...
Run Format

Source file src/html/template/escape.go

Documentation: html/template

  // Copyright 2011 The Go Authors. All rights reserved.
  // Use of this source code is governed by a BSD-style
  // license that can be found in the LICENSE file.
  
  package template
  
  import (
  	"bytes"
  	"fmt"
  	"html"
  	"io"
  	"text/template"
  	"text/template/parse"
  )
  
  // escapeTemplate rewrites the named template, which must be
  // associated with t, to guarantee that the output of any of the named
  // templates is properly escaped. If no error is returned, then the named templates have
  // been modified. Otherwise the named templates have been rendered
  // unusable.
  func escapeTemplate(tmpl *Template, node parse.Node, name string) error {
  	c, _ := tmpl.esc.escapeTree(context{}, node, name, 0)
  	var err error
  	if c.err != nil {
  		err, c.err.Name = c.err, name
  	} else if c.state != stateText {
  		err = &Error{ErrEndContext, nil, name, 0, fmt.Sprintf("ends in a non-text context: %v", c)}
  	}
  	if err != nil {
  		// Prevent execution of unsafe templates.
  		if t := tmpl.set[name]; t != nil {
  			t.escapeErr = err
  			t.text.Tree = nil
  			t.Tree = nil
  		}
  		return err
  	}
  	tmpl.esc.commit()
  	if t := tmpl.set[name]; t != nil {
  		t.escapeErr = escapeOK
  		t.Tree = t.text.Tree
  	}
  	return nil
  }
  
  // evalArgs formats the list of arguments into a string. It is equivalent to
  // fmt.Sprint(args...), except that it deferences all pointers.
  func evalArgs(args ...interface{}) string {
  	// Optimization for simple common case of a single string argument.
  	if len(args) == 1 {
  		if s, ok := args[0].(string); ok {
  			return s
  		}
  	}
  	for i, arg := range args {
  		args[i] = indirectToStringerOrError(arg)
  	}
  	return fmt.Sprint(args...)
  }
  
  // funcMap maps command names to functions that render their inputs safe.
  var funcMap = template.FuncMap{
  	"_html_template_attrescaper":     attrEscaper,
  	"_html_template_commentescaper":  commentEscaper,
  	"_html_template_cssescaper":      cssEscaper,
  	"_html_template_cssvaluefilter":  cssValueFilter,
  	"_html_template_htmlnamefilter":  htmlNameFilter,
  	"_html_template_htmlescaper":     htmlEscaper,
  	"_html_template_jsregexpescaper": jsRegexpEscaper,
  	"_html_template_jsstrescaper":    jsStrEscaper,
  	"_html_template_jsvalescaper":    jsValEscaper,
  	"_html_template_nospaceescaper":  htmlNospaceEscaper,
  	"_html_template_rcdataescaper":   rcdataEscaper,
  	"_html_template_urlescaper":      urlEscaper,
  	"_html_template_urlfilter":       urlFilter,
  	"_html_template_urlnormalizer":   urlNormalizer,
  	"_eval_args_":                    evalArgs,
  }
  
  // escaper collects type inferences about templates and changes needed to make
  // templates injection safe.
  type escaper struct {
  	// ns is the nameSpace that this escaper is associated with.
  	ns *nameSpace
  	// output[templateName] is the output context for a templateName that
  	// has been mangled to include its input context.
  	output map[string]context
  	// derived[c.mangle(name)] maps to a template derived from the template
  	// named name templateName for the start context c.
  	derived map[string]*template.Template
  	// called[templateName] is a set of called mangled template names.
  	called map[string]bool
  	// xxxNodeEdits are the accumulated edits to apply during commit.
  	// Such edits are not applied immediately in case a template set
  	// executes a given template in different escaping contexts.
  	actionNodeEdits   map[*parse.ActionNode][]string
  	templateNodeEdits map[*parse.TemplateNode]string
  	textNodeEdits     map[*parse.TextNode][]byte
  }
  
  // makeEscaper creates a blank escaper for the given set.
  func makeEscaper(n *nameSpace) escaper {
  	return escaper{
  		n,
  		map[string]context{},
  		map[string]*template.Template{},
  		map[string]bool{},
  		map[*parse.ActionNode][]string{},
  		map[*parse.TemplateNode]string{},
  		map[*parse.TextNode][]byte{},
  	}
  }
  
