...
Run Format

Source file src/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    81	// in which XML escape sequences have been replaced by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		offset         int64
   200		unmarshalDepth int
   201	}
   202	
   203	// NewDecoder creates a new XML parser reading from r.
   204	// If r does not implement io.ByteReader, NewDecoder will
   205	// do its own buffering.
   206	func NewDecoder(r io.Reader) *Decoder {
   207		d := &Decoder{
   208			ns:       make(map[string]string),
   209			nextByte: -1,
   210			line:     1,
   211			Strict:   true,
   212		}
   213		d.switchToReader(r)
   214		return d
   215	}
   216	
   217	// Token returns the next XML token in the input stream.
   218	// At the end of the input stream, Token returns nil, io.EOF.
   219	//
   220	// Slices of bytes in the returned token data refer to the
   221	// parser's internal buffer and remain valid only until the next
   222	// call to Token. To acquire a copy of the bytes, call CopyToken
   223	// or the token's Copy method.
   224	//
   225	// Token expands self-closing elements such as <br/>
   226	// into separate start and end elements returned by successive calls.
   227	//
   228	// Token guarantees that the StartElement and EndElement
   229	// tokens it returns are properly nested and matched:
   230	// if Token encounters an unexpected end element
   231	// or EOF before all expected end elements,
   232	// it will return an error.
   233	//
   234	// Token implements XML name spaces as described by
   235	// http://www.w3.org/TR/REC-xml-names/.  Each of the
   236	// Name structures contained in the Token has the Space
   237	// set to the URL identifying its name space when known.
   238	// If Token encounters an unrecognized name space prefix,
   239	// it uses the prefix as the Space rather than report an error.
   240	func (d *Decoder) Token() (Token, error) {
   241		var t Token
   242		var err error
   243		if d.stk != nil && d.stk.kind == stkEOF {
   244			return nil, io.EOF
   245		}
   246		if d.nextToken != nil {
   247			t = d.nextToken
   248			d.nextToken = nil
   249		} else if t, err = d.rawToken(); err != nil {
   250			if err == io.EOF && d.stk != nil && d.stk.kind != stkEOF {
   251				err = d.syntaxError("unexpected EOF")
   252			}
   253			return t, err
   254		}
   255	
   256		if !d.Strict {
   257			if t1, ok := d.autoClose(t); ok {
   258				d.nextToken = t
   259				t = t1
   260			}
   261		}
   262		switch t1 := t.(type) {
   263		case StartElement:
   264			// In XML name spaces, the translations listed in the
   265			// attributes apply to the element name and
   266			// to the other attribute names, so process
   267			// the translations first.
   268			for _, a := range t1.Attr {
   269				if a.Name.Space == "xmlns" {
   270					v, ok := d.ns[a.Name.Local]
   271					d.pushNs(a.Name.Local, v, ok)
   272					d.ns[a.Name.Local] = a.Value
   273				}
   274				if a.Name.Space == "" && a.Name.Local == "xmlns" {
   275					// Default space for untagged names
   276					v, ok := d.ns[""]
   277					d.pushNs("", v, ok)
   278					d.ns[""] = a.Value
   279				}
   280			}
   281	
   282			d.translate(&t1.Name, true)
   283			for i := range t1.Attr {
   284				d.translate(&t1.Attr[i].Name, false)
   285			}
   286			d.pushElement(t1.Name)
   287			t = t1
   288	
   289		case EndElement:
   290			d.translate(&t1.Name, true)
   291			if !d.popElement(&t1) {
   292				return nil, d.err
   293			}
   294			t = t1
   295		}
   296		return t, err
   297	}
   298	
   299	const xmlURL = "http://www.w3.org/XML/1998/namespace"
   300	
   301	// Apply name space translation to name n.
   302	// The default name space (for Space=="")
   303	// applies only to element names, not to attribute names.
   304	func (d *Decoder) translate(n *Name, isElementName bool) {
   305		switch {
   306		case n.Space == "xmlns":
   307			return
   308		case n.Space == "" && !isElementName:
   309			return
   310		case n.Space == "xml":
   311			n.Space = xmlURL
   312		case n.Space == "" && n.Local == "xmlns":
   313			return
   314		}
   315		if v, ok := d.ns[n.Space]; ok {
   316			n.Space = v
   317		} else if n.Space == "" {
   318			n.Space = d.DefaultSpace
   319		}
   320	}
   321	
   322	func (d *Decoder) switchToReader(r io.Reader) {
   323		// Get efficient byte at a time reader.
   324		// Assume that if reader has its own
   325		// ReadByte, it's efficient enough.
   326		// Otherwise, use bufio.
   327		if rb, ok := r.(io.ByteReader); ok {
   328			d.r = rb
   329		} else {
   330			d.r = bufio.NewReader(r)
   331		}
   332	}
   333	
   334	// Parsing state - stack holds old name space translations
   335	// and the current set of open elements. The translations to pop when
   336	// ending a given tag are *below* it on the stack, which is
   337	// more work but forced on us by XML.
   338	type stack struct {
   339		next *stack
   340		kind int
   341		name Name
   342		ok   bool
   343	}
   344	
   345	const (
   346		stkStart = iota
   347		stkNs
   348		stkEOF
   349	)
   350	
   351	func (d *Decoder) push(kind int) *stack {
   352		s := d.free
   353		if s != nil {
   354			d.free = s.next
   355		} else {
   356			s = new(stack)
   357		}
   358		s.next = d.stk
   359		s.kind = kind
   360		d.stk = s
   361		return s
   362	}
   363	
   364	func (d *Decoder) pop() *stack {
   365		s := d.stk
   366		if s != nil {
   367			d.stk = s.next
   368			s.next = d.free
   369			d.free = s
   370		}
   371		return s
   372	}
   373	
   374	// Record that after the current element is finished
   375	// (that element is already pushed on the stack)
   376	// Token should return EOF until popEOF is called.
   377	func (d *Decoder) pushEOF() {
   378		// Walk down stack to find Start.
   379		// It might not be the top, because there might be stkNs
   380		// entries above it.
   381		start := d.stk
   382		for start.kind != stkStart {
   383			start = start.next
   384		}
   385		// The stkNs entries below a start are associated with that
   386		// element too; skip over them.
   387		for start.next != nil && start.next.kind == stkNs {
   388			start = start.next
   389		}
   390		s := d.free
   391		if s != nil {
   392			d.free = s.next
   393		} else {
   394			s = new(stack)
   395		}
   396		s.kind = stkEOF
   397		s.next = start.next
   398		start.next = s
   399	}
   400	
   401	// Undo a pushEOF.
   402	// The element must have been finished, so the EOF should be at the top of the stack.
   403	func (d *Decoder) popEOF() bool {
   404		if d.stk == nil || d.stk.kind != stkEOF {
   405			return false
   406		}
   407		d.pop()
   408		return true
   409	}
   410	
   411	// Record that we are starting an element with the given name.
   412	func (d *Decoder) pushElement(name Name) {
   413		s := d.push(stkStart)
   414		s.name = name
   415	}
   416	
   417	// Record that we are changing the value of ns[local].
   418	// The old value is url, ok.
   419	func (d *Decoder) pushNs(local string, url string, ok bool) {
   420		s := d.push(stkNs)
   421		s.name.Local = local
   422		s.name.Space = url
   423		s.ok = ok
   424	}
   425	
   426	// Creates a SyntaxError with the current line number.
   427	func (d *Decoder) syntaxError(msg string) error {
   428		return &SyntaxError{Msg: msg, Line: d.line}
   429	}
   430	
   431	// Record that we are ending an element with the given name.
   432	// The name must match the record at the top of the stack,
   433	// which must be a pushElement record.
   434	// After popping the element, apply any undo records from
   435	// the stack to restore the name translations that existed
   436	// before we saw this element.
   437	func (d *Decoder) popElement(t *EndElement) bool {
   438		s := d.pop()
   439		name := t.Name
   440		switch {
   441		case s == nil || s.kind != stkStart:
   442			d.err = d.syntaxError("unexpected end element </" + name.Local + ">")
   443			return false
   444		case s.name.Local != name.Local:
   445			if !d.Strict {
   446				d.needClose = true
   447				d.toClose = t.Name
   448				t.Name = s.name
   449				return true
   450			}
   451			d.err = d.syntaxError("element <" + s.name.Local + "> closed by </" + name.Local + ">")
   452			return false
   453		case s.name.Space != name.Space:
   454			d.err = d.syntaxError("element <" + s.name.Local + "> in space " + s.name.Space +
   455				"closed by </" + name.Local + "> in space " + name.Space)
   456			return false
   457		}
   458	
   459		// Pop stack until a Start or EOF is on the top, undoing the
   460		// translations that were associated with the element we just closed.
