_lock(stdout, || stdout().lock()); +} + +// Helper trait to make lock testing function generic. +trait Stdio<'a>: 'static +where + Self::Lock: 'a, +{ + type Lock; + fn lock(&'a self) -> Self::Lock; +} +impl<'a> Stdio<'a> for Stderr { + type Lock = StderrLock<'a>; + fn lock(&'a self) -> StderrLock<'a> { + self.lock() + } +} +impl<'a> Stdio<'a> for Stdin { + type Lock = StdinLock<'a>; + fn lock(&'a self) -> StdinLock<'a> { + self.lock() + } +} +impl<'a> Stdio<'a> for Stdout { + type Lock = StdoutLock<'a>; + fn lock(&'a self) -> StdoutLock<'a> { + self.lock() + } +} + +// Helper trait to make lock testing function generic. +trait StdioOwnedLock: 'static {} +impl StdioOwnedLock for StderrLock<'static> {} +impl StdioOwnedLock for StdinLock<'static> {} +impl StdioOwnedLock for StdoutLock<'static> {} + +// Tests locking on stdio handles by starting two threads and checking that +// they block each other appropriately. +fn test_lock<T, U>(get_handle: fn() -> T, get_locked: fn() -> U) +where + T: for<'a> Stdio<'a>, + U: StdioOwnedLock, +{ + // State enum to track different phases of the test, primarily when + // each lock is acquired and released. + #[derive(Debug, PartialEq)] + enum State { + Start1, + Acquire1, + Start2, + Release1, + Acquire2, + Release2, + } + use State::*; + // Logging vector to be checked to make sure lock acquisitions and + // releases happened in the correct order. + let log = Arc::new(Mutex::new(Vec::new())); + let ((tx1, rx1), (tx2, rx2)) = (sync_channel(0), sync_channel(0)); + let th1 = { + let (log, tx) = (Arc::clone(&log), tx1); + thread::spawn(move || { + log.lock().unwrap().push(Start1); + let handle = get_handle(); + { + let locked = handle.lock(); + log.lock().unwrap().push(Acquire1); + tx.send(Acquire1).unwrap(); // notify of acquisition + tx.send(Release1).unwrap(); // wait for release command + log.lock().unwrap().push(Release1); + } + tx.send(Acquire1).unwrap(); // wait for th2 acquire + { + let locked = handle.lock(); + log.lock().unwrap().push(Acquire1); + } + log.lock().unwrap().push(Release1); + }) + }; + let th2 = { + let (log, tx) = (Arc::clone(&log), tx2); + thread::spawn(move || { + tx.send(Start2).unwrap(); // wait for start command + let locked = get_locked(); + log.lock().unwrap().push(Acquire2); + tx.send(Acquire2).unwrap(); // notify of acquisition + tx.send(Release2).unwrap(); // wait for release command + log.lock().unwrap().push(Release2); + }) + }; + assert_eq!(rx1.recv().unwrap(), Acquire1); // wait for th1 acquire + log.lock().unwrap().push(Start2); + assert_eq!(rx2.recv().unwrap(), Start2); // block th2 + assert_eq!(rx1.recv().unwrap(), Release1); // release th1 + assert_eq!(rx2.recv().unwrap(), Acquire2); // wait for th2 acquire + assert_eq!(rx1.recv().unwrap(), Acquire1); // block th1 + assert_eq!(rx2.recv().unwrap(), Release2); // release th2 + th2.join().unwrap(); + th1.join().unwrap(); + assert_eq!( + *log.lock().unwrap(), + [Start1, Acquire1, Start2, Release1, Acquire2, Release2, Acquire1, Release1] + ); +} diff --git a/library/std/src/io/tests.rs b/library/std/src/io/tests.rs new file mode 100644 index 000000000..f357f33ec --- /dev/null +++ b/library/std/src/io/tests.rs @@ -0,0 +1,623 @@ +use super::{repeat, Cursor, ReadBuf, SeekFrom}; +use crate::cmp::{self, min}; +use crate::io::{self, IoSlice, IoSliceMut}; +use crate::io::{BufRead, BufReader, Read, Seek, Write}; +use crate::mem::MaybeUninit; +use crate::ops::Deref; + +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn read_until() { + let mut buf = Cursor::new(&b"12"[..]); + let mut v = Vec::new(); + assert_eq!(buf.read_until(b'3', &mut v).unwrap(), 2); + assert_eq!(v, b"12"); + + let mut buf = Cursor::new(&b"1233"[..]); + let mut v = Vec::new(); + assert_eq!(buf.read_until(b'3', &mut v).unwrap(), 3); + assert_eq!(v, b"123"); + v.truncate(0); + assert_eq!(buf.read_until(b'3', &mut v).unwrap(), 1); + assert_eq!