  // filterFailsafe is an innocuous word that is emitted in place of unsafe values
  // by sanitizer functions. It is not a keyword in any programming language,
  // contains no special characters, is not empty, and when it appears in output
  // it is distinct enough that a developer can find the source of the problem
  // via a search engine.
  const filterFailsafe = "ZgotmplZ"
  
  // escape escapes a template node.
  func (e *escaper) escape(c context, n parse.Node) context {
  	switch n := n.(type) {
  	case *parse.ActionNode:
  		return e.escapeAction(c, n)
  	case *parse.IfNode:
  		return e.escapeBranch(c, &n.BranchNode, "if")
  	case *parse.ListNode:
  		return e.escapeList(c, n)
  	case *parse.RangeNode:
  		return e.escapeBranch(c, &n.BranchNode, "range")
  	case *parse.TemplateNode:
  		return e.escapeTemplate(c, n)
  	case *parse.TextNode:
  		return e.escapeText(c, n)
  	case *parse.WithNode:
  		return e.escapeBranch(c, &n.BranchNode, "with")
  	}
  	panic("escaping " + n.String() + " is unimplemented")
  }
  
  // escapeAction escapes an action template node.
  func (e *escaper) escapeAction(c context, n *parse.ActionNode) context {
  	if len(n.Pipe.Decl) != 0 {
  		// A local variable assignment, not an interpolation.
  		return c
  	}
  	c = nudge(c)
  	// Check for disallowed use of predefined escapers in the pipeline.
  	for pos, idNode := range n.Pipe.Cmds {
  		node, ok := idNode.Args[0].(*parse.IdentifierNode)
  		if !ok {
  			// A predefined escaper "esc" will never be found as an identifier in a
  			// Chain or Field node, since:
  			// - "esc.x ..." is invalid, since predefined escapers return strings, and
  			//   strings do not have methods, keys or fields.
  			// - "... .esc" is invalid, since predefined escapers are global functions,
  			//   not methods or fields of any types.
  			// Therefore, it is safe to ignore these two node types.
  			continue
  		}
  		ident := node.Ident
  		if _, ok := predefinedEscapers[ident]; ok {
  			if pos < len(n.Pipe.Cmds)-1 ||
  				c.state == stateAttr && c.delim == delimSpaceOrTagEnd && ident == "html" {
  				return context{
  					state: stateError,
  					err:   errorf(ErrPredefinedEscaper, n, n.Line, "predefined escaper %q disallowed in template", ident),
  				}
  			}
  		}
  	}
  	s := make([]string, 0, 3)
  	switch c.state {
  	case stateError:
  		return c
  	case stateURL, stateCSSDqStr, stateCSSSqStr, stateCSSDqURL, stateCSSSqURL, stateCSSURL:
  		switch c.urlPart {
  		case urlPartNone:
  			s = append(s, "_html_template_urlfilter")
  			fallthrough
  		case urlPartPreQuery:
  			switch c.state {
  			case stateCSSDqStr, stateCSSSqStr:
  				s = append(s, "_html_template_cssescaper")
  			default:
  				s = append(s, "_html_template_urlnormalizer")
  			}
  		case urlPartQueryOrFrag:
  			s = append(s, "_html_template_urlescaper")
  		case urlPartUnknown:
  			return context{
  				state: stateError,
  				err:   errorf(ErrAmbigContext, n, n.Line, "%s appears in an ambiguous context within a URL", n),
  			}
  		default:
  			panic(c.urlPart.String())
  		}
  	case stateJS:
  		s = append(s, "_html_template_jsvalescaper")
  		// A slash after a value starts a div operator.
  		c.jsCtx = jsCtxDivOp
  	case stateJSDqStr, stateJSSqStr:
  		s = append(s, "_html_template_jsstrescaper")
  	case stateJSRegexp:
  		s = append(s, "_html_template_jsregexpescaper")
  	case stateCSS:
  		s = append(s, "_html_template_cssvaluefilter")
  	case stateText:
  		s = append(s, "_html_template_htmlescaper")
  	case stateRCDATA:
  		s = append(s, "_html_template_rcdataescaper")
  	case stateAttr:
  		// Handled below in delim check.
  	case stateAttrName, stateTag:
  		c.state = stateAttrName
  		s = append(s, "_html_template_htmlnamefilter")
  	default:
  		if isComment(c.state) {
  			s = append(s, "_html_template_commentescaper")
  		} else {
  			panic("unexpected state " + c.state.String())
  		}
  	}
  	switch c.delim {
  	case delimNone:
  		// No extra-escaping needed for raw text content.
  	case delimSpaceOrTagEnd:
  		s = append(s, "_html_template_nospaceescaper")
  	default:
  		s = append(s, "_html_template_attrescaper")
  	}
  	e.editActionNode(n, s)
  	return c
  }
  