   461		for d.stk != nil && d.stk.kind != stkStart && d.stk.kind != stkEOF {
   462			s := d.pop()
   463			if s.ok {
   464				d.ns[s.name.Local] = s.name.Space
   465			} else {
   466				delete(d.ns, s.name.Local)
   467			}
   468		}
   469	
   470		return true
   471	}
   472	
   473	// If the top element on the stack is autoclosing and
   474	// t is not the end tag, invent the end tag.
   475	func (d *Decoder) autoClose(t Token) (Token, bool) {
   476		if d.stk == nil || d.stk.kind != stkStart {
   477			return nil, false
   478		}
   479		name := strings.ToLower(d.stk.name.Local)
   480		for _, s := range d.AutoClose {
   481			if strings.ToLower(s) == name {
   482				// This one should be auto closed if t doesn't close it.
   483				et, ok := t.(EndElement)
   484				if !ok || et.Name.Local != name {
   485					return EndElement{d.stk.name}, true
   486				}
   487				break
   488			}
   489		}
   490		return nil, false
   491	}
   492	
   493	var errRawToken = errors.New("xml: cannot use RawToken from UnmarshalXML method")
   494	
   495	// RawToken is like Token but does not verify that
   496	// start and end elements match and does not translate
   497	// name space prefixes to their corresponding URLs.
   498	func (d *Decoder) RawToken() (Token, error) {
   499		if d.unmarshalDepth > 0 {
   500			return nil, errRawToken
   501		}
   502		return d.rawToken()
   503	}
   504	
   505	func (d *Decoder) rawToken() (Token, error) {
   506		if d.err != nil {
   507			return nil, d.err
   508		}
   509		if d.needClose {
   510			// The last element we read was self-closing and
   511			// we returned just the StartElement half.
   512			// Return the EndElement half now.
   513			d.needClose = false
   514			return EndElement{d.toClose}, nil
   515		}
   516	
   517		b, ok := d.getc()
   518		if !ok {
   519			return nil, d.err
   520		}
   521	
   522		if b != '<' {
   523			// Text section.
   524			d.ungetc(b)
   525			data := d.text(-1, false)
   526			if data == nil {
   527				return nil, d.err
   528			}
   529			return CharData(data), nil
   530		}
   531	
   532		if b, ok = d.mustgetc(); !ok {
   533			return nil, d.err
   534		}
   535		switch b {
   536		case '/':
   537			// </: End element
   538			var name Name
   539			if name, ok = d.nsname(); !ok {
   540				if d.err == nil {
   541					d.err = d.syntaxError("expected element name after </")
   542				}
   543				return nil, d.err
   544			}
   545			d.space()
   546			if b, ok = d.mustgetc(); !ok {
   547				return nil, d.err
   548			}
   549			if b != '>' {
   550				d.err = d.syntaxError("invalid characters between </" + name.Local + " and >")
   551				return nil, d.err
   552			}
   553			return EndElement{name}, nil
   554	
   555		case '?':
   556			// <?: Processing instruction.
   557			var target string
   558			if target, ok = d.name(); !ok {
   559				if d.err == nil {
   560					d.err = d.syntaxError("expected target name after <?")
   561				}
   562				return nil, d.err
   563			}
   564			d.space()
   565			d.buf.Reset()
   566			var b0 byte
   567			for {
   568				if b, ok = d.mustgetc(); !ok {
   569					return nil, d.err
   570				}
   571				d.buf.WriteByte(b)
   572				if b0 == '?' && b == '>' {
   573					break
   574				}
   575				b0 = b
   576			}
   577			data := d.buf.Bytes()
   578			data = data[0 : len(data)-2] // chop ?>
   579	
   580			if target == "xml" {
   581				content := string(data)
   582				ver := procInst("version", content)
   583				if ver != "" && ver != "1.0" {
   584					d.err = fmt.Errorf("xml: unsupported version %q; only version 1.0 is supported", ver)
   585					return nil, d.err
   586				}
   587				enc := procInst("encoding", content)
   588				if enc != "" && enc != "utf-8" && enc != "UTF-8" && !strings.EqualFold(enc, "utf-8") {
   589					if d.CharsetReader == nil {
   590						d.err = fmt.Errorf("xml: encoding %q declared but Decoder.CharsetReader is nil", enc)
   591						return nil, d.err
   592					}
   593					newr, err := d.CharsetReader(enc, d.r.(io.Reader))
   594					if err != nil {
   595						d.err = fmt.Errorf("xml: opening charset %q: %v", enc, err)
   596						return nil, d.err
   597					}
   598					if newr == nil {
   599						panic("CharsetReader returned a nil Reader for charset " + enc)
   600					}
   601					d.switchToReader(newr)
   602				}
   603			}
   604			return ProcInst{target, data}, nil
   605	
   606		case '!':
   607			// <!: Maybe comment, maybe CDATA.
   608			if b, ok = d.mustgetc(); !ok {
   609				return nil, d.err
   610			}
   611			switch b {
   612			case '-': // <!-
   613				// Probably <!-- for a comment.
   614				if b, ok = d.mustgetc(); !ok {
   615					return nil, d.err
   616				}
   617				if b != '-' {
   618					d.err = d.syntaxError("invalid sequence <!- not part of <!--")
   619					return nil, d.err
   620				}
   621				// Look for terminator.
   622				d.buf.Reset()
   623				var b0, b1 byte
   624				for {
   625					if b, ok = d.mustgetc(); !ok {
   626						return nil, d.err
   627					}
   628					d.buf.WriteByte(b)
   629					if b0 == '-' && b1 == '-' {
   630						if b != '>' {
   631							d.err = d.syntaxError(
   632								`invalid sequence "--" not allowed in comments`)
   633							return nil, d.err
   634						}
   635						break
   636					}
   637					b0, b1 = b1, b
   638				}
   639				data := d.buf.Bytes()
   640				data = data[0 : len(data)-3] // chop -->
   641				return Comment(data), nil
   642	
   643			case '[': // <![
   644				// Probably <![CDATA[.
   645				for i := 0; i < 6; i++ {
   646					if b, ok = d.mustgetc(); !ok {
   647						return nil, d.err
   648					}
   649					if b != "CDATA["[i] {
   650						d.err = d.syntaxError("invalid <![ sequence")
   651						return nil, d.err
   652					}
   653				}
   654				// Have <![CDATA[.  Read text until ]]>.
   655				data := d.text(-1, true)
   656				if data == nil {
   657					return nil, d.err
   658				}
   659				return CharData(data), nil
   660			}
   661	
   662			// Probably a directive: <!DOCTYPE ...>, <!ENTITY ...>, etc.
   663			// We don't care, but accumulate for caller. Quoted angle
   664			// brackets do not count for nesting.
   665			d.buf.Reset()
   666			d.buf.WriteByte(b)
   667			inquote := uint8(0)
   668			depth := 0
   669			for {
   670				if b, ok = d.mustgetc(); !ok {
   671					return nil, d.err
   672				}
   673				if inquote == 0 && b == '>' && depth == 0 {
   674					break
   675				}
   676			HandleB:
   677				d.buf.WriteByte(b)
   678				switch {
   679				case b == inquote:
   680					inquote = 0
   681	
   682				case inquote != 0:
   683					// in quotes, no special action
   684	
   685				case b == '\'' || b == '"':
   686					inquote = b
   687	
   688				case b == '>' && inquote == 0:
   689					depth--
   690	
   691				case b == '<' && inquote == 0:
   692					// Look for <!-- to begin comment.
   693					s := "!--"
   694					for i := 0; i < len(s); i++ {
   695						if b, ok = d.mustgetc(); !ok {
   696							return nil, d.err
   697						}
   698						if b != s[i] {
   699							for j := 0; j < i; j++ {
   700								d.buf.WriteByte(s[j])
   701							}
   702							depth++
   703							goto HandleB
   704						}
   705					}
   706	
   707					// Remove < that was written above.
   708					d.buf.Truncate(d.buf.Len() - 1)
   709	
   710					// Look for terminator.
