io
Лекция 9

Арсений Балобанов

* io.Reader & io.Writer

  type Reader interface {
      Read(p []byte) (n int, err error)
  }

  type Writer interface {
      Write(p []byte) (n int, err error)
  }

*Conceptually:*

- *Reader* has data, and you read it out and make use of that data.
- You have data and you want to shove it into a *Writer* where something happens to that data.

* io.Reader

  Read(p []byte) (n int, err error)

- Reads up to len(p) bytes into p.
- Returns the number of bytes read (0 <= n <= len(p)) and any error encountered.
- If some data is available but not len(p) bytes, Read conventionally returns what is available instead of waiting for more.

* io.Reader

  Read(p []byte) (n int, err error)

- Non-zero number of bytes at the end may result in either err == EOF or err == nil. The next Read should return 0, EOF.
- Even if Read returns n < len(p), it may use all of p as scratch space during the call (e.g. decompression).
- Process the n > 0 bytes returned before considering the error err.
- Implementations must not retain p.

* Reader Variants

  type ReadWriter interface { Reader; Writer }
  type ReadCloser interface { Reader; Closer }
  type ReadSeeker interface { Reader; Seeker }
  type ReadWriteCloser interface { Reader; Writer; Closer }
  type ReadWriteSeeker interface { Reader; Writer; Seeker }

  type ByteReader interface { ReadByte() (byte, error) }                 // reads single byte
  type RuneReader interface { ReadRune() (r rune, size int, err error) } // reads single rune

  type LimitedReader struct // limited to N bytes
  type PipeReader struct    // read half of a pipe
  type SectionReader struct // read interior section

* SectionReader

.play -edit sectionreader/main.go /^func main/,/^}/

* Reader Composition
  // LimitReader returns a Reader that reads from r but stops with EOF after n bytes.
  func LimitReader(r Reader, n int64) Reader

  // MultiReader returns a Reader that's the logical concatenation of the provided input readers.
  func MultiReader(readers ...Reader) Reader

  // TeeReader returns a Reader that writes to w what it reads from r.
  func TeeReader(r Reader, w Writer) Reader

* LimitReader

.play -edit limitreader/main.go /^func main/,/^}/

* MultiReader

.play -edit multireader/main.go /^func main/,/^}/

* TeeReader

.play -edit teereader/main.go /^func main/,/^}/

* Example: http chunking

* Example: http chunking

Chunked transfer encoding is a streaming data transfer mechanism.

  4\r\n
  Wiki\r\n
  5\r\n
  pedia\r\n
  E\r\n
   in\r\n
  \r\n
  chunks.\r\n
  0\r\n
  Date: Sun, 06 Nov 1994 08:49:37 GMT\r\n
  Content-MD5: 1B2M2Y8AsgTpgAmY7PhCfg==\r\n
  \r\n

Actual body

  Wikipedia in

  chunks.

* Example: http chunking

Problem: proxy a chunked HTTP in a stream.

  func transfer(clientWriter io.Writer, responseBody io.Reader)

- send chunks as is
- validate MD5

* Example: http chunking

.play -edit httpchunking/solution1/main.go

* Example: http chunking

.play -edit httpchunking/solution2/main.go  /^func transfer/,/^}/

* Example: http chunking

.play -edit httpchunking/solution3/main.go  /^func transfer/,/^}/

* ioutil

Package *io/ioutil* implements some I/O utility functions.

  var Discard io.Writer = devNull(0)

  func NopCloser(r io.Reader) io.ReadCloser
  func ReadAll(r io.Reader) ([]byte, error)
  func ReadDir(dirname string) ([]os.FileInfo, error)
  func ReadFile(filename string) ([]byte, error)
  func TempDir(dir, pattern string) (name string, err error)
  func TempFile(dir, pattern string) (f *os.File, err error)
  func WriteFile(filename string, data []byte, perm os.FileMode) error

* ReadAll

Convenience method for Reader → []byte conversion.

.play -edit readall/main.go /^func main/,/^}/

* ReadAll misuse
  ReaderAll(Reader) → []byte → Writer

Consider *io.Copy* instead

  io.Copy(dst Writer, src Reader)

* io.Copy

  func Copy(dst Writer, src Reader) (written int64, err error)

- Allocates a 32KB buffer to read from src and then write to dst.

  func CopyBuffer(dst Writer, src Reader, buf []byte) (written int64, err error)

- Reuse your own buffer with CopyBuffer()

Can we avoid using an intermediate buffer entirely?