(v, b"3"); + v.truncate(0); + assert_eq!(buf.read_until(b'3', &mut v).unwrap(), 0); + assert_eq!(v, []); +} + +#[test] +fn split() { + let buf = Cursor::new(&b"12"[..]); + let mut s = buf.split(b'3'); + assert_eq!(s.next().unwrap().unwrap(), vec![b'1', b'2']); + assert!(s.next().is_none()); + + let buf = Cursor::new(&b"1233"[..]); + let mut s = buf.split(b'3'); + assert_eq!(s.next().unwrap().unwrap(), vec![b'1', b'2']); + assert_eq!(s.next().unwrap().unwrap(), vec![]); + assert!(s.next().is_none()); +} + +#[test] +fn read_line() { + let mut buf = Cursor::new(&b"12"[..]); + let mut v = String::new(); + assert_eq!(buf.read_line(&mut v).unwrap(), 2); + assert_eq!(v, "12"); + + let mut buf = Cursor::new(&b"12\n\n"[..]); + let mut v = String::new(); + assert_eq!(buf.read_line(&mut v).unwrap(), 3); + assert_eq!(v, "12\n"); + v.truncate(0); + assert_eq!(buf.read_line(&mut v).unwrap(), 1); + assert_eq!(v, "\n"); + v.truncate(0); + assert_eq!(buf.read_line(&mut v).unwrap(), 0); + assert_eq!(v, ""); +} + +#[test] +fn lines() { + let buf = Cursor::new(&b"12\r"[..]); + let mut s = buf.lines(); + assert_eq!(s.next().unwrap().unwrap(), "12\r".to_string()); + assert!(s.next().is_none()); + + let buf = Cursor::new(&b"12\r\n\n"[..]); + let mut s = buf.lines(); + assert_eq!(s.next().unwrap().unwrap(), "12".to_string()); + assert_eq!(s.next().unwrap().unwrap(), "".to_string()); + assert!(s.next().is_none()); +} + +#[test] +fn buf_read_has_data_left() { + let mut buf = Cursor::new(&b"abcd"[..]); + assert!(buf.has_data_left().unwrap()); + buf.read_exact(&mut [0; 2]).unwrap(); + assert!(buf.has_data_left().unwrap()); + buf.read_exact(&mut [0; 2]).unwrap(); + assert!(!buf.has_data_left().unwrap()); +} + +#[test] +fn read_to_end() { + let mut c = Cursor::new(&b""[..]); + let mut v = Vec::new(); + assert_eq!(c.read_to_end(&mut v).unwrap(), 0); + assert_eq!(v, []); + + let mut c = Cursor::new(&b"1"[..]); + let mut v = Vec::new(); + assert_eq!(c.read_to_end(&mut v).unwrap(), 1); + assert_eq!(v, b"1"); + + let cap = 1024 * 1024; + let data = (0..cap).map(|i| (i / 3) as u8).collect::<Vec<_>>(); + let mut v = Vec::new(); + let (a, b) = data.split_at(data.len() / 2); + assert_eq!(Cursor::new(a).read_to_end(&mut v).unwrap(), a.len()); + assert_eq!(Cursor::new(b).read_to_end(&mut v).unwrap(), b.len()); + assert_eq!(v, data); +} + +#[test] +fn read_to_string() { + let mut c = Cursor::new(&b""[..]); + let mut v = String::new(); + assert_eq!(c.read_to_string(&mut v).unwrap(), 0); + assert_eq!(v, ""); + + let mut c = Cursor::new(&b"1"[..]); + let mut v = String::new(); + assert_eq!(c.read_to_string(&mut v).unwrap(), 1); + assert_eq!(v, "1"); + + let mut c = Cursor::new(&b"\xff"[..]); + let mut v = String::new(); + assert!(c.read_to_string(&mut v).is_err()); +} + +#[test] +fn read_exact() { + let mut buf = [0; 4]; + + let mut c = Cursor::new(&b""[..]); + assert_eq!(c.read_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + + let mut c = Cursor::new(&b"123"[..]).chain(Cursor::new(&b"456789"[..])); + c.read_exact(&mut buf).unwrap(); + assert_eq!(&buf, b"1234"); + c.read_exact(&mut buf).unwrap(); + assert_eq!(&buf, b"5678"); + assert_eq!(c.read_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); +} + +#[test] +fn read_exact_slice() { + let mut buf = [0; 4]; + + let mut c = &b""[..]; + assert_eq!(c.read_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + + let mut c = &b"123"[..]; + assert_eq!(c.read_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + // make sure the optimized (early returning) method is being used + assert_eq!(&buf, &[0; 4]); + + let mut c = &b"1234"[..]; + c.read_exact(&mut buf).unwrap(); + assert_eq!(&buf, b"1234"); + + let mut c = &b"56789"[..]; + c.read_exact(&mut buf).unwrap(); + assert_eq!(&buf, b"5678"); + assert_eq!(c, b"9"); +} + +#[test] +fn read_buf_exact() { + let mut buf = [0; 4]; + let mut buf = ReadBuf::new(&mut buf); + + let mut c = Cursor::new(&b""[..]); + assert_eq!