   711					var b0, b1 byte
   712					for {
   713						if b, ok = d.mustgetc(); !ok {
   714							return nil, d.err
   715						}
   716						if b0 == '-' && b1 == '-' && b == '>' {
   717							break
   718						}
   719						b0, b1 = b1, b
   720					}
   721				}
   722			}
   723			return Directive(d.buf.Bytes()), nil
   724		}
   725	
   726		// Must be an open element like <a href="foo">
   727		d.ungetc(b)
   728	
   729		var (
   730			name  Name
   731			empty bool
   732			attr  []Attr
   733		)
   734		if name, ok = d.nsname(); !ok {
   735			if d.err == nil {
   736				d.err = d.syntaxError("expected element name after <")
   737			}
   738			return nil, d.err
   739		}
   740	
   741		attr = []Attr{}
   742		for {
   743			d.space()
   744			if b, ok = d.mustgetc(); !ok {
   745				return nil, d.err
   746			}
   747			if b == '/' {
   748				empty = true
   749				if b, ok = d.mustgetc(); !ok {
   750					return nil, d.err
   751				}
   752				if b != '>' {
   753					d.err = d.syntaxError("expected /> in element")
   754					return nil, d.err
   755				}
   756				break
   757			}
   758			if b == '>' {
   759				break
   760			}
   761			d.ungetc(b)
   762	
   763			n := len(attr)
   764			if n >= cap(attr) {
   765				nCap := 2 * cap(attr)
   766				if nCap == 0 {
   767					nCap = 4
   768				}
   769				nattr := make([]Attr, n, nCap)
   770				copy(nattr, attr)
   771				attr = nattr
   772			}
   773			attr = attr[0 : n+1]
   774			a := &attr[n]
   775			if a.Name, ok = d.nsname(); !ok {
   776				if d.err == nil {
   777					d.err = d.syntaxError("expected attribute name in element")
   778				}
   779				return nil, d.err
   780			}
   781			d.space()
   782			if b, ok = d.mustgetc(); !ok {
   783				return nil, d.err
   784			}
   785			if b != '=' {
   786				if d.Strict {
   787					d.err = d.syntaxError("attribute name without = in element")
   788					return nil, d.err
   789				} else {
   790					d.ungetc(b)
   791					a.Value = a.Name.Local
   792				}
   793			} else {
   794				d.space()
   795				data := d.attrval()
   796				if data == nil {
   797					return nil, d.err
   798				}
   799				a.Value = string(data)
   800			}
   801		}
   802		if empty {
   803			d.needClose = true
   804			d.toClose = name
   805		}
   806		return StartElement{name, attr}, nil
   807	}
   808	
   809	func (d *Decoder) attrval() []byte {
   810		b, ok := d.mustgetc()
   811		if !ok {
   812			return nil
   813		}
   814		// Handle quoted attribute values
   815		if b == '"' || b == '\'' {
   816			return d.text(int(b), false)
   817		}
   818		// Handle unquoted attribute values for strict parsers
   819		if d.Strict {
   820			d.err = d.syntaxError("unquoted or missing attribute value in element")
   821			return nil
   822		}
   823		// Handle unquoted attribute values for unstrict parsers
   824		d.ungetc(b)
   825		d.buf.Reset()
   826		for {
   827			b, ok = d.mustgetc()
   828			if !ok {
   829				return nil
   830			}
   831			// http://www.w3.org/TR/REC-html40/intro/sgmltut.html#h-3.2.2
   832			if 'a' <= b && b <= 'z' || 'A' <= b && b <= 'Z' ||
   833				'0' <= b && b <= '9' || b == '_' || b == ':' || b == '-' {
   834				d.buf.WriteByte(b)
   835			} else {
   836				d.ungetc(b)
   837				break
   838			}
   839		}
   840		return d.buf.Bytes()
   841	}
   842	
   843	// Skip spaces if any
   844	func (d *Decoder) space() {
   845		for {
   846			b, ok := d.getc()
   847			if !ok {
   848				return
   849			}
   850			switch b {
   851			case ' ', '\r', '\n', '\t':
   852			default:
   853				d.ungetc(b)
   854				return
   855			}
   856		}
   857	}
   858	
   859	// Read a single byte.
   860	// If there is no byte to read, return ok==false
   861	// and leave the error in d.err.
   862	// Maintain line number.
   863	func (d *Decoder) getc() (b byte, ok bool) {
   864		if d.err != nil {
   865			return 0, false
   866		}
   867		if d.nextByte >= 0 {
   868			b = byte(d.nextByte)
   869			d.nextByte = -1
   870		} else {
   871			b, d.err = d.r.ReadByte()
   872			if d.err != nil {
   873				return 0, false
   874			}
   875			if d.saved != nil {
   876				d.saved.WriteByte(b)
   877			}
   878		}
   879		if b == '\n' {
   880			d.line++
   881		}
   882		d.offset++
   883		return b, true
   884	}
   885	
   886	// InputOffset returns the input stream byte offset of the current decoder position.
   887	// The offset gives the location of the end of the most recently returned token
   888	// and the beginning of the next token.
   889	func (d *Decoder) InputOffset() int64 {
   890		return d.offset
   891	}
   892	
   893	// Return saved offset.
   894	// If we did ungetc (nextByte >= 0), have to back up one.
   895	func (d *Decoder) savedOffset() int {
   896		n := d.saved.Len()
   897		if d.nextByte >= 0 {
   898			n--
   899		}
   900		return n
   901	}
   902	
   903	// Must read a single byte.
   904	// If there is no byte to read,
   905	// set d.err to SyntaxError("unexpected EOF")
   906	// and return ok==false
   907	func (d *Decoder) mustgetc() (b byte, ok bool) {
   908		if b, ok = d.getc(); !ok {
   909			if d.err == io.EOF {
   910				d.err = d.syntaxError("unexpected EOF")
   911			}
   912		}
   913		return
   914	}
   915	
   916	// Unread a single byte.
   917	func (d *Decoder) ungetc(b byte) {
   918		if b == '\n' {
   919			d.line--
   920		}
   921		d.nextByte = int(b)
   922		d.offset--
   923	}
   924	
   925	var entity = map[string]int{
   926		"lt":   '<',
   927		"gt":   '>',
   928		"amp":  '&',
   929		"apos": '\'',
   930		"quot": '"',
   931	}
   932	
   933	// Read plain text section (XML calls it character data).
   934	// If quote >= 0, we are in a quoted string and need to find the matching quote.
   935	// If cdata == true, we are in a <![CDATA[ section and need to find ]]>.
   936	// On failure return nil and leave the error in d.err.
   937	func (d *Decoder) text(quote int, cdata bool) []byte {
   938		var b0, b1 byte
   939		var trunc int
   940		d.buf.Reset()
   941	Input:
   942		for {
   943			b, ok := d.getc()
   944			if !ok {
   945				if cdata {
   946					if d.err == io.EOF {
   947						d.err = d.syntaxError("unexpected EOF in CDATA section")
   948					}
   949					return nil
   950				}
   951				break Input
   952			}
   953	
   954			// <![CDATA[ section ends with ]]>.
   955			// It is an error for ]]> to appear in ordinary text.
   956			if b0 == ']' && b1 == ']' && b == '>' {
   957				if cdata {
   958					trunc = 2
   959					break Input
   960				}
   961				d.err = d.syntaxError("unescaped ]]> not in CDATA section")
   962				return nil
   963			}
   964	
   965			// Stop reading text if we see a <.
   966			if b == '<' && !cdata {
   967				if quote >= 0 {
   968					d.err = d.syntaxError("unescaped < inside quoted string")
   969					return nil
   970				}
   971				d.ungetc('<')
   972				break Input
   973			}
   974			if quote >= 0 && b == byte(quote) {
   975				break Input
   976			}
   977			if b == '&' && !cdata {
   978				// Read escaped character expression up to semicolon.
   979				// XML in all its glory allows a document to define and use
   980				// its own character names with <!ENTITY ...> directives.
   981				// Parsers are required to recognize lt, gt, amp, apos, and quot
   982				// even if they have not been declared.
