...
Run Format

Source file src/pkg/encoding/xml/xml.go

     1	// Copyright 2009 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 xml implements a simple XML 1.0 parser that
     6	// understands XML name spaces.
     7	package xml
     8	
     9	// References:
    10	//    Annotated XML spec: http://www.xml.com/axml/testaxml.htm
    11	//    XML name spaces: http://www.w3.org/TR/REC-xml-names/
    12	
    13	// TODO(rsc):
    14	//	Test error handling.
    15	
    16	import (
    17		"bufio"
    18		"bytes"
    19		"errors"
    20		"fmt"
    21		"io"
    22		"strconv"
    23		"strings"
    24		"unicode"
    25		"unicode/utf8"
    26	)
    27	
    28	// A SyntaxError represents a syntax error in the XML input stream.
    29	type SyntaxError struct {
    30		Msg  string
    31		Line int
    32	}
    33	
    34	func (e *SyntaxError) Error() string {
    35		return "XML syntax error on line " + strconv.Itoa(e.Line) + ": " + e.Msg
    36	}
    37	
    38	// A Name represents an XML name (Local) annotated
    39	// with a name space identifier (Space).
    40	// In tokens returned by Decoder.Token, the Space identifier
    41	// is given as a canonical URL, not the short prefix used
    42	// in the document being parsed.
    43	type Name struct {
    44		Space, Local string
    45	}
    46	
    47	// An Attr represents an attribute in an XML element (Name=Value).
    48	type Attr struct {
    49		Name  Name
    50		Value string
    51	}
    52	
    53	// A Token is an interface holding one of the token types:
    54	// StartElement, EndElement, CharData, Comment, ProcInst, or Directive.
    55	type Token interface{}
    56	
    57	// A StartElement represents an XML start element.
    58	type StartElement struct {
    59		Name Name
    60		Attr []Attr
    61	}
    62	
    63	func (e StartElement) Copy() StartElement {
    64		attrs := make([]Attr, len(e.Attr))
    65		copy(attrs, e.Attr)
    66		e.Attr = attrs
    67		return e
    68	}
    69	
    70	// End returns the corresponding XML end element.
    71	func (e StartElement) End() EndElement {
    72		return EndElement{e.Name}
    73	}
    74	
    75	// An EndElement represents an XML end element.
    76	type EndElement struct {
    77		Name Name
    78	}
    79	
    80	// A CharData represents XML character data (raw text),
    81	// in which XML escape sequences have been replaced by
    82	// the characters they represent.
    83	type CharData []byte
    84	
    85	func makeCopy(b []byte) []byte {
    86		b1 := make([]byte, len(b))
    87		copy(b1, b)
    88		return b1
    89	}
    90	
    91	func (c CharData) Copy() CharData { return CharData(makeCopy(c)) }
    92	
    93	// A Comment represents an XML comment of the form <!--comment-->.
    94	// The bytes do not include the <!-- and --> comment markers.
    95	type Comment []byte
    96	
    97	func (c Comment) Copy() Comment { return Comment(makeCopy(c)) }
    98	
    99	// A ProcInst represents an XML processing instruction of the form <?target inst?>
   100	type ProcInst struct {
   101		Target string
   102		Inst   []byte
   103	}
   104	
   105	func (p ProcInst) Copy() ProcInst {
   106		p.Inst = makeCopy(p.Inst)
   107		return p
   108	}
   109	
   110	// A Directive represents an XML directive of the form <!text>.
   111	// The bytes do not include the <! and > markers.
   112	type Directive []byte
   113	
   114	func (d Directive) Copy() Directive { return Directive(makeCopy(d)) }
   115	
   116	// CopyToken returns a copy of a Token.
   117	func CopyToken(t Token) Token {
   118		switch v := t.(type) {
   119		case CharData:
   120			return v.Copy()
   121		case Comment:
   122			return v.Copy()
   123		case Directive:
   124			return v.Copy()
   125		case ProcInst:
   126			return v.Copy()
   127		case StartElement:
   128			return v.Copy()
   129		}
   130		return t
   131	}
   132	
   133	// A Decoder represents an XML parser reading a particular input stream.
   134	// The parser assumes that its input is encoded in UTF-8.
   135	type Decoder struct {
   136		// Strict defaults to true, enforcing the requirements
   137		// of the XML specification.
   138		// If set to false, the parser allows input containing common
   139		// mistakes:
   140		//	* If an element is missing an end tag, the parser invents
   141		//	  end tags as necessary to keep the return values from Token
   142		//	  properly balanced.
   143		//	* In attribute values and character data, unknown or malformed
   144		//	  character entities (sequences beginning with &) are left alone.
   145		//
   146		// Setting:
   147		//
   148		//	d.Strict = false;
   149		//	d.AutoClose = HTMLAutoClose;
   150		//	d.Entity = HTMLEntity
   151		//
   152		// creates a parser that can handle typical HTML.
   153		//
   154		// Strict mode does not enforce the requirements of the XML name spaces TR.
   155		// In particular it does not reject name space tags using undefined prefixes.
   156		// Such tags are recorded with the unknown prefix as the name space URL.
   157		Strict bool
   158	
   159		// When Strict == false, AutoClose indicates a set of elements to
   160		// consider closed immediately after they are opened, regardless
   161		// of whether an end element is present.
   162		AutoClose []string
   163	
   164		// Entity can be used to map non-standard entity names to string replacements.
   165		// The parser behaves as if these standard mappings are present in the map,
   166		// regardless of the actual map content:
   167		//
   168		//	"lt": "<",
   169		//	"gt": ">",
   170		//	"amp": "&",
   171		//	"apos": "'",
   172		//	"quot": `"`,
   173		Entity map[string]string
   174	
   175		// CharsetReader, if non-nil, defines a function to generate
   176		// charset-conversion readers, converting from the provided
   177		// non-UTF-8 charset into UTF-8. If CharsetReader is nil or
   178		// returns an error, parsing stops with an error. One of the
   179		// the CharsetReader's result values must be non-nil.
   180		CharsetReader func(charset string, input io.Reader) (io.Reader, error)
   181	
   182		// DefaultSpace sets the default name space used for unadorned tags,
   183		// as if the entire XML stream were wrapped in an element containing
   184		// the attribute xmlns="DefaultSpace".
   185		DefaultSpace string
   186	
   187		r              io.ByteReader
   188		buf            bytes.Buffer
   189		saved          *bytes.Buffer
   190		stk            *stack
   191		free           *stack
   192		needClose      bool
   193		toClose        Name
   194		nextToken      Token
   195		nextByte       int
   196		ns             map[string]string
   197		err            error
   198		line           int
   199		unmarshalDepth int
   200	}
   201	
   202	// NewDecoder creates a new XML parser reading from r.
   203	// If r does not implement io.ByteReader, NewDecoder will
   204	// do its own buffering.
   205	func NewDecoder(r io.Reader) *Decoder {
   206		d := &Decoder{
   207			ns:       make(map[string]string),
   208			nextByte: -1,
   209			line:     1,
   210			Strict:   true,
   211		}
   212		d.switchToReader(r)
   213		return d
   214	}
   215	
   216	// Token returns the next XML token in the input stream.
   217	// At the end of the input stream, Token returns nil, io.EOF.
   218	//
   219	// Slices of bytes in the returned token data refer to the
   220	// parser's internal buffer and remain valid only until the next
   221	// call to Token.  To acquire a copy of the bytes, call CopyToken
   222	// or the token's Copy method.
   223	//
   224	// Token expands self-closing elements such as <br/>
   225	// into separate start and end elements returned by successive calls.
   226	//
   227	// Token guarantees that the StartElement and EndElement
   228	// tokens it returns are properly nested and matched:
   229	// if Token encounters an unexpected end element,
   230	// it will return an error.
   231	//
   232	// Token implements XML name spaces as described by
   233	// http://www.w3.org/TR/REC-xml-names/.  Each of the
   234	// Name structures contained in the Token has the Space
   235	// set to the URL identifying its name space when known.
   236	// If Token encounters an unrecognized name space prefix,
   237	// it uses the prefix as the Space rather than report an error.
   238	func (d *Decoder) Token() (t Token, err error) {
   239		if d.stk != nil && d.stk.kind == stkEOF {
   240			err = io.EOF
   241			return
   242		}
   243		if d.nextToken != nil {
   244			t = d.nextToken
   245			d.nextToken = nil
   246		} else if t, err = d.rawToken(); err != nil {
   247			return
   248		}
   249	
   250		if !d.Strict {
   251			if t1, ok := d.autoClose(t); ok {
   252				d.nextToken = t
   253				t = t1
   254			}
   255		}
   256		switch t1 := t.(type) {
   257		case StartElement:
   258			// In XML name spaces, the translations listed in the
   259			// attributes apply to the element name and
   260			// to the other attribute names, so process
   261			// the translations first.
