| 1 | /** |
| 2 | * A small, safe Markdown renderer for READMEs, issues, pull requests and job |
| 3 | * summaries. All raw HTML in the source is escaped, except HTML comments, which |
| 4 | * are dropped as GitHub drops them; only http(s), mailto-free relative and |
| 5 | * fragment URLs survive in links and images. |
| 6 | * |
| 7 | * Supported: ATX headings, paragraphs, hard breaks, fenced code, thematic |
| 8 | * breaks, blockquotes, ordered/unordered/task lists (nested), tables, inline |
| 9 | * code, bold, italic, strikethrough, links, images and bare/auto links. |
| 10 | */ |
| 11 | |
| 12 | export interface MarkdownBase { |
| 13 | /** Repository owner and name, for resolving relative links. */ |
| 14 | owner: string; |
| 15 | repo: string; |
| 16 | /** Ref (branch, tag or sha) that relative links point at. */ |
| 17 | ref: string; |
| 18 | /** Directory of the document inside the repository ("" for the root). */ |
| 19 | dir: string; |
| 20 | } |
| 21 | |
| 22 | export interface MarkdownOptions { |
| 23 | base?: MarkdownBase; |
| 24 | /** Syntax highlighter for fenced code: HTML for the code, or null for plain text. */ |
| 25 | highlight?: (code: string, lang: string) => string | null; |
| 26 | /** Prefix for heading ids and fragment links (default "h-", which keeps user content clear of the page's own ids). */ |
| 27 | idPrefix?: string; |
| 28 | /** |
| 29 | * Where an image's src (absolute, or relative to this site) should point; null shows its alt |
| 30 | * text instead. The web pages send images from other sites through Diffinery's image proxy |
| 31 | * this way (lib/camo.ts). Links keep their original targets. |
| 32 | */ |
| 33 | image?: (src: string) => string | null; |
| 34 | } |
| 35 | |
| 36 | const MAX_SOURCE = 512 * 1024; |
| 37 | const MAX_DEPTH = 12; |
| 38 | const MAX_INLINE = 60_000; |
| 39 | const MAX_SCAN = 2000; |
| 40 | /** |
| 41 | * Characters that link, code and emphasis lookahead may scan in one render: 4 per source |
| 42 | * character, where real documents scan well under 1. Each unmatched `[`, `*` or `_` scans up |
| 43 | * to MAX_SCAN ahead, so a crafted paragraph would cost MAX_SCAN times its length; past the |
| 44 | * budget the rest of the inline text is shown escaped, without formatting. |
| 45 | */ |
| 46 | const scanBudget = (sourceLength: number): number => 250_000 + 4 * sourceLength; |
| 47 | // Rendering is synchronous, so one counter per isolate is safe. |
| 48 | let scanned = 0; |
| 49 | let budget = 0; |
| 50 | |
| 51 | const ESC: Record<string, string> = { "&": "&", "<": "<", ">": ">", '"': """, "'": "'" }; |
| 52 | export const escapeHtml = (s: string): string => s.replace(/[&<>"']/g, (c) => ESC[c]!); |
| 53 | |
| 54 | const encPath = (p: string): string => p.split("/").map(encodeURIComponent).join("/"); |
| 55 | |
| 56 | export function slugify(text: string): string { |
| 57 | return text |
| 58 | .toLowerCase() |
| 59 | .replace(/<[^<>]*>/g, "") |
| 60 | .replace(/[^\p{L}\p{N}\s_-]/gu, "") |
| 61 | .trim() |
| 62 | .replace(/\s+/g, "-") |
| 63 | .slice(0, 80); |
| 64 | } |
| 65 | |
| 66 | /** Normalise "a/./b/../c" style paths; returns null when the path escapes the root. */ |