   983				before := d.buf.Len()
   984				d.buf.WriteByte('&')
   985				var ok bool
   986				var text string
   987				var haveText bool
   988				if b, ok = d.mustgetc(); !ok {
   989					return nil
   990				}
   991				if b == '#' {
   992					d.buf.WriteByte(b)
   993					if b, ok = d.mustgetc(); !ok {
   994						return nil
   995					}
   996					base := 10
   997					if b == 'x' {
   998						base = 16
   999						d.buf.WriteByte(b)
  1000						if b, ok = d.mustgetc(); !ok {
  1001							return nil
  1002						}
  1003					}
  1004					start := d.buf.Len()
  1005					for '0' <= b && b <= '9' ||
  1006						base == 16 && 'a' <= b && b <= 'f' ||
  1007						base == 16 && 'A' <= b && b <= 'F' {
  1008						d.buf.WriteByte(b)
  1009						if b, ok = d.mustgetc(); !ok {
  1010							return nil
  1011						}
  1012					}
  1013					if b != ';' {
  1014						d.ungetc(b)
  1015					} else {
  1016						s := string(d.buf.Bytes()[start:])
  1017						d.buf.WriteByte(';')
  1018						n, err := strconv.ParseUint(s, base, 64)
  1019						if err == nil && n <= unicode.MaxRune {
  1020							text = string(n)
  1021							haveText = true
  1022						}
  1023					}
  1024				} else {
  1025					d.ungetc(b)
  1026					if !d.readName() {
  1027						if d.err != nil {
  1028							return nil
  1029						}
  1030						ok = false
  1031					}
  1032					if b, ok = d.mustgetc(); !ok {
  1033						return nil
  1034					}
  1035					if b != ';' {
  1036						d.ungetc(b)
  1037					} else {
  1038						name := d.buf.Bytes()[before+1:]
  1039						d.buf.WriteByte(';')
  1040						if isName(name) {
  1041							s := string(name)
  1042							if r, ok := entity[s]; ok {
  1043								text = string(r)
  1044								haveText = true
  1045							} else if d.Entity != nil {
  1046								text, haveText = d.Entity[s]
  1047							}
  1048						}
  1049					}
  1050				}
  1051	
  1052				if haveText {
  1053					d.buf.Truncate(before)
  1054					d.buf.Write([]byte(text))
  1055					b0, b1 = 0, 0
  1056					continue Input
  1057				}
  1058				if !d.Strict {
  1059					b0, b1 = 0, 0
  1060					continue Input
  1061				}
  1062				ent := string(d.buf.Bytes()[before:])
  1063				if ent[len(ent)-1] != ';' {
  1064					ent += " (no semicolon)"
  1065				}
  1066				d.err = d.syntaxError("invalid character entity " + ent)
  1067				return nil
  1068			}
  1069	
  1070			// We must rewrite unescaped \r and \r\n into \n.
  1071			if b == '\r' {
  1072				d.buf.WriteByte('\n')
  1073			} else if b1 == '\r' && b == '\n' {
  1074				// Skip \r\n--we already wrote \n.
  1075			} else {
  1076				d.buf.WriteByte(b)
  1077			}
  1078	
  1079			b0, b1 = b1, b
  1080		}
  1081		data := d.buf.Bytes()
  1082		data = data[0 : len(data)-trunc]
  1083	
  1084		// Inspect each rune for being a disallowed character.
  1085		buf := data
  1086		for len(buf) > 0 {
  1087			r, size := utf8.DecodeRune(buf)
  1088			if r == utf8.RuneError && size == 1 {
  1089				d.err = d.syntaxError("invalid UTF-8")
  1090				return nil
  1091			}
  1092			buf = buf[size:]
  1093			if !isInCharacterRange(r) {
  1094				d.err = d.syntaxError(fmt.Sprintf("illegal character code %U", r))
  1095				return nil
  1096			}
  1097		}
  1098	
  1099		return data
  1100	}
  1101	
  1102	// Decide whether the given rune is in the XML Character Range, per
  1103	// the Char production of http://www.xml.com/axml/testaxml.htm,
  1104	// Section 2.2 Characters.
  1105	func isInCharacterRange(r rune) (inrange bool) {
  1106		return r == 0x09 ||
  1107			r == 0x0A ||
  1108			r == 0x0D ||
  1109			r >= 0x20 && r <= 0xDF77 ||
  1110			r >= 0xE000 && r <= 0xFFFD ||
  1111			r >= 0x10000 && r <= 0x10FFFF
  1112	}
  1113	
  1114	// Get name space name: name with a : stuck in the middle.
  1115	// The part before the : is the name space identifier.
  1116	func (d *Decoder) nsname() (name Name, ok bool) {
  1117		s, ok := d.name()
  1118		if !ok {
  1119			return
  1120		}
  1121		i := strings.Index(s, ":")
  1122		if i < 0 {
  1123			name.Local = s
  1124		} else {
  1125			name.Space = s[0:i]
  1126			name.Local = s[i+1:]
  1127		}
  1128		return name, true
  1129	}
  1130	
  1131	// Get name: /first(first|second)*/
  1132	// Do not set d.err if the name is missing (unless unexpected EOF is received):
  1133	// let the caller provide better context.
  1134	func (d *Decoder) name() (s string, ok bool) {
  1135		d.buf.Reset()
  1136		if !d.readName() {
  1137			return "", false
  1138		}
  1139	
  1140		// Now we check the characters.
  1141		b := d.buf.Bytes()
  1142		if !isName(b) {
  1143			d.err = d.syntaxError("invalid XML name: " + string(b))
  1144			return "", false
  1145		}
  1146		return string(b), true
  1147	}
  1148	
  1149	// Read a name and append its bytes to d.buf.
  1150	// The name is delimited by any single-byte character not valid in names.
  1151	// All multi-byte characters are accepted; the caller must check their validity.
  1152	func (d *Decoder) readName() (ok bool) {
  1153		var b byte
  1154		if b, ok = d.mustgetc(); !ok {
  1155			return
  1156		}
  1157		if b < utf8.RuneSelf && !isNameByte(b) {
  1158			d.ungetc(b)
  1159			return false
  1160		}
  1161		d.buf.WriteByte(b)
  1162	
  1163		for {
  1164			if b, ok = d.mustgetc(); !ok {
  1165				return
  1166			}
  1167			if b < utf8.RuneSelf && !isNameByte(b) {
  1168				d.ungetc(b)
  1169				break
  1170			}
  1171			d.buf.WriteByte(b)
  1172		}
  1173		return true
  1174	}
  1175	
  1176	func isNameByte(c byte) bool {
  1177		return 'A' <= c && c <= 'Z' ||
  1178			'a' <= c && c <= 'z' ||
  1179			'0' <= c && c <= '9' ||
  1180			c == '_' || c == ':' || c == '.' || c == '-'
  1181	}
  1182	
  1183	func isName(s []byte) bool {
  1184		if len(s) == 0 {
  1185			return false
  1186		}
  1187		c, n := utf8.DecodeRune(s)
  1188		if c == utf8.RuneError && n == 1 {
  1189			return false
  1190		}
  1191		if !unicode.Is(first, c) {
  1192			return false
  1193		}
  1194		for n < len(s) {
  1195			s = s[n:]
  1196			c, n = utf8.DecodeRune(s)
  1197			if c == utf8.RuneError && n == 1 {
  1198				return false
  1199			}
  1200			if !unicode.Is(first, c) && !unicode.Is(second, c) {
  1201				return false
  1202			}
  1203		}
  1204		return true
  1205	}
  1206	
  1207	func isNameString(s string) bool {
  1208		if len(s) == 0 {
  1209			return false
  1210		}
  1211		c, n := utf8.DecodeRuneInString(s)
  1212		if c == utf8.RuneError && n == 1 {
  1213			return false
  1214		}
  1215		if !unicode.Is(first, c) {
  1216			return false
  1217		}
  1218		for n < len(s) {
  1219			s = s[n:]
  1220			c, n = utf8.DecodeRuneInString(s)
  1221			if c == utf8.RuneError && n == 1 {
  1222				return false
  1223			}
  1224			if !unicode.Is(first, c) && !unicode.Is(second, c) {
  1225				return false
  1226			}
  1227		}
  1228		return true
  1229	}
  1230	
  1231	// These tables were generated by cut and paste from Appendix B of
  1232	// the XML spec at http://www.xml.com/axml/testaxml.htm
  1233	// and then reformatting. First corresponds to (Letter | '_' | ':')
  1234	// and second corresponds to NameChar.