   262			for _, a := range t1.Attr {
   263				if a.Name.Space == "xmlns" {
   264					v, ok := d.ns[a.Name.Local]
   265					d.pushNs(a.Name.Local, v, ok)
   266					d.ns[a.Name.Local] = a.Value
   267				}
   268				if a.Name.Space == "" && a.Name.Local == "xmlns" {
   269					// Default space for untagged names
   270					v, ok := d.ns[""]
   271					d.pushNs("", v, ok)
   272					d.ns[""] = a.Value
   273				}
   274			}
   275	
   276			d.translate(&t1.Name, true)
   277			for i := range t1.Attr {
   278				d.translate(&t1.Attr[i].Name, false)
   279			}
   280			d.pushElement(t1.Name)
   281			t = t1
   282	
   283		case EndElement:
   284			d.translate(&t1.Name, true)
   285			if !d.popElement(&t1) {
   286				return nil, d.err
   287			}
   288			t = t1
   289		}
   290		return
   291	}
   292	
   293	const xmlURL = "http://www.w3.org/XML/1998/namespace"
   294	
   295	// Apply name space translation to name n.
   296	// The default name space (for Space=="")
   297	// applies only to element names, not to attribute names.
   298	func (d *Decoder) translate(n *Name, isElementName bool) {
   299		switch {
   300		case n.Space == "xmlns":
   301			return
   302		case n.Space == "" && !isElementName:
   303			return
   304		case n.Space == "xml":
   305			n.Space = xmlURL
   306		case n.Space == "" && n.Local == "xmlns":
   307			return
   308		}
   309		if v, ok := d.ns[n.Space]; ok {
   310			n.Space = v
   311		} else if n.Space == "" {
   312			n.Space = d.DefaultSpace
   313		}
   314	}
   315	
   316	func (d *Decoder) switchToReader(r io.Reader) {
   317		// Get efficient byte at a time reader.
   318		// Assume that if reader has its own
   319		// ReadByte, it's efficient enough.
   320		// Otherwise, use bufio.
   321		if rb, ok := r.(io.ByteReader); ok {
   322			d.r = rb
   323		} else {
   324			d.r = bufio.NewReader(r)
   325		}
   326	}
   327	
   328	// Parsing state - stack holds old name space translations
   329	// and the current set of open elements.  The translations to pop when
   330	// ending a given tag are *below* it on the stack, which is
   331	// more work but forced on us by XML.
   332	type stack struct {
   333		next *stack
   334		kind int
   335		name Name
   336		ok   bool
   337	}
   338	
   339	const (
   340		stkStart = iota
   341		stkNs
   342		stkEOF
   343	)
   344	
   345	func (d *Decoder) push(kind int) *stack {
   346		s := d.free
   347		if s != nil {
   348			d.free = s.next
   349		} else {
   350			s = new(stack)
   351		}
   352		s.next = d.stk
   353		s.kind = kind
   354		d.stk = s
   355		return s
   356	}
   357	
   358	func (d *Decoder) pop() *stack {
   359		s := d.stk
   360		if s != nil {
   361			d.stk = s.next
   362			s.next = d.free
   363			d.free = s
   364		}
   365		return s
   366	}
   367	
   368	// Record that after the current element is finished
   369	// (that element is already pushed on the stack)
   370	// Token should return EOF until popEOF is called.
   371	func (d *Decoder) pushEOF() {
   372		// Walk down stack to find Start.
   373		// It might not be the top, because there might be stkNs
   374		// entries above it.
   375		start := d.stk
   376		for start.kind != stkStart {
   377			start = start.next
   378		}
   379		// The stkNs entries below a start are associated with that
   380		// element too; skip over them.
   381		for start.next != nil && start.next.kind == stkNs {
   382			start = start.next
   383		}
   384		s := d.free
   385		if s != nil {
   386			d.free = s.next
   387		} else {
   388			s = new(stack)
   389		}
   390		s.kind = stkEOF
   391		s.next = start.next
   392		start.next = s
   393	}
   394	
   395	// Undo a pushEOF.
   396	// The element must have been finished, so the EOF should be at the top of the stack.
   397	func (d *Decoder) popEOF() bool {
   398		if d.stk == nil || d.stk.kind != stkEOF {
   399			return false
   400		}
   401		d.pop()
   402		return true
   403	}
   404	
   405	// Record that we are starting an element with the given name.
   406	func (d *Decoder) pushElement(name Name) {
   407		s := d.push(stkStart)
   408		s.name = name
   409	}
   410	
   411	// Record that we are changing the value of ns[local].
   412	// The old value is url, ok.
   413	func (d *Decoder) pushNs(local string, url string, ok bool) {
   414		s := d.push(stkNs)
   415		s.name.Local = local
   416		s.name.Space = url
   417		s.ok = ok
   418	}
   419	
   420	// Creates a SyntaxError with the current line number.
   421	func (d *Decoder) syntaxError(msg string) error {
   422		return &SyntaxError{Msg: msg, Line: d.line}
   423	}
   424	
   425	// Record that we are ending an element with the given name.
   426	// The name must match the record at the top of the stack,
   427	// which must be a pushElement record.
   428	// After popping the element, apply any undo records from
   429	// the stack to restore the name translations that existed
   430	// before we saw this element.
   431	func (d *Decoder) popElement(t *EndElement) bool {
   432		s := d.pop()
   433		name := t.Name
   434		switch {
   435		case s == nil || s.kind != stkStart:
   436			d.err = d.syntaxError("unexpected end element </" + name.Local + ">")
   437			return false
   438		case s.name.Local != name.Local:
   439			if !d.Strict {
   440				d.needClose = true
   441				d.toClose = t.Name
   442				t.Name = s.name
   443				return true
   444			}
   445			d.err = d.syntaxError("element <" + s.name.Local + "> closed by </" + name.Local + ">")
   446			return false
   447		case s.name.Space != name.Space:
   448			d.err = d.syntaxError("element <" + s.name.Local + "> in space " + s.name.Space +
   449				"closed by </" + name.Local + "> in space " + name.Space)
   450			return false
   451		}
   452	
   453		// Pop stack until a Start or EOF is on the top, undoing the
   454		// translations that were associated with the element we just closed.
   455		for d.stk != nil && d.stk.kind != stkStart && d.stk.kind != stkEOF {
   456			s := d.pop()
   457			if s.ok {
   458				d.ns[s.name.Local] = s.name.Space
   459			} else {
   460				delete(d.ns, s.name.Local)
   461			}
   462		}
   463	
   464		return true
   465	}
   466	
   467	// If the top element on the stack is autoclosing and
   468	// t is not the end tag, invent the end tag.
   469	func (d *Decoder) autoClose(t Token) (Token, bool) {
   470		if d.stk == nil || d.stk.kind != stkStart {
   471			return nil, false
   472		}
   473		name := strings.ToLower(d.stk.name.Local)
   474		for _, s := range d.AutoClose {
   475			if strings.ToLower(s) == name {
   476				// This one should be auto closed if t doesn't close it.
   477				et, ok := t.(EndElement)
   478				if !ok || et.Name.Local != name {
   479					return EndElement{d.stk.name}, true
   480				}
   481				break
   482			}
   483		}
   484		return nil, false
   485	}
   486	
   487	var errRawToken = errors.New("xml: cannot use RawToken from UnmarshalXML method")
   488	
   489	// RawToken is like Token but does not verify that
   490	// start and end elements match and does not translate
   491	// name space prefixes to their corresponding URLs.
   492	func (d *Decoder) RawToken() (Token, error) {
   493		if d.unmarshalDepth > 0 {
   494			return nil, errRawToken
   495		}
   496		return d.rawToken()
   497	}
   498	
   499	func (d *Decoder) rawToken() (Token, error) {
   500		if d.err != nil {
   501			return nil, d.err
   502		}
   503		if d.needClose {
   504			// The last element we read was self-closing and
   505			// we returned just the StartElement half.
   506			// Return the EndElement half now.
   507			d.needClose = false
   508			return EndElement{d.toClose}, nil
   509		}
   510	
   511		b, ok := d.getc()
   512		if !ok {
   513			return nil, d.err
   514		}
   515	
   516		if b != '<' {
   517			// Text section.
   518			d.ungetc(b)
   519			data := d.text(-1, false)
   520			if data == nil {
   521				return nil, d.err
   522			}
   523			return CharData(data), nil
   524		}
   525	
   526		if b, ok = d.mustgetc(); !ok {
   527			return nil, d.err
   528		}
   529		switch b {
   530		case '/':
   531			// </: End element
   532			var name Name
   533			if name, ok = d.nsname(); !ok {
   534				if d.err == nil {
   535					d.err = d.syntaxError("expected element name after </")
   536				}
   537				return nil, d.err
   538			}
   539			d.space()
   540			if b, ok = d.mustgetc(); !ok {
   541				return nil, d.err
   542			}
   543			if b != '>' {
   544				d.err = d.syntaxError("invalid characters between </" + name.Local + " and >")
   545				return nil, d.err
   546			}
   547			return EndElement{name}, nil
   548	
   549		case '?':
   550			// <?: Processing instruction.
   551			// TODO(rsc): Should parse the <?xml declaration to make sure the version is 1.0.
   552			var target string
   553			if target, ok = d.name(); !ok {
   554				if d.err == nil {
   555					d.err = d.syntaxError("expected target name after <?")
