| 1 | package markdown |
| 2 | |
| 3 | import ( |
| 4 | "html" |
| 5 | "net/url" |
| 6 | "path" |
| 7 | "strings" |
| 8 | ) |
| 9 | |
| 10 | // Afbeeldingen in Markdown. Een relatief pad wijst naar het bestand in de repo |
| 11 | // en wordt een /-/raw/-adres op dezelfde ref. Een .svg krijgt as=img, zodat de |
| 12 | // server hem als image/svg+xml in een sandbox levert; een <img> voert nooit |
| 13 | // script uit. Ruwe HTML blijft geëscaped, behalve een kleine, herschreven set |
| 14 | // tags (img, a, br, en omhullende p/div/center/picture/span zonder attributen). |
| 15 | |
| 16 | var imageExts = map[string]bool{".png": true, ".jpg": true, ".jpeg": true, ".gif": true, ".webp": true, ".svg": true, ".avif": true, ".bmp": true, ".ico": true} |
| 17 | |
| 18 | // splitBlobBase splitst /owner/repo/-/blob/dir/ in /owner/repo/-/blob/ en dir/. |
| 19 | func splitBlobBase(linkBase string) (root, dir string, ok bool) { |
| 20 | const mark = "/-/blob/" |
| 21 | i := strings.Index(linkBase, mark) |
| 22 | if i < 0 { |
| 23 | return "", "", false |
| 24 | } |
| 25 | return linkBase[:i+len(mark)], linkBase[i+len(mark):], true |
| 26 | } |
| 27 | |
| 28 | // cleanRepoPath maakt van href een pad vanaf de repo-root, met dir als map van |
| 29 | // het document. Een leidende / is de repo-root; .. kan nooit boven de root uit. |
| 30 | // Een #anker blijft achteraan staan. |
| 31 | func cleanRepoPath(dir, href string) string { |
| 32 | p, _ := cleanRepoPathFrag(dir, href) |
| 33 | return p |
| 34 | } |
| 35 | |
| 36 | func cleanRepoPathFrag(dir, href string) (string, string) { |
| 37 | frag := "" |
| 38 | if j := strings.IndexByte(href, '#'); j >= 0 { |
| 39 | href, frag = href[:j], href[j:] |
| 40 | } |
| 41 | if j := strings.IndexByte(href, '?'); j >= 0 { |
| 42 | href = href[:j] |
| 43 | } |
| 44 | if un, err := url.PathUnescape(href); err == nil { |
| 45 | href = un |
| 46 | } |
| 47 | var p string |
| 48 | if strings.HasPrefix(href, "/") { |
| 49 | p = path.Clean(href) |
| 50 | } else { |
| 51 | p = path.Clean("/" + dir + "/" + href) |
| 52 | } |
| 53 | p = strings.TrimPrefix(p, "/") |
| 54 | segs := strings.Split(p, "/") |
| 55 | for k := range segs { |
| 56 | segs[k] = url.PathEscape(segs[k]) |
| 57 | } |
| 58 | return strings.Join(segs, "/"), frag |
| 59 | } |
| 60 | |
| 61 | // rawBase geeft het /-/raw/-voorvoegsel en de map voor relatieve afbeeldingen. |
| 62 | func rawBase(linkBase, repoPath string) (root, dir string, ok bool) { |
| 63 | if r, d, ok := splitBlobBase(linkBase); ok { |
| 64 | return strings.TrimSuffix(r, "/-/blob/") + "/-/raw/", d, true |
| 65 | } |
| 66 | if repoPath != "" { |
| 67 | return "/" + strings.Trim(repoPath, "/") + "/-/raw/", "", true |
| 68 | } |
| 69 | return "", "", false |
| 70 | } |
| 71 | |
| 72 | // imageSrc maakt een veilig src-adres, of "" als het niet mag. |
| 73 | func imageSrc(src, linkBase, rev, repoPath string) string { |
| 74 | src = strings.TrimSpace(src) |
| 75 | src = strings.TrimSuffix(strings.TrimPrefix(src, "<"), ">") |
| 76 | src = strings.ReplaceAll(src, " ", "%20") |
| 77 | if src == "" || strings.ContainsAny(src, "\"'<>\\`\x00\n\r\t ") { |
| 78 | return "" |
| 79 | } |
| 80 | lower := strings.ToLower(src) |