  // ensurePipelineContains ensures that the pipeline ends with the commands with
  // the identifiers in s in order. If the pipeline ends with a predefined escaper
  // (i.e. "html" or "urlquery"), merge it with the identifiers in s.
  func ensurePipelineContains(p *parse.PipeNode, s []string) {
  	if len(s) == 0 {
  		// Do not rewrite pipeline if we have no escapers to insert.
  		return
  	}
  	// Precondition: p.Cmds contains at most one predefined escaper and the
  	// escaper will be present at p.Cmds[len(p.Cmds)-1]. This precondition is
  	// always true because of the checks in escapeAction.
  	pipelineLen := len(p.Cmds)
  	if pipelineLen > 0 {
  		lastCmd := p.Cmds[pipelineLen-1]
  		if idNode, ok := lastCmd.Args[0].(*parse.IdentifierNode); ok {
  			if esc := idNode.Ident; predefinedEscapers[esc] {
  				// Pipeline ends with a predefined escaper.
  				if len(p.Cmds) == 1 && len(lastCmd.Args) > 1 {
  					// Special case: pipeline is of the form {{ esc arg1 arg2 ... argN }},
  					// where esc is the predefined escaper, and arg1...argN are its arguments.
  					// Convert this into the equivalent form
  					// {{ _eval_args_ arg1 arg2 ... argN | esc }}, so that esc can be easily
  					// merged with the escapers in s.
  					lastCmd.Args[0] = parse.NewIdentifier("_eval_args_").SetTree(nil).SetPos(lastCmd.Args[0].Position())
  					p.Cmds = appendCmd(p.Cmds, newIdentCmd(esc, p.Position()))
  					pipelineLen++
  				}
  				// If any of the commands in s that we are about to insert is equivalent
  				// to the predefined escaper, use the predefined escaper instead.
  				dup := false
  				for i, escaper := range s {
  					if escFnsEq(esc, escaper) {
  						s[i] = idNode.Ident
  						dup = true
  					}
  				}
  				if dup {
  					// The predefined escaper will already be inserted along with the
  					// escapers in s, so do not copy it to the rewritten pipeline.
  					pipelineLen--
  				}
  			}
  		}
  	}
  	// Rewrite the pipeline, creating the escapers in s at the end of the pipeline.
  	newCmds := make([]*parse.CommandNode, pipelineLen, pipelineLen+len(s))
  	copy(newCmds, p.Cmds)
  	for _, name := range s {
  		newCmds = appendCmd(newCmds, newIdentCmd(name, p.Position()))
  	}
  	p.Cmds = newCmds
  }
  
  // predefinedEscapers contains template predefined escapers that are equivalent
  // to some contextual escapers. Keep in sync with equivEscapers.
  var predefinedEscapers = map[string]bool{
  	"html":     true,
  	"urlquery": true,
  }
  
  // equivEscapers matches contextual escapers to equivalent predefined
  // template escapers.
  var equivEscapers = map[string]string{
  	// The following pairs of HTML escapers provide equivalent security
  	// guarantees, since they all escape '\000', '\'', '"', '&', '<', and '>'.
  	"_html_template_attrescaper":   "html",
  	"_html_template_htmlescaper":   "html",
  	"_html_template_rcdataescaper": "html",
  	// These two URL escapers produce URLs safe for embedding in a URL query by
  	// percent-encoding all the reserved characters specified in RFC 3986 Section
  	// 2.2
  	"_html_template_urlescaper": "urlquery",
  	// These two functions are not actually equivalent; urlquery is stricter as it
  	// escapes reserved characters (e.g. '#'), while _html_template_urlnormalizer
  	// does not. It is therefore only safe to replace _html_template_urlnormalizer
  	// with urlquery (this happens in ensurePipelineContains), but not the otherI've
  	// way around. We keep this entry around to preserve the behavior of templates
  	// written before Go 1.9, which might depend on this substitution taking place.
  	"_html_template_urlnormalizer": "urlquery",
  }
  