   556				}
   557				return nil, d.err
   558			}
   559			d.space()
   560			d.buf.Reset()
   561			var b0 byte
   562			for {
   563				if b, ok = d.mustgetc(); !ok {
   564					return nil, d.err
   565				}
   566				d.buf.WriteByte(b)
   567				if b0 == '?' && b == '>' {
   568					break
   569				}
   570				b0 = b
   571			}
   572			data := d.buf.Bytes()
   573			data = data[0 : len(data)-2] // chop ?>
   574	
   575			if target == "xml" {
   576				enc := procInstEncoding(string(data))
   577				if enc != "" && enc != "utf-8" && enc != "UTF-8" {
   578					if d.CharsetReader == nil {
   579						d.err = fmt.Errorf("xml: encoding %q declared but Decoder.CharsetReader is nil", enc)
   580						return nil, d.err
   581					}
   582					newr, err := d.CharsetReader(enc, d.r.(io.Reader))
   583					if err != nil {
   584						d.err = fmt.Errorf("xml: opening charset %q: %v", enc, err)
   585						return nil, d.err
   586					}
   587					if newr == nil {
   588						panic("CharsetReader returned a nil Reader for charset " + enc)
   589					}
   590					d.switchToReader(newr)
   591				}
   592			}
   593			return ProcInst{target, data}, nil
   594	
   595		case '!':
   596			// <!: Maybe comment, maybe CDATA.
   597			if b, ok = d.mustgetc(); !ok {
   598				return nil, d.err
   599			}
   600			switch b {
   601			case '-': // <!-
   602				// Probably <!-- for a comment.
   603				if b, ok = d.mustgetc(); !ok {
   604					return nil, d.err
   605				}
   606				if b != '-' {
   607					d.err = d.syntaxError("invalid sequence <!- not part of <!--")
   608					return nil, d.err
   609				}
   610				// Look for terminator.
   611				d.buf.Reset()
   612				var b0, b1 byte
   613				for {
   614					if b, ok = d.mustgetc(); !ok {
   615						return nil, d.err
   616					}
   617					d.buf.WriteByte(b)
   618					if b0 == '-' && b1 == '-' && b == '>' {
   619						break
   620					}
   621					b0, b1 = b1, b
   622				}
   623				data := d.buf.Bytes()
   624				data = data[0 : len(data)-3] // chop -->
   625				return Comment(data), nil
   626	
   627			case '[': // <![
   628				// Probably <![CDATA[.
   629				for i := 0; i < 6; i++ {
   630					if b, ok = d.mustgetc(); !ok {
   631						return nil, d.err
   632					}
   633					if b != "CDATA["[i] {
   634						d.err = d.syntaxError("invalid <![ sequence")
   635						return nil, d.err
   636					}
   637				}
   638				// Have <![CDATA[.  Read text until ]]>.
   639				data := d.text(-1, true)
   640				if data == nil {
   641					return nil, d.err
   642				}
   643				return CharData(data), nil
   644			}
   645	
   646			// Probably a directive: <!DOCTYPE ...>, <!ENTITY ...>, etc.
   647			// We don't care, but accumulate for caller. Quoted angle
   648			// brackets do not count for nesting.
   649			d.buf.Reset()
   650			d.buf.WriteByte(b)
   651			inquote := uint8(0)
   652			depth := 0
   653			for {
   654				if b, ok = d.mustgetc(); !ok {
   655					return nil, d.err
   656				}
   657				if inquote == 0 && b == '>' && depth == 0 {
   658					break
   659				}
   660			HandleB:
   661				d.buf.WriteByte(b)
   662				switch {
   663				case b == inquote:
   664					inquote = 0
   665	
   666				case inquote != 0:
   667					// in quotes, no special action
   668	
   669				case b == '\'' || b == '"':
   670					inquote = b
   671	
   672				case b == '>' && inquote == 0:
   673					depth--
   674	
   675				case b == '<' && inquote == 0:
   676					// Look for <!-- to begin comment.
   677					s := "!--"
   678					for i := 0; i < len(s); i++ {
   679						if b, ok = d.mustgetc(); !ok {
   680							return nil, d.err
   681						}
   682						if b != s[i] {
   683							for j := 0; j < i; j++ {
   684								d.buf.WriteByte(s[j])
   685							}
   686							depth++
   687							goto HandleB
   688						}
   689					}
   690	
   691					// Remove < that was written above.
   692					d.buf.Truncate(d.buf.Len() - 1)
   693	
   694					// Look for terminator.
   695					var b0, b1 byte
   696					for {
   697						if b, ok = d.mustgetc(); !ok {
   698							return nil, d.err
   699						}
   700						if b0 == '-' && b1 == '-' && b == '>' {
   701							break
   702						}
   703						b0, b1 = b1, b
   704					}
   705				}
   706			}
   707			return Directive(d.buf.Bytes()), nil
   708		}
   709	
   710		// Must be an open element like <a href="foo">
   711		d.ungetc(b)
   712	
   713		var (
   714			name  Name
   715			empty bool
   716			attr  []Attr
   717		)
   718		if name, ok = d.nsname(); !ok {
   719			if d.err == nil {
   720				d.err = d.syntaxError("expected element name after <")
   721			}
   722			return nil, d.err
   723		}
   724	
   725		attr = make([]Attr, 0, 4)
   726		for {
   727			d.space()
   728			if b, ok = d.mustgetc(); !ok {
   729				return nil, d.err
   730			}
   731			if b == '/' {
   732				empty = true
   733				if b, ok = d.mustgetc(); !ok {
   734					return nil, d.err
   735				}
   736				if b != '>' {
   737					d.err = d.syntaxError("expected /> in element")
   738					return nil, d.err
   739				}
   740				break
   741			}
   742			if b == '>' {
   743				break
   744			}
   745			d.ungetc(b)
   746	
   747			n := len(attr)
   748			if n >= cap(attr) {
   749				nattr := make([]Attr, n, 2*cap(attr))
   750				copy(nattr, attr)
   751				attr = nattr
   752			}
   753			attr = attr[0 : n+1]
   754			a := &attr[n]
   755			if a.Name, ok = d.nsname(); !ok {
   756				if d.err == nil {
   757					d.err = d.syntaxError("expected attribute name in element")
   758				}
   759				return nil, d.err
   760			}
   761			d.space()
   762			if b, ok = d.mustgetc(); !ok {
   763				return nil, d.err
   764			}
   765			if b != '=' {
   766				if d.Strict {
   767					d.err = d.syntaxError("attribute name without = in element")
   768					return nil, d.err
   769				} else {
   770					d.ungetc(b)
   771					a.Value = a.Name.Local
   772				}
   773			} else {
   774				d.space()
   775				data := d.attrval()
   776				if data == nil {
   777					return nil, d.err
   778				}
   779				a.Value = string(data)
   780			}
   781		}
   782		if empty {
   783			d.needClose = true
   784			d.toClose = name
   785		}
   786		return StartElement{name, attr}, nil
   787	}
   788	
   789	func (d *Decoder) attrval() []byte {
   790		b, ok := d.mustgetc()
   791		if !ok {
   792			return nil
   793		}
   794		// Handle quoted attribute values
   795		if b == '"' || b == '\'' {
   796			return d.text(int(b), false)
   797		}
   798		// Handle unquoted attribute values for strict parsers
   799		if d.Strict {
   800			d.err = d.syntaxError("unquoted or missing attribute value in element")
   801			return nil
   802		}
   803		// Handle unquoted attribute values for unstrict parsers
   804		d.ungetc(b)
   805		d.buf.Reset()
   806		for {
   807			b, ok = d.mustgetc()
   808			if !ok {
   809				return nil
   810			}
   811			// http://www.w3.org/TR/REC-html40/intro/sgmltut.html#h-3.2.2
   812			if 'a' <= b && b <= 'z' || 'A' <= b && b <= 'Z' ||
   813				'0' <= b && b <= '9' || b == '_' || b == ':' || b == '-' {
   814				d.buf.WriteByte(b)
   815			} else {
   816				d.ungetc(b)
   817				break
   818			}
   819		}
   820		return d.buf.Bytes()
   821	}
   822	
   823	// Skip spaces if any
   824	func (d *Decoder) space() {
   825		for {
   826			b, ok := d.getc()
   827			if !ok {
   828				return
   829			}
   830			switch b {
   831			case ' ', '\r', '\n', '\t':
   832			default:
   833				d.ungetc(b)
   834				return
   835			}
   836		}
   837	}
   838	
   839	// Read a single byte.
   840	// If there is no byte to read, return ok==false
   841	// and leave the error in d.err.
   842	// Maintain line number.
   843	func (d *Decoder) getc() (b byte, ok bool) {
   844		if d.err != nil {
   845			return 0, false
   846		}
   847		if d.nextByte >= 0 {
   848			b = byte(d.nextByte)
   849			d.nextByte = -1
   850		} else {
   851			b, d.err = d.r.ReadByte()
   852			if d.err != nil {
   853				return 0, false
   854			}
   855			if d.saved != nil {
   856				d.saved.WriteByte(b)
   857			}
   858		}
   859		if b == '\n' {
   860			d.line++
   861		}
   862		return b, true
   863	}
   864	
   865	// Return saved offset.