(c.read_buf_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + + let mut c = Cursor::new(&b"123456789"[..]); + c.read_buf_exact(&mut buf).unwrap(); + assert_eq!(buf.filled(), b"1234"); + + buf.clear(); + + c.read_buf_exact(&mut buf).unwrap(); + assert_eq!(buf.filled(), b"5678"); + + buf.clear(); + + assert_eq!(c.read_buf_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); +} + +#[test] +fn take_eof() { + struct R; + + impl Read for R { + fn read(&mut self, _: &mut [u8]) -> io::Result<usize> { + Err(io::const_io_error!(io::ErrorKind::Other, "")) + } + } + impl BufRead for R { + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Err(io::const_io_error!(io::ErrorKind::Other, "")) + } + fn consume(&mut self, _amt: usize) {} + } + + let mut buf = [0; 1]; + assert_eq!(0, R.take(0).read(&mut buf).unwrap()); + assert_eq!(b"", R.take(0).fill_buf().unwrap()); +} + +fn cmp_bufread<Br1: BufRead, Br2: BufRead>(mut br1: Br1, mut br2: Br2, exp: &[u8]) { + let mut cat = Vec::new(); + loop { + let consume = { + let buf1 = br1.fill_buf().unwrap(); + let buf2 = br2.fill_buf().unwrap(); + let minlen = if buf1.len() < buf2.len() { buf1.len() } else { buf2.len() }; + assert_eq!(buf1[..minlen], buf2[..minlen]); + cat.extend_from_slice(&buf1[..minlen]); + minlen + }; + if consume == 0 { + break; + } + br1.consume(consume); + br2.consume(consume); + } + assert_eq!(br1.fill_buf().unwrap().len(), 0); + assert_eq!(br2.fill_buf().unwrap().len(), 0); + assert_eq!(&cat[..], &exp[..]) +} + +#[test] +fn chain_bufread() { + let testdata = b"ABCDEFGHIJKL"; + let chain1 = + (&testdata[..3]).chain(&testdata[3..6]).chain(&testdata[6..9]).chain(&testdata[9..]); + let chain2 = (&testdata[..4]).chain(&testdata[4..8]).chain(&testdata[8..]); + cmp_bufread(chain1, chain2, &testdata[..]); +} + +#[test] +fn bufreader_size_hint() { + let testdata = b"ABCDEFGHIJKL"; + let mut buf_reader = BufReader::new(&testdata[..]); + assert_eq!(buf_reader.buffer().len(), 0); + + let buffer_length = testdata.len(); + buf_reader.fill_buf().unwrap(); + + // Check that size hint matches buffer contents + let mut buffered_bytes = buf_reader.bytes(); + let (lower_bound, _upper_bound) = buffered_bytes.size_hint(); + assert_eq!(lower_bound, buffer_length); + + // Check that size hint matches buffer contents after advancing + buffered_bytes.next().unwrap().unwrap(); + let (lower_bound, _upper_bound) = buffered_bytes.size_hint(); + assert_eq!(lower_bound, buffer_length - 1); +} + +#[test] +fn empty_size_hint() { + let size_hint = io::empty().bytes().size_hint(); + assert_eq!(size_hint, (0, Some(0))); +} + +#[test] +fn slice_size_hint() { + let size_hint = (&[1, 2, 3]).bytes().size_hint(); + assert_eq!(size_hint, (3, Some(3))); +} + +#[test] +fn take_size_hint() { + let size_hint = (&[1, 2, 3]).take(2).bytes().size_hint(); + assert_eq!(size_hint, (2, Some(2))); + + let size_hint = (&[1, 2, 3]).take(4).bytes().size_hint(); + assert_eq!(size_hint, (3, Some(3))); + + let size_hint = io::repeat(0).take(3).bytes().size_hint(); + assert_eq!(size_hint, (3, Some(3))); +} + +#[test] +fn chain_empty_size_hint() { + let chain = io::empty().chain(io::empty()); + let size_hint = chain.bytes().size_hint(); + assert_eq!(size_hint, (0, Some(0))); +} + +#[test] +fn chain_size_hint() { + let testdata = b"ABCDEFGHIJKL"; + let mut buf_reader_1 = BufReader::new(&testdata[..6]); + let mut buf_reader_2 = BufReader::new(&testdata[6..]); + + buf_reader_1.fill_buf().unwrap(); + buf_reader_2.fill_buf().unwrap(); + + let chain = buf_reader_1.chain(buf_reader_2); + let size_hint = chain.bytes().size_hint(); + assert_eq!(size_hint, (testdata.len(), Some(testdata.len()))); +} + +#[test] +fn chain_zero_length_read_is_not_eof() { + let a = b"A"; + let b = b"B"; + let mut s = String::new(); + let mut chain = (&a[..]).chain(&b[..]); + chain.read(&mut []).unwrap(); + chain.read_to_string(&mut s).unwrap(); + assert_eq!