commit e9d3edf0272502b0dd6c26e8a4869c2912657615 Author: Simon Renblad Date: Wed Dec 18 13:43:46 2024 +0800 add patch for 29ef6cf diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..96ef6c0 --- /dev/null +++ b/.gitignore @@ -0,0 +1,2 @@ +/target +Cargo.lock diff --git a/Cargo.toml b/Cargo.toml new file mode 100644 index 0000000..0a0c430 --- /dev/null +++ b/Cargo.toml @@ -0,0 +1,10 @@ +[package] +name = "core_io" +version = "0.1.0" +edition = "2018" + +# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html + +[features] +alloc = [] +collections = ["alloc"] diff --git a/src/buffered/bufreader.rs b/src/buffered/bufreader.rs new file mode 100644 index 0000000..da550fa --- /dev/null +++ b/src/buffered/bufreader.rs @@ -0,0 +1,438 @@ +use core::cmp; +use core::fmt; +use crate::{ + self as io, BufRead, Initializer, IoSliceMut, Read, Seek, SeekFrom, SizeHint, DEFAULT_BUF_SIZE, +}; +use crate::prelude::*; + +/// The `BufReader` struct adds buffering to any reader. +/// +/// It can be excessively inefficient to work directly with a [`Read`] instance. +/// For example, every call to [`read`][`TcpStream::read`] on [`TcpStream`] +/// results in a system call. A `BufReader` performs large, infrequent reads on +/// the underlying [`Read`] and maintains an in-memory buffer of the results. +/// +/// `BufReader` can improve the speed of programs that make *small* and +/// *repeated* read calls to the same file or network socket. It does not +/// help when reading very large amounts at once, or reading just one or a few +/// times. It also provides no advantage when reading from a source that is +/// already in memory, like a [`Vec`]``. +/// +/// When the `BufReader` is dropped, the contents of its buffer will be +/// discarded. Creating multiple instances of a `BufReader` on the same +/// stream can cause data loss. Reading from the underlying reader after +/// unwrapping the `BufReader` with [`BufReader::into_inner`] can also cause +/// data loss. +/// +// HACK(#78696): can't use `crate` for associated items +/// [`TcpStream::read`]: super::super::super::net::TcpStream::read +/// [`TcpStream`]: crate::net::TcpStream +/// +/// # Examples +/// +/// ```no_run +/// use std::io::prelude::*; +/// use std::io::BufReader; +/// use std::fs::File; +/// +/// fn main() -> std::io::Result<()> { +/// let f = File::open("log.txt")?; +/// let mut reader = BufReader::new(f); +/// +/// let mut line = String::new(); +/// let len = reader.read_line(&mut line)?; +/// println!("First line is {} bytes long", len); +/// Ok(()) +/// } +/// ``` +pub struct BufReader { + inner: R, + buf: Box<[u8]>, + pos: usize, + cap: usize, +} + +impl BufReader { + /// Creates a new `BufReader` with a default buffer capacity. The default is currently 8 KB, + /// but may change in the future. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let reader = BufReader::new(f); + /// Ok(()) + /// } + /// ``` + pub fn new(inner: R) -> BufReader { + BufReader::with_capacity(DEFAULT_BUF_SIZE, inner) + } + + /// Creates a new `BufReader` with the specified buffer capacity. + /// + /// # Examples + /// + /// Creating a buffer with ten bytes of capacity: + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let reader = BufReader::with_capacity(10, f); + /// Ok(()) + /// } + /// ``` + pub fn with_capacity(capacity: usize, inner: R) -> BufReader { + unsafe { + let mut buf = Box::new_uninit_slice(capacity).assume_init(); + inner.initializer().initialize(&mut buf); + BufReader { inner, buf, pos: 0, cap: 0 } + } + } +} + +impl BufReader { + /// Gets a reference to the underlying reader. + /// + /// It is inadvisable to directly read from the underlying reader. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f1 = File::open("log.txt")?; + /// let reader = BufReader::new(f1); + /// + /// let f2 = reader.get_ref(); + /// Ok(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + pub fn buffer(&self) -> &[u8] { + &self.buf[self.pos..self.cap] + } + + /// Returns the number of bytes the internal buffer can hold at once. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::{BufReader, BufRead}; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f = File::open("log.txt")?; + /// let mut reader = BufReader::new(f); + /// + /// let capacity = reader.capacity(); + /// let buffer = reader.fill_buf()?; + /// assert!(buffer.len() <= capacity); + /// Ok(()) + /// } + /// ``` + pub fn capacity(&self) -> usize { + self.buf.len() + } + + /// Unwraps this `BufReader`, returning the underlying reader. + /// + /// Note that any leftover data in the internal buffer is lost. Therefore, + /// a following read from the underlying reader may lead to data loss. + /// + /// # Examples + /// + /// ```no_run + /// use std::io::BufReader; + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let f1 = File::open("log.txt")?; + /// let reader = BufReader::new(f1); + /// + /// let f2 = reader.into_inner(); + /// Ok(()) + /// } + /// ``` + pub fn into_inner(self) -> R { + self.inner + } + + /// Invalidates all data in the internal buffer. + #[inline] + fn discard_buffer(&mut self) { + self.pos = 0; + self.cap = 0; + } +} + +impl BufReader { + /// Seeks relative to the current position. If the new position lies within the buffer, + /// the buffer will not be flushed, allowing for more efficient seeks. + /// This method does not return the location of the underlying reader, so the caller + /// must track this information themselves if it is required. + pub fn seek_relative(&mut self, offset: i64) -> io::Result<()> { + let pos = self.pos as u64; + if offset < 0 { + if let Some(new_pos) = pos.checked_sub((-offset) as u64) { + self.pos = new_pos as usize; + return Ok(()); + } + } else { + if let Some(new_pos) = pos.checked_add(offset as u64) { + if new_pos <= self.cap as u64 { + self.pos = new_pos as usize; + return Ok(()); + } + } + } + self.seek(SeekFrom::Current(offset)).map(drop) + } +} + +impl Read for BufReader { + fn read(&mut self, buf: &mut [u8]) -> io::Result { + // If we don't have any buffered data and we're doing a massive read + // (larger than our internal buffer), bypass our internal buffer + // entirely. + if self.pos == self.cap && buf.len() >= self.buf.len() { + self.discard_buffer(); + return self.inner.read(buf); + } + let nread = { + let mut rem = self.fill_buf()?; + rem.read(buf)? + }; + self.consume(nread); + Ok(nread) + } + + // Small read_exacts from a BufReader are extremely common when used with a deserializer. + // The default implementation calls read in a loop, which results in surprisingly poor code + // generation for the common path where the buffer has enough bytes to fill the passed-in + // buffer. + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + if self.buffer().len() >= buf.len() { + buf.copy_from_slice(&self.buffer()[..buf.len()]); + self.consume(buf.len()); + return Ok(()); + } + + crate::default_read_exact(self, buf) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + let total_len = bufs.iter().map(|b| b.len()).sum::(); + if self.pos == self.cap && total_len >= self.buf.len() { + self.discard_buffer(); + return self.inner.read_vectored(bufs); + } + let nread = { + let mut rem = self.fill_buf()?; + rem.read_vectored(bufs)? + }; + self.consume(nread); + Ok(nread) + } + + fn is_read_vectored(&self) -> bool { + self.inner.is_read_vectored() + } + + // we can't skip unconditionally because of the large buffer case in read. + unsafe fn initializer(&self) -> Initializer { + self.inner.initializer() + } +} + +impl BufRead for BufReader { + fn fill_buf(&mut self) -> io::Result<&[u8]> { + // If we've reached the end of our internal buffer then we need to fetch + // some more data from the underlying reader. + // Branch using `>=` instead of the more correct `==` + // to tell the compiler that the pos..cap slice is always valid. + if self.pos >= self.cap { + debug_assert!(self.pos == self.cap); + self.cap = self.inner.read(&mut self.buf)?; + self.pos = 0; + } + Ok(&self.buf[self.pos..self.cap]) + } + + fn consume(&mut self, amt: usize) { + self.pos = cmp::min(self.pos + amt, self.cap); + } +} + +impl fmt::Debug for BufReader +where + R: fmt::Debug, +{ + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.debug_struct("BufReader") + .field("reader", &self.inner) + .field("buffer", &format_args!("{}/{}", self.cap - self.pos, self.buf.len())) + .finish() + } +} + +impl Seek for BufReader { + /// Seek to an offset, in bytes, in the underlying reader. + /// + /// The position used for seeking with [`SeekFrom::Current`]`(_)` is the + /// position the underlying reader would be at if the `BufReader` had no + /// internal buffer. + /// + /// Seeking always discards the internal buffer, even if the seek position + /// would otherwise fall within it. This guarantees that calling + /// [`BufReader::into_inner()`] immediately after a seek yields the underlying reader + /// at the same position. + /// + /// To seek without discarding the internal buffer, use [`BufReader::seek_relative`]. + /// + /// See [`std::io::Seek`] for more details. + /// + /// Note: In the edge case where you're seeking with [`SeekFrom::Current`]`(n)` + /// where `n` minus the internal buffer length overflows an `i64`, two + /// seeks will be performed instead of one. If the second seek returns + /// [`Err`], the underlying reader will be left at the same position it would + /// have if you called `seek` with [`SeekFrom::Current`]`(0)`. + /// + /// [`std::io::Seek`]: Seek + fn seek(&mut self, pos: SeekFrom) -> io::Result { + let result: u64; + if let SeekFrom::Current(n) = pos { + let remainder = (self.cap - self.pos) as i64; + // it should be safe to assume that remainder fits within an i64 as the alternative + // means we managed to allocate 8 exbibytes and that's absurd. + // But it's not out of the realm of possibility for some weird underlying reader to + // support seeking by i64::MIN so we need to handle underflow when subtracting + // remainder. + if let Some(offset) = n.checked_sub(remainder) { + result = self.inner.seek(SeekFrom::Current(offset))?; + } else { + // seek backwards by our remainder, and then by the offset + self.inner.seek(SeekFrom::Current(-remainder))?; + self.discard_buffer(); + result = self.inner.seek(SeekFrom::Current(n))?; + } + } else { + // Seeking with Start/End doesn't care about our buffer length. + result = self.inner.seek(pos)?; + } + self.discard_buffer(); + Ok(result) + } + + /// Returns the current seek position from the start of the stream. + /// + /// The value returned is equivalent to `self.seek(SeekFrom::Current(0))` + /// but does not flush the internal buffer. Due to this optimization the + /// function does not guarantee that calling `.into_inner()` immediately + /// afterwards will yield the underlying reader at the same position. Use + /// [`BufReader::seek`] instead if you require that guarantee. + /// + /// # Panics + /// + /// This function will panic if the position of the inner reader is smaller + /// than the amount of buffered data. That can happen if the inner reader + /// has an incorrect implementation of [`Seek::stream_position`], or if the + /// position has gone out of sync due to calling [`Seek::seek`] directly on + /// the underlying reader. + /// + /// # Example + /// + /// ```no_run + /// use std::{ + /// io::{self, BufRead, BufReader, Seek}, + /// fs::File, + /// }; + /// + /// fn main() -> io::Result<()> { + /// let mut f = BufReader::new(File::open("foo.txt")?); + /// + /// let before = f.stream_position()?; + /// f.read_line(&mut String::new())?; + /// let after = f.stream_position()?; + /// + /// println!