refactor: render highlight language (#38793)

Avoid CSS injection

More details are in the comment of CodeBlockAttributes

modules/highlight/highlight.go

@@ -8,8 +8,10 @@ import (
 	"bytes"
 	gohtml "html"
 	"html/template"
+	"strings"
 	"sync"
 
+	"gitea.dev/modules/htmlutil"
 	"gitea.dev/modules/log"
 	"gitea.dev/modules/setting"
 	"gitea.dev/modules/util"
@@ -161,3 +163,51 @@ func formatLexerName(name string) string {
 	}
 	return util.ToTitleCaseNoLower(name)
 }
+
+func languageForCssAttrName(lang string) (forCSS, forAttr string) {
+	s := strings.ToLower(lang)
+	if s == "" || s == LanguagePlaintext || s == chromaLexerFallback {
+		return "text", "text"
+	}
+	isValid := func(c byte) bool {
+		// although "-" is valid in CSS name, it is used as a field separator, so we don't want to keep it in the name
+		return 'a' <= c && c <= 'z' || '0' <= c && c <= '9' || c == '_'
+	}
+	idx := 0
+	for ; idx < len(s); idx++ {
+		if !isValid(s[idx]) {
+			break
+		}
+	}
+	if idx == len(s) {
+		return s, lang
+	}
+	out := []byte(s)
+	for i := idx; i < len(s); i++ {
+		if !isValid(out[i]) {
+			out[i] = '_'
+		}
+	}
+	return string(out), lang
+}
+
+func CodeBlockAttributes(lang string) (preAttrs, codeAttrs template.HTML) {
+	// Code block's "chroma" class is used to highlight the code.
+	// "language-{LanguageName}" class is used as part of commonmark spec.
+	// It's unclear about how to handle special chars for a language name like "Visual Basic.NET" or "C++" or "F#".
+	// The commonmark spec seems wrong: https://spec.commonmark.org/0.31.2/#info-string, it just outputs invalid CSS class names.
+
+	cssName, attrLang := languageForCssAttrName(lang)
+	renderByFrontend := lang == "mermaid" || lang == "math"
+	preExtraClasses := ""
+	if renderByFrontend {
+		preExtraClasses = " is-loading"
+	}
+
+	// The "math.ts" strictly depends on the structure: <pre class="code-block"><code class="language-math">...</code></pre>
+	// * If "pre" exists, it is rendered as "block", otherwise, it is rendered as "inline"
+	// The "mermaid.ts" also strictly depends on the structure: "pre" must exist because it is always rendered as "block".
+	//
+	// Hint: "data-code-language" is not exposed in some cases due to the Markup sanitizer, the rules can be refactored in the future if the attribute is useful.
+	return htmlutil.HTMLFormat(`class="code-block%s"`, preExtraClasses), htmlutil.HTMLFormat(`class="chroma language-%s" data-code-language="%s"`, cssName, attrLang)
+}

modules/highlight/highlight_test.go

@@ -216,3 +216,19 @@ func TestUnsafeSplitHighlightedLines(t *testing.T) {
 	assert.Equal(t, "<span>a</span>\n", string(ret[0]))
 	assert.Equal(t, "<span>b\n</span>", string(ret[1]))
 }
+
+func TestCodeBlockAttributes(t *testing.T) {
+	test := func(t *testing.T, lang string, css, attr template.HTML) {
+		t.Helper()
+		cssActual, attrActual := CodeBlockAttributes(lang)
+		assert.Equal(t, css, cssActual)
+		assert.Equal(t, attr, attrActual)
+	}
+	for _, s := range []string{"", "FALLback", "plainTEXT"} {
+		test(t, s, `class="code-block"`, `class="chroma language-text" data-code-language="text"`)
+	}
+	test(t, "math", `class="code-block is-loading"`, `class="chroma language-math" data-code-language="math"`)
+	test(t, "mermaid", `class="code-block is-loading"`, `class="chroma language-mermaid" data-code-language="mermaid"`)
+	test(t, "Visual Basic.NET", `class="code-block"`, `class="chroma language-visual_basic_net" data-code-language="Visual Basic.NET"`)
+	test(t, "c++-x", `class="code-block"`, `class="chroma language-c___x" data-code-language="c++-x"`)
+}