("AB", s); +} + +#[bench] +#[cfg_attr(target_os = "emscripten", ignore)] +fn bench_read_to_end(b: &mut test::Bencher) { + b.iter(|| { + let mut lr = repeat(1).take(10000000); + let mut vec = Vec::with_capacity(1024); + super::default_read_to_end(&mut lr, &mut vec) + }); +} + +#[test] +fn seek_len() -> io::Result<()> { + let mut c = Cursor::new(vec![0; 15]); + assert_eq!(c.stream_len()?, 15); + + c.seek(SeekFrom::End(0))?; + let old_pos = c.stream_position()?; + assert_eq!(c.stream_len()?, 15); + assert_eq!(c.stream_position()?, old_pos); + + c.seek(SeekFrom::Start(7))?; + c.seek(SeekFrom::Current(2))?; + let old_pos = c.stream_position()?; + assert_eq!(c.stream_len()?, 15); + assert_eq!(c.stream_position()?, old_pos); + + Ok(()) +} + +#[test] +fn seek_position() -> io::Result<()> { + // All `asserts` are duplicated here to make sure the method does not + // change anything about the seek state. + let mut c = Cursor::new(vec![0; 15]); + assert_eq!(c.stream_position()?, 0); + assert_eq!(c.stream_position()?, 0); + + c.seek(SeekFrom::End(0))?; + assert_eq!(c.stream_position()?, 15); + assert_eq!(c.stream_position()?, 15); + + c.seek(SeekFrom::Start(7))?; + c.seek(SeekFrom::Current(2))?; + assert_eq!(c.stream_position()?, 9); + assert_eq!(c.stream_position()?, 9); + + c.seek(SeekFrom::End(-3))?; + c.seek(SeekFrom::Current(1))?; + c.seek(SeekFrom::Current(-5))?; + assert_eq!(c.stream_position()?, 8); + assert_eq!(c.stream_position()?, 8); + + c.rewind()?; + assert_eq!(c.stream_position()?, 0); + assert_eq!(c.stream_position()?, 0); + + Ok(()) +} + +// A simple example reader which uses the default implementation of +// read_to_end. +struct ExampleSliceReader<'a> { + slice: &'a [u8], +} + +impl<'a> Read for ExampleSliceReader<'a> { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + let len = cmp::min(self.slice.len(), buf.len()); + buf[..len].copy_from_slice(&self.slice[..len]); + self.slice = &self.slice[len..]; + Ok(len) + } +} + +#[test] +fn test_read_to_end_capacity() -> io::Result<()> { + let input = &b"foo"[..]; + + // read_to_end() takes care not to over-allocate when a buffer is the + // exact size needed. + let mut vec1 = Vec::with_capacity(input.len()); + ExampleSliceReader { slice: input }.read_to_end(&mut vec1)?; + assert_eq!(vec1.len(), input.len()); + assert_eq!(vec1.capacity(), input.len(), "did not allocate more"); + + Ok(()) +} + +#[test] +fn io_slice_mut_advance_slices() { + let mut buf1 = [1; 8]; + let mut buf2 = [2; 16]; + let mut buf3 = [3; 8]; + let mut bufs = &mut [ + IoSliceMut::new(&mut buf1), + IoSliceMut::new(&mut buf2), + IoSliceMut::new(&mut buf3), + ][..]; + + // Only in a single buffer.. + IoSliceMut::advance_slices(&mut bufs, 1); + assert_eq!(bufs[0].deref(), [1; 7].as_ref()); + assert_eq!(bufs[1].deref(), [2; 16].as_ref()); + assert_eq!(bufs[2].deref(), [3; 8].as_ref()); + + // Removing a buffer, leaving others as is. + IoSliceMut::advance_slices(&mut bufs, 7); + assert_eq!(bufs[0].deref(), [2; 16].as_ref()); + assert_eq!(bufs[1].deref(), [3; 8].as_ref()); + + // Removing a buffer and removing from the next buffer. + IoSliceMut::advance_slices(&mut bufs, 18); + assert_eq!(bufs[0].deref(), [3; 6].as_ref()); +} + +#[test] +#[should_panic] +fn io_slice_mut_advance_slices_empty_slice() { + let mut empty_bufs = &mut [][..]; + IoSliceMut::advance_slices(&mut empty_bufs, 1); +} + +#[test] +#[should_panic] +fn io_slice_mut_advance_slices_beyond_total_length() { + let mut buf1 = [1; 8]; + let mut bufs = &mut [IoSliceMut::new(&mut buf1)][..]; + + IoSliceMut::advance_slices(&mut bufs, 9); + assert!(bufs.is_empty()); +} + +#[test] +fn io_slice_advance_slices() { + let buf1 = [1; 8]; + let buf2 = [2; 16]; + let buf3 = [3; 8]; + let mut bufs = &mut [IoSlice::new(&buf1), IoSlice::new(&buf2), IoSlice::new(&buf3)][..]; + + // Only in a single buffer.. + IoSlice::advance_slices(&mut bufs, 1); + assert_eq!(bufs[0].deref(), [1; 7].as_ref()); + assert_eq!