   866	// If we did ungetc (nextByte >= 0), have to back up one.
   867	func (d *Decoder) savedOffset() int {
   868		n := d.saved.Len()
   869		if d.nextByte >= 0 {
   870			n--
   871		}
   872		return n
   873	}
   874	
   875	// Must read a single byte.
   876	// If there is no byte to read,
   877	// set d.err to SyntaxError("unexpected EOF")
   878	// and return ok==false
   879	func (d *Decoder) mustgetc() (b byte, ok bool) {
   880		if b, ok = d.getc(); !ok {
   881			if d.err == io.EOF {
   882				d.err = d.syntaxError("unexpected EOF")
   883			}
   884		}
   885		return
   886	}
   887	
   888	// Unread a single byte.
   889	func (d *Decoder) ungetc(b byte) {
   890		if b == '\n' {
   891			d.line--
   892		}
   893		d.nextByte = int(b)
   894	}
   895	
   896	var entity = map[string]int{
   897		"lt":   '<',
   898		"gt":   '>',
   899		"amp":  '&',
   900		"apos": '\'',
   901		"quot": '"',
   902	}
   903	
   904	// Read plain text section (XML calls it character data).
   905	// If quote >= 0, we are in a quoted string and need to find the matching quote.
   906	// If cdata == true, we are in a <![CDATA[ section and need to find ]]>.
   907	// On failure return nil and leave the error in d.err.
   908	func (d *Decoder) text(quote int, cdata bool) []byte {
   909		var b0, b1 byte
   910		var trunc int
   911		d.buf.Reset()
   912	Input:
   913		for {
   914			b, ok := d.getc()
   915			if !ok {
   916				if cdata {
   917					if d.err == io.EOF {
   918						d.err = d.syntaxError("unexpected EOF in CDATA section")
   919					}
   920					return nil
   921				}
   922				break Input
   923			}
   924	
   925			// <![CDATA[ section ends with ]]>.
   926			// It is an error for ]]> to appear in ordinary text.
   927			if b0 == ']' && b1 == ']' && b == '>' {
   928				if cdata {
   929					trunc = 2
   930					break Input
   931				}
   932				d.err = d.syntaxError("unescaped ]]> not in CDATA section")
   933				return nil
   934			}
   935	
   936			// Stop reading text if we see a <.
   937			if b == '<' && !cdata {
   938				if quote >= 0 {
   939					d.err = d.syntaxError("unescaped < inside quoted string")
   940					return nil
   941				}
   942				d.ungetc('<')
   943				break Input
   944			}
   945			if quote >= 0 && b == byte(quote) {
   946				break Input
   947			}
   948			if b == '&' && !cdata {
   949				// Read escaped character expression up to semicolon.
   950				// XML in all its glory allows a document to define and use
   951				// its own character names with <!ENTITY ...> directives.
   952				// Parsers are required to recognize lt, gt, amp, apos, and quot
   953				// even if they have not been declared.
   954				before := d.buf.Len()
   955				d.buf.WriteByte('&')
   956				var ok bool
   957				var text string
   958				var haveText bool
   959				if b, ok = d.mustgetc(); !ok {
   960					return nil
   961				}
   962				if b == '#' {
   963					d.buf.WriteByte(b)
   964					if b, ok = d.mustgetc(); !ok {
   965						return nil
   966					}
   967					base := 10
   968					if b == 'x' {
   969						base = 16
   970						d.buf.WriteByte(b)
   971						if b, ok = d.mustgetc(); !ok {
   972							return nil
   973						}
   974					}
   975					start := d.buf.Len()
   976					for '0' <= b && b <= '9' ||
   977						base == 16 && 'a' <= b && b <= 'f' ||
   978						base == 16 && 'A' <= b && b <= 'F' {
   979						d.buf.WriteByte(b)
   980						if b, ok = d.mustgetc(); !ok {
   981							return nil
   982						}
   983					}
   984					if b != ';' {
   985						d.ungetc(b)
   986					} else {
   987						s := string(d.buf.Bytes()[start:])
   988						d.buf.WriteByte(';')
   989						n, err := strconv.ParseUint(s, base, 64)
   990						if err == nil && n <= unicode.MaxRune {
   991							text = string(n)
   992							haveText = true
   993						}
   994					}
   995				} else {
   996					d.ungetc(b)
   997					if !d.readName() {
   998						if d.err != nil {
   999							return nil
  1000						}
  1001						ok = false
  1002					}
  1003					if b, ok = d.mustgetc(); !ok {
  1004						return nil
  1005					}
  1006					if b != ';' {
  1007						d.ungetc(b)
  1008					} else {
  1009						name := d.buf.Bytes()[before+1:]
  1010						d.buf.WriteByte(';')
  1011						if isName(name) {
  1012							s := string(name)
  1013							if r, ok := entity[s]; ok {
  1014								text = string(r)
  1015								haveText = true
  1016							} else if d.Entity != nil {
  1017								text, haveText = d.Entity[s]
  1018							}
  1019						}
  1020					}
  1021				}
  1022	
  1023				if haveText {
  1024					d.buf.Truncate(before)
  1025					d.buf.Write([]byte(text))
  1026					b0, b1 = 0, 0
  1027					continue Input
  1028				}
  1029				if !d.Strict {
  1030					b0, b1 = 0, 0
  1031					continue Input
  1032				}
  1033				ent := string(d.buf.Bytes()[before:])
  1034				if ent[len(ent)-1] != ';' {
  1035					ent += " (no semicolon)"
  1036				}
  1037				d.err = d.syntaxError("invalid character entity " + ent)
  1038				return nil
  1039			}
  1040	
  1041			// We must rewrite unescaped \r and \r\n into \n.
  1042			if b == '\r' {
  1043				d.buf.WriteByte('\n')
  1044			} else if b1 == '\r' && b == '\n' {
  1045				// Skip \r\n--we already wrote \n.
  1046			} else {
  1047				d.buf.WriteByte(b)
  1048			}
  1049	
  1050			b0, b1 = b1, b
  1051		}
  1052		data := d.buf.Bytes()
  1053		data = data[0 : len(data)-trunc]
  1054	
  1055		// Inspect each rune for being a disallowed character.
  1056		buf := data
  1057		for len(buf) > 0 {
  1058			r, size := utf8.DecodeRune(buf)
  1059			if r == utf8.RuneError && size == 1 {
  1060				d.err = d.syntaxError("invalid UTF-8")
  1061				return nil
  1062			}
  1063			buf = buf[size:]
  1064			if !isInCharacterRange(r) {
  1065				d.err = d.syntaxError(fmt.Sprintf("illegal character code %U", r))
  1066				return nil
  1067			}
  1068		}
  1069	
  1070		return data
  1071	}
  1072	
  1073	// Decide whether the given rune is in the XML Character Range, per
  1074	// the Char production of http://www.xml.com/axml/testaxml.htm,
  1075	// Section 2.2 Characters.
  1076	func isInCharacterRange(r rune) (inrange bool) {
  1077		return r == 0x09 ||
  1078			r == 0x0A ||
  1079			r == 0x0D ||
  1080			r >= 0x20 && r <= 0xDF77 ||
  1081			r >= 0xE000 && r <= 0xFFFD ||
  1082			r >= 0x10000 && r <= 0x10FFFF
  1083	}
  1084	
  1085	// Get name space name: name with a : stuck in the middle.
  1086	// The part before the : is the name space identifier.
  1087	func (d *Decoder) nsname() (name Name, ok bool) {
  1088		s, ok := d.name()
  1089		if !ok {
  1090			return
  1091		}
  1092		i := strings.Index(s, ":")
  1093		if i < 0 {
  1094			name.Local = s
  1095		} else {
  1096			name.Space = s[0:i]
  1097			name.Local = s[i+1:]
  1098		}
  1099		return name, true
  1100	}
  1101	
  1102	// Get name: /first(first|second)*/
  1103	// Do not set d.err if the name is missing (unless unexpected EOF is received):
  1104	// let the caller provide better context.
  1105	func (d *Decoder) name() (s string, ok bool) {
  1106		d.buf.Reset()
  1107		if !d.readName() {
  1108			return "", false
  1109		}
  1110	
  1111		// Now we check the characters.
  1112		s = d.buf.String()
  1113		if !isName([]byte(s)) {
  1114			d.err = d.syntaxError("invalid XML name: " + s)
  1115			return "", false
  1116		}
  1117		return s, true
  1118	}
  1119	
  1120	// Read a name and append its bytes to d.buf.
  1121	// The name is delimited by any single-byte character not valid in names.
  1122	// All multi-byte characters are accepted; the caller must check their validity.