("The first line was {} bytes long", after - before); + /// Ok(()) + /// } + /// ``` + fn stream_position(&mut self) -> io::Result { + let remainder = (self.cap - self.pos) as u64; + self.inner.stream_position().map(|pos| { + pos.checked_sub(remainder).expect( + "overflow when subtracting remaining buffer size from inner stream position", + ) + }) + } +} + +impl SizeHint for BufReader { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(self.get_ref()) + self.buffer().len() + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(self.get_ref()).and_then(|up| self.buffer().len().checked_add(up)) + } +} diff --git a/src/buffered/bufwriter.rs b/src/buffered/bufwriter.rs new file mode 100644 index 0000000..59f3b52 --- /dev/null +++ b/src/buffered/bufwriter.rs @@ -0,0 +1,648 @@ +use core::fmt; +use crate::{ + self as io, Error, ErrorKind, IntoInnerError, IoSlice, Seek, SeekFrom, Write, DEFAULT_BUF_SIZE, +}; +use core::ptr; +use core::mem; +use crate::Vec; + +/// 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` keeps an in-memory buffer of data and writes it to an underlying +/// writer in large, infrequent batches. +/// +/// `BufWriter` 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 [`Vec`]``. +/// +/// It is critical to call [`flush`] before `BufWriter` 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`: +/// +/// ```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`, 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 +pub struct BufWriter { + 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, + // #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 BufWriter { + /// Creates a new `BufWriter` 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()); + /// ``` + pub fn new(inner: W) -> BufWriter { + BufWriter::with_capacity(DEFAULT_BUF_SIZE, inner) + } + + /// Creates a new `BufWriter` with the specified buffer capacity. + /// + /// # Examples + /// + /// Creating a buffer with a buffer of 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); + /// ``` + pub fn with_capacity(capacity: usize, inner: W) -> BufWriter { + 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(crate) 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, + written: usize, + } + + impl<'a> BufGuard<'a> { + fn new(buffer: &'a mut Vec) -> 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(Error::new_const( + 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(); + /// ``` + 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(); + /// ``` + 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(); + /// ``` + 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(crate) fn buffer_mut(&mut self) -> &mut Vec { + &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(); + /// ``` + pub fn capacity(&self) -> usize { + self.buf.capacity() + } + + /// Unwraps this `BufWriter`, 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(); + /// ``` + pub fn into_inner(mut self) -> Result>> { + match self.flush_buf() { + Err(e) => Err(IntoInnerError::new(self, e)), + Ok(()) => Ok(self.into_parts().0), + } + } + + /// Disassembles this `BufWriter`, 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"); + /// ``` + pub fn into_parts(mut self) -> (W, Result, 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 { + 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() + } +} + +/// 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 { 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, +} + +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. + pub fn into_inner(self) -> Vec { + self.buf + } + + const DESCRIPTION: &'static str = + "BufWriter inner writer panicked, what data remains unwritten is not known"; +} + +impl fmt::Display for WriterPanicked { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "{}", Self::DESCRIPTION) + } +} + +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() + } +} + +impl Write for BufWriter { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + // 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 { + // 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()) + } +} + +impl fmt::Debug for BufWriter +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() + } +} + +impl Seek for BufWriter { + /// 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 { + self.flush_buf()?; + self.get_mut().seek(pos) + } +} + +impl Drop for BufWriter { + 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/src/buffered/linewriter.rs b/src/buffered/linewriter.rs new file mode 100644 index 0000000..5f154f2 --- /dev/null +++ b/src/buffered/linewriter.rs @@ -0,0 +1,224 @@ +use core::fmt; +use crate::{self as io, 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(()) +/// } +/// ``` +pub struct LineWriter { + inner: BufWriter, +} + +impl LineWriter { + /// 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(()) + /// } + /// ``` + pub fn new(inner: W) -> LineWriter { + // Lines typically aren't that long, don't use a giant buffer + LineWriter::with_capacity(1024, inner) + } + + /// Creates a new `LineWriter` with a 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(()) + /// } + /// ``` + pub fn with_capacity(capacity: usize, inner: W) -> LineWriter { + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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 = LineWriter::new(file); + /// + /// let file: File = writer.into_inner()?; + /// Ok(()) + /// } + /// ``` + pub fn into_inner(self) -> Result>> { + self.inner.into_inner().map_err(|err| err.new_wrapped(|inner| LineWriter { inner })) + } +} + +impl Write for LineWriter { + fn write(&mut self, buf: &[u8]) -> io::Result { + 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 { + 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) + } +} + +impl fmt::Debug for LineWriter +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/src/buffered/linewritershim.rs b/src/buffered/linewritershim.rs new file mode 100644 index 0000000..8dd7022 --- /dev/null +++ b/src/buffered/linewritershim.rs @@ -0,0 +1,276 @@ +use crate::{self as io, BufWriter, IoSlice, Write}; +use core::slice::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, +} + +impl<'a, W: Write> LineWriterShim<'a, W> { + pub fn new(buffer: &'a mut BufWriter) -> 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 { + 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 { + // 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/src/buffered/mod.rs b/src/buffered/mod.rs new file mode 100644 index 0000000..dffbc94 --- /dev/null +++ b/src/buffered/mod.rs @@ -0,0 +1,179 @@ +//! Buffering wrappers for I/O traits + +mod bufreader; +mod bufwriter; +mod linewriter; +mod linewritershim; + +#[cfg(test)] +mod tests; + +use core::fmt; +use crate::Error; + +pub use self::bufreader::BufReader; +pub use self::bufwriter::BufWriter; +pub use self::bufwriter::WriterPanicked; +pub use self::linewriter::LineWriter; +use self::linewritershim::LineWriterShim; + +/// 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)] +pub struct IntoInnerError(W, Error); + +impl IntoInnerError { + /// 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(self, f: impl FnOnce(W) -> W2) -> IntoInnerError { + 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."); + /// } + /// }; + /// ``` + 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() + /// } + /// }; + /// ``` + 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); + /// ``` + 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"); + /// ``` + pub fn into_parts(self) -> (Error, W) { + (self.1, self.0) + } +} + +impl From> for Error { + fn from(iie: IntoInnerError) -> Error { + iie.1 + } +} + +impl fmt::Display for IntoInnerError { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.error().fmt(f) + } +} diff --git a/src/buffered/tests.rs b/src/buffered/tests.rs new file mode 100644 index 0000000..f6c2b49 --- /dev/null +++ b/src/buffered/tests.rs @@ -0,0 +1,970 @@ +use crate::io::prelude::*; +use crate::io::{self, BufReader, BufWriter, ErrorKind, IoSlice, LineWriter, SeekFrom}; +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, +} + +// 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 { + 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_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 { + 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 { + 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 { + for x in &mut *buf { + *x = 0; + } + Ok(buf.len()) + } + } + impl Seek for ErrAfterFirstSeekReader { + fn seek(&mut self, _: SeekFrom) -> io::Result { + 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_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 { + 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 { + 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::>(); + b.iter(|| { + let mut reader = BufReader::new(&data[..]); + let mut buf = [0u8; 4]; + for _ in 0..1024 { + reader.read_exact(&mut buf).unwrap(); + } + }); +} + +#[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, + + // Flush sets this flag + pub flushed: bool, + + // 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, + + // If set, counts down with each write, and writes return an error + // when it hits 0 + pub max_writes: Option, + + // 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 { + 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 { + self.flushed = true; + 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 }, + Flush { output: io::Result<()> }, + } + + #[derive(Default)] + struct Writer { + calls: VecDeque, + } + + impl Write for Writer { + fn write(&mut self, buf: &[u8]) -> io::Result { + self.write_vectored(&[IoSlice::new(buf)]) + } + + fn write_vectored(&mut self, buf: &[IoSlice<'_>]) -> io::Result { + match self.calls.pop_front().expect("unexpected call to write") { + Call::Write { inputs, output } => { + assert_eq!(inputs, buf.iter().map(|b| &**b).collect::>()); + 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, +} + +impl Write for WriteRecorder { + fn write(&mut self, buf: &[u8]) -> io::Result { + 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/src/copy.rs b/src/copy.rs new file mode 100644 index 0000000..df99a47 --- /dev/null +++ b/src/copy.rs @@ -0,0 +1,151 @@ +use super::{ErrorKind, Read, Result, Write, DEFAULT_BUF_SIZE}; +#[cfg(feature = "collections")] use super::BufWriter; +use core::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 = vec![]; +/// +/// io::copy(&mut reader, &mut writer)?; +/// +/// assert_eq!(&b"hello"[..], &writer[..]); +/// Ok(()) +/// } +/// ``` +pub fn copy(reader: &mut R, writer: &mut W) -> Result +where + R: Read, + W: Write, +{ + 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(reader: &mut R, writer: &mut W) -> Result +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(reader: &mut R, writer: &mut Self) -> Result; +} + +impl BufferedCopySpec for W { + default fn copy_to(reader: &mut R, writer: &mut Self) -> Result { + stack_buffer_copy(reader, writer) + } +} + +#[cfg(feature = "collections")] +impl BufferedCopySpec for BufWriter { + fn copy_to(reader: &mut R, writer: &mut Self) -> Result { + if writer.capacity() < DEFAULT_BUF_SIZE { + return stack_buffer_copy(reader, writer); + } + + // FIXME: #42788 + // + // - This creates a (mut) reference to a slice of + // _uninitialized_ integers, which is **undefined behavior** + // + // - Only the standard library gets to soundly "ignore" this, + // based on its privileged knowledge of unstable rustc + // internals; + unsafe { + let spare_cap = writer.buffer_mut().spare_capacity_mut(); + reader.initializer().initialize(MaybeUninit::slice_assume_init_mut(spare_cap)); + } + + let mut len = 0; + + loop { + let buf = writer.buffer_mut(); + let spare_cap = buf.spare_capacity_mut(); + + if spare_cap.len() >= DEFAULT_BUF_SIZE { + match reader.read(unsafe { MaybeUninit::slice_assume_init_mut(spare_cap) }) { + Ok(0) => return Ok(len), // EOF reached + Ok(bytes_read) => { + assert!