(bufs[1].deref(), [2; 16].as_ref()); + assert_eq!(bufs[2].deref(), [3; 8].as_ref()); + + // Removing a buffer, leaving others as is. + IoSlice::advance_slices(&mut bufs, 7); + assert_eq!(bufs[0].deref(), [2; 16].as_ref()); + assert_eq!(bufs[1].deref(), [3; 8].as_ref()); + + // Removing a buffer and removing from the next buffer. + IoSlice::advance_slices(&mut bufs, 18); + assert_eq!(bufs[0].deref(), [3; 6].as_ref()); +} + +#[test] +#[should_panic] +fn io_slice_advance_slices_empty_slice() { + let mut empty_bufs = &mut [][..]; + IoSlice::advance_slices(&mut empty_bufs, 1); +} + +#[test] +#[should_panic] +fn io_slice_advance_slices_beyond_total_length() { + let buf1 = [1; 8]; + let mut bufs = &mut [IoSlice::new(&buf1)][..]; + + IoSlice::advance_slices(&mut bufs, 9); + assert!(bufs.is_empty()); +} + +/// Create a new writer that reads from at most `n_bufs` and reads +/// `per_call` bytes (in total) per call to write. +fn test_writer(n_bufs: usize, per_call: usize) -> TestWriter { + TestWriter { n_bufs, per_call, written: Vec::new() } +} + +struct TestWriter { + n_bufs: usize, + per_call: usize, + written: Vec<u8>, +} + +impl Write for TestWriter { + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + self.write_vectored(&[IoSlice::new(buf)]) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let mut left = self.per_call; + let mut written = 0; + for buf in bufs.iter().take(self.n_bufs) { + let n = min(left, buf.len()); + self.written.extend_from_slice(&buf[0..n]); + left -= n; + written += n; + } + Ok(written) + } + + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[test] +fn test_writer_read_from_one_buf() { + let mut writer = test_writer(1, 2); + + assert_eq!(writer.write(&[]).unwrap(), 0); + assert_eq!(writer.write_vectored(&[]).unwrap(), 0); + + // Read at most 2 bytes. + assert_eq!(writer.write(&[1, 1, 1]).unwrap(), 2); + let bufs = &[IoSlice::new(&[2, 2, 2])]; + assert_eq!(writer.write_vectored(bufs).unwrap(), 2); + + // Only read from first buf. + let bufs = &[IoSlice::new(&[3]), IoSlice::new(&[4, 4])]; + assert_eq!(writer.write_vectored(bufs).unwrap(), 1); + + assert_eq!(writer.written, &[1, 1, 2, 2, 3]); +} + +#[test] +fn test_writer_read_from_multiple_bufs() { + let mut writer = test_writer(3, 3); + + // Read at most 3 bytes from two buffers. + let bufs = &[IoSlice::new(&[1]), IoSlice::new(&[2, 2, 2])]; + assert_eq!(writer.write_vectored(bufs).unwrap(), 3); + + // Read at most 3 bytes from three buffers. + let bufs = &[IoSlice::new(&[3]), IoSlice::new(&[4]), IoSlice::new(&[5, 5])]; + assert_eq!(writer.write_vectored(bufs).unwrap(), 3); + + assert_eq!(writer.written, &[1, 2, 2, 3, 4, 5]); +} + +#[test] +fn test_write_all_vectored() { + #[rustfmt::skip] // Becomes unreadable otherwise. + let tests: Vec<(_, &'static [u8])> = vec![ + (vec![], &[]), + (vec![IoSlice::new(&[]), IoSlice::new(&[])], &[]), + (vec![IoSlice::new(&[1])], &[1]), + (vec![IoSlice::new(&[1, 2])], &[1, 2]), + (vec![IoSlice::new(&[1, 2, 3])], &[1, 2, 3]), + (vec![IoSlice::new(&[1, 2, 3, 4])], &[1, 2, 3, 4]), + (vec![IoSlice::new(&[1, 2, 3, 4, 5])], &[1, 2, 3, 4, 5]), + (vec![IoSlice::new(&[1]), IoSlice::new(&[2])], &[1, 2]), + (vec![IoSlice::new(&[1]), IoSlice::new(&[2, 2])], &[1, 2, 2]), + (vec![IoSlice::new(&[1, 1]), IoSlice::new(&[2, 2])], &[1, 1, 2, 2]), + (vec![IoSlice::new(&[1, 1]), IoSlice::new(&[2, 2, 2])], &[1, 1, 2, 2, 2]), + (vec![IoSlice::new(&[1, 1]), IoSlice::new(&[2, 2, 2])], &[1, 1, 2, 2, 2]), + (vec![IoSlice::new(&[1, 1, 1]), IoSlice::new(&[2, 2, 2])], &[1, 1, 1, 2, 2, 2]), + (vec![IoSlice::new(&[1, 1, 1]), IoSlice::new(&[2, 2, 2, 2])], &[1, 1, 1, 2, 2, 2, 2]), + (vec![IoSlice::new(&[1, 1, 1, 1]), IoSlice::new(&[2, 2, 2, 2])], &[1, 1, 1, 1, 2, 2, 2, 2]), + (vec![IoSlice::new(&[1]), IoSlice::new(&[2]), IoSlice::new(&[3])], &[1, 2, 3]), + (vec![IoSlice::new(&[1, 1]), IoSlice::new(&[2, 2]), IoSlice::new(&[3, 3])], &[1, 1, 2, 2, 3, 3]), + (vec![IoSlice::new(&[1]), IoSlice::new(&[2, 2]), IoSlice::new(&[3, 3, 3])], &[1, 2, 2, 3, 3, 3]), + (vec![IoSlice::new(&[1, 1, 1]), IoSlice::new(&[2, 2, 2]), IoSlice::new(&[3, 3, 3])], &[1, 1, 1, 2, 2, 2, 3, 3, 3]), + ]; + + let writer_configs = &[(1, 1), (1, 2), (1, 3), (2, 2), (2, 3), (3, 3)]; + + for (n_bufs, per_call) in writer_configs.iter().copied() { + for (mut input, wanted) in tests.clone().into_iter() { + let mut writer = test_writer(n_bufs, per_call); + assert!