  1235	
  1236	var first = &unicode.RangeTable{
  1237		R16: []unicode.Range16{
  1238			{0x003A, 0x003A, 1},
  1239			{0x0041, 0x005A, 1},
  1240			{0x005F, 0x005F, 1},
  1241			{0x0061, 0x007A, 1},
  1242			{0x00C0, 0x00D6, 1},
  1243			{0x00D8, 0x00F6, 1},
  1244			{0x00F8, 0x00FF, 1},
  1245			{0x0100, 0x0131, 1},
  1246			{0x0134, 0x013E, 1},
  1247			{0x0141, 0x0148, 1},
  1248			{0x014A, 0x017E, 1},
  1249			{0x0180, 0x01C3, 1},
  1250			{0x01CD, 0x01F0, 1},
  1251			{0x01F4, 0x01F5, 1},
  1252			{0x01FA, 0x0217, 1},
  1253			{0x0250, 0x02A8, 1},
  1254			{0x02BB, 0x02C1, 1},
  1255			{0x0386, 0x0386, 1},
  1256			{0x0388, 0x038A, 1},
  1257			{0x038C, 0x038C, 1},
  1258			{0x038E, 0x03A1, 1},
  1259			{0x03A3, 0x03CE, 1},
  1260			{0x03D0, 0x03D6, 1},
  1261			{0x03DA, 0x03E0, 2},
  1262			{0x03E2, 0x03F3, 1},
  1263			{0x0401, 0x040C, 1},
  1264			{0x040E, 0x044F, 1},
  1265			{0x0451, 0x045C, 1},
  1266			{0x045E, 0x0481, 1},
  1267			{0x0490, 0x04C4, 1},
  1268			{0x04C7, 0x04C8, 1},
  1269			{0x04CB, 0x04CC, 1},
  1270			{0x04D0, 0x04EB, 1},
  1271			{0x04EE, 0x04F5, 1},
  1272			{0x04F8, 0x04F9, 1},
  1273			{0x0531, 0x0556, 1},
  1274			{0x0559, 0x0559, 1},
  1275			{0x0561, 0x0586, 1},
  1276			{0x05D0, 0x05EA, 1},
  1277			{0x05F0, 0x05F2, 1},
  1278			{0x0621, 0x063A, 1},
  1279			{0x0641, 0x064A, 1},
  1280			{0x0671, 0x06B7, 1},
  1281			{0x06BA, 0x06BE, 1},
  1282			{0x06C0, 0x06CE, 1},
  1283			{0x06D0, 0x06D3, 1},
  1284			{0x06D5, 0x06D5, 1},
  1285			{0x06E5, 0x06E6, 1},
  1286			{0x0905, 0x0939, 1},
  1287			{0x093D, 0x093D, 1},
  1288			{0x0958, 0x0961, 1},
  1289			{0x0985, 0x098C, 1},
  1290			{0x098F, 0x0990, 1},
  1291			{0x0993, 0x09A8, 1},
  1292			{0x09AA, 0x09B0, 1},
  1293			{0x09B2, 0x09B2, 1},
  1294			{0x09B6, 0x09B9, 1},
  1295			{0x09DC, 0x09DD, 1},
  1296			{0x09DF, 0x09E1, 1},
  1297			{0x09F0, 0x09F1, 1},
  1298			{0x0A05, 0x0A0A, 1},
  1299			{0x0A0F, 0x0A10, 1},
  1300			{0x0A13, 0x0A28, 1},
  1301			{0x0A2A, 0x0A30, 1},
  1302			{0x0A32, 0x0A33, 1},
  1303			{0x0A35, 0x0A36, 1},
  1304			{0x0A38, 0x0A39, 1},
  1305			{0x0A59, 0x0A5C, 1},
  1306			{0x0A5E, 0x0A5E, 1},
  1307			{0x0A72, 0x0A74, 1},
  1308			{0x0A85, 0x0A8B, 1},
  1309			{0x0A8D, 0x0A8D, 1},
  1310			{0x0A8F, 0x0A91, 1},
  1311			{0x0A93, 0x0AA8, 1},
  1312			{0x0AAA, 0x0AB0, 1},
  1313			{0x0AB2, 0x0AB3, 1},
  1314			{0x0AB5, 0x0AB9, 1},
  1315			{0x0ABD, 0x0AE0, 0x23},
  1316			{0x0B05, 0x0B0C, 1},
  1317			{0x0B0F, 0x0B10, 1},
  1318			{0x0B13, 0x0B28, 1},
  1319			{0x0B2A, 0x0B30, 1},
  1320			{0x0B32, 0x0B33, 1},
  1321			{0x0B36, 0x0B39, 1},
  1322			{0x0B3D, 0x0B3D, 1},
  1323			{0x0B5C, 0x0B5D, 1},
  1324			{0x0B5F, 0x0B61, 1},
  1325			{0x0B85, 0x0B8A, 1},
  1326			{0x0B8E, 0x0B90, 1},
  1327			{0x0B92, 0x0B95, 1},
  1328			{0x0B99, 0x0B9A, 1},
  1329			{0x0B9C, 0x0B9C, 1},
  1330			{0x0B9E, 0x0B9F, 1},
  1331			{0x0BA3, 0x0BA4, 1},
  1332			{0x0BA8, 0x0BAA, 1},
  1333			{0x0BAE, 0x0BB5, 1},
  1334			{0x0BB7, 0x0BB9, 1},
  1335			{0x0C05, 0x0C0C, 1},
  1336			{0x0C0E, 0x0C10, 1},
  1337			{0x0C12, 0x0C28, 1},
  1338			{0x0C2A, 0x0C33, 1},
  1339			{0x0C35, 0x0C39, 1},
  1340			{0x0C60, 0x0C61, 1},
  1341			{0x0C85, 0x0C8C, 1},
  1342			{0x0C8E, 0x0C90, 1},
  1343			{0x0C92, 0x0CA8, 1},
  1344			{0x0CAA, 0x0CB3, 1},
  1345			{0x0CB5, 0x0CB9, 1},
  1346			{0x0CDE, 0x0CDE, 1},
  1347			{0x0CE0, 0x0CE1, 1},
  1348			{0x0D05, 0x0D0C, 1},
  1349			{0x0D0E, 0x0D10, 1},
  1350			{0x0D12, 0x0D28, 1},
  1351			{0x0D2A, 0x0D39, 1},
  1352			{0x0D60, 0x0D61, 1},
  1353			{0x0E01, 0x0E2E, 1},
  1354			{0x0E30, 0x0E30, 1},
  1355			{0x0E32, 0x0E33, 1},
  1356			{0x0E40, 0x0E45, 1},
  1357			{0x0E81, 0x0E82, 1},
  1358			{0x0E84, 0x0E84, 1},
  1359			{0x0E87, 0x0E88, 1},
  1360			{0x0E8A, 0x0E8D, 3},
  1361			{0x0E94, 0x0E97, 1},
  1362			{0x0E99, 0x0E9F, 1},
  1363			{0x0EA1, 0x0EA3, 1},
  1364			{0x0EA5, 0x0EA7, 2},
  1365			{0x0EAA, 0x0EAB, 1},
  1366			{0x0EAD, 0x0EAE, 1},
  1367			{0x0EB0, 0x0EB0, 1},
  1368			{0x0EB2, 0x0EB3, 1},
  1369			{0x0EBD, 0x0EBD, 1},
  1370			{0x0EC0, 0x0EC4, 1},
  1371			{0x0F40, 0x0F47, 1},
  1372			{0x0F49, 0x0F69, 1},
  1373			{0x10A0, 0x10C5, 1},
  1374			{0x10D0, 0x10F6, 1},
  1375			{0x1100, 0x1100, 1},
  1376			{0x1102, 0x1103, 1},
  1377			{0x1105, 0x1107, 1},
  1378			{0x1109, 0x1109, 1},
  1379			{0x110B, 0x110C, 1},
  1380			{0x110E, 0x1112, 1},
  1381			{0x113C, 0x1140, 2},
  1382			{0x114C, 0x1150, 2},
  1383			{0x1154, 0x1155, 1},
  1384			{0x1159, 0x1159, 1},
  1385			{0x115F, 0x1161, 1},
  1386			{0x1163, 0x1169, 2},
  1387			{0x116D, 0x116E, 1},
  1388			{0x1172, 0x1173, 1},
  1389			{0x1175, 0x119E, 0x119E - 0x1175},
  1390			{0x11A8, 0x11AB, 0x11AB - 0x11A8},
  1391			{0x11AE, 0x11AF, 1},