| 67 | function normalise(path: string): string | null { |
| 68 | const out: string[] = []; |
| 69 | for (const seg of path.split("/")) { |
| 70 | if (!seg || seg === ".") continue; |
| 71 | if (seg === "..") { |
| 72 | if (!out.length) return null; |
| 73 | out.pop(); |
| 74 | } else out.push(seg); |
| 75 | } |
| 76 | return out.join("/"); |
| 77 | } |
| 78 | |
| 79 | /** |
| 80 | * Validate and resolve a URL from Markdown source. Returns null for anything |
| 81 | * that is not http(s), a fragment, or a relative path. |
| 82 | */ |
| 83 | export function safeUrl(url: string, kind: "link" | "image", opts: MarkdownOptions): string | null { |
| 84 | const u = url.trim(); |
| 85 | if (!u) return null; |
| 86 | // Browsers ignore control characters and whitespace when parsing schemes. |
| 87 | const probe = u.replace(/[\x00-\x20\x7f]/g, ""); |
| 88 | if (/^https?:\/\//i.test(probe)) return u; |
| 89 | if (/^[a-z][a-z0-9+.-]*:/i.test(probe) || /^[/\\]{2}/.test(probe)) return null; |
| 90 | if (u.startsWith("#")) return kind === "link" ? `#${opts.idPrefix ?? "h-"}${slugify(decodeSafe(u.slice(1)))}` : null; |
| 91 | const base = opts.base; |
| 92 | if (!base) return u.startsWith("/") ? u : null; |
| 93 | const hashAt = u.search(/[?#]/); |
| 94 | const pathPart = hashAt >= 0 ? u.slice(0, hashAt) : u; |
| 95 | const suffix = hashAt >= 0 ? u.slice(hashAt) : ""; |
| 96 | const joined = pathPart.startsWith("/") ? pathPart : `${base.dir ? `${base.dir}/` : ""}${pathPart}`; |
| 97 | const norm = normalise(decodeSafe(joined)); |
| 98 | if (norm === null) return null; |
| 99 | const root = `/${base.owner}/${base.repo}`; |
| 100 | if (kind === "image") return `${root}/raw/${encPath(base.ref)}/${encPath(norm)}`; |
| 101 | const isDir = pathPart.endsWith("/") || norm === ""; |
| 102 | return `${root}/${isDir ? "tree" : "blob"}/${encPath(base.ref)}${norm ? `/${encPath(norm)}` : ""}${suffix.replace(/[<>"']/g, "")}`; |
| 103 | } |
| 104 | |
| 105 | function decodeSafe(s: string): string { |
| 106 | try { |
| 107 | return decodeURIComponent(s); |
| 108 | } catch { |
| 109 | return s; |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | // ---------------------------------------------------------------- inline |
| 114 | |
| 115 | /** |
| 116 | * Text with its HTML comments (`<!-- … -->`, also `<!-->` and `<!--->`) removed. Code spans and |
| 117 | * backslash-escaped `<` are kept as written, and an unclosed `<!--` stays as text. |
| 118 | */ |
| 119 | function stripComments(src: string): string { |
| 120 | if (!src.includes("<!--")) return src; |
| 121 | let out = ""; |
| 122 | let i = 0; |
| 123 | const n = src.length; |
| 124 | // Every comment closes at or before the last "-->", so an opener past it is text: no rescans. |
| 125 | const lastClose = src.lastIndexOf("-->"); |
| 126 | while (i < n) { |
| 127 | if (scanned > budget) return out + src.slice(i); |
| 128 | const ch = src[i]!; |
| 129 | if (ch === "\\") { |
| 130 | out += src.slice(i, i + 2); |
| 131 | i += 2; |
| 132 | continue; |
| 133 | } |
| 134 | if (ch === "`") { |
| 135 | let run = 1; |
| 136 | while (src[i + run] === "`") run++; |
| 137 | const fence = "`".repeat(run); |
| 138 | let close = src.indexOf(fence, i + run); |