(writer.write_all_vectored(&mut *input).is_ok()); + assert_eq!(&*writer.written, &*wanted); + } + } +} + +// Issue 94981 +#[test] +#[should_panic = "number of read bytes exceeds limit"] +fn test_take_wrong_length() { + struct LieAboutSize(bool); + + impl Read for LieAboutSize { + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + // Lie about the read size at first time of read. + if core::mem::take(&mut self.0) { Ok(buf.len() + 1) } else { Ok(buf.len()) } + } + } + + let mut buffer = vec![0; 4]; + let mut reader = LieAboutSize(true).take(4); + // Primed the `Limit` by lying about the read size. + let _ = reader.read(&mut buffer[..]); +} + +#[bench] +fn bench_take_read(b: &mut test::Bencher) { + b.iter(|| { + let mut buf = [0; 64]; + + [255; 128].take(64).read(&mut buf).unwrap(); + }); +} + +#[bench] +fn bench_take_read_buf(b: &mut test::Bencher) { + b.iter(|| { + let mut buf = [MaybeUninit::uninit(); 64]; + + let mut rbuf = ReadBuf::uninit(&mut buf); + + [255; 128].take(64).read_buf(&mut rbuf).unwrap(); + }); +} diff --git a/library/std/src/io/util.rs b/library/std/src/io/util.rs new file mode 100644 index 000000000..c1300cd67 --- /dev/null +++ b/library/std/src/io/util.rs @@ -0,0 +1,270 @@ +#![allow(missing_copy_implementations)] + +#[cfg(test)] +mod tests; + +use crate::fmt; +use crate::io::{ + self, BufRead, IoSlice, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, SizeHint, Write, +}; + +/// A reader which is always at EOF. +/// +/// This struct is generally created by calling [`empty()`]. Please see +/// the documentation of [`empty()`] for more details. +#[stable(feature = "rust1", since = "1.0.0")] +#[non_exhaustive] +#[derive(Copy, Clone, Default)] +pub struct Empty; + +/// Constructs a new handle to an empty reader. +/// +/// All reads from the returned reader will return <code>[Ok]\(0)</code>. +/// +/// # Examples +/// +/// A slightly sad example of not reading anything into a buffer: +/// +/// ``` +/// use std::io::{self, Read}; +/// +/// let mut buffer = String::new(); +/// io::empty().read_to_string(&mut buffer).unwrap(); +/// assert!(buffer.is_empty()); +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] +pub const fn empty() -> Empty { + Empty +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for Empty { + #[inline] + fn read(&mut self, _buf: &mut [u8]) -> io::Result<usize> { + Ok(0) + } + + #[inline] + fn read_buf(&mut self, _buf: &mut ReadBuf<'_>) -> io::Result<()> { + Ok(()) + } +} +#[stable(feature = "rust1", since = "1.0.0")] +impl BufRead for Empty { + #[inline] + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Ok(&[]) + } + #[inline] + fn consume(&mut self, _n: usize) {} +} + +#[stable(feature = "empty_seek", since = "1.51.0")] +impl Seek for Empty { + fn seek(&mut self, _pos: SeekFrom) -> io::Result<u64> { + Ok(0) + } + + fn stream_len(&mut self) -> io::Result<u64> { + Ok(0) + } + + fn stream_position(&mut self) -> io::Result<u64> { + Ok(0) + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Empty { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Empty").finish_non_exhaustive() + } +} + +impl SizeHint for Empty { + #[inline] + fn upper_bound(&self) -> Option<usize> { + Some(0) + } +} + +/// A reader which yields one byte over and over and over and over and over and... +/// +/// This struct is generally created by calling [`repeat()`]. Please +/// see the documentation of [`repeat()`] for more details. +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Repeat { + byte: u8, +} + +/// Creates an instance of a reader that infinitely repeats one byte. +/// +/// All reads from this reader will succeed by filling the specified buffer with +/// the given byte. +/// +/// # Examples +/// +/// ``` +/// use std::io::{self, Read}; +/// +/// let mut buffer = [0; 3]; +/// io::repeat(0b101).read_exact(&mut buffer).unwrap(); +/// assert_eq!