(bytes_read <= spare_cap.len()); + // SAFETY: The initializer contract guarantees that either it or `read` + // will have initialized these bytes. And we just checked that the number + // of bytes is within the buffer capacity. + 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( + reader: &mut R, + writer: &mut W, +) -> Result { + let mut buf = MaybeUninit::<[u8; DEFAULT_BUF_SIZE]>::uninit(); + // FIXME: #42788 + // + // - This creates a (mut) reference to a slice of + // _uninitialized_ integers, which is **undefined behavior** + // + // - Only the standard library gets to soundly "ignore" this, + // based on its privileged knowledge of unstable rustc + // internals; + unsafe { + reader.initializer().initialize(buf.assume_init_mut()); + } + + let mut written = 0; + loop { + let len = match reader.read(unsafe { buf.assume_init_mut() }) { + Ok(0) => return Ok(written), + Ok(len) => len, + Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(e) => return Err(e), + }; + writer.write_all(unsafe { &buf.assume_init_ref()[..len] })?; + written += len as u64; + } +} diff --git a/src/cursor.rs b/src/cursor.rs new file mode 100644 index 0000000..7bea077 --- /dev/null +++ b/src/cursor.rs @@ -0,0 +1,515 @@ +#[cfg(test)] +mod tests; + +use crate::prelude::*; + +use core::cmp; +use crate::{self as io, Error, ErrorKind, Initializer, IoSlice, IoSliceMut, SeekFrom}; + +#[cfg(feature="collections")] use core::convert::TryInto; +#[cfg(feature="collections")] use crate::Vec; + + +/// A `Cursor` wraps an in-memory buffer and provides it with a +/// [`Seek`] implementation. +/// +/// `Cursor`s are used with in-memory buffers, anything implementing +/// [`AsRef`]`<[u8]>`, 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 `Cursor<`[`Vec`]`>` and +/// `Cursor<`[`&[u8]`][bytes]`>`. +/// +/// # 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 +/// [`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(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]); +/// } +/// ``` +#[derive(Debug, Default, Eq, PartialEq)] +pub struct Cursor { + inner: T, + pos: u64, +} + +impl Cursor { + /// 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>) {} + /// # force_inference(&buff); + /// ``` + pub const fn new(inner: T) -> Cursor { + 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>) {} + /// # force_inference(&buff); + /// + /// let vec = buff.into_inner(); + /// ``` + 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>) {} + /// # force_inference(&buff); + /// + /// let reference = buff.get_ref(); + /// ``` + 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>) {} + /// # force_inference(&buff); + /// + /// let reference = buff.get_mut(); + /// ``` + 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); + /// ``` + 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); + /// ``` + pub fn set_position(&mut self, pos: u64) { + self.pos = pos; + } +} + +impl Cursor +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(), &[]); + /// ``` + 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()); + /// ``` + pub fn is_empty(&self) -> bool { + self.pos >= self.inner.as_ref().len() as u64 + } +} + +impl Clone for Cursor +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; + } +} + +impl io::Seek for Cursor +where + T: AsRef<[u8]>, +{ + fn seek(&mut self, style: SeekFrom) -> io::Result { + 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), + }; + let new_pos = if offset >= 0 { + base_pos.checked_add(offset as u64) + } else { + base_pos.checked_sub((offset.wrapping_neg()) as u64) + }; + match new_pos { + Some(n) => { + self.pos = n; + Ok(self.pos) + } + None => Err(Error::new_const( + ErrorKind::InvalidInput, + &"invalid seek to a negative or overflowing position", + )), + } + } + + fn stream_len(&mut self) -> io::Result { + Ok(self.inner.as_ref().len() as u64) + } + + fn stream_position(&mut self) -> io::Result { + Ok(self.pos) + } +} + +impl Read for Cursor +where + T: AsRef<[u8]>, +{ + fn read(&mut self, buf: &mut [u8]) -> io::Result { + let n = Read::read(&mut self.remaining_slice(), buf)?; + self.pos += n as u64; + Ok(n) + } + + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + 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(()) + } + + #[inline] + unsafe fn initializer(&self) -> Initializer { + Initializer::nop() + } +} + +#[cfg(feature="collections")] +impl BufRead for Cursor +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 { + 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 { + 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) +} + +// Resizing write implementation +#[cfg(feature="collections")] +fn vec_write(pos_mut: &mut u64, vec: &mut Vec, buf: &[u8]) -> io::Result { + let pos: usize = (*pos_mut).try_into().map_err(|_| { + Error::new_const( + ErrorKind::InvalidInput, + &"cursor position exceeds maximum possible vector length", + ) + })?; + // Make sure the internal buffer is as least as big as where we + // currently are + let len = vec.len(); + if len < pos { + // use `resize` so that the zero filling is as efficient as possible + vec.resize(pos, 0); + } + // Figure out what bytes will be used to overwrite what's currently + // there (left), and what will be appended on the end (right) + { + let space = vec.len() - pos; + let (left, right) = buf.split_at(cmp::min(space, buf.len())); + vec[pos..pos + left.len()].copy_from_slice(left); + vec.extend_from_slice(right); + } + + // Bump us forward + *pos_mut = (pos + buf.len()) as u64; + Ok(buf.len()) +} + +#[cfg(feature="collections")] +fn vec_write_vectored( + pos_mut: &mut u64, + vec: &mut Vec, + bufs: &[IoSlice<'_>], +) -> io::Result { + let mut nwritten = 0; + for buf in bufs { + nwritten += vec_write(pos_mut, vec, buf)?; + } + Ok(nwritten) +} + +impl Write for Cursor<&mut [u8]> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + slice_write(&mut self.pos, self.inner, buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} + +#[cfg(feature="collections")] +impl Write for Cursor<&mut Vec> { + fn write(&mut self, buf: &[u8]) -> io::Result { + vec_write(&mut self.pos, self.inner, buf) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} + +#[cfg(feature = "collections")] +impl Write for Cursor> { + fn write(&mut self, buf: &[u8]) -> io::Result { + vec_write(&mut self.pos, &mut self.inner, buf) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} + +#[cfg(feature = "alloc")] +impl Write for Cursor<::alloc::boxed::Box<[u8]>> { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + slice_write(&mut self.pos, &mut self.inner, buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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/src/cursor/tests.rs b/src/cursor/tests.rs new file mode 100644 index 0000000..5da31ce --- /dev/null +++ b/src/cursor/tests.rs @@ -0,0 +1,523 @@ +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()); + 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); +} + +#[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); +} + +#[test] +fn test_box_slice_writer() { + let mut writer = Cursor::new(vec![0u8; 9].into_boxed_slice()); + 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(), b); +} + +#[test] +fn test_box_slice_writer_vectored() { + let mut writer = Cursor::new(vec![0u8; 9].into_boxed_slice()); + 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(), b); +} + +#[test] +fn test_buf_writer() { + let mut buf = [0 as u8; 9]; + { + let mut writer = Cursor::new(&mut buf[..]); + 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!(buf, b); +} + +#[test] +fn test_buf_writer_vectored() { + let mut buf = [0 as u8; 9]; + { + let mut writer = Cursor::new(&mut buf[..]); + 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!(buf, b); +} + +#[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::::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::::new()), Cursor::new(Vec::::new())); +} + +#[test] +fn test_eq() { + struct AssertEq(pub T); + + let _: AssertEq>> = 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(); +} diff --git a/src/error.rs b/src/error.rs new file mode 100644 index 0000000..133621b --- /dev/null +++ b/src/error.rs @@ -0,0 +1,662 @@ +#[cfg(test)] +mod tests; + +use core::convert::From; +use core::fmt; +use core::result; + +use core::convert::Into; +use core::marker::{Send, Sync}; +use core::option::Option::{self, Some, None}; +#[cfg(feature="alloc")] use alloc::boxed::Box; +#[cfg(not(feature="alloc"))] use crate::FakeBox as Box; +#[cfg(feature="collections")] use alloc::string::String; +#[cfg(not(feature="collections"))] use crate::ErrorString as String; + +/// 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 { +/// let mut buffer = String::new(); +/// +/// io::stdin().read_line(&mut buffer)?; +/// +/// Ok(buffer) +/// } +/// ``` +pub type Result = result::Result; + +/// 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 +pub struct Error { + repr: Repr, +} + +impl fmt::Debug for Error { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt::Debug::fmt(&self.repr, f) + } +} + +enum Repr { + Os(i32), + Simple(ErrorKind), + // &str is a fat pointer, but &&str is a thin pointer. + SimpleMessage(ErrorKind, &'static &'static str), + #[cfg(feature="alloc")] + Custom(Box), + #[cfg(not(feature="alloc"))] + Custom(Custom), +} + +#[derive(Debug)] +struct Custom { + kind: ErrorKind, + error: String, +} + +/// 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 +#[derive(Clone, Copy, Debug, Eq, Hash, Ord, PartialEq, PartialOrd)] +#[allow(deprecated)] +#[non_exhaustive] +pub enum ErrorKind { + /// An entity was not found, often a file. + NotFound, + /// The operation lacked the necessary privileges to complete. + PermissionDenied, + /// The connection was refused by the remote server. + ConnectionRefused, + /// The connection was reset by the remote server. + ConnectionReset, + /// The remote host is not reachable. + HostUnreachable, + /// The network containing the remote host is not reachable. + NetworkUnreachable, + /// The connection was aborted (terminated) by the remote server. + ConnectionAborted, + /// The network operation failed because it was not connected yet. + NotConnected, + /// A socket address could not be bound because the address is already in + /// use elsewhere. + AddrInUse, + /// A nonexistent interface was requested or the requested address was not + /// local. + AddrNotAvailable, + /// The system's networking is down. + NetworkDown, + /// The operation failed because a pipe was closed. + BrokenPipe, + /// An entity already exists, often a file. + AlreadyExists, + /// The operation needs to block to complete, but the blocking operation was + /// requested to not occur. + WouldBlock, + /// For example, a filesystem path was specified where one of the intermediate directory + /// components was, in fact, a plain file. + NotADirectory, + /// The filesystem object is, unexpectedly, a directory. + /// + /// A directory was specified when a non-directory was expected. + IsADirectory, + /// A non-empty directory was specified where an empty directory was expected. + DirectoryNotEmpty, + /// The filesystem or storage medium is read-only, but a write operation was attempted. + 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. + 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. + StaleNetworkFileHandle, + /// A parameter was incorrect. + 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 + InvalidData, + /// The I/O operation's timeout expired, causing it to be canceled. + 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 + WriteZero, + /// + /// This does not include out of quota errors. + 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`. + NotSeekable, + /// Filesystem quota was exceeded. + 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. + FileTooLarge, + /// Resource is busy. + 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.) + ExecutableFileBusy, + /// Deadlock (avoided). + /// + /// A file locking operation would result in deadlock. This situation is typically detected, if + /// at all, on a best-effort basis. + Deadlock, + /// Cross-device or cross-filesystem (hard) link or rename. + CrossesDevices, + /// Too many (hard) links to the same filesystem object. + /// + /// The filesystem does not support making so many hardlinks to the same file. + TooManyLinks, + /// Filename too long. + /// + /// The