  1123	func (d *Decoder) readName() (ok bool) {
  1124		var b byte
  1125		if b, ok = d.mustgetc(); !ok {
  1126			return
  1127		}
  1128		if b < utf8.RuneSelf && !isNameByte(b) {
  1129			d.ungetc(b)
  1130			return false
  1131		}
  1132		d.buf.WriteByte(b)
  1133	
  1134		for {
  1135			if b, ok = d.mustgetc(); !ok {
  1136				return
  1137			}
  1138			if b < utf8.RuneSelf && !isNameByte(b) {
  1139				d.ungetc(b)
  1140				break
  1141			}
  1142			d.buf.WriteByte(b)
  1143		}
  1144		return true
  1145	}
  1146	
  1147	func isNameByte(c byte) bool {
  1148		return 'A' <= c && c <= 'Z' ||
  1149			'a' <= c && c <= 'z' ||
  1150			'0' <= c && c <= '9' ||
  1151			c == '_' || c == ':' || c == '.' || c == '-'
  1152	}
  1153	
  1154	func isName(s []byte) bool {
  1155		if len(s) == 0 {
  1156			return false
  1157		}
  1158		c, n := utf8.DecodeRune(s)
  1159		if c == utf8.RuneError && n == 1 {
  1160			return false
  1161		}
  1162		if !unicode.Is(first, c) {
  1163			return false
  1164		}
  1165		for n < len(s) {
  1166			s = s[n:]
  1167			c, n = utf8.DecodeRune(s)
  1168			if c == utf8.RuneError && n == 1 {
  1169				return false
  1170			}
  1171			if !unicode.Is(first, c) && !unicode.Is(second, c) {
  1172				return false
  1173			}
  1174		}
  1175		return true
  1176	}
  1177	
  1178	func isNameString(s string) bool {
  1179		if len(s) == 0 {
  1180			return false
  1181		}
  1182		c, n := utf8.DecodeRuneInString(s)
  1183		if c == utf8.RuneError && n == 1 {
  1184			return false
  1185		}
  1186		if !unicode.Is(first, c) {
  1187			return false
  1188		}
  1189		for n < len(s) {
  1190			s = s[n:]
  1191			c, n = utf8.DecodeRuneInString(s)
  1192			if c == utf8.RuneError && n == 1 {
  1193				return false
  1194			}
  1195			if !unicode.Is(first, c) && !unicode.Is(second, c) {
  1196				return false
  1197			}
  1198		}
  1199		return true
  1200	}
  1201	
  1202	// These tables were generated by cut and paste from Appendix B of
  1203	// the XML spec at http://www.xml.com/axml/testaxml.htm
  1204	// and then reformatting.  First corresponds to (Letter | '_' | ':')
  1205	// and second corresponds to NameChar.
  1206	
  1207	var first = &unicode.RangeTable{
  1208		R16: []unicode.Range16{
  1209			{0x003A, 0x003A, 1},
  1210			{0x0041, 0x005A, 1},
  1211			{0x005F, 0x005F, 1},
  1212			{0x0061, 0x007A, 1},
  1213			{0x00C0, 0x00D6, 1},
  1214			{0x00D8, 0x00F6, 1},
  1215			{0x00F8, 0x00FF, 1},
  1216			{0x0100, 0x0131, 1},
  1217			{0x0134, 0x013E, 1},
  1218			{0x0141, 0x0148, 1},
  1219			{0x014A, 0x017E, 1},
  1220			{0x0180, 0x01C3, 1},
  1221			{0x01CD, 0x01F0, 1},
  1222			{0x01F4, 0x01F5, 1},
  1223			{0x01FA, 0x0217, 1},
  1224			{0x0250, 0x02A8, 1},
  1225			{0x02BB, 0x02C1, 1},
  1226			{0x0386, 0x0386, 1},
  1227			{0x0388, 0x038A, 1},
  1228			{0x038C, 0x038C, 1},
  1229			{0x038E, 0x03A1, 1},
  1230			{0x03A3, 0x03CE, 1},
  1231			{0x03D0, 0x03D6, 1},
  1232			{0x03DA, 0x03E0, 2},
  1233			{0x03E2, 0x03F3, 1},
  1234			{0x0401, 0x040C, 1},
  1235			{0x040E, 0x044F, 1},
  1236			{0x0451, 0x045C, 1},
  1237			{0x045E, 0x0481, 1},
  1238			{0x0490, 0x04C4, 1},
  1239			{0x04C7, 0x04C8, 1},
  1240			{0x04CB, 0x04CC, 1},
  1241			{0x04D0, 0x04EB, 1},
  1242			{0x04EE, 0x04F5, 1},
  1243			{0x04F8, 0x04F9, 1},
  1244			{0x0531, 0x0556, 1},
  1245			{0x0559, 0x0559, 1},
  1246			{0x0561, 0x0586, 1},
  1247			{0x05D0, 0x05EA, 1},
  1248			{0x05F0, 0x05F2, 1},
  1249			{0x0621, 0x063A, 1},
  1250			{0x0641, 0x064A, 1},
  1251			{0x0671, 0x06B7, 1},
  1252			{0x06BA, 0x06BE, 1},
  1253			{0x06C0, 0x06CE, 1},
  1254			{0x06D0, 0x06D3, 1},
  1255			{0x06D5, 0x06D5, 1},
  1256			{0x06E5, 0x06E6, 1},
  1257			{0x0905, 0x0939, 1},
  1258			{0x093D, 0x093D, 1},
  1259			{0x0958, 0x0961, 1},
  1260			{0x0985, 0x098C, 1},
  1261			{0x098F, 0x0990, 1},
  1262			{0x0993, 0x09A8, 1},
  1263			{0x09AA, 0x09B0, 1},
  1264			{0x09B2, 0x09B2, 1},
  1265			{0x09B6, 0x09B9, 1},
  1266			{0x09DC, 0x09DD, 1},
  1267			{0x09DF, 0x09E1, 1},
  1268			{0x09F0, 0x09F1, 1},
  1269			{0x0A05, 0x0A0A, 1},
  1270			{0x0A0F, 0x0A10, 1},
  1271			{0x0A13, 0x0A28, 1},
  1272			{0x0A2A, 0x0A30, 1},
  1273			{0x0A32, 0x0A33, 1},
  1274			{0x0A35, 0x0A36, 1},
  1275			{0x0A38, 0x0A39, 1},
  1276			{0x0A59, 0x0A5C, 1},
  1277			{0x0A5E, 0x0A5E, 1},
  1278			{0x0A72, 0x0A74, 1},
  1279			{0x0A85, 0x0A8B, 1},
  1280			{0x0A8D, 0x0A8D, 1},
  1281			{0x0A8F, 0x0A91, 1},
  1282			{0x0A93, 0x0AA8, 1},
  1283			{0x0AAA, 0x0AB0, 1},
  1284			{0x0AB2, 0x0AB3, 1},
  1285			{0x0AB5, 0x0AB9, 1},
  1286			{0x0ABD, 0x0AE0, 0x23},
  1287			{0x0B05, 0x0B0C, 1},
  1288			{0x0B0F, 0x0B10, 1},
  1289			{0x0B13, 0x0B28, 1},
  1290			{0x0B2A, 0x0B30, 1},
  1291			{0x0B32, 0x0B33, 1},
  1292			{0x0B36, 0x0B39, 1},
  1293			{0x0B3D, 0x0B3D, 1},
  1294			{0x0B5C, 0x0B5D, 1},
  1295			{0x0B5F, 0x0B61, 1},
  1296			{0x0B85, 0x0B8A, 1},
  1297			{0x0B8E, 0x0B90, 1},
  1298			{0x0B92, 0x0B95, 1},
  1299			{0x0B99, 0x0B9A, 1},
  1300			{0x0B9C, 0x0B9C, 1},
  1301			{0x0B9E, 0x0B9F, 1},
  1302			{0x0BA3, 0x0BA4, 1},
  1303			{0x0BA8, 0x0BAA, 1},
  1304			{0x0BAE, 0x0BB5, 1},
  1305			{0x0BB7, 0x0BB9, 1},
  1306			{0x0C05, 0x0C0C, 1},
  1307			{0x0C0E, 0x0C10, 1},
  1308			{0x0C12, 0x0C28, 1},
  1309			{0x0C2A, 0x0C33, 1},
  1310			{0x0C35, 0x0C39, 1},
  1311			{0x0C60, 0x0C61, 1},
  1312			{0x0C85, 0x0C8C, 1},
  1313			{0x0C8E, 0x0C90, 1},
  1314			{0x0C92, 0x0CA8, 1},
  1315			{0x0CAA, 0x0CB3, 1},
  1316			{0x0CB5, 0x0CB9, 1},
  1317			{0x0CDE, 0x0CDE, 1},
  1318			{0x0CE0, 0x0CE1, 1},
  1319			{0x0D05, 0x0D0C, 1},
  1320			{0x0D0E, 0x0D10, 1},
  1321			{0x0D12, 0x0D28, 1},
  1322			{0x0D2A, 0x0D39, 1},
  1323			{0x0D60, 0x0D61, 1},
  1324			{0x0E01, 0x0E2E, 1},
  1325			{0x0E30, 0x0E30, 1},
  1326			{0x0E32, 0x0E33, 1},
  1327			{0x0E40, 0x0E45, 1},
