override.go raw

   1  // Copyright 2015 The Go Authors. All rights reserved.
   2  // Use of this source code is governed by a BSD-style
   3  // license that can be found in the LICENSE file.
   4  
   5  package unicode
   6  
   7  import (
   8  	"golang.org/x/text/transform"
   9  )
  10  
  11  // BOMOverride returns a new decoder transformer that is identical to fallback,
  12  // except that the presence of a Byte Order Mark at the start of the input
  13  // causes it to switch to the corresponding Unicode decoding. It will only
  14  // consider BOMs for UTF-8, UTF-16BE, and UTF-16LE.
  15  //
  16  // This differs from using ExpectBOM by allowing a BOM to switch to UTF-8, not
  17  // just UTF-16 variants, and allowing falling back to any encoding scheme.
  18  //
  19  // This technique is recommended by the W3C for use in HTML 5: "For
  20  // compatibility with deployed content, the byte order mark (also known as BOM)
  21  // is considered more authoritative than anything else."
  22  // http://www.w3.org/TR/encoding/#specification-hooks
  23  //
  24  // Using BOMOverride is mostly intended for use cases where the first characters
  25  // of a fallback encoding are known to not be a BOM, for example, for valid HTML
  26  // and most encodings.
  27  func BOMOverride(fallback transform.Transformer) transform.Transformer {
  28  	// TODO: possibly allow a variadic argument of unicode encodings to allow
  29  	// specifying details of which fallbacks are supported as well as
  30  	// specifying the details of the implementations. This would also allow for
  31  	// support for UTF-32, which should not be supported by default.
  32  	return &bomOverride{fallback: fallback}
  33  }
  34  
  35  type bomOverride struct {
  36  	fallback transform.Transformer
  37  	current  transform.Transformer
  38  }
  39  
  40  func (d *bomOverride) Reset() {
  41  	d.current = nil
  42  	d.fallback.Reset()
  43  }
  44  
  45  var (
  46  	// TODO: we could use decode functions here, instead of allocating a new
  47  	// decoder on every NewDecoder as IgnoreBOM decoders can be stateless.
  48  	utf16le = UTF16(LittleEndian, IgnoreBOM)
  49  	utf16be = UTF16(BigEndian, IgnoreBOM)
  50  )
  51  
  52  const utf8BOM = "\ufeff"
  53  
  54  func (d *bomOverride) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
  55  	if d.current != nil {
  56  		return d.current.Transform(dst, src, atEOF)
  57  	}
  58  	if len(src) < 3 && !atEOF {
  59  		return 0, 0, transform.ErrShortSrc
  60  	}
  61  	d.current = d.fallback
  62  	bomSize := 0
  63  	if len(src) >= 2 {
  64  		if src[0] == 0xFF && src[1] == 0xFE {
  65  			d.current = utf16le.NewDecoder()
  66  			bomSize = 2
  67  		} else if src[0] == 0xFE && src[1] == 0xFF {
  68  			d.current = utf16be.NewDecoder()
  69  			bomSize = 2
  70  		} else if len(src) >= 3 &&
  71  			src[0] == utf8BOM[0] &&
  72  			src[1] == utf8BOM[1] &&
  73  			src[2] == utf8BOM[2] {
  74  			d.current = transform.Nop
  75  			bomSize = 3
  76  		}
  77  	}
  78  	if bomSize < len(src) {
  79  		nDst, nSrc, err = d.current.Transform(dst, src[bomSize:], atEOF)
  80  	}
  81  	return nDst, nSrc + bomSize, err
  82  }
  83