| 139 | while (close >= 0 && src[close + run] === "`") { |
| 140 | let skip = close; |
| 141 | while (src[skip] === "`") skip++; |
| 142 | close = src.indexOf(fence, skip); |
| 143 | } |
| 144 | scanned += (close < 0 ? n : close) - i; |
| 145 | const end = close < 0 ? i + run : close + run; |
| 146 | out += src.slice(i, end); |
| 147 | i = end; |
| 148 | continue; |
| 149 | } |
| 150 | if (ch === "<" && src.startsWith("<!--", i) && lastClose >= i + 2) { |
| 151 | const close = src.indexOf("-->", i + 2); |
| 152 | scanned += close - i; |
| 153 | i = close + 3; |
| 154 | continue; |
| 155 | } |
| 156 | out += ch; |
| 157 | i++; |
| 158 | } |
| 159 | return out; |
| 160 | } |
| 161 | |
| 162 | function findClosing(src: string, delim: string, from: number, wordy: boolean): number { |
| 163 | let at = from; |
| 164 | const limit = Math.min(src.length, from + MAX_SCAN * 2); |
| 165 | const window = src.slice(0, limit); |
| 166 | for (;;) { |
| 167 | const idx = window.indexOf(delim, at); |
| 168 | scanned += (idx < 0 ? limit : idx) - at + 1; |
| 169 | if (idx < 0) return -1; |
| 170 | const before = src[idx - 1]; |
| 171 | const after = src[idx + delim.length]; |
| 172 | if (idx > from && before !== undefined && !/\s/.test(before) && (!wordy || after === undefined || !/[\p{L}\p{N}]/u.test(after))) { |
| 173 | // When closing a single delimiter, skip over a double run ("**") that belongs to nested strong text. |
| 174 | if (delim.length === 1 && after === delim && src[idx + 2] !== delim) { |
| 175 | at = idx + 2; |
| 176 | continue; |
| 177 | } |
| 178 | return idx; |
| 179 | } |
| 180 | at = idx + 1; |
| 181 | } |
| 182 | } |
| 183 | |
| 184 | interface LinkMatch { |
| 185 | text: string; |
| 186 | url: string; |
| 187 | title: string | null; |
| 188 | end: number; |
| 189 | } |
| 190 | |
| 191 | /** Parse `[text](url "title")` starting at the "[" at `i`. */ |
| 192 | function parseLink(src: string, i: number): LinkMatch | null { |
| 193 | let depth = 0; |
| 194 | let j = i; |
| 195 | const limit = Math.min(src.length, i + MAX_SCAN); |
| 196 | for (; j < limit; j++) { |
| 197 | const ch = src[j]; |
| 198 | if (ch === "\\") { |
| 199 | j++; |
| 200 | continue; |
| 201 | } |
| 202 | if (ch === "`") { |
| 203 | const close = src.indexOf("`", j + 1); |
| 204 | if (close > 0 && close < limit) j = close; |
| 205 | continue; |
| 206 | } |
| 207 | if (ch === "[") depth++; |
| 208 | else if (ch === "]") { |
| 209 | depth--; |
| 210 | if (depth === 0) break; |
| 211 | } |
| 212 | } |
| 213 | scanned += j - i; |
| 214 | if (depth !== 0 || src[j] !== "]" || src[j + 1] !== "(") return null; |
| 215 | const text = src.slice(i + 1, j); |
| 216 | let k = j + 2; |
| 217 | while (src[k] === " ") k++; |
| 218 | let url = ""; |
| 219 | if (src[k] === "<") { |
| 220 | const close = src.indexOf(">", k); |
| 221 | scanned += (close < 0 ? src.length : close) - k; |
| 222 | if (close < 0 || close - k > MAX_SCAN) return null; |
| 223 | url = src.slice(k + 1, close); |
| 224 | k = close + 1; |
| 225 | } else { |
| 226 | let parens = 0; |
| 227 | const start = k; |
| 228 | for (; k < src.length && k - start < MAX_SCAN; k++) { |
| 229 | const ch = src[k]!; |
| 230 | if (ch === "\\") { |
