Fix non-ASCII heading anchors not navigating from links

internal/markdown/convert.go

@@ -59,9 +59,10 @@ func Convert(source []byte, plantumlServer string) ([]byte, error) {
 	// Pre-process: replace svg, mermaid, plantuml code blocks with raw HTML
 	source = PreprocessCodeBlocks(source, plantumlServer)
 
-	// Parse into AST
+	// Parse into AST using GFM-compatible heading ID generation.
 	reader := text.NewReader(source)
-	doc := md.Parser().Parse(reader)
+	ctx := parser.NewContext(parser.WithIDs(newGFMIDs()))
+	doc := md.Parser().Parse(reader, parser.WithContext(ctx))
 
 	// Insert line anchors into AST
 	source = insertLineAnchors(doc, source)

internal/markdown/convert_test.go

@@ -5,6 +5,45 @@ import (
 	"testing"
 )
 
+func TestConvert_NonASCIIHeadingAnchor(t *testing.T) {
+	// Headings with non-ASCII (e.g. Japanese) characters must produce an id
+	// matching GitHub GFM behavior so that links like [text](#ヘディング) work.
+	tests := []struct {
+		input    string
+		wantID   string
+		wantLink string
+	}{
+		{
+			// goldmark percent-encodes non-ASCII in href; browsers decode before
+			// fragment matching, so "#%E3%83%98..." navigates to id="ヘディング".
+			input:    "# ヘディング\n\n[link](#ヘディング)\n",
+			wantID:   `id="ヘディング"`,
+			wantLink: `href="#%E3%83%98%E3%83%87%E3%82%A3%E3%83%B3%E3%82%B0"`,
+		},
+		{
+			input:  "# Hello World\n",
+			wantID: `id="hello-world"`,
+		},
+		{
+			input:  "# 見出し Test\n",
+			wantID: `id="見出し-test"`,
+		},
+	}
+	for _, tt := range tests {
+		html, err := Convert([]byte(tt.input), "")
+		if err != nil {
+			t.Fatalf("Convert error: %v", err)
+		}
+		s := string(html)
+		if !strings.Contains(s, tt.wantID) {
+			t.Errorf("expected %q in output:\n%s", tt.wantID, s)
+		}
+		if tt.wantLink != "" && !strings.Contains(s, tt.wantLink) {
+			t.Errorf("expected %q in output:\n%s", tt.wantLink, s)
+		}
+	}
+}
+
 func TestConvert_CRLFMathBlock(t *testing.T) {
 	// Regression: CRLF line endings must not break $$ math block rendering.
 	input := []byte("$$\r\ny = x^2\r\n$$\r\n")

internal/markdown/headingid.go

@@ -0,0 +1,60 @@
+package markdown
+
+import (
+	"bytes"
+	"fmt"
+	"unicode"
+
+	"github.com/yuin/goldmark/ast"
+	"github.com/yuin/goldmark/parser"
+)
+
+// newGFMIDs returns a parser.IDs that generates heading IDs following the
+// GitHub Flavored Markdown convention: Unicode letters and digits are kept
+// (lowercased where applicable), spaces/hyphens become "-", others are
+// dropped. This preserves non-ASCII characters such as Japanese kana/kanji,
+// unlike goldmark's default which strips all multi-byte characters.
+func newGFMIDs() parser.IDs {
+	return &gfmIDs{values: map[string]bool{}}
+}
+
+type gfmIDs struct {
+	values map[string]bool
+}
+
+func (s *gfmIDs) Generate(value []byte, kind ast.NodeKind) []byte {
+	value = bytes.TrimSpace(value)
+	var result []rune
+	for _, r := range string(value) {
+		switch {
+		case unicode.IsLetter(r) || unicode.IsDigit(r):
+			result = append(result, unicode.ToLower(r))
+		case unicode.IsSpace(r) || r == '-' || r == '_':
+			result = append(result, '-')
+		}
+	}
+	res := []byte(string(result))
+	if len(res) == 0 {
+		if kind == ast.KindHeading {
+			res = []byte("heading")
+		} else {
+			res = []byte("id")
+		}
+	}
+	key := string(res)
+	if !s.values[key] {
+		s.values[key] = true
+		return res
+	}
+	for i := 1; ; i++ {
+		newKey := fmt.Sprintf("%s-%d", key, i)
+		if !s.values[newKey] {
+			s.values[newKey] = true
+			return []byte(newKey)
+		}
+	}
+}
+
+func (s *gfmIDs) Put(value []byte) {
+	s.values[string(value)] = true
+}