  // escFnsEq reports whether the two escaping functions are equivalent.
  func escFnsEq(a, b string) bool {
  	if e := equivEscapers[a]; e != "" {
  		a = e
  	}
  	if e := equivEscapers[b]; e != "" {
  		b = e
  	}
  	return a == b
  }
  
  // redundantFuncs[a][b] implies that funcMap[b](funcMap[a](x)) == funcMap[a](x)
  // for all x.
  var redundantFuncs = map[string]map[string]bool{
  	"_html_template_commentescaper": {
  		"_html_template_attrescaper":    true,
  		"_html_template_nospaceescaper": true,
  		"_html_template_htmlescaper":    true,
  	},
  	"_html_template_cssescaper": {
  		"_html_template_attrescaper": true,
  	},
  	"_html_template_jsregexpescaper": {
  		"_html_template_attrescaper": true,
  	},
  	"_html_template_jsstrescaper": {
  		"_html_template_attrescaper": true,
  	},
  	"_html_template_urlescaper": {
  		"_html_template_urlnormalizer": true,
  	},
  }
  
  // appendCmd appends the given command to the end of the command pipeline
  // unless it is redundant with the last command.
  func appendCmd(cmds []*parse.CommandNode, cmd *parse.CommandNode) []*parse.CommandNode {
  	if n := len(cmds); n != 0 {
  		last, okLast := cmds[n-1].Args[0].(*parse.IdentifierNode)
  		next, okNext := cmd.Args[0].(*parse.IdentifierNode)
  		if okLast && okNext && redundantFuncs[last.Ident][next.Ident] {
  			return cmds
  		}
  	}
  	return append(cmds, cmd)
  }
  
  // indexOfStr is the first i such that eq(s, strs[i]) or -1 if s was not found.
  func indexOfStr(s string, strs []string, eq func(a, b string) bool) int {
  	for i, t := range strs {
  		if eq(s, t) {
  			return i
  		}
  	}
  	return -1
  }
  
  // newIdentCmd produces a command containing a single identifier node.
  func newIdentCmd(identifier string, pos parse.Pos) *parse.CommandNode {
  	return &parse.CommandNode{
  		NodeType: parse.NodeCommand,
  		Args:     []parse.Node{parse.NewIdentifier(identifier).SetTree(nil).SetPos(pos)}, // TODO: SetTree.
  	}
  }
  
  // nudge returns the context that would result from following empty string
  // transitions from the input context.
  // For example, parsing:
  //     `<a href=`
  // will end in context{stateBeforeValue, attrURL}, but parsing one extra rune:
  //     `<a href=x`
  // will end in context{stateURL, delimSpaceOrTagEnd, ...}.
  // There are two transitions that happen when the 'x' is seen:
  // (1) Transition from a before-value state to a start-of-value state without
  //     consuming any character.
  // (2) Consume 'x' and transition past the first value character.
  // In this case, nudging produces the context after (1) happens.
  func nudge(c context) context {
  	switch c.state {
  	case stateTag:
  		// In `<foo {{.}}`, the action should emit an attribute.
  		c.state = stateAttrName
  	case stateBeforeValue:
  		// In `<foo bar={{.}}`, the action is an undelimited value.
  		c.state, c.delim, c.attr = attrStartStates[c.attr], delimSpaceOrTagEnd, attrNone
  	case stateAfterName:
  		// In `<foo bar {{.}}`, the action is an attribute name.
  		c.state, c.attr = stateAttrName, attrNone
  	}
  	return c
  }
  
  // join joins the two contexts of a branch template node. The result is an
  // error context if either of the input contexts are error contexts, or if the
  // the input contexts differ.
  func join(a, b context, node parse.Node, nodeName string) context {
  	if a.state == stateError {
  		return a
  	}
  	if b.state == stateError {
  		return b
  	}
  	if a.eq(b) {
  		return a
  	}
  
  	c := a
  	c.urlPart = b.urlPart
  	if c.eq(b) {
  		// The contexts differ only by urlPart.
  		c.urlPart = urlPartUnknown
  		return c
  	}
  
  	c = a
  	c.jsCtx = b.jsCtx
  	if c.eq(b) {
  		// The contexts differ only by jsCtx.
  		c.jsCtx = jsCtxUnknown
  		return c
  	}
  