  type ReaderFrom interface {
      ReadFrom(r Reader) (n int64, err error)
  }

  type WriterTo interface {
      WriteTo(w Writer) (n int64, err error)
  }

* Example: sendfile

* Example: sendfile

.code -edit sendfile/main.go /^func readFileHandler/,/^}/

* Example: sendfile

  ✗ strace ./sendfile
  fstat(6, {st_mode=S_IFREG|0644, st_size=1338, ...}) = 0
  read(6, "{\"id\":\"hello\",\"type"..., 1850) = 1338
  read(6, "", 512)                        = 0
  close(6)                                = 0
  write(4, "HTTP/1.1 200 OK\r\nContent-Disposi"..., 1515) = 1515

* Example: sendfile

.code -edit sendfile/main.go /^func copyHandler/,/^}/

* Example: sendfile

  ✗ strace ./sendfile
  read(6, "{\"id\":\"hello\",\"type"..., 512) = 512
  fstat(6, {st_mode=S_IFREG|0644, st_size=1338, ...}) = 0
  lseek(6, 0, SEEK_SET)                   = 0
  fstat(6, {st_mode=S_IFREG|0644, st_size=1338, ...}) = 0
  write(4, "HTTP/1.1 200 OK\r\nContent-Disposi"..., 177) = 177
  sendfile(4, 6, NULL, 4194304)           = 1338
  sendfile(4, 6, NULL, 4194304)           = 0
  close(6)                                = 0

* Example: sendfile

- http.ResponseWriter's is an io.ReaderFrom that uses the implementation of underlying tcp conn.

  // ReadFrom is here to optimize copying from an *os.File regular file
  // to a *net.TCPConn with sendfile.
  func (w *response) ReadFrom(src io.Reader) (n int64, err error) {

- net.TCPConn also implements io.ReaderFrom that uses *sendfile* system call.

  func (c *TCPConn) ReadFrom(r io.Reader) (int64, error) {

* ioutil.Discard

  // Discard is an io.Writer on which all Write calls succeed
  // without doing anything.
  var Discard io.Writer = devNull(0)

- Implements io.ReaderFrom!

.play -edit discard/main.go

* Reader implementations

Readers are all over the standard library:

- bufio iotest bytes strings crypto debug packet
- archive/...
- image/...
- compress/...
- encoding/...
- text/...
- // and many more...

* io.Writer

  Write(p []byte) (n int, err error)

- Write must return a non-nil error if it returns n < len(p)
- Write must not modify the slice data, even temporarily
- Implementations must not retain p

* Writer Variants

  type ReadWriter interface { Reader; Writer }
  type StringWriter interface { WriteString(s string) (n int, err error) }
  type ByteWriter interface { WriteByte(c byte) error }
  type PipeWriter struct // the write half of a pipe

*Composition*

  // MultiWriter creates a writer that duplicates its writes to all the provided writers, similar to the Unix tee(1) command.
  func MultiWriter(writers ...Writer) Writer

* Writer implementations

- iotest/...
- archive/...
- compress/...
- text/...
- net/http/ResponseWriter
- // and many more

* io.Pipe

.play -edit pipe/main.go /^func main/,/^}/

- synchronous
- in-memory
- no internal buffering

* iotest

Package iotest implements Readers and Writers useful mainly for testing.

  // DataErrReader creates a reader that returns (n > 0, EOF) at the end.
  func DataErrReader(r io.Reader) io.Reader

  // HalfReader returns a Reader that implements Read
  // by reading half as many requested bytes from r.
  func HalfReader(r io.Reader) io.Reader

  // OneByteReader returns a Reader that
  // implements each non-empty Read by reading one byte from r.
  func OneByteReader(r io.Reader) io.Reader

  // TimeoutReader returns ErrTimeout on the second read with no data.
  // Subsequent calls to read succeed.
  func TimeoutReader(r io.Reader) io.Reader

  // TruncateWriter returns a Writer that writes to w but stops silently after n bytes.
  func TruncateWriter(w io.Writer, n int64) io.Writer

* bufio

Package bufio implements buffered I/O. It wraps an io.Reader or io.Writer objects.

  type Reader
      func NewReader(rd io.Reader) *Reader
      func (b *Reader) Discard(n int) (discarded int, err error)
      func (b *Reader) Peek(n int) ([]byte, error)
      func (b *Reader) Read(p []byte) (n int, err error)
      func (b *Reader) ReadByte() (byte, error)
      func (b *Reader) ReadBytes(delim byte) ([]byte, error)
      func (b *Reader) ReadLine() (line []byte, isPrefix bool, err error)
      func (b *Reader) ReadRune() (r rune, size int, err error)
      func (b *Reader) ReadSlice(delim byte) (line []byte, err error)
      func (b *Reader) ReadString(delim byte) (string, error)

  type Writer
      func NewWriter(w io.Writer) *Writer
      func (b *Writer) Flush() error
      func (b *Writer) ReadFrom(r io.Reader) (n int64, err error)
      func (b *Writer) Write(p []byte) (nn int, err error)
      func (b *Writer) WriteByte(c byte) error
      func (b *Writer) WriteRune(r rune) (size int, err error)
      func (b *Writer) WriteString(s string) (int, error)

* bufio.Scanner

Utility type to efficiently read independent lines of text from an io.Reader.