limit might be from the underlying filesystem or API, or an administratively imposed + /// resource limit. + FilenameTooLong, + /// 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. + ArgumentListTooLong, + /// This operation was interrupted. + /// + /// Interrupted operations can typically be retried. + Interrupted, + + /// This operation is unsupported on this platform. + /// + /// This means that the operation can never succeed. + 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. + UnexpectedEof, + + /// An operation could not be completed, because it failed + /// to allocate enough memory. + 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. + 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. + #[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", + FilenameTooLong => "filename too long", + FilesystemLoop => "filesystem loop or indirection limit (e.g. symlink loop)", + FilesystemQuotaExceeded => "filesystem quota exceeded", + HostUnreachable => "host unreachable", + Interrupted => "operation interrupted", + InvalidData => "invalid data", + 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", + } + } +} + +/// Intended for use for errors not exposed to the user, where allocating onto +/// the heap (for normal construction via Error::new) is too costly. +impl From for Error { + /// Converts an [`ErrorKind`] into an [`Error`]. + /// + /// This conversion allocates 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::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`]. + /// + /// # 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); + /// ``` + pub fn new(kind: ErrorKind, error: E) -> Error + where + E: Into, + { + Self::_new(kind, error.into()) + } + + fn _new(kind: ErrorKind, error: String) -> Error { + Error { repr: Repr::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. + /// + /// This function should maybe change to + /// `new_const(kind: ErrorKind)` + /// in the future, when const generics allow that. + #[inline] + pub(crate) const fn new_const(kind: ErrorKind, message: &'static &'static str) -> Error { + Self { repr: Repr::SimpleMessage(kind, message) } + } + + /// 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); + /// # } + /// ``` + #[inline] + pub fn from_raw_os_error(code: i32) -> Error { + Error { repr: Repr::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!")); + /// } + /// ``` + #[inline] + pub fn raw_os_error(&self) -> Option { + match self.repr { + Repr::Os(i) => Some(i), + Repr::Custom(..) => None, + Repr::Simple(..) => None, + Repr::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!")); + /// } + /// ``` + #[inline] + pub fn get_ref(&self) -> Option<&String> { + match self.repr { + Repr::Os(..) => None, + Repr::Simple(..) => None, + Repr::SimpleMessage(..) => None, + Repr::Custom(ref 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::().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()))); + /// } + /// ``` + #[inline] + pub fn get_mut(&mut self) -> Option<&mut String> { + match self.repr { + Repr::Os(..) => None, + Repr::Simple(..) => None, + Repr::SimpleMessage(..) => None, + Repr::Custom(ref mut 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!")); + /// } + /// ``` + #[inline] + pub fn into_inner(self) -> Option { + match self.repr { + Repr::Os(..) => None, + Repr::Simple(..) => None, + Repr::SimpleMessage(..) => None, + Repr::Custom(c) => Some(c.error), + } + } + + /// 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!")); + /// } + /// ``` + #[inline] + pub fn kind(&self) -> ErrorKind { + match self.repr { + Repr::Os(_code) => ErrorKind::Other, + Repr::Custom(ref c) => c.kind, + Repr::Simple(kind) => kind, + Repr::SimpleMessage(kind, _) => kind, + } + } +} + +impl fmt::Debug for Repr { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + Repr::Os(code) => fmt + .debug_struct("Os") + .field("code", &code) + .finish(), + Repr::Custom(ref c) => fmt::Debug::fmt(&c, fmt), + Repr::Simple(kind) => fmt.debug_tuple("Kind").field(&kind).finish(), + Repr::SimpleMessage(kind, &message) => { + fmt.debug_struct("Error").field("kind", &kind).field("message", &message).finish() + } + } + } +} + +impl fmt::Display for Error { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + match self.repr { + Repr::Os(code) => { + write!(fmt, "os error {}", code) + } + Repr::Custom(ref c) => c.error.fmt(fmt), + Repr::Simple(kind) => write!(fmt, "{}", kind.as_str()), + Repr::SimpleMessage(_, &msg) => msg.fmt(fmt), + } + } +} + +fn _assert_error_is_sync_send() { + fn _is_sync_send() {} + _is_sync_send::(); +} diff --git a/src/error/tests.rs b/src/error/tests.rs new file mode 100644 index 0000000..5098a46 --- /dev/null +++ b/src/error/tests.rs @@ -0,0 +1,69 @@ +use super::{Custom, Error, ErrorKind, Repr}; +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::() <= 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::Custom(box Custom { + kind: ErrorKind::InvalidInput, + error: box Error { repr: super::Repr::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::()); + assert_eq!("asdf", err.get_ref().unwrap().to_string()); + assert!(err.get_mut().unwrap().is::()); + let extracted = err.into_inner().unwrap(); + extracted.downcast::().unwrap(); +} + +#[test] +fn test_const() { + const E: Error = Error::new_const(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")); +} diff --git a/src/impls.rs b/src/impls.rs new file mode 100644 index 0000000..e789f02 --- /dev/null +++ b/src/impls.rs @@ -0,0 +1,408 @@ +#[cfg(test)] +mod tests; + +#[cfg(feature="collections")] use core::alloc::Allocator; +use core::cmp; +use core::fmt; +use crate::{ + self as io, Error, ErrorKind, Initializer, IoSlice, IoSliceMut, Read, Seek, SeekFrom, Write, +}; +#[cfg(feature="collections")] use crate::BufRead; +use core::mem; + +#[cfg(feature="alloc")] use alloc::boxed::Box; +#[cfg(feature="collections")] use alloc::string::String; +#[cfg(feature="collections")] use alloc::vec::Vec; + +// ============================================================================= +// Forwarding implementations + +impl Read for &mut R { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result { + (**self).read(buf) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + (**self).read_vectored(bufs) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + (**self).is_read_vectored() + } + + #[inline] + unsafe fn initializer(&self) -> Initializer { + (**self).initializer() + } + + #[cfg(feature="collections")] + #[inline] + fn read_to_end(&mut self, buf: &mut Vec) -> io::Result { + (**self).read_to_end(buf) + } + + #[cfg(feature="collections")] + #[inline] + fn read_to_string(&mut self, buf: &mut String) -> io::Result { + (**self).read_to_string(buf) + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + (**self).read_exact(buf) + } +} +impl Write for &mut W { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + (**self).write(buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + (**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) + } +} +impl Seek for &mut S { + #[inline] + fn seek(&mut self, pos: SeekFrom) -> io::Result { + (**self).seek(pos) + } + + #[inline] + fn stream_position(&mut self) -> io::Result { + (**self).stream_position() + } +} +#[cfg(feature="collections")] +impl 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) -> io::Result { + (**self).read_until(byte, buf) + } + + #[inline] + fn read_line(&mut self, buf: &mut String) -> io::Result { + (**self).read_line(buf) + } +} + +#[cfg(feature="alloc")] +impl Read for Box { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result { + (**self).read(buf) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + (**self).read_vectored(bufs) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + (**self).is_read_vectored() + } + + #[inline] + unsafe fn initializer(&self) -> Initializer { + (**self).initializer() + } + + #[cfg(feature="collections")] + #[inline] + fn read_to_end(&mut self, buf: &mut Vec) -> io::Result { + (**self).read_to_end(buf) + } + + #[cfg(feature="collections")] + #[inline] + fn read_to_string(&mut self, buf: &mut String) -> io::Result { + (**self).read_to_string(buf) + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + (**self).read_exact(buf) + } +} +#[cfg(feature="alloc")] +impl Write for Box { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + (**self).write(buf) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + (**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) + } +} +#[cfg(feature="alloc")] +impl Seek for Box { + #[inline] + fn seek(&mut self, pos: SeekFrom) -> io::Result { + (**self).seek(pos) + } + + #[inline] + fn stream_position(&mut self) -> io::Result { + (**self).stream_position() + } +} +#[cfg(feature="collections")] +impl BufRead for Box { + #[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) -> io::Result { + (**self).read_until(byte, buf) + } + + #[inline] + fn read_line(&mut self, buf: &mut String) -> io::Result { + (**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. +impl Read for &[u8] { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result { + 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_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + 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] + unsafe fn initializer(&self) -> Initializer { + Initializer::nop() + } + + #[inline] + fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> { + if buf.len() > self.len() { + return Err(Error::new_const(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(()) + } + + #[cfg(feature="collections")] + #[inline] + fn read_to_end(&mut self, buf: &mut Vec) -> io::Result { + buf.extend_from_slice(*self); + let len = self.len(); + *self = &self[len..]; + Ok(len) + } +} + +#[cfg(feature="collections")] +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`. +impl Write for &mut [u8] { + #[inline] + fn write(&mut self, data: &[u8]) -> io::Result { + 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 { + 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(Error::new_const(ErrorKind::WriteZero, &"failed to write whole buffer")) + } + } + + #[inline] + fn flush(&mut self) -> io::Result<()> { + Ok(()) + } +} + +/// Write is implemented for `Vec` by appending to the vector. +/// The vector will grow as needed. +#[cfg(feature="collections")] +impl Write for Vec { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + self.extend_from_slice(buf); + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} diff --git a/src/impls/tests.rs b/src/impls/tests.rs new file mode 100644 index 0000000..d1cd84a --- /dev/null +++ b/src/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/src/io.rs b/src/io.rs new file mode 100644 index 0000000..ba9587b --- /dev/null +++ b/src/io.rs @@ -0,0 +1,2747 @@ +//! 