  	// Allow a nudged context to join with an unnudged one.
  	// This means that
  	//   <p title={{if .C}}{{.}}{{end}}
  	// ends in an unquoted value state even though the else branch
  	// ends in stateBeforeValue.
  	if c, d := nudge(a), nudge(b); !(c.eq(a) && d.eq(b)) {
  		if e := join(c, d, node, nodeName); e.state != stateError {
  			return e
  		}
  	}
  
  	return context{
  		state: stateError,
  		err:   errorf(ErrBranchEnd, node, 0, "{{%s}} branches end in different contexts: %v, %v", nodeName, a, b),
  	}
  }
  
  // escapeBranch escapes a branch template node: "if", "range" and "with".
  func (e *escaper) escapeBranch(c context, n *parse.BranchNode, nodeName string) context {
  	c0 := e.escapeList(c, n.List)
  	if nodeName == "range" && c0.state != stateError {
  		// The "true" branch of a "range" node can execute multiple times.
  		// We check that executing n.List once results in the same context
  		// as executing n.List twice.
  		c1, _ := e.escapeListConditionally(c0, n.List, nil)
  		c0 = join(c0, c1, n, nodeName)
  		if c0.state == stateError {
  			// Make clear that this is a problem on loop re-entry
  			// since developers tend to overlook that branch when
  			// debugging templates.
  			c0.err.Line = n.Line
  			c0.err.Description = "on range loop re-entry: " + c0.err.Description
  			return c0
  		}
  	}
  	c1 := e.escapeList(c, n.ElseList)
  	return join(c0, c1, n, nodeName)
  }
  
  // escapeList escapes a list template node.
  func (e *escaper) escapeList(c context, n *parse.ListNode) context {
  	if n == nil {
  		return c
  	}
  	for _, m := range n.Nodes {
  		c = e.escape(c, m)
  	}
  	return c
  }
  
  // escapeListConditionally escapes a list node but only preserves edits and
  // inferences in e if the inferences and output context satisfy filter.
  // It returns the best guess at an output context, and the result of the filter
  // which is the same as whether e was updated.
  func (e *escaper) escapeListConditionally(c context, n *parse.ListNode, filter func(*escaper, context) bool) (context, bool) {
  	e1 := makeEscaper(e.ns)
  	// Make type inferences available to f.
  	for k, v := range e.output {
  		e1.output[k] = v
  	}
  	c = e1.escapeList(c, n)
  	ok := filter != nil && filter(&e1, c)
  	if ok {
  		// Copy inferences and edits from e1 back into e.
  		for k, v := range e1.output {
  			e.output[k] = v
  		}
  		for k, v := range e1.derived {
  			e.derived[k] = v
  		}
  		for k, v := range e1.called {
  			e.called[k] = v
  		}
  		for k, v := range e1.actionNodeEdits {
  			e.editActionNode(k, v)
  		}
  		for k, v := range e1.templateNodeEdits {
  			e.editTemplateNode(k, v)
  		}
  		for k, v := range e1.textNodeEdits {
  			e.editTextNode(k, v)
  		}
  	}
  	return c, ok
  }
  
  // escapeTemplate escapes a {{template}} call node.
  func (e *escaper) escapeTemplate(c context, n *parse.TemplateNode) context {
  	c, name := e.escapeTree(c, n, n.Name, n.Line)
  	if name != n.Name {
  		e.editTemplateNode(n, name)
  	}
  	return c
  }
  