(buffer, [0b101, 0b101, 0b101]); +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] +pub const fn repeat(byte: u8) -> Repeat { + Repeat { byte } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Read for Repeat { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> { + for slot in &mut *buf { + *slot = self.byte; + } + Ok(buf.len()) + } + + fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + // SAFETY: No uninit bytes are being written + for slot in unsafe { buf.unfilled_mut() } { + slot.write(self.byte); + } + + let remaining = buf.remaining(); + + // SAFETY: the entire unfilled portion of buf has been initialized + unsafe { + buf.assume_init(remaining); + } + + buf.add_filled(remaining); + + Ok(()) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { + let mut nwritten = 0; + for buf in bufs { + nwritten += self.read(buf)?; + } + Ok(nwritten) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + true + } +} + +impl SizeHint for Repeat { + #[inline] + fn lower_bound(&self) -> usize { + usize::MAX + } + + #[inline] + fn upper_bound(&self) -> Option<usize> { + None + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Repeat { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Repeat").finish_non_exhaustive() + } +} + +/// A writer which will move data into the void. +/// +/// This struct is generally created by calling [`sink`]. Please +/// see the documentation of [`sink()`] for more details. +#[stable(feature = "rust1", since = "1.0.0")] +#[non_exhaustive] +#[derive(Copy, Clone, Default)] +pub struct Sink; + +/// Creates an instance of a writer which will successfully consume all data. +/// +/// All calls to [`write`] on the returned instance will return `Ok(buf.len())` +/// and the contents of the buffer will not be inspected. +/// +/// [`write`]: Write::write +/// +/// # Examples +/// +/// ```rust +/// use std::io::{self, Write}; +/// +/// let buffer = vec![1, 2, 3, 5, 8]; +/// let num_bytes = io::sink().write(&buffer).unwrap(); +/// assert_eq!(num_bytes, 5); +/// ``` +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_io_structs", issue = "78812")] +pub const fn sink() -> Sink { + Sink +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Write for Sink { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let total_len = bufs.iter().map(|b| b.len()).sum(); + Ok(total_len) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "write_mt", since = "1.48.0")] +impl Write for &Sink { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result<usize> { + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result<usize> { + let total_len = bufs.iter().map(|b| b.len()).sum(); + Ok(total_len) + } + + #[inline] + fn is_write_vectored(&self) -> bool { + true + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Sink { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Sink").finish_non_exhaustive() + } +} diff --git a/library/std/src/io/util/tests.rs b/library/std/src/io/util/tests.rs new file mode 100644 index 000000000..08972a59a --- /dev/null +++ b/library/std/src/io/util/tests.rs @@ -0,0 +1,147 @@ +use crate::cmp::{max, min}; +use crate::io::prelude::*; +use crate::io::{ + copy, empty, repeat, sink, BufWriter, Empty, ReadBuf, Repeat, Result, SeekFrom, Sink, + DEFAULT_BUF_SIZE, +}; + +use crate::mem::MaybeUninit; + +#[test] +fn copy_copies() { + let mut r = repeat(0).take(4); + let mut w = sink(); + assert_eq!(copy(&mut r, &mut w).unwrap(), 4); + + let mut r = repeat(0).take(1 << 17); + assert_eq!