Implement a custom parser for <ac:*/> tags

This replaces the workaround to replace colons in <ac:*/> tags with a
magic string with a custom parser for these tags to parse them as
ast.KindRawHtml.

The custom parser is a stripped down version of goldmark's rawHTMLParser.

pkg/mark/ac_tag_parser.go

@@ -0,0 +1,112 @@
+package mark
+
+import (
+	"bytes"
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/text"
+	"github.com/yuin/goldmark/util"
+	"regexp"
+)
+
+// NewACTagParser returns an inline parser that parses <ac:* /> tags to ensure that Confluence specific tags are parsed
+// as ast.KindRawHtml so they are not escaped at render time. The parser must be registered with a higher priority
+// than goldmark's linkParser. Otherwise, the linkParser would parse the <ac:* /> tags.
+func NewACTagParser() parser.InlineParser {
+	return &acTagParser{}
+}
+
+var _ parser.InlineParser = (*acTagParser)(nil)
+
+// acTagParser is a stripped down version of goldmark's rawHTMLParser.
+// See: https://github.com/yuin/goldmark/blob/master/parser/raw_html.go
+type acTagParser struct {
+}
+
+func (s *acTagParser) Trigger() []byte {
+	return []byte{'<'}
+}
+
+func (s *acTagParser) Parse(_ ast.Node, block text.Reader, pc parser.Context) ast.Node {
+	line, _ := block.PeekLine()
+	if len(line) > 1 && util.IsAlphaNumeric(line[1]) {
+		return s.parseMultiLineRegexp(openTagRegexp, block, pc)
+	}
+	if len(line) > 2 && line[1] == '/' && util.IsAlphaNumeric(line[2]) {
+		return s.parseMultiLineRegexp(closeTagRegexp, block, pc)
+	}
+	if len(line) > 2 && line[1] == '!' && line[2] >= 'A' && line[2] <= 'Z' {
+		return s.parseUntil(block, closeDecl, pc)
+	}
+	if bytes.HasPrefix(line, openCDATA) {
+		return s.parseUntil(block, closeCDATA, pc)
+	}
+	return nil
+}
+
+var tagnamePattern = `([A-Za-z][A-Za-z0-9-]*)`
+
+var attributePattern = `(?:[\r\n \t]+[a-zA-Z_:][a-zA-Z0-9:._-]*(?:[\r\n \t]*=[\r\n \t]*(?:[^\"'=<>` + "`" + `\x00-\x20]+|'[^']*'|"[^"]*"))?)`
+
+// Only match <ac:*/> tags
+var openTagRegexp = regexp.MustCompile("^<ac:" + tagnamePattern + attributePattern + `*[ \t]*/?>`)
+var closeTagRegexp = regexp.MustCompile("^</ac:" + tagnamePattern + `\s*>`)
+
+var openCDATA = []byte("<![CDATA[")
+var closeCDATA = []byte("]]>")
+var closeDecl = []byte(">")
+
+func (s *acTagParser) parseUntil(block text.Reader, closer []byte, _ parser.Context) ast.Node {
+	savedLine, savedSegment := block.Position()
+	node := ast.NewRawHTML()
+	for {
+		line, segment := block.PeekLine()
+		if line == nil {
+			break
+		}
+		index := bytes.Index(line, closer)
+		if index > -1 {
+			node.Segments.Append(segment.WithStop(segment.Start + index + len(closer)))
+			block.Advance(index + len(closer))
+			return node
+		}
+		node.Segments.Append(segment)
+		block.AdvanceLine()
+	}
+	block.SetPosition(savedLine, savedSegment)
+	return nil
+}
+
+func (s *acTagParser) parseMultiLineRegexp(reg *regexp.Regexp, block text.Reader, _ parser.Context) ast.Node {
+	sline, ssegment := block.Position()
+	if block.Match(reg) {
+		node := ast.NewRawHTML()
+		eline, esegment := block.Position()
+		block.SetPosition(sline, ssegment)
+		for {
+			line, segment := block.PeekLine()
+			if line == nil {
+				break
+			}
+			l, _ := block.Position()
+			start := segment.Start
+			if l == sline {
+				start = ssegment.Start
+			}
+			end := segment.Stop
+			if l == eline {
+				end = esegment.Start
+			}
+
+			node.Segments.Append(text.NewSegment(start, end))
+			if l == eline {
+				block.Advance(end - start)
+				break
+			} else {
+				block.AdvanceLine()
+			}
+		}
+		return node
+	}
+	return nil
+}

pkg/mark/markdown.go

@@ -430,22 +430,9 @@ func (r *ConfluenceRenderer) renderCodeBlock(writer util.BufWriter, source []byt
 	return ast.WalkContinue, nil
 }
 
-// compileMarkdown will replace tags like <ac:rich-tech-body> with escaped
-// equivalent, because goldmark markdown parser replaces that tags with
-// <a href="ac:rich-text-body">ac:rich-text-body</a> because of the autolink
-// rule.
 func CompileMarkdown(markdown []byte, stdlib *stdlib.Lib) string {
 	log.Tracef(nil, "rendering markdown:\n%s", string(markdown))
 
-	colon := []byte("---bf-COLON---")
-
-	tags := regexp.MustCompile(`</?ac:[^>]+>`)
-
-	for _, match := range tags.FindAll(markdown, -1) {
-		// Replace the colon in all "<ac:*>" tags with the colon bytes to avoid having Goldmark escape the HTML output.
-		markdown = bytes.ReplaceAll(markdown, match, bytes.ReplaceAll(match, []byte(":"), colon))
-	}
-
 	converter := goldmark.New(
 		goldmark.WithExtensions(
 			extension.GFM,
@@ -461,6 +448,12 @@ func CompileMarkdown(markdown []byte, stdlib *stdlib.Lib) string {
 			html.WithUnsafe(),
 		))
 
+	converter.Parser().AddOptions(parser.WithInlineParsers(
+		// Must be registered with a higher priority than goldmark's linkParser to make sure goldmark doesn't parse
+		// the <ac:*/> tags.
+		util.Prioritized(NewACTagParser(), 199),
+	))
+
 	converter.Renderer().AddOptions(renderer.WithNodeRenderers(
 		util.Prioritized(NewConfluenceRenderer(stdlib), 100),
 	))
@@ -472,8 +465,7 @@ func CompileMarkdown(markdown []byte, stdlib *stdlib.Lib) string {
 		panic(err)
 	}
 
-	// Restore all the colons we previously replaced.
-	html := bytes.ReplaceAll(buf.Bytes(), colon, []byte(":"))
+	html := buf.Bytes()
 
 	log.Tracef(nil, "rendered markdown to html:\n%s", string(html))
 

pkg/mark/testdata/macro-include.html

@@ -1,6 +1,6 @@
 <p><foo>bar</foo></p>
-<ac:structured-macro ac:name="info">
+<p><ac:structured-macro ac:name="info">
 <ac:parameter ac:name="icon">true</ac:parameter>
 <ac:parameter ac:name="title">Attention</ac:parameter>
 <ac:rich-text-body>This is an info!</ac:rich-text-body>
-</ac:structured-macro>
+</ac:structured-macro></p>