  // escapeTree escapes the named template starting in the given context as
  // necessary and returns its output context.
  func (e *escaper) escapeTree(c context, node parse.Node, name string, line int) (context, string) {
  	// Mangle the template name with the input context to produce a reliable
  	// identifier.
  	dname := c.mangle(name)
  	e.called[dname] = true
  	if out, ok := e.output[dname]; ok {
  		// Already escaped.
  		return out, dname
  	}
  	t := e.template(name)
  	if t == nil {
  		// Two cases: The template exists but is empty, or has never been mentioned at
  		// all. Distinguish the cases in the error messages.
  		if e.ns.set[name] != nil {
  			return context{
  				state: stateError,
  				err:   errorf(ErrNoSuchTemplate, node, line, "%q is an incomplete or empty template", name),
  			}, dname
  		}
  		return context{
  			state: stateError,
  			err:   errorf(ErrNoSuchTemplate, node, line, "no such template %q", name),
  		}, dname
  	}
  	if dname != name {
  		// Use any template derived during an earlier call to escapeTemplate
  		// with different top level templates, or clone if necessary.
  		dt := e.template(dname)
  		if dt == nil {
  			dt = template.New(dname)
  			dt.Tree = &parse.Tree{Name: dname, Root: t.Root.CopyList()}
  			e.derived[dname] = dt
  		}
  		t = dt
  	}
  	return e.computeOutCtx(c, t), dname
  }
  
  // computeOutCtx takes a template and its start context and computes the output
  // context while storing any inferences in e.
  func (e *escaper) computeOutCtx(c context, t *template.Template) context {
  	// Propagate context over the body.
  	c1, ok := e.escapeTemplateBody(c, t)
  	if !ok {
  		// Look for a fixed point by assuming c1 as the output context.
  		if c2, ok2 := e.escapeTemplateBody(c1, t); ok2 {
  			c1, ok = c2, true
  		}
  		// Use c1 as the error context if neither assumption worked.
  	}
  	if !ok && c1.state != stateError {
  		return context{
  			state: stateError,
  			err:   errorf(ErrOutputContext, t.Tree.Root, 0, "cannot compute output context for template %s", t.Name()),
  		}
  	}
  	return c1
  }
  
  // escapeTemplateBody escapes the given template assuming the given output
  // context, and returns the best guess at the output context and whether the
  // assumption was correct.
  func (e *escaper) escapeTemplateBody(c context, t *template.Template) (context, bool) {
  	filter := func(e1 *escaper, c1 context) bool {
  		if c1.state == stateError {
  			// Do not update the input escaper, e.
  			return false
  		}
  		if !e1.called[t.Name()] {
  			// If t is not recursively called, then c1 is an
  			// accurate output context.
  			return true
  		}
  		// c1 is accurate if it matches our assumed output context.
  		return c.eq(c1)
  	}
  	// We need to assume an output context so that recursive template calls
  	// take the fast path out of escapeTree instead of infinitely recursing.
  	// Naively assuming that the input context is the same as the output
  	// works >90% of the time.
  	e.output[t.Name()] = c
  	return e.escapeListConditionally(c, t.Tree.Root, filter)
  }
  
  // delimEnds maps each delim to a string of characters that terminate it.
  var delimEnds = [...]string{
  	delimDoubleQuote: `"`,
  	delimSingleQuote: "'",
  	// Determined empirically by running the below in various browsers.
  	// var div = document.createElement("DIV");
  	// for (var i = 0; i < 0x10000; ++i) {
  	//   div.innerHTML = "<span title=x" + String.fromCharCode(i) + "-bar>";
  	//   if (div.getElementsByTagName("SPAN")[0].title.indexOf("bar") < 0)
  	//     document.write("<p>U+" + i.toString(16));
  	// }
  	delimSpaceOrTagEnd: " \t\n\f\r>",
  }
  
  var doctypeBytes = []byte("<!DOCTYPE")
  