  1392			{0x11B7, 0x11B8, 1},
  1393			{0x11BA, 0x11BA, 1},
  1394			{0x11BC, 0x11C2, 1},
  1395			{0x11EB, 0x11F0, 0x11F0 - 0x11EB},
  1396			{0x11F9, 0x11F9, 1},
  1397			{0x1E00, 0x1E9B, 1},
  1398			{0x1EA0, 0x1EF9, 1},
  1399			{0x1F00, 0x1F15, 1},
  1400			{0x1F18, 0x1F1D, 1},
  1401			{0x1F20, 0x1F45, 1},
  1402			{0x1F48, 0x1F4D, 1},
  1403			{0x1F50, 0x1F57, 1},
  1404			{0x1F59, 0x1F5B, 0x1F5B - 0x1F59},
  1405			{0x1F5D, 0x1F5D, 1},
  1406			{0x1F5F, 0x1F7D, 1},
  1407			{0x1F80, 0x1FB4, 1},
  1408			{0x1FB6, 0x1FBC, 1},
  1409			{0x1FBE, 0x1FBE, 1},
  1410			{0x1FC2, 0x1FC4, 1},
  1411			{0x1FC6, 0x1FCC, 1},
  1412			{0x1FD0, 0x1FD3, 1},
  1413			{0x1FD6, 0x1FDB, 1},
  1414			{0x1FE0, 0x1FEC, 1},
  1415			{0x1FF2, 0x1FF4, 1},
  1416			{0x1FF6, 0x1FFC, 1},
  1417			{0x2126, 0x2126, 1},
  1418			{0x212A, 0x212B, 1},
  1419			{0x212E, 0x212E, 1},
  1420			{0x2180, 0x2182, 1},
  1421			{0x3007, 0x3007, 1},
  1422			{0x3021, 0x3029, 1},
  1423			{0x3041, 0x3094, 1},
  1424			{0x30A1, 0x30FA, 1},
  1425			{0x3105, 0x312C, 1},
  1426			{0x4E00, 0x9FA5, 1},
  1427			{0xAC00, 0xD7A3, 1},
  1428		},
  1429	}
  1430	
  1431	var second = &unicode.RangeTable{
  1432		R16: []unicode.Range16{
  1433			{0x002D, 0x002E, 1},
  1434			{0x0030, 0x0039, 1},
  1435			{0x00B7, 0x00B7, 1},
  1436			{0x02D0, 0x02D1, 1},
  1437			{0x0300, 0x0345, 1},
  1438			{0x0360, 0x0361, 1},
  1439			{0x0387, 0x0387, 1},
  1440			{0x0483, 0x0486, 1},
  1441			{0x0591, 0x05A1, 1},
  1442			{0x05A3, 0x05B9, 1},
  1443			{0x05BB, 0x05BD, 1},
  1444			{0x05BF, 0x05BF, 1},
  1445			{0x05C1, 0x05C2, 1},
  1446			{0x05C4, 0x0640, 0x0640 - 0x05C4},
  1447			{0x064B, 0x0652, 1},
  1448			{0x0660, 0x0669, 1},
  1449			{0x0670, 0x0670, 1},
  1450			{0x06D6, 0x06DC, 1},
  1451			{0x06DD, 0x06DF, 1},
  1452			{0x06E0, 0x06E4, 1},
  1453			{0x06E7, 0x06E8, 1},
  1454			{0x06EA, 0x06ED, 1},
  1455			{0x06F0, 0x06F9, 1},
  1456			{0x0901, 0x0903, 1},
  1457			{0x093C, 0x093C, 1},
  1458			{0x093E, 0x094C, 1},
  1459			{0x094D, 0x094D, 1},
  1460			{0x0951, 0x0954, 1},
  1461			{0x0962, 0x0963, 1},
  1462			{0x0966, 0x096F, 1},
  1463			{0x0981, 0x0983, 1},
  1464			{0x09BC, 0x09BC, 1},
  1465			{0x09BE, 0x09BF, 1},
  1466			{0x09C0, 0x09C4, 1},
  1467			{0x09C7, 0x09C8, 1},
  1468			{0x09CB, 0x09CD, 1},
  1469			{0x09D7, 0x09D7, 1},
  1470			{0x09E2, 0x09E3, 1},
  1471			{0x09E6, 0x09EF, 1},
  1472			{0x0A02, 0x0A3C, 0x3A},
  1473			{0x0A3E, 0x0A3F, 1},
  1474			{0x0A40, 0x0A42, 1},
  1475			{0x0A47, 0x0A48, 1},
  1476			{0x0A4B, 0x0A4D, 1},
  1477			{0x0A66, 0x0A6F, 1},
  1478			{0x0A70, 0x0A71, 1},
  1479			{0x0A81, 0x0A83, 1},
  1480			{0x0ABC, 0x0ABC, 1},
  1481			{0x0ABE, 0x0AC5, 1},
  1482			{0x0AC7, 0x0AC9, 1},
  1483			{0x0ACB, 0x0ACD, 1},
  1484			{0x0AE6, 0x0AEF, 1},
  1485			{0x0B01, 0x0B03, 1},
  1486			{0x0B3C, 0x0B3C, 1},
  1487			{0x0B3E, 0x0B43, 1},
  1488			{0x0B47, 0x0B48, 1},
  1489			{0x0B4B, 0x0B4D, 1},
  1490			{0x0B56, 0x0B57, 1},
  1491			{0x0B66, 0x0B6F, 1},
  1492			{0x0B82, 0x0B83, 1},
  1493			{0x0BBE, 0x0BC2, 1},
  1494			{0x0BC6, 0x0BC8, 1},
  1495			{0x0BCA, 0x0BCD, 1},
  1496			{0x0BD7, 0x0BD7, 1},
  1497			{0x0BE7, 0x0BEF, 1},
  1498			{0x0C01, 0x0C03, 1},
  1499			{0x0C3E, 0x0C44, 1},
  1500			{0x0C46, 0x0C48, 1},
  1501			{0x0C4A, 0x0C4D, 1},
  1502			{0x0C55, 0x0C56, 1},
  1503			{0x0C66, 0x0C6F, 1},
  1504			{0x0C82, 0x0C83, 1},
  1505			{0x0CBE, 0x0CC4, 1},
  1506			{0x0CC6, 0x0CC8, 1},
  1507			{0x0CCA, 0x0CCD, 1},
  1508			{0x0CD5, 0x0CD6, 1},
  1509			{0x0CE6, 0x0CEF, 1},
  1510			{0x0D02, 0x0D03, 1},
  1511			{0x0D3E, 0x0D43, 1},
  1512			{0x0D46, 0x0D48, 1},
  1513			{0x0D4A, 0x0D4D, 1},
  1514			{0x0D57, 0x0D57, 1},
  1515			{0x0D66, 0x0D6F, 1},
  1516			{0x0E31, 0x0E31, 1},
  1517			{0x0E34, 0x0E3A, 1},
  1518			{0x0E46, 0x0E46, 1},
  1519			{0x0E47, 0x0E4E, 1},
  1520			{0x0E50, 0x0E59, 1},
  1521			{0x0EB1, 0x0EB1, 1},
  1522			{0x0EB4, 0x0EB9, 1},
  1523			{0x0EBB, 0x0EBC, 1},
  1524			{0x0EC6, 0x0EC6, 1},
  1525			{0x0EC8, 0x0ECD, 1},
  1526			{0x0ED0, 0x0ED9, 1},
  1527			{0x0F18, 0x0F19, 1},
  1528			{0x0F20, 0x0F29, 1},
  1529			{0x0F35, 0x0F39, 2},
  1530			{0x0F3E, 0x0F3F, 1},
  1531			{0x0F71, 0x0F84, 1},
  1532			{0x0F86, 0x0F8B, 1},
  1533			{0x0F90, 0x0F95, 1},
  1534			{0x0F97, 0x0F97, 1},
  1535			{0x0F99, 0x0FAD, 1},
  1536			{0x0FB1, 0x0FB7, 1},
  1537			{0x0FB9, 0x0FB9, 1},
  1538			{0x20D0, 0x20DC, 1},
  1539			{0x20E1, 0x3005, 0x3005 - 0x20E1},
  1540			{0x302A, 0x302F, 1},
  1541			{0x3031, 0x3035, 1},
  1542			{0x3099, 0x309A, 1},
  1543			{0x309D, 0x309E, 1},
  1544			{0x30FC, 0x30FE, 1},
  1545		},
  1546	}
  1547	
  1548	// HTMLEntity is an entity map containing translations for the
  1549	// standard HTML entity characters.