  1328			{0x0E81, 0x0E82, 1},
  1329			{0x0E84, 0x0E84, 1},
  1330			{0x0E87, 0x0E88, 1},
  1331			{0x0E8A, 0x0E8D, 3},
  1332			{0x0E94, 0x0E97, 1},
  1333			{0x0E99, 0x0E9F, 1},
  1334			{0x0EA1, 0x0EA3, 1},
  1335			{0x0EA5, 0x0EA7, 2},
  1336			{0x0EAA, 0x0EAB, 1},
  1337			{0x0EAD, 0x0EAE, 1},
  1338			{0x0EB0, 0x0EB0, 1},
  1339			{0x0EB2, 0x0EB3, 1},
  1340			{0x0EBD, 0x0EBD, 1},
  1341			{0x0EC0, 0x0EC4, 1},
  1342			{0x0F40, 0x0F47, 1},
  1343			{0x0F49, 0x0F69, 1},
  1344			{0x10A0, 0x10C5, 1},
  1345			{0x10D0, 0x10F6, 1},
  1346			{0x1100, 0x1100, 1},
  1347			{0x1102, 0x1103, 1},
  1348			{0x1105, 0x1107, 1},
  1349			{0x1109, 0x1109, 1},
  1350			{0x110B, 0x110C, 1},
  1351			{0x110E, 0x1112, 1},
  1352			{0x113C, 0x1140, 2},
  1353			{0x114C, 0x1150, 2},
  1354			{0x1154, 0x1155, 1},
  1355			{0x1159, 0x1159, 1},
  1356			{0x115F, 0x1161, 1},
  1357			{0x1163, 0x1169, 2},
  1358			{0x116D, 0x116E, 1},
  1359			{0x1172, 0x1173, 1},
  1360			{0x1175, 0x119E, 0x119E - 0x1175},
  1361			{0x11A8, 0x11AB, 0x11AB - 0x11A8},
  1362			{0x11AE, 0x11AF, 1},
  1363			{0x11B7, 0x11B8, 1},
  1364			{0x11BA, 0x11BA, 1},
  1365			{0x11BC, 0x11C2, 1},
  1366			{0x11EB, 0x11F0, 0x11F0 - 0x11EB},
  1367			{0x11F9, 0x11F9, 1},
  1368			{0x1E00, 0x1E9B, 1},
  1369			{0x1EA0, 0x1EF9, 1},
  1370			{0x1F00, 0x1F15, 1},
  1371			{0x1F18, 0x1F1D, 1},
  1372			{0x1F20, 0x1F45, 1},
  1373			{0x1F48, 0x1F4D, 1},
  1374			{0x1F50, 0x1F57, 1},
  1375			{0x1F59, 0x1F5B, 0x1F5B - 0x1F59},
  1376			{0x1F5D, 0x1F5D, 1},
  1377			{0x1F5F, 0x1F7D, 1},
  1378			{0x1F80, 0x1FB4, 1},
  1379			{0x1FB6, 0x1FBC, 1},
  1380			{0x1FBE, 0x1FBE, 1},
  1381			{0x1FC2, 0x1FC4, 1},
  1382			{0x1FC6, 0x1FCC, 1},
  1383			{0x1FD0, 0x1FD3, 1},
  1384			{0x1FD6, 0x1FDB, 1},
  1385			{0x1FE0, 0x1FEC, 1},
  1386			{0x1FF2, 0x1FF4, 1},
  1387			{0x1FF6, 0x1FFC, 1},
  1388			{0x2126, 0x2126, 1},
  1389			{0x212A, 0x212B, 1},
  1390			{0x212E, 0x212E, 1},
  1391			{0x2180, 0x2182, 1},
  1392			{0x3007, 0x3007, 1},
  1393			{0x3021, 0x3029, 1},
  1394			{0x3041, 0x3094, 1},
  1395			{0x30A1, 0x30FA, 1},
  1396			{0x3105, 0x312C, 1},
  1397			{0x4E00, 0x9FA5, 1},
  1398			{0xAC00, 0xD7A3, 1},
  1399		},
  1400	}
  1401	
  1402	var second = &unicode.RangeTable{
  1403		R16: []unicode.Range16{
  1404			{0x002D, 0x002E, 1},
  1405			{0x0030, 0x0039, 1},
  1406			{0x00B7, 0x00B7, 1},
  1407			{0x02D0, 0x02D1, 1},
  1408			{0x0300, 0x0345, 1},
  1409			{0x0360, 0x0361, 1},
  1410			{0x0387, 0x0387, 1},
  1411			{0x0483, 0x0486, 1},
  1412			{0x0591, 0x05A1, 1},
  1413			{0x05A3, 0x05B9, 1},
  1414			{0x05BB, 0x05BD, 1},
  1415			{0x05BF, 0x05BF, 1},
  1416			{0x05C1, 0x05C2, 1},
  1417			{0x05C4, 0x0640, 0x0640 - 0x05C4},
  1418			{0x064B, 0x0652, 1},
  1419			{0x0660, 0x0669, 1},
  1420			{0x0670, 0x0670, 1},
  1421			{0x06D6, 0x06DC, 1},
  1422			{0x06DD, 0x06DF, 1},
  1423			{0x06E0, 0x06E4, 1},
  1424			{0x06E7, 0x06E8, 1},
  1425			{0x06EA, 0x06ED, 1},
  1426			{0x06F0, 0x06F9, 1},
  1427			{0x0901, 0x0903, 1},
  1428			{0x093C, 0x093C, 1},
  1429			{0x093E, 0x094C, 1},
  1430			{0x094D, 0x094D, 1},
  1431			{0x0951, 0x0954, 1},
  1432			{0x0962, 0x0963, 1},
  1433			{0x0966, 0x096F, 1},
  1434			{0x0981, 0x0983, 1},
  1435			{0x09BC, 0x09BC, 1},
  1436			{0x09BE, 0x09BF, 1},
  1437			{0x09C0, 0x09C4, 1},
  1438			{0x09C7, 0x09C8, 1},
  1439			{0x09CB, 0x09CD, 1},
  1440			{0x09D7, 0x09D7, 1},
  1441			{0x09E2, 0x09E3, 1},
  1442			{0x09E6, 0x09EF, 1},
  1443			{0x0A02, 0x0A3C, 0x3A},
  1444			{0x0A3E, 0x0A3F, 1},
  1445			{0x0A40, 0x0A42, 1},
  1446			{0x0A47, 0x0A48, 1},
  1447			{0x0A4B, 0x0A4D, 1},
  1448			{0x0A66, 0x0A6F, 1},
  1449			{0x0A70, 0x0A71, 1},
  1450			{0x0A81, 0x0A83, 1},
  1451			{0x0ABC, 0x0ABC, 1},
  1452			{0x0ABE, 0x0AC5, 1},
  1453			{0x0AC7, 0x0AC9, 1},
  1454			{0x0ACB, 0x0ACD, 1},
  1455			{0x0AE6, 0x0AEF, 1},
  1456			{0x0B01, 0x0B03, 1},
  1457			{0x0B3C, 0x0B3C, 1},
  1458			{0x0B3E, 0x0B43, 1},
  1459			{0x0B47, 0x0B48, 1},
  1460			{0x0B4B, 0x0B4D, 1},
  1461			{0x0B56, 0x0B57, 1},
  1462			{0x0B66, 0x0B6F, 1},
  1463			{0x0B82, 0x0B83, 1},
  1464			{0x0BBE, 0x0BC2, 1},
  1465			{0x0BC6, 0x0BC8, 1},
  1466			{0x0BCA, 0x0BCD, 1},
  1467			{0x0BD7, 0x0BD7, 1},
  1468			{0x0BE7, 0x0BEF, 1},
  1469			{0x0C01, 0x0C03, 1},
  1470			{0x0C3E, 0x0C44, 1},
  1471			{0x0C46, 0x0C48, 1},
  1472			{0x0C4A, 0x0C4D, 1},
  1473			{0x0C55, 0x0C56, 1},
  1474			{0x0C66, 0x0C6F, 1},
  1475			{0x0C82, 0x0C83, 1},
  1476			{0x0CBE, 0x0CC4, 1},
  1477			{0x0CC6, 0x0CC8, 1},
  1478			{0x0CCA, 0x0CCD, 1},
  1479			{0x0CD5, 0x0CD6, 1},
  1480			{0x0CE6, 0x0CEF, 1},
  1481			{0x0D02, 0x0D03, 1},
  1482			{0x0D3E, 0x0D43, 1},
  1483			{0x0D46, 0x0D48, 1},
  1484			{0x0D4A, 0x0D4D, 1},
  1485			{0x0D57, 0x0D57, 1},
  1486			{0x0D66, 0x0D6F, 1},
  1487			{0x0E31, 0x0E31, 1},
  1488			{0x0E34, 0x0E3A, 1},
  1489			{0x0E46, 0x0E46, 1},
  1490			{0x0E47, 0x0E4E, 1},
  1491			{0x0E50, 0x0E59, 1},
  1492			{0x0EB1, 0x0EB1, 1},
  1493			{0x0EB4, 0x0EB9, 1},
  1494			{0x0EBB, 0x0EBC, 1},
  1495			{0x0EC6, 0x0EC6, 1},
  1496			{0x0EC8, 0x0ECD, 1},
  1497			{0x0ED0, 0x0ED9, 1},
  1498			{0x0F18, 0x0F19, 1},
  1499			{0x0F20, 0x0F29, 1},
  1500			{0x0F35, 0x0F39, 2},
  1501			{0x0F3E, 0x0F3F, 1},
  1502			{0x0F71, 0x0F84, 1},
  1503			{0x0F86, 0x0F8B, 1},
  1504			{0x0F90, 0x0F95, 1},
  1505			{0x0F97, 0x0F97, 1},
  1506			{0x0F99, 0x0FAD, 1},
  1507			{0x0FB1, 0x0FB7, 1},
  1508			{0x0FB9, 0x0FB9, 1},
  1509			{0x20D0, 0x20DC, 1},
  1510			{0x20E1, 0x3005, 0x3005 - 0x20E1},
  1511			{0x302A, 0x302F, 1},
  1512			{0x3031, 0x3035, 1},
  1513			{0x3099, 0x309A, 1},
  1514			{0x309D, 0x309E, 1},
  1515			{0x30FC, 0x30FE, 1},
  1516		},
  1517	}
  1518	
  1519	// HTMLEntity is an entity map containing translations for the
  1520	// standard HTML entity characters.