| 81 | switch { |
| 82 | case strings.HasPrefix(lower, "https://"): |
| 83 | return src |
| 84 | case strings.HasPrefix(lower, "data:image/") && !strings.HasPrefix(lower, "data:image/svg"): |
| 85 | return src |
| 86 | case strings.HasPrefix(lower, "//"), hasScheme(lower): |
| 87 | // http:, javascript:, data:text/..., schema-relatief: niet laden. |
| 88 | return "" |
| 89 | } |
| 90 | if strings.HasPrefix(src, "#") { |
| 91 | return "" |
| 92 | } |
| 93 | root, dir, ok := rawBase(linkBase, repoPath) |
| 94 | if !ok { |
| 95 | // Geen repo (bv. /help): alleen relatief aan linkBase. |
| 96 | if linkBase == "" { |
| 97 | return "" |
| 98 | } |
| 99 | return linkBase + strings.TrimPrefix(src, "./") |
| 100 | } |
| 101 | rel := cleanRepoPath(dir, src) |
| 102 | if rel == "" || rel == "." { |
| 103 | return "" |
| 104 | } |
| 105 | q := url.Values{} |
| 106 | if rev != "" { |
| 107 | q.Set("rev", rev) |
| 108 | } |
| 109 | if strings.EqualFold(path.Ext(rel), ".svg") { |
| 110 | q.Set("as", "img") |
| 111 | } |
| 112 | out := root + rel |
| 113 | if len(q) > 0 { |
| 114 | out += "?" + encodeOrdered(q) |
| 115 | } |
| 116 | return out |
| 117 | } |
| 118 | |
| 119 | // encodeOrdered zet rev vóór as, zodat adressen voorspelbaar zijn. |
| 120 | func encodeOrdered(q url.Values) string { |
| 121 | var parts []string |
| 122 | for _, k := range []string{"rev", "as"} { |
| 123 | if v := q.Get(k); v != "" { |
| 124 | parts = append(parts, k+"="+url.QueryEscape(v)) |
| 125 | } |
| 126 | } |
| 127 | return strings.Join(parts, "&") |
| 128 | } |
| 129 | |
| 130 | // imageTag schrijft een <img>; zonder geldige bron alleen de alt-tekst. |
| 131 | func imageTag(src, alt, title, width, height, linkBase, rev, repoPath string) string { |
| 132 | u := imageSrc(src, linkBase, rev, repoPath) |
| 133 | if u == "" { |
| 134 | return html.EscapeString(alt) |
| 135 | } |
| 136 | var b strings.Builder |
| 137 | b.WriteString(`<img class="mdimg" src="`) |
| 138 | b.WriteString(html.EscapeString(u)) |
| 139 | b.WriteString(`" alt="`) |
| 140 | b.WriteString(html.EscapeString(alt)) |
| 141 | b.WriteString(`"`) |
| 142 | if title != "" { |
| 143 | b.WriteString(` title="`) |
| 144 | b.WriteString(html.EscapeString(title)) |
| 145 | b.WriteString(`"`) |
| 146 | } |
| 147 | if w := dimension(width); w != "" { |
| 148 | b.WriteString(` width="` + w + `"`) |
| 149 | } |
| 150 | if h := dimension(height); h != "" { |
| 151 | b.WriteString(` height="` + h + `"`) |
| 152 | } |
| 153 | b.WriteString(` loading="lazy" decoding="async" referrerpolicy="no-referrer">`) |
| 154 | return b.String() |
| 155 | } |
| 156 | |
| 157 | // dimension laat alleen een getal (pixels) door, maximaal vier cijfers. |
| 158 | func dimension(v string) string { |
| 159 | v = strings.TrimSuffix(strings.TrimSpace(v), "px") |
| 160 | if v == "" || len(v) > 4 { |
| 161 | return "" |
| 162 | } |
| 163 | for _, c := range v { |
| 164 | if c < '0' || c > '9' { |
| 165 | return "" |
| 166 | } |
| 167 | } |
| 168 | return v |
| 169 | } |
| 170 | |
| 171 | // bracketLink leest [tekst](doel) vanaf s[i] == '[' met geneste haken. |
| 172 | func bracketLink(s string, i int) (text, dest string, n int, ok bool) { |