| 231 | k++; |
| 232 | continue; |
| 233 | } |
| 234 | if (/\s/.test(ch)) break; |
| 235 | if (ch === "(") parens++; |
| 236 | else if (ch === ")") { |
| 237 | if (parens === 0) break; |
| 238 | parens--; |
| 239 | } |
| 240 | } |
| 241 | scanned += k - start; |
| 242 | url = src.slice(start, k).replace(/\\([()])/g, "$1"); |
| 243 | } |
| 244 | while (src[k] === " ") k++; |
| 245 | let title: string | null = null; |
| 246 | const q = src[k]; |
| 247 | if (q === '"' || q === "'" || q === "(") { |
| 248 | const closeCh = q === "(" ? ")" : q; |
| 249 | const close = src.indexOf(closeCh, k + 1); |
| 250 | scanned += (close < 0 ? src.length : close) - k; |
| 251 | if (close < 0) return null; |
| 252 | title = src.slice(k + 1, close); |
| 253 | k = close + 1; |
| 254 | while (src[k] === " ") k++; |
| 255 | } |
| 256 | if (src[k] !== ")") return null; |
| 257 | return { text, url, title, end: k + 1 }; |
| 258 | } |
| 259 | |
| 260 | const BARE_URL = /^https?:\/\/[^\s<>"'`]*[^\s<>"'`.,:;!?)\]}*_~]/i; |
| 261 | |
| 262 | function inline(text: string, opts: MarkdownOptions, depth = 0): string { |
| 263 | if (text.length > MAX_INLINE || depth > MAX_DEPTH) return escapeHtml(text); |
| 264 | const src = stripComments(text); |
| 265 | let out = ""; |
| 266 | let i = 0; |
| 267 | const n = src.length; |
| 268 | while (i < n) { |
| 269 | if (scanned > budget) return out + escapeHtml(src.slice(i)); |
| 270 | const ch = src[i]!; |
| 271 | // Backslash escapes and hard line breaks. |
| 272 | if (ch === "\\") { |
| 273 | const next = src[i + 1]; |
| 274 | if (next === "\n") { |
| 275 | out += "<br>\n"; |
| 276 | i += 2; |
| 277 | continue; |
| 278 | } |
| 279 | if (next !== undefined && /[!-/:-@[-`{-~]/.test(next)) { |
| 280 | out += escapeHtml(next); |
| 281 | i += 2; |
| 282 | continue; |
| 283 | } |
| 284 | } |
| 285 | // Two or more spaces before a newline are a hard line break. Decided from the source: |
| 286 | // looking at the end of `out` would flatten the whole string on every line. |
| 287 | if (ch === " ") { |
| 288 | let run = 1; |
| 289 | while (src[i + run] === " ") run++; |
| 290 | if (run >= 2 && src[i + run] === "\n") { |
| 291 | out += "<br>\n"; |
| 292 | i += run + 1; |
| 293 | continue; |
| 294 | } |
| 295 | out += " ".repeat(run); |
| 296 | i += run; |
| 297 | continue; |
| 298 | } |
| 299 | // Code spans. |
| 300 | if (ch === "`") { |
| 301 | let run = 1; |
| 302 | while (src[i + run] === "`") run++; |
| 303 | const fence = "`".repeat(run); |
| 304 | let close = src.indexOf(fence, i + run); |
| 305 | while (close >= 0 && src[close + run] === "`") { |
| 306 | let skip = close; |
| 307 | while (src[skip] === "`") skip++; |
| 308 | close = src.indexOf(fence, skip); |
| 309 | } |
| 310 | scanned += (close < 0 ? n : close) - i; |
| 311 | if (close >= 0) { |
| 312 | let code = src.slice(i + run, close).replace(/\n/g, " "); |
| 313 | if (code.length > 2 && code.startsWith(" ") && code.endsWith(" ") && code.trim()) code = code.slice(1, -1); |
| 314 | out += `<code>${escapeHtml(code)}</code>`; |
| 315 | i = close + run; |
| 316 | continue; |
| 317 | } |
| 318 | out += fence; |
| 319 | i += run; |
| 320 | continue; |
| 321 | } |
| 322 | // Images. |