  // escapeText escapes a text template node.
  func (e *escaper) escapeText(c context, n *parse.TextNode) context {
  	s, written, i, b := n.Text, 0, 0, new(bytes.Buffer)
  	for i != len(s) {
  		c1, nread := contextAfterText(c, s[i:])
  		i1 := i + nread
  		if c.state == stateText || c.state == stateRCDATA {
  			end := i1
  			if c1.state != c.state {
  				for j := end - 1; j >= i; j-- {
  					if s[j] == '<' {
  						end = j
  						break
  					}
  				}
  			}
  			for j := i; j < end; j++ {
  				if s[j] == '<' && !bytes.HasPrefix(bytes.ToUpper(s[j:]), doctypeBytes) {
  					b.Write(s[written:j])
  					b.WriteString("&lt;")
  					written = j + 1
  				}
  			}
  		} else if isComment(c.state) && c.delim == delimNone {
  			switch c.state {
  			case stateJSBlockCmt:
  				// http://es5.github.com/#x7.4:
  				// "Comments behave like white space and are
  				// discarded except that, if a MultiLineComment
  				// contains a line terminator character, then
  				// the entire comment is considered to be a
  				// LineTerminator for purposes of parsing by
  				// the syntactic grammar."
  				if bytes.ContainsAny(s[written:i1], "\n\r\u2028\u2029") {
  					b.WriteByte('\n')
  				} else {
  					b.WriteByte(' ')
  				}
  			case stateCSSBlockCmt:
  				b.WriteByte(' ')
  			}
  			written = i1
  		}
  		if c.state != c1.state && isComment(c1.state) && c1.delim == delimNone {
  			// Preserve the portion between written and the comment start.
  			cs := i1 - 2
  			if c1.state == stateHTMLCmt {
  				// "<!--" instead of "/*" or "//"
  				cs -= 2
  			}
  			b.Write(s[written:cs])
  			written = i1
  		}
  		if i == i1 && c.state == c1.state {
  			panic(fmt.Sprintf("infinite loop from %v to %v on %q..%q", c, c1, s[:i], s[i:]))
  		}
  		c, i = c1, i1
  	}
  
  	if written != 0 && c.state != stateError {
  		if !isComment(c.state) || c.delim != delimNone {
  			b.Write(n.Text[written:])
  		}
  		e.editTextNode(n, b.Bytes())
  	}
  	return c
  }
  
  // contextAfterText starts in context c, consumes some tokens from the front of
  // s, then returns the context after those tokens and the unprocessed suffix.
  func contextAfterText(c context, s []byte) (context, int) {
  	if c.delim == delimNone {
  		c1, i := tSpecialTagEnd(c, s)
  		if i == 0 {
  			// A special end tag (`</script>`) has been seen and
  			// all content preceding it has been consumed.
  			return c1, 0
  		}
  		// Consider all content up to any end tag.
  		return transitionFunc[c.state](c, s[:i])
  	}
  
  	// We are at the beginning of an attribute value.
  
  	i := bytes.IndexAny(s, delimEnds[c.delim])
  	if i == -1 {
  		i = len(s)
  	}
  	if c.delim == delimSpaceOrTagEnd {
  		// http://www.w3.org/TR/html5/syntax.html#attribute-value-(unquoted)-state
  		// lists the runes below as error characters.
  		// Error out because HTML parsers may differ on whether
  		// "<a id= onclick=f("     ends inside id's or onclick's value,
  		// "<a class=`foo "        ends inside a value,
  		// "<a style=font:'Arial'" needs open-quote fixup.
  		// IE treats '`' as a quotation character.
  		if j := bytes.IndexAny(s[:i], "\"'<=`"); j >= 0 {
  			return context{
  				state: stateError,
  				err:   errorf(ErrBadHTML, nil, 0, "%q in unquoted attr: %q", s[j:j+1], s[:i]),
  			}, len(s)
  		}
  	}
  	if i == len(s) {
  		// Remain inside the attribute.
  		// Decode the value so non-HTML rules can easily handle
  		//     <button onclick="alert(&quot;Hi!&quot;)">
  		// without having to entity decode token boundaries.
  		for u := []byte(html.UnescapeString(string(s))); len(u) != 0; {
  			c1, i1 := transitionFunc[c.state](c, u)
  			c, u = c1, u[i1:]
  		}
  		return c, len(s)
  	}
  
  	element := c.element
  
  	// If this is a non-JS "type" attribute inside "script" tag, do not treat the contents as JS.
  	if c.state == stateAttr && c.element == elementScript && c.attr == attrScriptType && !isJSType(string(s[:i])) {
  		element = elementNone
  	}
  
  	if c.delim != delimSpaceOrTagEnd {
  		// Consume any quote.
  		i++
  	}
  	// On exiting an attribute, we discard all state information
  	// except the state and element.
  	return context{state: stateTag, element: element}, i
  }
  