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`]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`][`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 + +#[cfg(test)] +mod tests; + +use core::cmp; +use core::fmt; +use core::slice::memchr; +use core::ops::{Deref, DerefMut}; +use core::ptr; +use core::slice; +use core::str; +use core::convert::TryInto; +use core::mem::replace; + +#[cfg(feature="collections")] pub use self::buffered::IntoInnerError; +#[cfg(feature="collections")] pub use self::buffered::WriterPanicked; +#[cfg(feature="collections")] pub use self::buffered::{BufReader, BufWriter, LineWriter}; + +pub use self::copy::copy; +pub use self::cursor::Cursor; +pub use self::error::{Error, ErrorKind, Result}; +pub use self::util::{empty, repeat, sink, Empty, Repeat, Sink}; + +#[cfg(feature="collections")] use collections::string::String; +#[cfg(feature="collections")] use collections::vec::Vec; +#[cfg(feature="alloc")] use alloc::boxed::Box; + +#[cfg(feature="collections")] mod buffered; +pub(crate) mod copy; +mod cursor; +mod error; +mod impls; +pub mod prelude; +mod util; + +const DEFAULT_BUF_SIZE: usize = 8 * 1024; + +#[cfg(feature="collections")] +struct Guard<'a> { + buf: &'a mut Vec, + len: usize, +} + +#[cfg(feature="collections")] +impl Drop for Guard<'_> { + fn drop(&mut self) { + unsafe { + self.buf.set_len(self.len); + } + } +} + +// A few methods below (read_to_string, read_line) 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. +#[cfg(feature="collections")] +fn append_to_string(buf: &mut String, f: F) -> Result +where + F: FnOnce(&mut Vec) -> Result, +{ + unsafe { + 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::new_const(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. +// +// Because we're extending the buffer with uninitialized data for trusted +// readers, we need to make sure to truncate that if any of this panics. +#[cfg(feature="collections")] +fn read_to_end(r: &mut R, buf: &mut Vec) -> Result { + read_to_end_with_reservation(r, buf, |_| 32) +} + +#[cfg(feature="collections")] +fn read_to_end_with_reservation( + r: &mut R, + buf: &mut Vec, + mut reservation_size: F, +) -> Result +where + R: Read + ?Sized, + F: FnMut(&R) -> usize, +{ + let start_len = buf.len(); + let mut g = Guard { len: buf.len(), buf }; + loop { + if g.len == g.buf.len() { + unsafe { + // FIXME(danielhenrymantilla): #42788 + // + // - This creates a (mut) reference to a slice of + // _uninitialized_ integers, which is **undefined behavior** + // + // - Only the standard library gets to soundly "ignore" this, + // based on its privileged knowledge of unstable rustc + // internals; + g.buf.reserve(reservation_size(r)); + let capacity = g.buf.capacity(); + g.buf.set_len(capacity); + r.initializer().initialize(&mut g.buf[g.len..]); + } + } + + let buf = &mut g.buf[g.len..]; + match r.read(buf) { + Ok(0) => return Ok(g.len - start_len), + Ok(n) => { + // We can't allow bogus values from read. If it is too large, the returned vec could have its length + // set past its capacity, or if it overflows the vec could be shortened which could create an invalid + // string if this is called via read_to_string. + assert!(n <= buf.len()); + g.len += n; + } + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } +} + +pub(crate) fn default_read_vectored(read: F, bufs: &mut [IoSliceMut<'_>]) -> Result +where + F: FnOnce(&mut [u8]) -> Result, +{ + let buf = bufs.iter_mut().find(|b| !b.is_empty()).map_or(&mut [][..], |b| &mut **b); + read(buf) +} + +pub(crate) fn default_write_vectored(write: F, bufs: &[IoSlice<'_>]) -> Result +where + F: FnOnce(&[u8]) -> Result, +{ + let buf = bufs.iter().find(|b| !b.is_empty()).map_or(&[][..], |b| &**b); + write(buf) +} + +pub(crate) fn default_read_exact(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::new_const(ErrorKind::UnexpectedEof, &"failed to fill whole buffer")) + } else { + 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 +#[doc(notable_trait)] +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`]) is not safe, and can lead to undefined behavior. + /// + /// [`MaybeUninit`]: 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(()) + /// } + /// ``` + fn read(&mut self, buf: &mut [u8]) -> Result; + + /// 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. + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> Result { + 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`. + fn is_read_vectored(&self) -> bool { + false + } + + /// Determines if this `Read`er can work with buffers of uninitialized + /// memory. + /// + /// The default implementation returns an initializer which will zero + /// buffers. + /// + /// If a `Read`er guarantees that it can work properly with uninitialized + /// memory, it should call [`Initializer::nop()`]. See the documentation for + /// [`Initializer`] for details. + /// + /// The behavior of this method must be independent of the state of the + /// `Read`er - the method only takes `&self` so that it can be used through + /// trait objects. + /// + /// # Safety + /// + /// This method is unsafe because a `Read`er could otherwise return a + /// non-zeroing `Initializer` from another `Read` type without an `unsafe` + /// block. + #[inline] + unsafe fn initializer(&self) -> Initializer { + Initializer::zeroing() + } + + /// 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 + #[cfg(feature="collections")] + fn read_to_end(&mut self, buf: &mut Vec) -> Result { + 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 + #[cfg(feature="collections")] + fn read_to_string(&mut self, buf: &mut String) -> Result { + // Note that we do *not* call `.read_to_end()` here. We are passing + // `&mut Vec` (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 `read_to_end` implementation which we + // know is guaranteed to only read data into the end of the buffer. + append_to_string(buf, |b| read_to_end(self, b)) + } + + /// 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(()) + /// } + /// ``` + fn read_exact(&mut self, buf: &mut [u8]) -> Result<()> { + default_read_exact(self, buf) + } + + /// Creates a "by reference" adapter 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(()) + /// } + /// ``` + 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 + /// [`Result`]`<`[`u8`]`, `[`io::Error`]`>`. + /// 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`: + /// + /// [`File`]: crate::fs::File + /// [`Result`]: crate::result::Result + /// [`io::Error`]: self::Error + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// + /// for byte in f.bytes() { + /// println!("{}", byte.unwrap()); + /// } + /// Ok(()) + /// } + /// ``` + fn bytes(self) -> Bytes + 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 mut f1 = File::open("foo.txt")?; + /// let mut 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(()) + /// } + /// ``` + fn chain(self, next: R) -> Chain + 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 mut 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(()) + /// } + /// ``` + fn take(self, limit: u64) -> Take + 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. +/// +/// # 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(&mut io::stdin())?; +/// println!("Stdin was:"); +/// println!("{}", stdin); +/// Ok(()) +/// } +/// ``` +#[cfg(feature="collections")] +pub fn read_to_string(reader: &mut R) -> Result { + let mut buf = String::new(); + reader.read_to_string(&mut buf)?; + Ok(buf) +} + +#[derive(Copy, Clone)] +pub struct IoVecBuffer<'a>(&'a [u8]); + +impl<'a> IoVecBuffer<'a> { + #[inline] + pub fn new(buf: &'a [u8]) -> IoVecBuffer<'a> { + IoVecBuffer(buf) + } + + #[inline] + pub fn advance(&mut self, n: usize) { + self.0 = &self.0[n..] + } + + #[inline] + pub fn as_slice(&self) -> &[u8] { + self.0 + } +} + +pub struct IoVecMutBuffer<'a>(&'a mut [u8]); + +impl<'a> IoVecMutBuffer<'a> { + #[inline] + pub fn new(buf: &'a mut [u8]) -> IoVecMutBuffer<'a> { + IoVecMutBuffer(buf) + } + + #[inline] + pub fn advance(&mut self, n: usize) { + let slice = core::mem::replace(&mut self.0, &mut []); + let (_, remaining) = slice.split_at_mut(n); + self.0 = remaining; + } + + #[inline] + pub fn as_slice(&self) -> &[u8] { + self.0 + } + + #[inline] + pub fn as_mut_slice(&mut self) -> &mut [u8] { + self.0 + } +} + +/// 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. +#[repr(transparent)] +pub struct IoSliceMut<'a>(IoVecMutBuffer<'a>); + +unsafe impl<'a> Send for IoSliceMut<'a> {} + +unsafe impl<'a> Sync for IoSliceMut<'a> {} + +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. + #[inline] + pub fn new(buf: &'a mut [u8]) -> IoSliceMut<'a> { + IoSliceMut(IoVecMutBuffer::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSliceMut::advance_slices`] to advance the cursors of + /// multiple buffers. + /// + /// # 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()); + /// ``` + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance the internal cursor of the slices. + /// + /// # Notes + /// + /// Elements in the slice may be modified if the cursor is not advanced to + /// the end of the slice. For example if we have a slice of buffers with 2 + /// `IoSliceMut`s, both of length 8, and we advance the cursor by 10 bytes + /// the first `IoSliceMut` will be untouched however the second will be + /// modified to remove the first 2 bytes (10 - 8). + /// + /// # 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()); + /// ``` + #[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() { + bufs[0].advance(n - accumulated_len) + } + } +} + +impl<'a> Deref for IoSliceMut<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +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. +#[derive(Copy, Clone)] +#[repr(transparent)] +pub struct IoSlice<'a>(IoVecBuffer<'a>); + +unsafe impl<'a> Send for IoSlice<'a> {} + +unsafe impl<'a> Sync for IoSlice<'a> {} + +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. + #[inline] + pub fn new(buf: &'a [u8]) -> IoSlice<'a> { + IoSlice(IoVecBuffer::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSlice::advance_slices`] to advance the cursors of multiple + /// buffers. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSlice; + /// use std::ops::Deref; + /// + /// let mut data = [1; 8]; + /// let mut buf = IoSlice::new(&mut data); + /// + /// // Mark 3 bytes as read. + /// buf.advance(3); + /// assert_eq!(buf.deref(), [1; 5].as_ref()); + /// ``` + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance the internal cursor of the slices. + /// + /// # Notes + /// + /// Elements in the slice may be modified if the cursor is not advanced to + /// the end of the slice. For example if we have a slice of buffers with 2 + /// `IoSlice`s, both of length 8, and we advance the cursor by 10 bytes the + /// first `IoSlice` will be untouched however the second will be modified to + /// remove the first 2 bytes (10 - 8). + /// + /// # 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()); + #[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() { + bufs[0].advance(n - accumulated_len) + } + } +} + +impl<'a> Deref for IoSlice<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +/// A type used to conditionally initialize buffers passed to `Read` methods. +#[derive(Debug)] +pub struct Initializer(bool); + +impl Initializer { + /// Returns a new `Initializer` which will zero out buffers. + #[inline] + pub fn zeroing() -> Initializer { + Initializer(true) + } + + /// Returns a new `Initializer` which will not zero out buffers. + /// + /// # Safety + /// + /// This may only be called by `Read`ers which guarantee that they will not + /// read from buffers passed to `Read` methods, and that the return value of + /// the method accurately reflects the number of bytes that have been + /// written to the head of the buffer. + #[inline] + pub unsafe fn nop() -> Initializer { + Initializer(false) + } + + /// Indicates if a buffer should be initialized. + #[inline] + pub fn should_initialize(&self) -> bool { + self.0 + } + + /// Initializes a buffer if necessary. + #[inline] + pub fn initialize(&self, buf: &mut [u8]) { + if self.should_initialize() { + unsafe { ptr::write_bytes(buf.as_mut_ptr(), 0, buf.len()) } + } + } +} + +/// 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 +#[doc(notable_trait)] +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 + fn write(&mut self, buf: &[u8]) -> Result; + + /// 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 mut data1 = [1; 8]; + /// let mut data2 = [15; 8]; + /// let io_slice1 = IoSlice::new(&mut data1); + /// let io_slice2 = IoSlice::new(&mut 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 + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> Result { + 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 + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + fn write_all(&mut self, mut buf: &[u8]) -> Result<()> { + while !buf.is_empty() { + match self.write(buf) { + Ok(0) => { + return Err(Error::new_const( + 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(()) } + /// ``` + 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::new_const( + 