| 323 | if (ch === "!" && src[i + 1] === "[") { |
| 324 | const m = parseLink(src, i + 1); |
| 325 | if (m) { |
| 326 | const safe = safeUrl(m.url, "image", opts); |
| 327 | const url = safe && opts.image ? opts.image(safe) : safe; |
| 328 | const alt = escapeHtml(m.text.replace(/[\\*_`[\]]/g, "")); |
| 329 | if (url) out += `<img src="${escapeHtml(url)}" alt="${alt}"${m.title ? ` title="${escapeHtml(m.title)}"` : ""} loading="lazy" referrerpolicy="no-referrer">`; |
| 330 | else out += alt; |
| 331 | i = m.end; |
| 332 | continue; |
| 333 | } |
| 334 | } |
| 335 | // Links. |
| 336 | if (ch === "[") { |
| 337 | const m = parseLink(src, i); |
| 338 | if (m) { |
| 339 | const url = safeUrl(m.url, "link", opts); |
| 340 | const label = inline(m.text, opts, depth + 1); |
| 341 | if (url) { |
| 342 | const external = /^https?:/i.test(url); |
| 343 | out += `<a href="${escapeHtml(url)}"${m.title ? ` title="${escapeHtml(m.title)}"` : ""}${external ? ' rel="nofollow noopener noreferrer"' : ""}>${label}</a>`; |
| 344 | } else out += label; |
| 345 | i = m.end; |
| 346 | continue; |
| 347 | } |
| 348 | } |
| 349 | // Autolinks <https://…>. |
| 350 | if (ch === "<") { |
| 351 | const m = /^<(https?:\/\/[^\s<>]+)>/i.exec(src.slice(i, i + MAX_SCAN)); |
| 352 | if (m) { |
| 353 | const url = escapeHtml(m[1]!); |
| 354 | out += `<a href="${url}" rel="nofollow noopener noreferrer">${url}</a>`; |
| 355 | i += m[0].length; |
| 356 | continue; |
| 357 | } |
| 358 | } |
| 359 | // Bare URLs. |
| 360 | if ((ch === "h" || ch === "H") && (i === 0 || /[\s(]/.test(src[i - 1]!))) { |
| 361 | const m = BARE_URL.exec(src.slice(i, i + MAX_SCAN)); |
| 362 | if (m) { |
| 363 | const url = escapeHtml(m[0]); |
| 364 | out += `<a href="${url}" rel="nofollow noopener noreferrer">${url}</a>`; |
| 365 | i += m[0].length; |
| 366 | continue; |
| 367 | } |
| 368 | } |
| 369 | // Strikethrough. |
| 370 | if (ch === "~" && src[i + 1] === "~" && src[i + 2] && !/\s/.test(src[i + 2]!)) { |
| 371 | const close = findClosing(src, "~~", i + 2, false); |
| 372 | if (close > i + 2) { |
| 373 | out += `<del>${inline(src.slice(i + 2, close), opts, depth + 1)}</del>`; |
| 374 | i = close + 2; |
| 375 | continue; |
| 376 | } |
| 377 | } |
| 378 | // Emphasis. |
| 379 | if (ch === "*" || ch === "_") { |
| 380 | let run = 1; |
| 381 | while (src[i + run] === ch) run++; |
| 382 | const wordy = ch === "_"; |
| 383 | const prev = src[i - 1]; |
| 384 | const nextCh = src[i + run]; |
| 385 | const canOpen = nextCh !== undefined && !/\s/.test(nextCh) && !(wordy && prev !== undefined && /[\p{L}\p{N}]/u.test(prev)); |
| 386 | let matched = false; |
| 387 | if (canOpen && run <= 3) { |
| 388 | for (const size of run === 3 ? [3, 2, 1] : run === 2 ? [2, 1] : [1]) { |
| 389 | const delim = ch.repeat(size); |
| 390 | const start = i + run; |
| 391 | const close = findClosing(src, delim, start, wordy); |
| 392 | if (close > start) { |
| 393 | const inner = inline(src.slice(start, close), opts, depth + 1); |
| 394 | out += escapeHtml(ch.repeat(run - size)); |
| 395 | out += size === 3 ? `<strong><em>${inner}</em></strong>` : size === 2 ? `<strong>${inner}</strong>` : `<em>${inner}</em>`; |
| 396 | i = close + size; |