  // editActionNode records a change to an action pipeline for later commit.
  func (e *escaper) editActionNode(n *parse.ActionNode, cmds []string) {
  	if _, ok := e.actionNodeEdits[n]; ok {
  		panic(fmt.Sprintf("node %s shared between templates", n))
  	}
  	e.actionNodeEdits[n] = cmds
  }
  
  // editTemplateNode records a change to a {{template}} callee for later commit.
  func (e *escaper) editTemplateNode(n *parse.TemplateNode, callee string) {
  	if _, ok := e.templateNodeEdits[n]; ok {
  		panic(fmt.Sprintf("node %s shared between templates", n))
  	}
  	e.templateNodeEdits[n] = callee
  }
  
  // editTextNode records a change to a text node for later commit.
  func (e *escaper) editTextNode(n *parse.TextNode, text []byte) {
  	if _, ok := e.textNodeEdits[n]; ok {
  		panic(fmt.Sprintf("node %s shared between templates", n))
  	}
  	e.textNodeEdits[n] = text
  }
  
  // commit applies changes to actions and template calls needed to contextually
  // autoescape content and adds any derived templates to the set.
  func (e *escaper) commit() {
  	for name := range e.output {
  		e.template(name).Funcs(funcMap)
  	}
  	// Any template from the name space associated with this escaper can be used
  	// to add derived templates to the underlying text/template name space.
  	tmpl := e.arbitraryTemplate()
  	for _, t := range e.derived {
  		if _, err := tmpl.text.AddParseTree(t.Name(), t.Tree); err != nil {
  			panic("error adding derived template")
  		}
  	}
  	for n, s := range e.actionNodeEdits {
  		ensurePipelineContains(n.Pipe, s)
  	}
  	for n, name := range e.templateNodeEdits {
  		n.Name = name
  	}
  	for n, s := range e.textNodeEdits {
  		n.Text = s
  	}
  	// Reset state that is specific to this commit so that the same changes are
  	// not re-applied to the template on subsequent calls to commit.
  	e.called = make(map[string]bool)
  	e.actionNodeEdits = make(map[*parse.ActionNode][]string)
  	e.templateNodeEdits = make(map[*parse.TemplateNode]string)
  	e.textNodeEdits = make(map[*parse.TextNode][]byte)
  }
  
  // template returns the named template given a mangled template name.
  func (e *escaper) template(name string) *template.Template {
  	// Any template from the name space associated with this escaper can be used
  	// to look up templates in the underlying text/template name space.
  	t := e.arbitraryTemplate().text.Lookup(name)
  	if t == nil {
  		t = e.derived[name]
  	}
  	return t
  }
  
  // arbitraryTemplate returns an arbitrary template from the name space
  // associated with e and panics if no templates are found.
  func (e *escaper) arbitraryTemplate() *Template {
  	for _, t := range e.ns.set {
  		return t
  	}
  	panic("no templates in name space")
  }
  
  // Forwarding functions so that clients need only import this package
  // to reach the general escaping functions of text/template.
  
  // HTMLEscape writes to w the escaped HTML equivalent of the plain text data b.
  func HTMLEscape(w io.Writer, b []byte) {
  	template.HTMLEscape(w, b)
  }
  
  // HTMLEscapeString returns the escaped HTML equivalent of the plain text data s.
  func HTMLEscapeString(s string) string {
  	return template.HTMLEscapeString(s)
  }
  
  // HTMLEscaper returns the escaped HTML equivalent of the textual
  // representation of its arguments.
  func HTMLEscaper(args ...interface{}) string {
  	return template.HTMLEscaper(args...)
  }
  
  // JSEscape writes to w the escaped JavaScript equivalent of the plain text data b.
  func JSEscape(w io.Writer, b []byte) {
  	template.JSEscape(w, b)
  }
  
  // JSEscapeString returns the escaped JavaScript equivalent of the plain text data s.
  func JSEscapeString(s string) string {
  	return template.JSEscapeString(s)
  }
  
  // JSEscaper returns the escaped JavaScript equivalent of the textual
  // representation of its arguments.
  func JSEscaper(args ...interface{}) string {
  	return template.JSEscaper(args...)
  }
  
  // URLQueryEscaper returns the escaped value of the textual representation of
  // its arguments in a form suitable for embedding in a URL query.
  func URLQueryEscaper(args ...interface{}) string {
  	return template.URLQueryEscaper(args...)
  }
  

View as plain text