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, but 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(()) + /// } + /// ``` + 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 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::new_const(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(()) + /// } + /// ``` + 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(()) +/// } +/// ``` +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. + fn seek(&mut self, pos: SeekFrom) -> Result; + + /// 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); + /// ``` + 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 {} bytes long", len); + /// Ok(()) + /// } + /// ``` + fn stream_len(&mut self) -> Result { + 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(()) + /// } + /// ``` + fn stream_position(&mut self) -> Result { + 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)] +pub enum SeekFrom { + /// Sets the offset to the provided number of bytes. + Start(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. + End(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. + Current(i64), +} + +#[cfg(feature="collections")] +fn read_until(r: &mut R, delim: u8, buf: &mut Vec) -> Result { + 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(()) +/// } +/// ``` +#[cfg(feature="collections")] +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); + /// ``` + 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 + 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`, 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); + /// } + /// ``` + fn has_data_left(&mut self) -> Result { + 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""); + /// ``` + fn read_until(&mut self, byte: u8, buf: &mut Vec) -> Result { + read_until(self, byte, buf) + } + + /// Read all bytes until a newline (the `0xA` byte) is reached, and append + /// them to the provided buffer. + /// + /// 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, ""); + /// ``` + fn read_line(&mut self, buf: &mut String) -> Result { + // 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`. + 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 + /// [`io::Result`]`<`[`Vec`]`>`. 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 + /// [`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); + /// ``` + fn split(self, byte: u8) -> Split + 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 + /// [`io::Result`]`<`[`String`]`>`. Each string returned will *not* have a newline + /// byte (the `0xA` byte) or `CRLF` (`0xD`, `0xA` bytes) at the end. + /// + /// [`io::Result`]: self::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`]. + fn lines(self) -> Lines + 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 +#[derive(Debug)] +pub struct Chain { + first: T, + second: U, + done_first: bool, +} + +impl Chain { + /// 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + pub fn get_mut(&mut self) -> (&mut T, &mut U) { + (&mut self.first, &mut self.second) + } +} + +impl Read for Chain { + fn read(&mut self, buf: &mut [u8]) -> Result { + 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 { + 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) + } + + unsafe fn initializer(&self) -> Initializer { + let initializer = self.first.initializer(); + if initializer.should_initialize() { initializer } else { self.second.initializer() } + } +} + +#[cfg(feature="collections")] +impl BufRead for Chain { + 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 SizeHint for Chain { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&self.first) + SizeHint::lower_bound(&self.second) + } + + #[inline] + fn upper_bound(&self) -> Option { + 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 +#[derive(Debug)] +pub struct Take { + inner: T, + limit: u64, +} + +impl Take { + /// 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + pub fn get_mut(&mut self) -> &mut T { + &mut self.inner + } +} + +impl Read for Take { + fn read(&mut self, buf: &mut [u8]) -> Result { + // 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])?; + self.limit -= n as u64; + Ok(n) + } + + unsafe fn initializer(&self) -> Initializer { + self.inner.initializer() + } + + #[cfg(feature="collections")] + fn read_to_end(&mut self, buf: &mut Vec) -> Result { + // Pass in a reservation_size closure that respects the current value + // of limit for each read. If we hit the read limit, this prevents the + // final zero-byte read from allocating again. + read_to_end_with_reservation(self, buf, |self_| cmp::min(self_.limit, 32) as usize) + } +} + +#[cfg(feature="collections")] +impl BufRead for Take { + 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 SizeHint for Take { + #[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 { + 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 +#[derive(Debug)] +pub struct Bytes { + inner: R, +} + +impl Iterator for Bytes { + type Item = Result; + + fn next(&mut self) -> Option> { + 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) { + SizeHint::size_hint(&self.inner) + } +} + +trait SizeHint { + fn lower_bound(&self) -> usize; + + fn upper_bound(&self) -> Option; + + fn size_hint(&self) -> (usize, Option) { + (self.lower_bound(), self.upper_bound()) + } +} + +impl SizeHint for T { + #[inline] + default fn lower_bound(&self) -> usize { + 0 + } + + #[inline] + default fn upper_bound(&self) -> Option { + None + } +} + +impl SizeHint for &mut T { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(*self) + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(*self) + } +} + +#[cfg(feature="alloc")] +impl SizeHint for Box { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&**self) + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(&**self) + } +} + +impl SizeHint for &[u8] { + #[inline] + fn lower_bound(&self) -> usize { + self.len() + } + + #[inline] + fn upper_bound(&self) -> Option { + 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 +#[cfg(feature="collections")] +#[derive(Debug)] +pub struct Split { + buf: B, + delim: u8, +} + +#[cfg(feature="collections")] +impl Iterator for Split { + type Item = Result>; + + fn next(&mut self) -> Option>> { + 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 +#[cfg(feature="collections")] +#[derive(Debug)] +pub struct Lines { + buf: B, +} + +#[cfg(feature="collections")] +impl Iterator for Lines { + type Item = Result; + + fn next(&mut self) -> Option> { + 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/src/lib.rs b/src/lib.rs new file mode 100644 index 0000000..4cecd7c --- /dev/null +++ b/src/lib.rs @@ -0,0 +1,2548 @@ +//!

+//! This is just a listing of the functionality available in this crate. See +//! the [std documentation](https://doc.rust-lang.org/nightly/std/io/index.html) +//! for a full description of the functionality. +#![allow(stable_features,unused_features,incomplete_features)] +#![feature(question_mark,const_fn_trait_bound,copy_from_slice,try_from,str_internals,align_offset, + doc_notable_trait,slice_internals,maybe_uninit_ref,mem_take,specialization)] +#![cfg_attr(any(feature="alloc",feature="collections"),feature(alloc,allocator_api))] +#![cfg_attr(feature="collections",feature(vec_spare_capacity,maybe_uninit_slice, + new_uninit,debug_non_exhaustive))] +#![cfg_attr(pattern_guards,feature(bind_by_move_pattern_guards,nll))] +#![cfg_attr(non_exhaustive,feature(non_exhaustive))] +#![cfg_attr(unicode,feature(str_char))] +#![cfg_attr(unicode,feature(unicode))] +#![no_std] + +#[cfg(feature="alloc")] extern crate alloc; +#[cfg(rustc_unicode)] +extern crate rustc_unicode; +#[cfg(std_unicode)] +extern crate std_unicode; + +#[cfg(not(feature="collections"))] +pub type ErrorString = &'static str; + +// Provide Box::new wrapper +#[cfg(not(feature="alloc"))] +struct FakeBox(core::marker::PhantomData); +#[cfg(not(feature="alloc"))] +impl FakeBox { + fn new(val: T) -> T { + val + } +} + +// Needed for older compilers, to ignore vec!/format! macros in tests +#[cfg(not(feature="collections"))] +#[allow(unused)] +macro_rules! vec ( + ( $ elem : expr ; $ n : expr ) => { () }; + ( $ ( $ x : expr ) , * ) => { () }; + ( $ ( $ x : expr , ) * ) => { () }; +); +#[cfg(not(feature="collections"))] +#[allow(unused)] +macro_rules! format { + ( $ ( $ arg : tt ) * ) => { () }; +} + +#[cfg(test)] +mod tests; + +use core::cmp; +use core::fmt; +#[cfg(feature="collections")] use core::slice::memchr; +use core::ops::{Deref, DerefMut}; +use core::ptr; +use core::slice; +use core::str; +use core::convert::TryInto; +use core::mem::replace; + +#[cfg(feature="collections")] pub use self::buffered::IntoInnerError; +#[cfg(feature="collections")] pub use self::buffered::WriterPanicked; +#[cfg(feature="collections")] pub use self::buffered::{BufReader, BufWriter, LineWriter}; + +pub use self::copy::copy; +pub use self::cursor::Cursor; +pub use self::error::{Error, ErrorKind, Result}; +pub use self::util::{empty, repeat, sink, Empty, Repeat, Sink}; + +#[cfg(feature="collections")] use alloc::string::String; +#[cfg(feature="collections")] use alloc::vec::Vec; +#[cfg(feature="alloc")] use alloc::boxed::Box; + +#[cfg(feature="collections")] mod buffered; +pub(crate) mod copy; +mod cursor; +mod error; +mod impls; +pub mod prelude; +mod util; + +const DEFAULT_BUF_SIZE: usize = 8 * 1024; + +#[cfg(feature="collections")] +struct Guard<'a> { + buf: &'a mut Vec, + len: usize, +} + +#[cfg(feature="collections")] +impl Drop for Guard<'_> { + fn drop(&mut self) { + unsafe { + self.buf.set_len(self.len); + } + } +} + +// A few methods below (read_to_string, read_line) 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. +#[cfg(feature="collections")] +fn append_to_string(buf: &mut String, f: F) -> Result +where + F: FnOnce(&mut Vec) -> Result, +{ + unsafe { + 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::new_const(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. +// +// Because we're extending the buffer with uninitialized data for trusted +// readers, we need to make sure to truncate that if any of this panics. +#[cfg(feature="collections")] +fn read_to_end(r: &mut R, buf: &mut Vec) -> Result { + read_to_end_with_reservation(r, buf, |_| 32) +} + +#[cfg(feature="collections")] +fn read_to_end_with_reservation( + r: &mut R, + buf: &mut Vec, + mut reservation_size: F, +) -> Result +where + R: Read + ?Sized, + F: FnMut(&R) -> usize, +{ + let start_len = buf.len(); + let mut g = Guard { len: buf.len(), buf }; + loop { + if g.len == g.buf.len() { + unsafe { + // FIXME(danielhenrymantilla): #42788 + // + // - This creates a (mut) reference to a slice of + // _uninitialized_ integers, which is **undefined behavior** + // + // - Only the standard library gets to soundly "ignore" this, + // based on its privileged knowledge of unstable rustc + // internals; + g.buf.reserve(reservation_size(r)); + let capacity = g.buf.capacity(); + g.buf.set_len(capacity); + r.initializer().initialize(&mut g.buf[g.len..]); + } + } + + let buf = &mut g.buf[g.len..]; + match r.read(buf) { + Ok(0) => return Ok(g.len - start_len), + Ok(n) => { + // We can't allow bogus values from read. If it is too large, the returned vec could have its length + // set past its capacity, or if it overflows the vec could be shortened which could create an invalid + // string if this is called via read_to_string. + assert!