  1521	var HTMLEntity = htmlEntity
  1522	
  1523	var htmlEntity = map[string]string{
  1524		/*
  1525			hget http://www.w3.org/TR/html4/sgml/entities.html |
  1526			ssam '
  1527				,y /\&gt;/ x/\&lt;(.|\n)+/ s/\n/ /g
  1528				,x v/^\&lt;!ENTITY/d
  1529				,s/\&lt;!ENTITY ([^ ]+) .*U\+([0-9A-F][0-9A-F][0-9A-F][0-9A-F]) .+/	"\1": "\\u\2",/g
  1530			'
  1531		*/
  1532		"nbsp":     "\u00A0",
  1533		"iexcl":    "\u00A1",
  1534		"cent":     "\u00A2",
  1535		"pound":    "\u00A3",
  1536		"curren":   "\u00A4",
  1537		"yen":      "\u00A5",
  1538		"brvbar":   "\u00A6",
  1539		"sect":     "\u00A7",
  1540		"uml":      "\u00A8",
  1541		"copy":     "\u00A9",
  1542		"ordf":     "\u00AA",
  1543		"laquo":    "\u00AB",
  1544		"not":      "\u00AC",
  1545		"shy":      "\u00AD",
  1546		"reg":      "\u00AE",
  1547		"macr":     "\u00AF",
  1548		"deg":      "\u00B0",
  1549		"plusmn":   "\u00B1",
  1550		"sup2":     "\u00B2",
  1551		"sup3":     "\u00B3",
  1552		"acute":    "\u00B4",
  1553		"micro":    "\u00B5",
  1554		"para":     "\u00B6",
  1555		"middot":   "\u00B7",
  1556		"cedil":    "\u00B8",
  1557		"sup1":     "\u00B9",
  1558		"ordm":     "\u00BA",
  1559		"raquo":    "\u00BB",
  1560		"frac14":   "\u00BC",
  1561		"frac12":   "\u00BD",
  1562		"frac34":   "\u00BE",
  1563		"iquest":   "\u00BF",
  1564		"Agrave":   "\u00C0",
  1565		"Aacute":   "\u00C1",
  1566		"Acirc":    "\u00C2",
  1567		"Atilde":   "\u00C3",
  1568		"Auml":     "\u00C4",
  1569		"Aring":    "\u00C5",
  1570		"AElig":    "\u00C6",
  1571		"Ccedil":   "\u00C7",
  1572		"Egrave":   "\u00C8",
  1573		"Eacute":   "\u00C9",
  1574		"Ecirc":    "\u00CA",
  1575		"Euml":     "\u00CB",
  1576		"Igrave":   "\u00CC",
  1577		"Iacute":   "\u00CD",
  1578		"Icirc":    "\u00CE",
  1579		"Iuml":     "\u00CF",
  1580		"ETH":      "\u00D0",
  1581		"Ntilde":   "\u00D1",
  1582		"Ograve":   "\u00D2",
  1583		"Oacute":   "\u00D3",
  1584		"Ocirc":    "\u00D4",
  1585		"Otilde":   "\u00D5",
  1586		"Ouml":     "\u00D6",
  1587		"times":    "\u00D7",
  1588		"Oslash":   "\u00D8",
  1589		"Ugrave":   "\u00D9",
  1590		"Uacute":   "\u00DA",
  1591		"Ucirc":    "\u00DB",
  1592		"Uuml":     "\u00DC",
  1593		"Yacute":   "\u00DD",
  1594		"THORN":    "\u00DE",
  1595		"szlig":    "\u00DF",
  1596		"agrave":   "\u00E0",
  1597		"aacute":   "\u00E1",
  1598		"acirc":    "\u00E2",
  1599		"atilde":   "\u00E3",
  1600		"auml":     "\u00E4",
  1601		"aring":    "\u00E5",
  1602		"aelig":    "\u00E6",
  1603		"ccedil":   "\u00E7",
  1604		"egrave":   "\u00E8",
  1605		"eacute":   "\u00E9",
  1606		"ecirc":    "\u00EA",
  1607		"euml":     "\u00EB",
  1608		"igrave":   "\u00EC",
  1609		"iacute":   "\u00ED",
  1610		"icirc":    "\u00EE",
  1611		"iuml":     "\u00EF",
  1612		"eth":      "\u00F0",
  1613		"ntilde":   "\u00F1",
  1614		"ograve":   "\u00F2",
  1615		"oacute":   "\u00F3",
  1616		"ocirc":    "\u00F4",
  1617		"otilde":   "\u00F5",
  1618		"ouml":     "\u00F6",
  1619		"divide":   "\u00F7",
  1620		"oslash":   "\u00F8",
  1621		"ugrave":   "\u00F9",
  1622		"uacute":   "\u00FA",
  1623		"ucirc":    "\u00FB",
  1624		"uuml":     "\u00FC",
  1625		"yacute":   "\u00FD",
  1626		"thorn":    "\u00FE",
  1627		"yuml":     "\u00FF",
  1628		"fnof":     "\u0192",
  1629		"Alpha":    "\u0391",
  1630		"Beta":     "\u0392",
  1631		"Gamma":    "\u0393",
  1632		"Delta":    "\u0394",
  1633		"Epsilon":  "\u0395",
  1634		"Zeta":     "\u0396",
  1635		"Eta":      "\u0397",
  1636		"Theta":    "\u0398",
  1637		"Iota":     "\u0399",
  1638		"Kappa":    "\u039A",
  1639		"Lambda":   "\u039B",
  1640		"Mu":       "\u039C",
  1641		"Nu":       "\u039D",
  1642		"Xi":       "\u039E",
  1643		"Omicron":  "\u039F",
  1644		"Pi":       "\u03A0",
  1645		"Rho":      "\u03A1",
  1646		"Sigma":    "\u03A3",
  1647		"Tau":      "\u03A4",
  1648		"Upsilon":  "\u03A5",
  1649		"Phi":      "\u03A6",
  1650		"Chi":      "\u03A7",
  1651		"Psi":      "\u03A8",
  1652		"Omega":    "\u03A9",
  1653		"alpha":    "\u03B1",
  1654		"beta":     "\u03B2",
  1655		"gamma":    "\u03B3",
  1656		"delta":    "\u03B4",
  1657		"epsilon":  "\u03B5",
  1658		"zeta":     "\u03B6",
  1659		"eta":      "\u03B7",
  1660		"theta":    "\u03B8",
  1661		"iota":     "\u03B9",
  1662		"kappa":    "\u03BA",
  1663		"lambda":   "\u03BB",
  1664		"mu":       "\u03BC",
  1665		"nu":       "\u03BD",
  1666		"xi":       "\u03BE",
  1667		"omicron":  "\u03BF",
  1668		"pi":       "\u03C0",
  1669		"rho":      "\u03C1",
  1670		"sigmaf":   "\u03C2",
  1671		"sigma":    "\u03C3",
  1672		"tau":      "\u03C4",
  1673		"upsilon":  "\u03C5",
  1674		"phi":      "\u03C6",
  1675		"chi":      "\u03C7",
  1676		"psi":      "\u03C8",
  1677		"omega":    "\u03C9",
  1678		"thetasym": "\u03D1",
  1679		"upsih":    "\u03D2",
  1680		"piv":      "\u03D6",
  1681		"bull":     "\u2022",
  1682		"hellip":   "\u2026",
  1683		"prime":    "\u2032",
  1684		"Prime":    "\u2033",
  1685		"oline":    "\u203E",
  1686		"frasl":    "\u2044",
  1687		"weierp":   "\u2118",
  1688		"image":    "\u2111",
  1689		"real":     "\u211C",
  1690		"trade":    "\u2122",
  1691		"alefsym":  "\u2135",
  1692		"larr":     "\u2190",
  1693		"uarr":     "\u2191",
  1694		"rarr":     "\u2192",
  1695		"darr":     "\u2193",
  1696		"harr":     "\u2194",
  1697		"crarr":    "\u21B5",
  1698		"lArr":     "\u21D0",
  1699		"uArr":     "\u21D1",
  1700		"rArr":     "\u21D2",
  1701		"dArr":     "\u21D3",
  1702		"hArr":     "\u21D4",
  1703		"forall":   "\u2200",
  1704		"part":     "\u2202",
  1705		"exist":    "\u2203",
  1706		"empty":    "\u2205",
  1707		"nabla":    "\u2207",
  1708		"isin":     "\u2208",
  1709		"notin":    "\u2209",
  1710		"ni":       "\u220B",
  1711		"prod":     "\u220F",
  1712		"sum":      "\u2211",
  1713		"minus":    "\u2212",
  1714		"lowast":   "\u2217",
  1715		"radic":    "\u221A",
  1716		"prop":     "\u221D",
  1717		"infin":    "\u221E",
  1718		"ang":      "\u2220",
  1719		"and":      "\u2227",
  1720		"or":       "\u2228",
