...
Run Format

Source file src/mime/quotedprintable/reader.go

Documentation: mime/quotedprintable

  // Copyright 2012 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  // Package quotedprintable implements quoted-printable encoding as specified by
  // RFC 2045.
  package quotedprintable
  
  import (
  	"bufio"
  	"bytes"
  	"fmt"
  	"io"
  )
  
  // Reader is a quoted-printable decoder.
  type Reader struct {
  	br   *bufio.Reader
  	rerr error  // last read error
  	line []byte // to be consumed before more of br
  }
  
  // NewReader returns a quoted-printable reader, decoding from r.
  func NewReader(r io.Reader) *Reader {
  	return &Reader{
  		br: bufio.NewReader(r),
  	}
  }
  
  func fromHex(b byte) (byte, error) {
  	switch {
  	case b >= '0' && b <= '9':
  		return b - '0', nil
  	case b >= 'A' && b <= 'F':
  		return b - 'A' + 10, nil
  	// Accept badly encoded bytes.
  	case b >= 'a' && b <= 'f':
  		return b - 'a' + 10, nil
  	}
  	return 0, fmt.Errorf("quotedprintable: invalid hex byte 0x%02x", b)
  }
  
  func readHexByte(v []byte) (b byte, err error) {
  	if len(v) < 2 {
  		return 0, io.ErrUnexpectedEOF
  	}
  	var hb, lb byte
  	if hb, err = fromHex(v[0]); err != nil {
  		return 0, err
  	}
  	if lb, err = fromHex(v[1]); err != nil {
  		return 0, err
  	}
  	return hb<<4 | lb, nil
  }
  
  func isQPDiscardWhitespace(r rune) bool {
  	switch r {
  	case '\n', '\r', ' ', '\t':
  		return true
  	}
  	return false
  }
  
  var (
  	crlf       = []byte("\r\n")
  	lf         = []byte("\n")
  	softSuffix = []byte("=")
  )
  
  // Read reads and decodes quoted-printable data from the underlying reader.
  func (r *Reader) Read(p []byte) (n int, err error) {
  	// Deviations from RFC 2045:
  	// 1. in addition to "=\r\n", "=\n" is also treated as soft line break.
  	// 2. it will pass through a '\r' or '\n' not preceded by '=', consistent
  	//    with other broken QP encoders & decoders.
  	// 3. it accepts soft line-break (=) at end of message (issue 15486); i.e.
  	//    the final byte read from the underlying reader is allowed to be '=',
  	//    and it will be silently ignored.
  	// 4. it takes = as literal = if not followed by two hex digits
  	//    but not at end of line (issue 13219).
  	for len(p) > 0 {
  		if len(r.line) == 0 {
  			if r.rerr != nil {
  				return n, r.rerr
  			}
  			r.line, r.rerr = r.br.ReadSlice('\n')
  
  			// Does the line end in CRLF instead of just LF?
  			hasLF := bytes.HasSuffix(r.line, lf)
  			hasCR := bytes.HasSuffix(r.line, crlf)
  			wholeLine := r.line
  			r.line = bytes.TrimRightFunc(wholeLine, isQPDiscardWhitespace)
  			if bytes.HasSuffix(r.line, softSuffix) {
  				rightStripped := wholeLine[len(r.line):]
  				r.line = r.line[:len(r.line)-1]
  				if !bytes.HasPrefix(rightStripped, lf) && !bytes.HasPrefix(rightStripped, crlf) &&
  					!(len(rightStripped) == 0 && len(r.line) > 0 && r.rerr == io.EOF) {
  					r.rerr = fmt.Errorf("quotedprintable: invalid bytes after =: %q", rightStripped)
  				}
  			} else if hasLF {
  				if hasCR {
  					r.line = append(r.line, '\r', '\n')
  				} else {
  					r.line = append(r.line, '\n')
  				}
  			}
  			continue
  		}
  		b := r.line[0]
  
  		switch {
  		case b == '=':
  			b, err = readHexByte(r.line[1:])
  			if err != nil {
  				if len(r.line) >= 2 && r.line[1] != '\r' && r.line[1] != '\n' {
  					// Take the = as a literal =.
  					b = '='
  					break
  				}
  				return n, err
  			}
  			r.line = r.line[2:] // 2 of the 3; other 1 is done below
  		case b == '\t' || b == '\r' || b == '\n':
  			break
  		case b < ' ' || b > '~':
  			return n, fmt.Errorf("quotedprintable: invalid unescaped byte 0x%02x in body", b)
  		}
  		p[0] = b
  		p = p[1:]
  		r.line = r.line[1:]
  		n++
  	}
  	return n, nil
  }
  

View as plain text