(n <= buf.len()); + g.len += n; + } + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} + Err(e) => return Err(e), + } + } +} + +pub(crate) fn default_read_vectored(read: F, bufs: &mut [IoSliceMut<'_>]) -> Result +where + F: FnOnce(&mut [u8]) -> Result, +{ + let buf = bufs.iter_mut().find(|b| !b.is_empty()).map_or(&mut [][..], |b| &mut **b); + read(buf) +} + +pub(crate) fn default_write_vectored(write: F, bufs: &[IoSlice<'_>]) -> Result +where + F: FnOnce(&[u8]) -> Result, +{ + let buf = bufs.iter().find(|b| !b.is_empty()).map_or(&[][..], |b| &**b); + write(buf) +} + +pub(crate) fn default_read_exact(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::new_const(ErrorKind::UnexpectedEof, &"failed to fill whole buffer")) + } else { + 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 +#[doc(notable_trait)] +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`]) is not safe, and can lead to undefined behavior. + /// + /// [`MaybeUninit`]: 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(()) + /// } + /// ``` + fn read(&mut self, buf: &mut [u8]) -> Result; + + /// 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. + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> Result { + 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`. + fn is_read_vectored(&self) -> bool { + false + } + + /// Determines if this `Read`er can work with buffers of uninitialized + /// memory. + /// + /// The default implementation returns an initializer which will zero + /// buffers. + /// + /// If a `Read`er guarantees that it can work properly with uninitialized + /// memory, it should call [`Initializer::nop()`]. See the documentation for + /// [`Initializer`] for details. + /// + /// The behavior of this method must be independent of the state of the + /// `Read`er - the method only takes `&self` so that it can be used through + /// trait objects. + /// + /// # Safety + /// + /// This method is unsafe because a `Read`er could otherwise return a + /// non-zeroing `Initializer` from another `Read` type without an `unsafe` + /// block. + #[inline] + unsafe fn initializer(&self) -> Initializer { + Initializer::zeroing() + } + + /// 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 + #[cfg(feature="collections")] + fn read_to_end(&mut self, buf: &mut Vec) -> Result { + 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 + #[cfg(feature="collections")] + fn read_to_string(&mut self, buf: &mut String) -> Result { + // Note that we do *not* call `.read_to_end()` here. We are passing + // `&mut Vec` (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 `read_to_end` implementation which we + // know is guaranteed to only read data into the end of the buffer. + append_to_string(buf, |b| read_to_end(self, b)) + } + + /// 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(()) + /// } + /// ``` + fn read_exact(&mut self, buf: &mut [u8]) -> Result<()> { + default_read_exact(self, buf) + } + + /// Creates a "by reference" adapter 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(()) + /// } + /// ``` + 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 + /// [`Result`]`<`[`u8`]`, `[`io::Error`]`>`. + /// 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`: + /// + /// [`File`]: crate::fs::File + /// [`Result`]: crate::result::Result + /// [`io::Error`]: self::Error + /// + /// ```no_run + /// use std::io; + /// use std::io::prelude::*; + /// use std::fs::File; + /// + /// fn main() -> io::Result<()> { + /// let mut f = File::open("foo.txt")?; + /// + /// for byte in f.bytes() { + /// println!("{}", byte.unwrap()); + /// } + /// Ok(()) + /// } + /// ``` + fn bytes(self) -> Bytes + 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 mut f1 = File::open("foo.txt")?; + /// let mut 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(()) + /// } + /// ``` + fn chain(self, next: R) -> Chain + 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 mut 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(()) + /// } + /// ``` + fn take(self, limit: u64) -> Take + 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. +/// +/// # 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(&mut io::stdin())?; +/// println!("Stdin was:"); +/// println!("{}", stdin); +/// Ok(()) +/// } +/// ``` +#[cfg(feature="collections")] +pub fn read_to_string(reader: &mut R) -> Result { + let mut buf = String::new(); + reader.read_to_string(&mut buf)?; + Ok(buf) +} + +#[derive(Copy, Clone)] +pub struct IoVecBuffer<'a>(&'a [u8]); + +impl<'a> IoVecBuffer<'a> { + #[inline] + pub fn new(buf: &'a [u8]) -> IoVecBuffer<'a> { + IoVecBuffer(buf) + } + + #[inline] + pub fn advance(&mut self, n: usize) { + self.0 = &self.0[n..] + } + + #[inline] + pub fn as_slice(&self) -> &[u8] { + self.0 + } +} + +pub struct IoVecMutBuffer<'a>(&'a mut [u8]); + +impl<'a> IoVecMutBuffer<'a> { + #[inline] + pub fn new(buf: &'a mut [u8]) -> IoVecMutBuffer<'a> { + IoVecMutBuffer(buf) + } + + #[inline] + pub fn advance(&mut self, n: usize) { + let slice = core::mem::replace(&mut self.0, &mut []); + let (_, remaining) = slice.split_at_mut(n); + self.0 = remaining; + } + + #[inline] + pub fn as_slice(&self) -> &[u8] { + self.0 + } + + #[inline] + pub fn as_mut_slice(&mut self) -> &mut [u8] { + self.0 + } +} + +/// 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. +#[repr(transparent)] +pub struct IoSliceMut<'a>(IoVecMutBuffer<'a>); + +unsafe impl<'a> Send for IoSliceMut<'a> {} + +unsafe impl<'a> Sync for IoSliceMut<'a> {} + +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. + #[inline] + pub fn new(buf: &'a mut [u8]) -> IoSliceMut<'a> { + IoSliceMut(IoVecMutBuffer::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSliceMut::advance_slices`] to advance the cursors of + /// multiple buffers. + /// + /// # 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()); + /// ``` + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance the internal cursor of the slices. + /// + /// # Notes + /// + /// Elements in the slice may be modified if the cursor is not advanced to + /// the end of the slice. For example if we have a slice of buffers with 2 + /// `IoSliceMut`s, both of length 8, and we advance the cursor by 10 bytes + /// the first `IoSliceMut` will be untouched however the second will be + /// modified to remove the first 2 bytes (10 - 8). + /// + /// # 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()); + /// ``` + #[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() { + bufs[0].advance(n - accumulated_len) + } + } +} + +impl<'a> Deref for IoSliceMut<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +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. +#[derive(Copy, Clone)] +#[repr(transparent)] +pub struct IoSlice<'a>(IoVecBuffer<'a>); + +unsafe impl<'a> Send for IoSlice<'a> {} + +unsafe impl<'a> Sync for IoSlice<'a> {} + +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. + #[inline] + pub fn new(buf: &'a [u8]) -> IoSlice<'a> { + IoSlice(IoVecBuffer::new(buf)) + } + + /// Advance the internal cursor of the slice. + /// + /// Also see [`IoSlice::advance_slices`] to advance the cursors of multiple + /// buffers. + /// + /// # Examples + /// + /// ``` + /// #![feature(io_slice_advance)] + /// + /// use std::io::IoSlice; + /// use std::ops::Deref; + /// + /// let mut data = [1; 8]; + /// let mut buf = IoSlice::new(&mut data); + /// + /// // Mark 3 bytes as read. + /// buf.advance(3); + /// assert_eq!(buf.deref(), [1; 5].as_ref()); + /// ``` + #[inline] + pub fn advance(&mut self, n: usize) { + self.0.advance(n) + } + + /// Advance the internal cursor of the slices. + /// + /// # Notes + /// + /// Elements in the slice may be modified if the cursor is not advanced to + /// the end of the slice. For example if we have a slice of buffers with 2 + /// `IoSlice`s, both of length 8, and we advance the cursor by 10 bytes the + /// first `IoSlice` will be untouched however the second will be modified to + /// remove the first 2 bytes (10 - 8). + /// + /// # 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()); + #[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() { + bufs[0].advance(n - accumulated_len) + } + } +} + +impl<'a> Deref for IoSlice<'a> { + type Target = [u8]; + + #[inline] + fn deref(&self) -> &[u8] { + self.0.as_slice() + } +} + +/// A type used to conditionally initialize buffers passed to `Read` methods. +#[derive(Debug)] +pub struct Initializer(bool); + +impl Initializer { + /// Returns a new `Initializer` which will zero out buffers. + #[inline] + pub fn zeroing() -> Initializer { + Initializer(true) + } + + /// Returns a new `Initializer` which will not zero out buffers. + /// + /// # Safety + /// + /// This may only be called by `Read`ers which guarantee that they will not + /// read from buffers passed to `Read` methods, and that the return value of + /// the method accurately reflects the number of bytes that have been + /// written to the head of the buffer. + #[inline] + pub unsafe fn nop() -> Initializer { + Initializer(false) + } + + /// Indicates if a buffer should be initialized. + #[inline] + pub fn should_initialize(&self) -> bool { + self.0 + } + + /// Initializes a buffer if necessary. + #[inline] + pub fn initialize(&self, buf: &mut [u8]) { + if self.should_initialize() { + unsafe { ptr::write_bytes(buf.as_mut_ptr(), 0, buf.len()) } + } + } +} + +/// 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 +#[doc(notable_trait)] +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 + fn write(&mut self, buf: &[u8]) -> Result; + + /// 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 mut data1 = [1; 8]; + /// let mut data2 = [15; 8]; + /// let io_slice1 = IoSlice::new(&mut data1); + /// let io_slice2 = IoSlice::new(&mut 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 + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> Result { + 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 + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + fn write_all(&mut self, mut buf: &[u8]) -> Result<()> { + while !buf.is_empty() { + match self.write(buf) { + Ok(0) => { + return Err(Error::new_const( + 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(()) } + /// ``` + 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::new_const( + 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, but 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(()) + /// } + /// ``` + 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 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::new_const(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(()) + /// } + /// ``` + 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(()) +/// } +/// ``` +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. + fn seek(&mut self, pos: SeekFrom) -> Result; + + /// 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); + /// ``` + 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 {} bytes long", len); + /// Ok(()) + /// } + /// ``` + fn stream_len(&mut self) -> Result { + 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(()) + /// } + /// ``` + fn stream_position(&mut self) -> Result { + 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)] +pub enum SeekFrom { + /// Sets the offset to the provided number of bytes. + Start(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. + End(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. + Current(i64), +} + +#[cfg(feature="collections")] +fn read_until(r: &mut R, delim: u8, buf: &mut Vec) -> Result { + 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(()) +/// } +/// ``` +#[cfg(feature="collections")] +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); + /// ``` + 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 + 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`, 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); + /// } + /// ``` + fn has_data_left(&mut self) -> Result { + 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""); + /// ``` + fn read_until(&mut self, byte: u8, buf: &mut Vec) -> Result { + read_until(self, byte, buf) + } + + /// Read all bytes until a newline (the `0xA` byte) is reached, and append + /// them to the provided buffer. + /// + /// 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, ""); + /// ``` + fn read_line(&mut self, buf: &mut String) -> Result { + // 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`. + 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 + /// [`io::Result`]`<`[`Vec`]`>`. 