.code -edit scanner/main.go

* bufio read line

There are multiple options to read single line. Which one to use
*ReadBytes('\n')* or *ReadString('\n')* or *ReadLine* or *Scanner*?

- *ReadBytes* returns the slice together with delimiter
- *ReadLine* doesn’t handle lines longer than internal buffer (default size 4096)
- *Scanner* has limited max size of the token (64 * 1024)
- *ReadLine* needs to be called for the second time to retrieve rest of the stream
- *ReadBytes* doesn’t have any limit
- *Scanner* has the simplest API and provides nicest abstraction for common cases

* bufio

- The net/http package already buffers data (using bufio itself) so you don't need this package for that
- If you are reading a file in one or a few large steps, you probably don't need buffering

* bytes.Buffer

A handy wrapper around byte slice implementing `io.Reader` and `io.Writer`.

Useful when you want to use code that takes an io.Writer, and store the results in memory for the later use.

.play -edit bytesbuffer/main.go

* bytes.Buffer vs strings.Builder

  bytes.Buffer
      func NewBufferString(s string) *Buffer
      func (b *Buffer) Bytes() []byte
      func (b *Buffer) Grow(n int)
      func (b *Buffer) Read(p []byte) (n int, err error)
      func (b *Buffer) ReadByte() (byte, error)
      // other read methods
      func (b *Buffer) Reset()
      func (b *Buffer) String() string
      func (b *Buffer) Write(p []byte) (n int, err error)
      func (b *Buffer) WriteByte(c byte) error
      func (b *Buffer) WriteRune(r rune) (n int, err error)
      func (b *Buffer) WriteString(s string) (n int, err error)

  strings.Builder
      func (b *Builder) Grow(n int)
      func (b *Builder) Reset()
      func (b *Builder) String() string
      func (b *Builder) Write(p []byte) (int, error)
      func (b *Builder) WriteByte(c byte) error
      func (b *Builder) WriteRune(r rune) (int, error)
      func (b *Builder) WriteString(s string) (int, error)

* bytes.Buffer vs strings.Builder

- *strings.Builder* is immutable and can only grow or reset
- *bytes.Buffer*'s internal byte slice can escape: (*Buffer).Bytes().
- strings.Builder.String() does not allocate/copy

  // String returns the accumulated string.
  func (b *Builder) String() string {
      return *(*string)(unsafe.Pointer(&b.buf))
  }

- bytes.Buffer.String() does

  func (b *Buffer) String() string {
      if b == nil {
          // Special case, useful in debugging.
          return "<nil>"
      }
      return string(b.buf[b.off:])
  }

* strings.Builder vs strings.Builder

.play -edit stringsbuilder/main.go

* bytes.Buffer vs strings.Builder

- *strings.Builder* has a copy check

  var b1 strings.Builder
  b1.WriteString("ABC")
  b2 := b1
  b2.WriteString("DEF")
  // illegal use of non-zero Builder copied by value

- Use pointer to share.

* *os.File

- portable
- implements io.Reader, and io.Writer which stream bytes to or from a file on disk
- useful if you don't want to read the whole file into memory
- has no internal buffers

*ioutil.ReadFile* reads an entire file into memory (as a []byte) in a single call

- allocates a byte slice of the correct size (no need to Read + append in a loop)
- closes the file
- returns the first error that prevented it from working

There are also

  func WriteFile(filename string, data []byte, perm os.FileMode) error
  func TempFile(dir, pattern string) (f *os.File, err error)

* Summary

- *io* defines interfaces that handle streams of bytes (Reader, Writer, etc...) as well as functions that work generically with types implement these interfaces (e.g. io.Copy)
- *io/ioutil* provides helper functions for some non-trivial file and io tasks
- *testing/iotest* implements Readers and Writers useful mainly for testing
- *bufio* provides buffering wrapper for io.Reader and io.Writer that can improve efficiency
- *bytes* provides helper functions and types for interacting with byte slices
- ***os.File* implements both io.Reader and io.Writer (among others)

* Links

.link https://medium.com/go-walkthrough/go-walkthrough-io-package-8ac5e95a9fbd io walkthrough
.link https://medium.com/golangspec/introduction-to-bufio-package-in-golang-ad7d1877f762 bufio walkthrough
.link https://www.youtube.com/watch?v=kTAsciVuZLQ advanced patterns with io.ReadWriter