  1721		"cap":      "\u2229",
  1722		"cup":      "\u222A",
  1723		"int":      "\u222B",
  1724		"there4":   "\u2234",
  1725		"sim":      "\u223C",
  1726		"cong":     "\u2245",
  1727		"asymp":    "\u2248",
  1728		"ne":       "\u2260",
  1729		"equiv":    "\u2261",
  1730		"le":       "\u2264",
  1731		"ge":       "\u2265",
  1732		"sub":      "\u2282",
  1733		"sup":      "\u2283",
  1734		"nsub":     "\u2284",
  1735		"sube":     "\u2286",
  1736		"supe":     "\u2287",
  1737		"oplus":    "\u2295",
  1738		"otimes":   "\u2297",
  1739		"perp":     "\u22A5",
  1740		"sdot":     "\u22C5",
  1741		"lceil":    "\u2308",
  1742		"rceil":    "\u2309",
  1743		"lfloor":   "\u230A",
  1744		"rfloor":   "\u230B",
  1745		"lang":     "\u2329",
  1746		"rang":     "\u232A",
  1747		"loz":      "\u25CA",
  1748		"spades":   "\u2660",
  1749		"clubs":    "\u2663",
  1750		"hearts":   "\u2665",
  1751		"diams":    "\u2666",
  1752		"quot":     "\u0022",
  1753		"amp":      "\u0026",
  1754		"lt":       "\u003C",
  1755		"gt":       "\u003E",
  1756		"OElig":    "\u0152",
  1757		"oelig":    "\u0153",
  1758		"Scaron":   "\u0160",
  1759		"scaron":   "\u0161",
  1760		"Yuml":     "\u0178",
  1761		"circ":     "\u02C6",
  1762		"tilde":    "\u02DC",
  1763		"ensp":     "\u2002",
  1764		"emsp":     "\u2003",
  1765		"thinsp":   "\u2009",
  1766		"zwnj":     "\u200C",
  1767		"zwj":      "\u200D",
  1768		"lrm":      "\u200E",
  1769		"rlm":      "\u200F",
  1770		"ndash":    "\u2013",
  1771		"mdash":    "\u2014",
  1772		"lsquo":    "\u2018",
  1773		"rsquo":    "\u2019",
  1774		"sbquo":    "\u201A",
  1775		"ldquo":    "\u201C",
  1776		"rdquo":    "\u201D",
  1777		"bdquo":    "\u201E",
  1778		"dagger":   "\u2020",
  1779		"Dagger":   "\u2021",
  1780		"permil":   "\u2030",
  1781		"lsaquo":   "\u2039",
  1782		"rsaquo":   "\u203A",
  1783		"euro":     "\u20AC",
  1784	}
  1785	
  1786	// HTMLAutoClose is the set of HTML elements that
  1787	// should be considered to close automatically.
  1788	var HTMLAutoClose = htmlAutoClose
  1789	
  1790	var htmlAutoClose = []string{
  1791		/*
  1792			hget http://www.w3.org/TR/html4/loose.dtd |
  1793			9 sed -n 's/<!ELEMENT ([^ ]*) +- O EMPTY.+/	"\1",/p' | tr A-Z a-z
  1794		*/
  1795		"basefont",
  1796		"br",
  1797		"area",
  1798		"link",
  1799		"img",
  1800		"param",
  1801		"hr",
  1802		"input",
  1803		"col",
  1804		"frame",
  1805		"isindex",
  1806		"base",
  1807		"meta",
  1808	}
  1809	
  1810	var (
  1811		esc_quot = []byte("&#34;") // shorter than "&quot;"
  1812		esc_apos = []byte("&#39;") // shorter than "&apos;"
  1813		esc_amp  = []byte("&amp;")
  1814		esc_lt   = []byte("&lt;")
  1815		esc_gt   = []byte("&gt;")
  1816		esc_tab  = []byte("&#x9;")
  1817		esc_nl   = []byte("&#xA;")
  1818		esc_cr   = []byte("&#xD;")
  1819		esc_fffd = []byte("\uFFFD") // Unicode replacement character
  1820	)
  1821	
  1822	// EscapeText writes to w the properly escaped XML equivalent
  1823	// of the plain text data s.
  1824	func EscapeText(w io.Writer, s []byte) error {
  1825		var esc []byte
  1826		last := 0
  1827		for i := 0; i < len(s); {
  1828			r, width := utf8.DecodeRune(s[i:])
  1829			i += width
  1830			switch r {
  1831			case '"':
  1832				esc = esc_quot
  1833			case '\'':
  1834				esc = esc_apos
  1835			case '&':
  1836				esc = esc_amp
  1837			case '<':
  1838				esc = esc_lt
  1839			case '>':
  1840				esc = esc_gt
  1841			case '\t':
  1842				esc = esc_tab
  1843			case '\n':
  1844				esc = esc_nl
  1845			case '\r':
  1846				esc = esc_cr
  1847			default:
  1848				if !isInCharacterRange(r) || (r == 0xFFFD && width == 1) {
  1849					esc = esc_fffd
  1850					break
  1851				}
  1852				continue
  1853			}
  1854			if _, err := w.Write(s[last : i-width]); err != nil {
  1855				return err
  1856			}
  1857			if _, err := w.Write(esc); err != nil {
  1858				return err
  1859			}
  1860			last = i
  1861		}
  1862		if _, err := w.Write(s[last:]); err != nil {
  1863			return err
  1864		}
  1865		return nil
  1866	}
  1867	
  1868	// EscapeString writes to p the properly escaped XML equivalent
  1869	// of the plain text data s.
  1870	func (p *printer) EscapeString(s string) {
  1871		var esc []byte
  1872		last := 0
  1873		for i := 0; i < len(s); {
  1874			r, width := utf8.DecodeRuneInString(s[i:])
  1875			i += width
  1876			switch r {
  1877			case '"':
  1878				esc = esc_quot
  1879			case '\'':
  1880				esc = esc_apos
  1881			case '&':
  1882				esc = esc_amp
  1883			case '<':
  1884				esc = esc_lt
  1885			case '>':
  1886				esc = esc_gt
  1887			case '\t':
  1888				esc = esc_tab
  1889			case '\n':
  1890				esc = esc_nl
  1891			case '\r':
  1892				esc = esc_cr
  1893			default:
  1894				if !isInCharacterRange(r) || (r == 0xFFFD && width == 1) {
  1895					esc = esc_fffd
  1896					break
  1897				}
  1898				continue
  1899			}
  1900			p.WriteString(s[last : i-width])
  1901			p.Write(esc)
  1902			last = i
  1903		}
  1904		p.WriteString(s[last:])
  1905	}
  1906	
  1907	// Escape is like EscapeText but omits the error return value.
  1908	// It is provided for backwards compatibility with Go 1.0.
  1909	// Code targeting Go 1.1 or later should use EscapeText.
  1910	func Escape(w io.Writer, s []byte) {
  1911		EscapeText(w, s)
  1912	}
  1913	
  1914	// procInstEncoding parses the `encoding="..."` or `encoding='...'`
  1915	// value out of the provided string, returning "" if not found.
  1916	func procInstEncoding(s string) string {
  1917		// TODO: this parsing is somewhat lame and not exact.
  1918		// It works for all actual cases, though.
  1919		idx := strings.Index(s, "encoding=")
  1920		if idx == -1 {
  1921			return ""
  1922		}
  1923		v := s[idx+len("encoding="):]
  1924		if v == "" {
  1925			return ""
  1926		}
  1927		if v[0] != '\'' && v[0] != '"' {
  1928			return ""
  1929		}
  1930		idx = strings.IndexRune(v[1:], rune(v[0]))
  1931		if idx == -1 {
  1932			return ""
  1933		}
  1934		return v[1 : idx+1]
  1935	}
  1936	

View as plain text