| 173 | if i >= len(s) || s[i] != '[' { |
| 174 | return "", "", 0, false |
| 175 | } |
| 176 | depth := 0 |
| 177 | j := i |
| 178 | for ; j < len(s); j++ { |
| 179 | switch s[j] { |
| 180 | case '\\': |
| 181 | j++ |
| 182 | case '[': |
| 183 | depth++ |
| 184 | case ']': |
| 185 | depth-- |
| 186 | } |
| 187 | if depth == 0 { |
| 188 | break |
| 189 | } |
| 190 | } |
| 191 | if j >= len(s) || j+1 >= len(s) || s[j+1] != '(' { |
| 192 | return "", "", 0, false |
| 193 | } |
| 194 | text = s[i+1 : j] |
| 195 | k := j + 2 |
| 196 | pd := 1 |
| 197 | for ; k < len(s); k++ { |
| 198 | if s[k] == '(' { |
| 199 | pd++ |
| 200 | } else if s[k] == ')' { |
| 201 | pd-- |
| 202 | if pd == 0 { |
| 203 | break |
| 204 | } |
| 205 | } |
| 206 | } |
| 207 | if k >= len(s) { |
| 208 | return "", "", 0, false |
| 209 | } |
| 210 | dest = strings.TrimSpace(s[j+2 : k]) |
| 211 | if dest == "" { |
| 212 | return "", "", 0, false |
| 213 | } |
| 214 | return text, dest, k + 1 - i, true |
| 215 | } |
| 216 | |
| 217 | // splitTitle haalt een "titel" achter het doel weg. |
| 218 | func splitTitle(dest string) (href, title string) { |
| 219 | dest = strings.TrimSpace(dest) |
| 220 | if sp := strings.IndexAny(dest, " \t"); sp > 0 { |
| 221 | rest := strings.TrimSpace(dest[sp:]) |
| 222 | if len(rest) >= 2 && (rest[0] == '"' && rest[len(rest)-1] == '"' || rest[0] == '\'' && rest[len(rest)-1] == '\'') { |
| 223 | return dest[:sp], rest[1 : len(rest)-1] |
| 224 | } |
| 225 | return dest[:sp], "" |
| 226 | } |
| 227 | return dest, "" |
| 228 | } |
| 229 | |
| 230 | // plainText haalt eenvoudige Markdown-tekens uit een alt-tekst. |
| 231 | func plainText(s string) string { |
| 232 | return strings.NewReplacer("**", "", "*", "", "`", "").Replace(s) |
| 233 | } |
| 234 | |
| 235 | // bareImage maakt van een los https-adres naar een afbeelding een <img> met link. |
| 236 | func bareImage(s, linkBase, rev, repoPath string) (string, int, bool) { |
| 237 | if !strings.HasPrefix(strings.ToLower(s), "https://") { |
| 238 | return "", 0, false |
| 239 | } |
| 240 | end := strings.IndexAny(s, " \t<>\"'`") |
| 241 | if end < 0 { |
| 242 | end = len(s) |
| 243 | } |
| 244 | u := strings.TrimRight(s[:end], ".,;:!?)") |
| 245 | p := u |
| 246 | if q := strings.IndexAny(p, "?#"); q >= 0 { |
| 247 | p = p[:q] |
| 248 | } |
| 249 | if !imageExts[strings.ToLower(path.Ext(p))] { |
| 250 | return "", 0, false |
| 251 | } |
| 252 | parsed, err := url.Parse(u) |
| 253 | if err != nil || parsed.Host == "" { |
| 254 | return "", 0, false |
| 255 | } |
| 256 | img := imageTag(u, path.Base(parsed.Path), "", "", "", linkBase, rev, repoPath) |
| 257 | return `<a href="` + html.EscapeString(u) + `" rel="noopener noreferrer">` + img + `</a>`, len(u), true |
| 258 | } |
| 259 | |
| 260 | // htmlTag herschrijft een toegestane HTML-tag. Al het andere blijft tekst. |
| 261 | func htmlTag(s, linkBase, rev, repoPath string, anchors *int) (string, int, bool) { |
| 262 | end := strings.IndexByte(s, '>') |
| 263 | if end < 0 || end > 4000 { |
| 264 | return "", 0, false |
| 265 | } |
| 266 | raw := s[1:end] |
| 267 | closing := strings.HasPrefix(raw, "/") |
| 268 | raw = strings.TrimPrefix(raw, "/") |