  1550	var HTMLEntity = htmlEntity
  1551	
  1552	var htmlEntity = map[string]string{
  1553		/*
  1554			hget http://www.w3.org/TR/html4/sgml/entities.html |
  1555			ssam '
  1556				,y /\&gt;/ x/\&lt;(.|\n)+/ s/\n/ /g
  1557				,x v/^\&lt;!ENTITY/d
  1558				,s/\&lt;!ENTITY ([^ ]+) .*U\+([0-9A-F][0-9A-F][0-9A-F][0-9A-F]) .+/	"\1": "\\u\2",/g
  1559			'
  1560		*/
  1561		"nbsp":     "\u00A0",
  1562		"iexcl":    "\u00A1",
  1563		"cent":     "\u00A2",
  1564		"pound":    "\u00A3",
  1565		"curren":   "\u00A4",
  1566		"yen":      "\u00A5",
  1567		"brvbar":   "\u00A6",
  1568		"sect":     "\u00A7",
  1569		"uml":      "\u00A8",
  1570		"copy":     "\u00A9",
  1571		"ordf":     "\u00AA",
  1572		"laquo":    "\u00AB",
  1573		"not":      "\u00AC",
  1574		"shy":      "\u00AD",
  1575		"reg":      "\u00AE",
  1576		"macr":     "\u00AF",
  1577		"deg":      "\u00B0",
  1578		"plusmn":   "\u00B1",
  1579		"sup2":     "\u00B2",
  1580		"sup3":     "\u00B3",
  1581		"acute":    "\u00B4",
  1582		"micro":    "\u00B5",
  1583		"para":     "\u00B6",
  1584		"middot":   "\u00B7",
  1585		"cedil":    "\u00B8",
  1586		"sup1":     "\u00B9",
  1587		"ordm":     "\u00BA",
  1588		"raquo":    "\u00BB",
  1589		"frac14":   "\u00BC",
  1590		"frac12":   "\u00BD",
  1591		"frac34":   "\u00BE",
  1592		"iquest":   "\u00BF",
  1593		"Agrave":   "\u00C0",
  1594		"Aacute":   "\u00C1",
  1595		"Acirc":    "\u00C2",
  1596		"Atilde":   "\u00C3",
  1597		"Auml":     "\u00C4",
  1598		"Aring":    "\u00C5",
  1599		"AElig":    "\u00C6",
  1600		"Ccedil":   "\u00C7",
  1601		"Egrave":   "\u00C8",
  1602		"Eacute":   "\u00C9",
  1603		"Ecirc":    "\u00CA",
  1604		"Euml":     "\u00CB",
  1605		"Igrave":   "\u00CC",
  1606		"Iacute":   "\u00CD",
  1607		"Icirc":    "\u00CE",
  1608		"Iuml":     "\u00CF",
  1609		"ETH":      "\u00D0",
  1610		"Ntilde":   "\u00D1",
  1611		"Ograve":   "\u00D2",
  1612		"Oacute":   "\u00D3",
  1613		"Ocirc":    "\u00D4",
  1614		"Otilde":   "\u00D5",
  1615		"Ouml":     "\u00D6",
  1616		"times":    "\u00D7",
  1617		"Oslash":   "\u00D8",
  1618		"Ugrave":   "\u00D9",
  1619		"Uacute":   "\u00DA",
  1620		"Ucirc":    "\u00DB",
  1621		"Uuml":     "\u00DC",
  1622		"Yacute":   "\u00DD",
  1623		"THORN":    "\u00DE",
  1624		"szlig":    "\u00DF",
  1625		"agrave":   "\u00E0",
  1626		"aacute":   "\u00E1",
  1627		"acirc":    "\u00E2",
  1628		"atilde":   "\u00E3",
  1629		"auml":     "\u00E4",
  1630		"aring":    "\u00E5",
  1631		"aelig":    "\u00E6",
  1632		"ccedil":   "\u00E7",
  1633		"egrave":   "\u00E8",
  1634		"eacute":   "\u00E9",
  1635		"ecirc":    "\u00EA",
  1636		"euml":     "\u00EB",
  1637		"igrave":   "\u00EC",
  1638		"iacute":   "\u00ED",
  1639		"icirc":    "\u00EE",
  1640		"iuml":     "\u00EF",
  1641		"eth":      "\u00F0",
  1642		"ntilde":   "\u00F1",
  1643		"ograve":   "\u00F2",
  1644		"oacute":   "\u00F3",
  1645		"ocirc":    "\u00F4",
  1646		"otilde":   "\u00F5",
  1647		"ouml":     "\u00F6",
  1648		"divide":   "\u00F7",
  1649		"oslash":   "\u00F8",
  1650		"ugrave":   "\u00F9",
  1651		"uacute":   "\u00FA",
  1652		"ucirc":    "\u00FB",
  1653		"uuml":     "\u00FC",
  1654		"yacute":   "\u00FD",
  1655		"thorn":    "\u00FE",
  1656		"yuml":     "\u00FF",
  1657		"fnof":     "\u0192",
  1658		"Alpha":    "\u0391",
  1659		"Beta":     "\u0392",
  1660		"Gamma":    "\u0393",
  1661		"Delta":    "\u0394",
  1662		"Epsilon":  "\u0395",
  1663		"Zeta":     "\u0396",
  1664		"Eta":      "\u0397",
  1665		"Theta":    "\u0398",
  1666		"Iota":     "\u0399",
  1667		"Kappa":    "\u039A",
  1668		"Lambda":   "\u039B",
  1669		"Mu":       "\u039C",
  1670		"Nu":       "\u039D",
  1671		"Xi":       "\u039E",
  1672		"Omicron":  "\u039F",
  1673		"Pi":       "\u03A0",
  1674		"Rho":      "\u03A1",
  1675		"Sigma":    "\u03A3",
  1676		"Tau":      "\u03A4",
  1677		"Upsilon":  "\u03A5",
  1678		"Phi":      "\u03A6",
  1679		"Chi":      "\u03A7",
  1680		"Psi":      "\u03A8",
  1681		"Omega":    "\u03A9",
  1682		"alpha":    "\u03B1",
  1683		"beta":     "\u03B2",
  1684		"gamma":    "\u03B3",
  1685		"delta":    "\u03B4",
  1686		"epsilon":  "\u03B5",
  1687		"zeta":     "\u03B6",
  1688		"eta":      "\u03B7",
  1689		"theta":    "\u03B8",
  1690		"iota":     "\u03B9",
  1691		"kappa":    "\u03BA",
  1692		"lambda":   "\u03BB",
  1693		"mu":       "\u03BC",
  1694		"nu":       "\u03BD",
  1695		"xi":       "\u03BE",
  1696		"omicron":  "\u03BF",
  1697		"pi":       "\u03C0",
  1698		"rho":      "\u03C1",
  1699		"sigmaf":   "\u03C2",
  1700		"sigma":    "\u03C3",
  1701		"tau":      "\u03C4",
  1702		"upsilon":  "\u03C5",
  1703		"phi":      "\u03C6",
  1704		"chi":      "\u03C7",
  1705		"psi":      "\u03C8",
  1706		"omega":    "\u03C9",
  1707		"thetasym": "\u03D1",
  1708		"upsih":    "\u03D2",
  1709		"piv":      "\u03D6",
  1710		"bull":     "\u2022",
  1711		"hellip":   "\u2026",
  1712		"prime":    "\u2032",
  1713		"Prime":    "\u2033",
  1714		"oline":    "\u203E",
  1715		"frasl":    "\u2044",
  1716		"weierp":   "\u2118",
  1717		"image":    "\u2111",
  1718		"real":     "\u211C",
  1719		"trade":    "\u2122",
  1720		"alefsym":  "\u2135",
  1721		"larr":     "\u2190",
  1722		"uarr":     "\u2191",
  1723		"rarr":     "\u2192",
  1724		"darr":     "\u2193",
  1725		"harr":     "\u2194",
  1726		"crarr":    "\u21B5",
  1727		"lArr":     "\u21D0",
  1728		"uArr":     "\u21D1",
  1729		"rArr":     "\u21D2",
  1730		"dArr":     "\u21D3",
  1731		"hArr":     "\u21D4",
  1732		"forall":   "\u2200",
  1733		"part":     "\u2202",
  1734		"exist":    "\u2203",
  1735		"empty":    "\u2205",
  1736		"nabla":    "\u2207",
  1737		"isin":     "\u2208",
  1738		"notin":    "\u2209",
  1739		"ni":       "\u220B",
  1740		"prod":     "\u220F",
  1741		"sum":      "\u2211",
  1742		"minus":    "\u2212",
  1743		"lowast":   "\u2217",
  1744		"radic":    "\u221A",
  1745		"prop":     "\u221D",
  1746		"infin":    "\u221E",
  1747		"ang":      "\u2220",
  1748		"and":      "\u2227",
  1749		"or":       "\u2228",
  1750		"cap":      "\u2229",
  1751		"cup":      "\u222A",
  1752		"int":      "\u222B",
  1753		"there4":   "\u2234",
  1754		"sim":      "\u223C",
  1755		"cong":     "\u2245",
  1756		"asymp":    "\u2248",
  1757		"ne":       "\u2260",
  1758		"equiv":    "\u2261",
  1759		"le":       "\u2264",
  1760		"ge":       "\u2265",
  1761		"sub":      "\u2282",
  1762		"sup":      "\u2283",
  1763		"nsub":     "\u2284",
  1764		"sube":     "\u2286",
  1765		"supe":     "\u2287",
  1766		"oplus":    "\u2295",
  1767		"otimes":   "\u2297",
  1768		"perp":     "\u22A5",
  1769		"sdot":     "\u22C5",
  1770		"lceil":    "\u2308",
  1771		"rceil":    "\u2309",
  1772		"lfloor":   "\u230A",
  1773		"rfloor":   "\u230B",
  1774		"lang":     "\u2329",
  1775		"rang":     "\u232A",
  1776		"loz":      "\u25CA",
  1777		"spades":   "\u2660",
  1778		"clubs":    "\u2663",
  1779		"hearts":   "\u2665",
  1780		"diams":    "\u2666",
  1781		"quot":     "\u0022",
  1782		"amp":      "\u0026",
  1783		"lt":       "\u003C",
  1784		"gt":       "\u003E",
  1785		"OElig":    "\u0152",
  1786		"oelig":    "\u0153",
  1787		"Scaron":   "\u0160",
  1788		"scaron":   "\u0161",
  1789		"Yuml":     "\u0178",
  1790		"circ":     "\u02C6",
  1791		"tilde":    "\u02DC",
  1792		"ensp":     "\u2002",
  1793		"emsp":     "\u2003",
  1794		"thinsp":   "\u2009",
  1795		"zwnj":     "\u200C",
  1796		"zwj":      "\u200D",
  1797		"lrm":      "\u200E",
  1798		"rlm":      "\u200F",
  1799		"ndash":    "\u2013",
  1800		"mdash":    "\u2014",
  1801		"lsquo":    "\u2018",
  1802		"rsquo":    "\u2019",
  1803		"sbquo":    "\u201A",
  1804		"ldquo":    "\u201C",
  1805		"rdquo":    "\u201D",
  1806		"bdquo":    "\u201E",
  1807		"dagger":   "\u2020",
  1808		"Dagger":   "\u2021",
  1809		"permil":   "\u2030",
  1810		"lsaquo":   "\u2039",
  1811		"rsaquo":   "\u203A",
  1812		"euro":     "\u20AC",
  1813	}
  1814	
  1815	// HTMLAutoClose is the set of HTML elements that
  1816	// should be considered to close automatically.