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 + /// [`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); + /// ``` + fn split(self, byte: u8) -> Split + 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 + /// [`io::Result`]`<`[`String`]`>`. Each string returned will *not* have a newline + /// byte (the `0xA` byte) or `CRLF` (`0xD`, `0xA` bytes) at the end. + /// + /// [`io::Result`]: self::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`]. + fn lines(self) -> Lines + 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 +#[derive(Debug)] +pub struct Chain { + first: T, + second: U, + done_first: bool, +} + +impl Chain { + /// 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + pub fn get_mut(&mut self) -> (&mut T, &mut U) { + (&mut self.first, &mut self.second) + } +} + +impl Read for Chain { + fn read(&mut self, buf: &mut [u8]) -> Result { + 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 { + 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) + } + + unsafe fn initializer(&self) -> Initializer { + let initializer = self.first.initializer(); + if initializer.should_initialize() { initializer } else { self.second.initializer() } + } +} + +#[cfg(feature="collections")] +impl BufRead for Chain { + 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 SizeHint for Chain { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&self.first) + SizeHint::lower_bound(&self.second) + } + + #[inline] + fn upper_bound(&self) -> Option { + 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 +#[derive(Debug)] +pub struct Take { + inner: T, + limit: u64, +} + +impl Take { + /// 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + 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(()) + /// } + /// ``` + pub fn get_mut(&mut self) -> &mut T { + &mut self.inner + } +} + +impl Read for Take { + fn read(&mut self, buf: &mut [u8]) -> Result { + // 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])?; + self.limit -= n as u64; + Ok(n) + } + + unsafe fn initializer(&self) -> Initializer { + self.inner.initializer() + } + + #[cfg(feature="collections")] + fn read_to_end(&mut self, buf: &mut Vec) -> Result { + // Pass in a reservation_size closure that respects the current value + // of limit for each read. If we hit the read limit, this prevents the + // final zero-byte read from allocating again. + read_to_end_with_reservation(self, buf, |self_| cmp::min(self_.limit, 32) as usize) + } +} + +#[cfg(feature="collections")] +impl BufRead for Take { + 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 SizeHint for Take { + #[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 { + 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 +#[derive(Debug)] +pub struct Bytes { + inner: R, +} + +impl Iterator for Bytes { + type Item = Result; + + fn next(&mut self) -> Option> { + 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) { + SizeHint::size_hint(&self.inner) + } +} + +trait SizeHint { + fn lower_bound(&self) -> usize; + + fn upper_bound(&self) -> Option; + + fn size_hint(&self) -> (usize, Option) { + (self.lower_bound(), self.upper_bound()) + } +} + +impl SizeHint for T { + #[inline] + default fn lower_bound(&self) -> usize { + 0 + } + + #[inline] + default fn upper_bound(&self) -> Option { + None + } +} + +impl SizeHint for &mut T { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(*self) + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(*self) + } +} + +#[cfg(feature="alloc")] +impl SizeHint for Box { + #[inline] + fn lower_bound(&self) -> usize { + SizeHint::lower_bound(&**self) + } + + #[inline] + fn upper_bound(&self) -> Option { + SizeHint::upper_bound(&**self) + } +} + +impl SizeHint for &[u8] { + #[inline] + fn lower_bound(&self) -> usize { + self.len() + } + + #[inline] + fn upper_bound(&self) -> Option { + 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 +#[cfg(feature="collections")] +#[derive(Debug)] +pub struct Split { + buf: B, + delim: u8, +} + +#[cfg(feature="collections")] +impl Iterator for Split { + type Item = Result>; + + fn next(&mut self) -> Option>> { + 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 +#[cfg(feature="collections")] +#[derive(Debug)] +pub struct Lines { + buf: B, +} + +#[cfg(feature="collections")] +impl Iterator for Lines { + type Item = Result; + + fn next(&mut self) -> Option> { + 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/src/prelude.rs b/src/prelude.rs new file mode 100644 index 0000000..ec471a9 --- /dev/null +++ b/src/prelude.rs @@ -0,0 +1,15 @@ +//! 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::*; +//! ``` + +pub use super::{Read, Seek, Write}; +#[cfg(feature="collections")] pub use super::BufRead; + +#[cfg(feature="collections")] pub use alloc::boxed::Box; +#[cfg(feature="collections")] pub use alloc::vec::Vec; diff --git a/src/stdio/tests.rs b/src/stdio/tests.rs new file mode 100644 index 0000000..b1df6b7 --- /dev/null +++ b/src/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::(); +} +#[test] +fn stdoutlock_unwind_safe() { + assert_unwind_safe::>(); + assert_unwind_safe::>(); +} +#[test] +fn stderr_unwind_safe() { + assert_unwind_safe::(); +} +#[test] +fn stderrlock_unwind_safe() { + assert_unwind_safe::>(); + assert_unwind_safe::>(); +} + +fn assert_unwind_safe() {} + +#[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_locked); +} +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn test_lock_stdin() { + test_lock(stdin, stdin_locked); +} +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] +fn test_lock_stdout() { + test_lock(stdout, stdout_locked); +} + +// 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(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/src/tests.rs b/src/tests.rs new file mode 100644 index 0000000..1beb72a --- /dev/null +++ b/src/tests.rs @@ -0,0 +1,573 @@ +use super::{repeat, Cursor, SeekFrom}; +use crate::cmp::{self, min}; +use crate::io::{self, IoSlice, IoSliceMut}; +use crate::io::{BufRead, BufReader, Read, Seek, Write}; +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::>(); + 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 take_eof() { + struct R; + + impl Read for R { + fn read(&mut self, _: &mut [u8]) -> io::Result { + Err(io::Error::new_const(io::ErrorKind::Other, &"")) + } + } + impl BufRead for R { + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Err(io::Error::new_const(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(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::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 { + 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() generally needs to over-allocate, both for efficiency + // and so that it can distinguish EOF. Assert that this is the case + // with this simple ExampleSliceReader struct, which uses the default + // implementation of read_to_end. Even though vec1 is allocated with + // exactly enough capacity for the read, read_to_end will allocate more + // space here. + let mut vec1 = Vec::with_capacity(input.len()); + ExampleSliceReader { slice: input }.read_to_end(&mut vec1)?; + assert_eq!(vec1.len(), input.len()); + assert!(vec1.capacity() > input.len(), "allocated more"); + + // However, std::io::Take includes an implementation of read_to_end + // that will not allocate when the limit has already been reached. In + // this case, vec2 never grows. + let mut vec2 = Vec::with_capacity(input.len()); + ExampleSliceReader { slice: input }.take(input.len() as u64).read_to_end(&mut vec2)?; + assert_eq!(vec2.len(), input.len()); + assert_eq!(vec2.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] +fn io_slice_mut_advance_slices_empty_slice() { + let mut empty_bufs = &mut [][..]; + // Shouldn't panic. + IoSliceMut::advance_slices(&mut empty_bufs, 1); +} + +#[test] +fn io_slice_mut_advance_slices_beyond_total_length() { + let mut buf1 = [1; 8]; + let mut bufs = &mut [IoSliceMut::new(&mut buf1)][..]; + + // Going beyond the total length should be ok. + 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] +fn io_slice_advance_slices_empty_slice() { + let mut empty_bufs = &mut [][..]; + // Shouldn't panic. + IoSlice::advance_slices(&mut empty_bufs, 1); +} + +#[test] +fn io_slice_advance_slices_beyond_total_length() { + let buf1 = [1; 8]; + let mut bufs = &mut [IoSlice::new(&buf1)][..]; + + // Going beyond the total length should be ok. + 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, +} + +impl Write for TestWriter { + fn write(&mut self, buf: &[u8]) -> io::Result { + self.write_vectored(&[IoSlice::new(buf)]) + } + + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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); + } + } +} diff --git a/src/util.rs b/src/util.rs new file mode 100644 index 0000000..f681d74 --- /dev/null +++ b/src/util.rs @@ -0,0 +1,239 @@ +#![allow(missing_copy_implementations)] + +#[cfg(test)] +mod tests; + +use core::fmt; +use crate::{ + self as io, Initializer, IoSlice, IoSliceMut, Read, Seek, SeekFrom, SizeHint, Write, +}; +#[cfg(feature="collections")] use crate::BufRead; + +/// A reader which is always at EOF. +/// +/// This struct is generally created by calling [`empty()`]. Please see +/// the documentation of [`empty()`] for more details. +#[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 [`Ok`]`(0)`. +/// +/// # 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()); +/// ``` +pub const fn empty() -> Empty { + Empty +} + +impl Read for Empty { + #[inline] + fn read(&mut self, _buf: &mut [u8]) -> io::Result { + Ok(0) + } + + #[inline] + unsafe fn initializer(&self) -> Initializer { + Initializer::nop() + } +} + +#[cfg(feature="collections")] +impl BufRead for Empty { + #[inline] + fn fill_buf(&mut self) -> io::Result<&[u8]> { + Ok(&[]) + } + #[inline] + fn consume(&mut self, _n: usize) {} +} + +impl Seek for Empty { + fn seek(&mut self, _pos: SeekFrom) -> io::Result { + Ok(0) + } + + fn stream_len(&mut self) -> io::Result { + Ok(0) + } + + fn stream_position(&mut self) -> io::Result { + Ok(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 { + 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. +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]); +/// ``` +pub const fn repeat(byte: u8) -> Repeat { + Repeat { byte } +} + +impl Read for Repeat { + #[inline] + fn read(&mut self, buf: &mut [u8]) -> io::Result { + for slot in &mut *buf { + *slot = self.byte; + } + Ok(buf.len()) + } + + #[inline] + fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result { + let mut nwritten = 0; + for buf in bufs { + nwritten += self.read(buf)?; + } + Ok(nwritten) + } + + #[inline] + fn is_read_vectored(&self) -> bool { + true + } + + #[inline] + unsafe fn initializer(&self) -> Initializer { + Initializer::nop() + } +} + +impl SizeHint for Repeat { + #[inline] + fn lower_bound(&self) -> usize { + usize::MAX + } + + #[inline] + fn upper_bound(&self) -> Option { + None + } +} + +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. +#[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); +/// ``` +pub const fn sink() -> Sink { + Sink +} + +impl Write for Sink { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} + +impl Write for &Sink { + #[inline] + fn write(&mut self, buf: &[u8]) -> io::Result { + Ok(buf.len()) + } + + #[inline] + fn write_vectored(&mut self, bufs: &[IoSlice<'_>]) -> io::Result { + 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(()) + } +} + +impl fmt::Debug for Sink { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Sink").finish_non_exhaustive() + } +} diff --git a/src/util/tests.rs b/src/util/tests.rs new file mode 100644 index 0000000..7632eaf --- /dev/null +++ b/src/util/tests.rs @@ -0,0 +1,120 @@ +use crate::cmp::{max, min}; +use crate::io::prelude::*; +use crate::io::{ + copy, empty, repeat, sink, BufWriter, Empty, Repeat, Result, SeekFrom, Sink, DEFAULT_BUF_SIZE, +}; + +#[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 { + 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 { + 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); +} + +#[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(); +}