| 269 | raw = strings.TrimSuffix(strings.TrimSpace(raw), "/") |
| 270 | name := raw |
| 271 | if sp := strings.IndexAny(raw, " \t\n"); sp >= 0 { |
| 272 | name = raw[:sp] |
| 273 | } |
| 274 | name = strings.ToLower(name) |
| 275 | attrs, okAttrs := parseAttrs(raw[len(name):]) |
| 276 | if !okAttrs { |
| 277 | return "", 0, false |
| 278 | } |
| 279 | n := end + 1 |
| 280 | switch name { |
| 281 | case "img": |
| 282 | if closing { |
| 283 | return "", n, true |
| 284 | } |
| 285 | return imageTag(attrs["src"], attrs["alt"], attrs["title"], attrs["width"], attrs["height"], linkBase, rev, repoPath), n, true |
| 286 | case "br": |
| 287 | return "<br>", n, true |
| 288 | case "a": |
| 289 | if closing { |
| 290 | if *anchors == 0 { |
| 291 | return "", n, true |
| 292 | } |
| 293 | *anchors-- |
| 294 | return "</a>", n, true |
| 295 | } |
| 296 | href, ok := attrs["href"] |
| 297 | if !ok { |
| 298 | return "", 0, false |
| 299 | } |
| 300 | h := strings.TrimSpace(href) |
| 301 | lower := strings.ToLower(h) |
| 302 | if strings.Contains(lower, ":") && !strings.HasPrefix(lower, "https://") && !strings.HasPrefix(lower, "http://") && !strings.HasPrefix(lower, "mailto:") { |
| 303 | return "", 0, false |
| 304 | } |
| 305 | *anchors++ |
| 306 | return `<a href="` + html.EscapeString(resolveLink(h, linkBase, rev)) + `" rel="noopener noreferrer">`, n, true |
| 307 | case "p", "div", "center", "picture", "span", "source", "sup", "sub": |
| 308 | // Omhulsels (vaak <p align="center">) vallen weg; de inhoud blijft. |
| 309 | return "", n, true |
| 310 | } |
| 311 | return "", 0, false |
| 312 | } |
| 313 | |
| 314 | // parseAttrs leest name="waarde"-paren. Ongeldige syntaxis geeft ok=false. |
| 315 | func parseAttrs(s string) (map[string]string, bool) { |
| 316 | out := map[string]string{} |
| 317 | i := 0 |
| 318 | for { |
| 319 | for i < len(s) && (s[i] == ' ' || s[i] == '\t' || s[i] == '\n') { |
| 320 | i++ |
| 321 | } |
| 322 | if i >= len(s) { |
| 323 | return out, true |
| 324 | } |
| 325 | j := i |
| 326 | for j < len(s) && (isIdent(s[j]) || s[j] == '-' || s[j] == ':') { |
| 327 | j++ |
| 328 | } |
| 329 | if j == i { |
| 330 | return nil, false |
| 331 | } |
| 332 | key := strings.ToLower(s[i:j]) |
| 333 | i = j |
| 334 | if i < len(s) && s[i] == '=' { |
| 335 | i++ |
| 336 | if i < len(s) && (s[i] == '"' || s[i] == '\'') { |
| 337 | q := s[i] |
| 338 | k := strings.IndexByte(s[i+1:], q) |
| 339 | if k < 0 { |
| 340 | return nil, false |
| 341 | } |
| 342 | out[key] = html.UnescapeString(s[i+1 : i+1+k]) |
| 343 | i += k + 2 |
| 344 | } else { |
| 345 | k := i |
| 346 | for k < len(s) && s[k] != ' ' && s[k] != '\t' && s[k] != '\n' { |
| 347 | k++ |
| 348 | } |
| 349 | out[key] = html.UnescapeString(s[i:k]) |
| 350 | i = k |
| 351 | } |
| 352 | } else { |
| 353 | out[key] = "" |
| 354 | } |
| 355 | } |
| 356 | } |
| 357 | |
| 358 | // hasScheme ziet een "schema:" vóór de eerste /, ? of #. |
| 359 | func hasScheme(s string) bool { |
| 360 | c := strings.IndexByte(s, ':') |
| 361 | if c < 0 { |
| 362 | return false |
| 363 | } |
| 364 | if k := strings.IndexAny(s, "/?#"); k >= 0 && k < c { |
| 365 | return false |
| 366 | } |
| 367 | return true |
| 368 | } |