| 1 | diff --git a/internal/webui/markdown.go b/internal/webui/markdown.go |
| 2 | new file mode 100644 |
| 3 | index 0000000..a4dd7bf |
| 4 | --- /dev/null |
| 5 | +++ b/internal/webui/markdown.go |
| 6 | @@ -0,0 +1,98 @@ |
| 7 | +package webui |
| 8 | + |
| 9 | +import ( |
| 10 | + "html" |
| 11 | + "html/template" |
| 12 | + "regexp" |
| 13 | + "strings" |
| 14 | +) |
| 15 | + |
| 16 | +// renderMarkdown turns a small, common subset of Markdown into HTML: ATX |
| 17 | +// headings, paragraphs, fenced code, unordered and ordered lists, inline |
| 18 | +// code, emphasis, and links. Every piece of text is escaped first and only |
| 19 | +// the tags this function emits reach the page, so untrusted README files and |
| 20 | +// relay descriptions cannot inject markup. Anything else stays plain text. |
| 21 | +func renderMarkdown(source string) template.HTML { |
| 22 | + lines := strings.Split(strings.ReplaceAll(source, "\r\n", "\n"), "\n") |
| 23 | + var out strings.Builder |
| 24 | + var paragraph []string |
| 25 | + flush := func() { |
| 26 | + if len(paragraph) == 0 { |
| 27 | + return |
| 28 | + } |
| 29 | + out.WriteString("<p>" + inlineMarkdown(strings.Join(paragraph, " ")) + "</p>\n") |
| 30 | + paragraph = paragraph[:0] |
| 31 | + } |
| 32 | + for i := 0; i < len(lines); i++ { |
| 33 | + line := lines[i] |
| 34 | + switch { |
| 35 | + case strings.HasPrefix(line, "```"): |
| 36 | + flush() |
| 37 | + var code []string |
| 38 | + for i++; i < len(lines) && !strings.HasPrefix(lines[i], "```"); i++ { |
| 39 | + code = append(code, lines[i]) |
| 40 | + } |
| 41 | + out.WriteString("<pre><code>" + html.EscapeString(strings.Join(code, "\n")) + "</code></pre>\n") |
| 42 | + case strings.HasPrefix(line, "#"): |
| 43 | + flush() |
| 44 | + level := len(line) - len(strings.TrimLeft(line, "#")) |
| 45 | + if level > 6 { |
| 46 | + level = 6 |
| 47 | + } |
| 48 | + text := strings.TrimSpace(line[level:]) |
| 49 | + tag := "h" + string(rune('0'+level)) |
| 50 | + out.WriteString("<" + tag + ">" + inlineMarkdown(text) + "</" + tag + ">\n") |
| 51 | + case isListItem(line): |
| 52 | + flush() |
| 53 | + ordered := orderedItem.MatchString(line) |
| 54 | + tag := "ul" |
| 55 | + if ordered { |
| 56 | + tag = "ol" |
| 57 | + } |
| 58 | + out.WriteString("<" + tag + ">") |
| 59 | + for ; i < len(lines) && isListItem(lines[i]); i++ { |
| 60 | + out.WriteString("<li>" + inlineMarkdown(listText(lines[i])) + "</li>") |
| 61 | + } |
| 62 | + i-- |
| 63 | + out.WriteString("</" + tag + ">\n") |
| 64 | + case strings.TrimSpace(line) == "": |
| 65 | + flush() |
| 66 | + default: |
| 67 | + paragraph = append(paragraph, strings.TrimSpace(line)) |
| 68 | + } |
| 69 | + } |
| 70 | + flush() |
| 71 | + return template.HTML(out.String()) |
| 72 | +} |
| 73 | + |
| 74 | +var ( |
| 75 | + orderedItem = regexp.MustCompile(`^\s*\d+[.)]\s+`) |
| 76 | + bulletItem = regexp.MustCompile(`^\s*[-*+]\s+`) |
| 77 | + codeSpan = regexp.MustCompile("`([^`]+)`") |
| 78 | + linkSpan = regexp.MustCompile(`\[([^\]]+)\]\(((?:https?://|/|\./|#|mailto:)[^)\s]*)\)`) |
| 79 | + strongSpan = regexp.MustCompile(`\*\*([^*]+)\*\*`) |
| 80 | + emSpan = regexp.MustCompile(`(^|[^*\w])\*([^*]+)\*`) |
| 81 | +) |
| 82 | + |
| 83 | +func isListItem(line string) bool { |
| 84 | + return bulletItem.MatchString(line) || orderedItem.MatchString(line) |
| 85 | +} |
| 86 | + |
| 87 | +func listText(line string) string { |
| 88 | + if loc := bulletItem.FindStringIndex(line); loc != nil { |
| 89 | + return line[loc[1]:] |
| 90 | + } |
| 91 | + return line[orderedItem.FindStringIndex(line)[1]:] |
| 92 | +} |
| 93 | + |
| 94 | +// inlineMarkdown escapes text, then restores the few inline forms it |
| 95 | +// recognises. Link targets are restricted to http, https, relative paths, |
| 96 | +// fragments and mailto so scripts and data URLs stay text. |
| 97 | +func inlineMarkdown(text string) string { |
| 98 | + escaped := html.EscapeString(text) |
| 99 | + escaped = codeSpan.ReplaceAllString(escaped, "<code>$1</code>") |
| 100 | + escaped = linkSpan.ReplaceAllString(escaped, `<a href="$2">$1</a>`) |
| 101 | + escaped = strongSpan.ReplaceAllString(escaped, "<strong>$1</strong>") |
| 102 | + escaped = emSpan.ReplaceAllString(escaped, "$1<em>$2</em>") |
| 103 | + return escaped |
| 104 | +} |
| 105 | diff --git a/internal/webui/markdown_test.go b/internal/webui/markdown_test.go |
| 106 | new file mode 100644 |
| 107 | index 0000000..860ccd7 |
| 108 | --- /dev/null |
| 109 | +++ b/internal/webui/markdown_test.go |
| 110 | @@ -0,0 +1,18 @@ |
| 111 | +package webui |
| 112 | + |
| 113 | +import ( |
| 114 | + "strings" |
| 115 | + "testing" |
| 116 | +) |
| 117 | + |
| 118 | +func TestMarkdownRendersCommonFormsAndEscapesTheRest(t *testing.T) { |
| 119 | + got := string(renderMarkdown("# Title\n\nA `code` span with **bold** and *em* and [a link](https://example.com).\n\n```sh\ngo build <x>\n```\n\n- one\n- two\n\n1. first\n2. second\n\n<script>alert(1)</script> and [bad](javascript:alert(1))")) |
| 120 | + for _, want := range []string{"<h1>Title</h1>", "<code>code</code>", "<strong>bold</strong>", "<em>em</em>", `<a href="https://example.com">a link</a>`, "<pre><code>go build <x></code></pre>", "<ul><li>one</li><li>two</li></ul>", "<ol><li>first</li><li>second</li></ol>", "<script>alert(1)</script>"} { |
| 121 | + if !strings.Contains(got, want) { |
| 122 | + t.Errorf("missing %q in %s", want, got) |
| 123 | + } |
| 124 | + } |
| 125 | + if strings.Contains(got, "<script>") || strings.Contains(got, `href="javascript:`) { |
| 126 | + t.Fatalf("unsafe markup leaked: %s", got) |
| 127 | + } |
| 128 | +} |