  1817	var HTMLAutoClose = htmlAutoClose
  1818	
  1819	var htmlAutoClose = []string{
  1820		/*
  1821			hget http://www.w3.org/TR/html4/loose.dtd |
  1822			9 sed -n 's/<!ELEMENT ([^ ]*) +- O EMPTY.+/	"\1",/p' | tr A-Z a-z
  1823		*/
  1824		"basefont",
  1825		"br",
  1826		"area",
  1827		"link",
  1828		"img",
  1829		"param",
  1830		"hr",
  1831		"input",
  1832		"col",
  1833		"frame",
  1834		"isindex",
  1835		"base",
  1836		"meta",
  1837	}
  1838	
  1839	var (
  1840		esc_quot = []byte("&#34;") // shorter than "&quot;"
  1841		esc_apos = []byte("&#39;") // shorter than "&apos;"
  1842		esc_amp  = []byte("&amp;")
  1843		esc_lt   = []byte("&lt;")
  1844		esc_gt   = []byte("&gt;")
  1845		esc_tab  = []byte("&#x9;")
  1846		esc_nl   = []byte("&#xA;")
  1847		esc_cr   = []byte("&#xD;")
  1848		esc_fffd = []byte("\uFFFD") // Unicode replacement character
  1849	)
  1850	
  1851	// EscapeText writes to w the properly escaped XML equivalent
  1852	// of the plain text data s.
  1853	func EscapeText(w io.Writer, s []byte) error {
  1854		return escapeText(w, s, true)
  1855	}
  1856	
  1857	// escapeText writes to w the properly escaped XML equivalent
  1858	// of the plain text data s. If escapeNewline is true, newline
  1859	// characters will be escaped.
  1860	func escapeText(w io.Writer, s []byte, escapeNewline bool) error {
  1861		var esc []byte
  1862		last := 0
  1863		for i := 0; i < len(s); {
  1864			r, width := utf8.DecodeRune(s[i:])
  1865			i += width
  1866			switch r {
  1867			case '"':
  1868				esc = esc_quot
  1869			case '\'':
  1870				esc = esc_apos
  1871			case '&':
  1872				esc = esc_amp
  1873			case '<':
  1874				esc = esc_lt
  1875			case '>':
  1876				esc = esc_gt
  1877			case '\t':
  1878				esc = esc_tab
  1879			case '\n':
  1880				if !escapeNewline {
  1881					continue
  1882				}
  1883				esc = esc_nl
  1884			case '\r':
  1885				esc = esc_cr
  1886			default:
  1887				if !isInCharacterRange(r) || (r == 0xFFFD && width == 1) {
  1888					esc = esc_fffd
  1889					break
  1890				}
  1891				continue
  1892			}
  1893			if _, err := w.Write(s[last : i-width]); err != nil {
  1894				return err
  1895			}
  1896			if _, err := w.Write(esc); err != nil {
  1897				return err
  1898			}
  1899			last = i
  1900		}
  1901		if _, err := w.Write(s[last:]); err != nil {
  1902			return err
  1903		}
  1904		return nil
  1905	}
  1906	
  1907	// EscapeString writes to p the properly escaped XML equivalent
  1908	// of the plain text data s.
  1909	func (p *printer) EscapeString(s string) {
  1910		var esc []byte
  1911		last := 0
  1912		for i := 0; i < len(s); {
  1913			r, width := utf8.DecodeRuneInString(s[i:])
  1914			i += width
  1915			switch r {
  1916			case '"':
  1917				esc = esc_quot
  1918			case '\'':
  1919				esc = esc_apos
  1920			case '&':
  1921				esc = esc_amp
  1922			case '<':
  1923				esc = esc_lt
  1924			case '>':
  1925				esc = esc_gt
  1926			case '\t':
  1927				esc = esc_tab
  1928			case '\n':
  1929				esc = esc_nl
  1930			case '\r':
  1931				esc = esc_cr
  1932			default:
  1933				if !isInCharacterRange(r) || (r == 0xFFFD && width == 1) {
  1934					esc = esc_fffd
  1935					break
  1936				}
  1937				continue
  1938			}
  1939			p.WriteString(s[last : i-width])
  1940			p.Write(esc)
  1941			last = i
  1942		}
  1943		p.WriteString(s[last:])
  1944	}
  1945	
  1946	// Escape is like EscapeText but omits the error return value.
  1947	// It is provided for backwards compatibility with Go 1.0.
  1948	// Code targeting Go 1.1 or later should use EscapeText.
  1949	func Escape(w io.Writer, s []byte) {
  1950		EscapeText(w, s)
  1951	}
  1952	
  1953	var (
  1954		cdataStart  = []byte("<![CDATA[")
  1955		cdataEnd    = []byte("]]>")
  1956		cdataEscape = []byte("]]]]><![CDATA[>")
  1957	)
  1958	
  1959	// emitCDATA writes to w the CDATA-wrapped plain text data s.
  1960	// It escapes CDATA directives nested in s.
  1961	func emitCDATA(w io.Writer, s []byte) error {
  1962		if len(s) == 0 {
  1963			return nil
  1964		}
  1965		if _, err := w.Write(cdataStart); err != nil {
  1966			return err
  1967		}
  1968		for {
  1969			i := bytes.Index(s, cdataEnd)
  1970			if i >= 0 && i+len(cdataEnd) <= len(s) {
  1971				// Found a nested CDATA directive end.
  1972				if _, err := w.Write(s[:i]); err != nil {
  1973					return err
  1974				}
  1975				if _, err := w.Write(cdataEscape); err != nil {
  1976					return err
  1977				}
  1978				i += len(cdataEnd)
  1979			} else {
  1980				if _, err := w.Write(s); err != nil {
  1981					return err
  1982				}
  1983				break
  1984			}
  1985			s = s[i:]
  1986		}
  1987		if _, err := w.Write(cdataEnd); err != nil {
  1988			return err
  1989		}
  1990		return nil
  1991	}
  1992	
  1993	// procInst parses the `param="..."` or `param='...'`
  1994	// value out of the provided string, returning "" if not found.
  1995	func procInst(param, s string) string {
  1996		// TODO: this parsing is somewhat lame and not exact.
  1997		// It works for all actual cases, though.
  1998		param = param + "="
  1999		idx := strings.Index(s, param)
  2000		if idx == -1 {
  2001			return ""
  2002		}
  2003		v := s[idx+len(param):]
  2004		if v == "" {
  2005			return ""
  2006		}
  2007		if v[0] != '\'' && v[0] != '"' {
  2008			return ""
  2009		}
  2010		idx = strings.IndexRune(v[1:], rune(v[0]))
  2011		if idx == -1 {
  2012			return ""
  2013		}
  2014		return v[1 : idx+1]
  2015	}
  2016	

View as plain text