(copy(&mut r as &mut dyn Read, &mut w as &mut dyn Write).unwrap(), 1 << 17); +} + +struct ShortReader { + cap: usize, + read_size: usize, + observed_buffer: usize, +} + +impl Read for ShortReader { + fn read(&mut self, buf: &mut [u8]) -> Result<usize> { + let bytes = min(self.cap, self.read_size); + self.cap -= bytes; + self.observed_buffer = max(self.observed_buffer, buf.len()); + Ok(bytes) + } +} + +struct WriteObserver { + observed_buffer: usize, +} + +impl Write for WriteObserver { + fn write(&mut self, buf: &[u8]) -> Result<usize> { + self.observed_buffer = max(self.observed_buffer, buf.len()); + Ok(buf.len()) + } + + fn flush(&mut self) -> Result<()> { + Ok(()) + } +} + +#[test] +fn copy_specializes_bufwriter() { + let cap = 117 * 1024; + let buf_sz = 16 * 1024; + let mut r = ShortReader { cap, observed_buffer: 0, read_size: 1337 }; + let mut w = BufWriter::with_capacity(buf_sz, WriteObserver { observed_buffer: 0 }); + assert_eq!( + copy(&mut r, &mut w).unwrap(), + cap as u64, + "expected the whole capacity to be copied" + ); + assert_eq!(r.observed_buffer, buf_sz, "expected a large buffer to be provided to the reader"); + assert!(w.get_mut().observed_buffer > DEFAULT_BUF_SIZE, "expected coalesced writes"); +} + +#[test] +fn sink_sinks() { + let mut s = sink(); + assert_eq!(s.write(&[]).unwrap(), 0); + assert_eq!(s.write(&[0]).unwrap(), 1); + assert_eq!(s.write(&[0; 1024]).unwrap(), 1024); + assert_eq!(s.by_ref().write(&[0; 1024]).unwrap(), 1024); +} + +#[test] +fn empty_reads() { + let mut e = empty(); + assert_eq!(e.read(&mut []).unwrap(), 0); + assert_eq!(e.read(&mut [0]).unwrap(), 0); + assert_eq!(e.read(&mut [0; 1024]).unwrap(), 0); + assert_eq!(e.by_ref().read(&mut [0; 1024]).unwrap(), 0); + + let mut buf = []; + let mut buf = ReadBuf::uninit(&mut buf); + e.read_buf(&mut buf).unwrap(); + assert_eq!(buf.filled_len(), 0); + assert_eq!(buf.initialized_len(), 0); + + let mut buf = [MaybeUninit::uninit()]; + let mut buf = ReadBuf::uninit(&mut buf); + e.read_buf(&mut buf).unwrap(); + assert_eq!(buf.filled_len(), 0); + assert_eq!(buf.initialized_len(), 0); + + let mut buf = [MaybeUninit::uninit(); 1024]; + let mut buf = ReadBuf::uninit(&mut buf); + e.read_buf(&mut buf).unwrap(); + assert_eq!(buf.filled_len(), 0); + assert_eq!(buf.initialized_len(), 0); + + let mut buf = [MaybeUninit::uninit(); 1024]; + let mut buf = ReadBuf::uninit(&mut buf); + e.by_ref().read_buf(&mut buf).unwrap(); + assert_eq!(buf.filled_len(), 0); + assert_eq!(buf.initialized_len(), 0); +} + +#[test] +fn empty_seeks() { + let mut e = empty(); + assert!(matches!(e.seek(SeekFrom::Start(0)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Start(1)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Start(u64::MAX)), Ok(0))); + + assert!(matches!(e.seek(SeekFrom::End(i64::MIN)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::End(-1)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::End(0)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::End(1)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::End(i64::MAX)), Ok(0))); + + assert!(matches!(e.seek(SeekFrom::Current(i64::MIN)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Current(-1)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Current(0)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Current(1)), Ok(0))); + assert!(matches!(e.seek(SeekFrom::Current(i64::MAX)), Ok(0))); +} + +#[test] +fn repeat_repeats() { + let mut r = repeat(4); + let mut b = [0; 1024]; + assert_eq!(r.read(&mut b).unwrap(), 1024); + assert!(b.iter().all(|b| *b == 4)); +} + +#[test] +fn take_some_bytes() { + assert_eq!(repeat(4).take(100).bytes().count(), 100); + assert_eq!(repeat(4).take(100).bytes().next().unwrap().unwrap(), 4); + assert_eq!(repeat(1).take(10).chain(repeat(2).take(10)).bytes().count(), 20); +} + +#[allow(dead_code)] +fn const_utils() { + const _: Empty = empty(); + const _: Repeat = repeat(b'c'); + const _: Sink = sink(); +} |