| 397 | matched = true; |
| 398 | break; |
| 399 | } |
| 400 | } |
| 401 | } |
| 402 | if (matched) continue; |
| 403 | out += escapeHtml(ch.repeat(run)); |
| 404 | i += run; |
| 405 | continue; |
| 406 | } |
| 407 | out += ESC[ch] ?? ch; |
| 408 | i++; |
| 409 | } |
| 410 | return out; |
| 411 | } |
| 412 | |
| 413 | // ---------------------------------------------------------------- blocks |
| 414 | |
| 415 | // Block patterns run on every line, so none may backtrack: no two adjacent quantifiers that |
| 416 | // can match the same characters, and `[^]` rather than `.`, which stops at U+2028 and fails |
| 417 | // the `$` after it. Trailing spaces and closing `#`s are trimmed by hand (see headingText). |
| 418 | const FENCE = /^ {0,3}(`{3,}(?!`)|~{3,}(?!~))(?![^`]*`)\s*([^`\s]*)/; |
| 419 | const HEADING = /^ {0,3}(#{1,6})(?:[ \t]([^]*))?$/; |
| 420 | const HR = /^ {0,3}([-*_])(?:[ \t]*\1){2,}[ \t]*$/; |
| 421 | const QUOTE = /^ {0,3}> ?([^]*)$/; |
| 422 | const ALERT = /^\s*\[!(NOTE|TIP|IMPORTANT|WARNING|CAUTION)\]\s*$/i; |
| 423 | const COMMENT = /^ {0,3}<!--/; |
| 424 | const LIST_ITEM = /^( {0,12})([*+-]|\d{1,9}[.)])([ \t]+([^]*))?$/; |
| 425 | const SEP_CELL = /^:?-+:?$/; |
| 426 | |
| 427 | function trimBlank(s: string): string { |
| 428 | let a = 0; |
| 429 | let b = s.length; |
| 430 | while (a < b && (s[a] === " " || s[a] === "\t")) a++; |
| 431 | while (b > a && (s[b - 1] === " " || s[b - 1] === "\t")) b--; |
| 432 | return s.slice(a, b); |
| 433 | } |
| 434 | |
| 435 | /** A heading's text without its surrounding spaces and optional closing `#`s. */ |
| 436 | function headingText(raw: string): string { |
| 437 | const t = trimBlank(raw); |
| 438 | let j = t.length; |
| 439 | while (j > 0 && t[j - 1] === "#") j--; |
| 440 | if (j === t.length) return t; |
| 441 | if (j === 0) return ""; |
| 442 | return t[j - 1] === " " || t[j - 1] === "\t" ? trimBlank(t.slice(0, j)) : t; |
| 443 | } |
| 444 | |
| 445 | /** A table's delimiter row, like `| :-- | --: |`, checked cell by cell. */ |
| 446 | function tableSep(line: string): boolean { |
| 447 | let s = trimBlank(line); |
| 448 | if (s.startsWith("|")) s = s.slice(1); |
| 449 | if (s.endsWith("|")) s = s.slice(0, -1); |
| 450 | return s.split("|").every((cell) => SEP_CELL.test(trimBlank(cell))); |
| 451 | } |
| 452 | |
| 453 | function splitRow(line: string): string[] { |
| 454 | let s = line.trim(); |
| 455 | if (s.startsWith("|")) s = s.slice(1); |
| 456 | if (s.endsWith("|") && !s.endsWith("\\|")) s = s.slice(0, -1); |
| 457 | const cells: string[] = []; |
| 458 | let cur = ""; |
| 459 | let inCode = false; |
| 460 | for (let i = 0; i < s.length; i++) { |
| 461 | const ch = s[i]!; |
| 462 | if (ch === "\\" && s[i + 1] === "|") { |
| 463 | cur += "|"; |
| 464 | i++; |
| 465 | continue; |
| 466 | } |
| 467 | if (ch === "`") inCode = !inCode; |
| 468 | if (ch === "|" && !inCode) { |
| 469 | cells.push(cur.trim()); |
| 470 | cur = ""; |
| 471 | continue; |
| 472 | } |
| 473 | cur += ch; |
| 474 | } |
| 475 | cells.push(cur.trim()); |
| 476 | return cells; |
| 477 | } |
| 478 | |
| 479 | function startsBlock(line: string, next: string | undefined): boolean { |
| 480 | return ( |
| 481 | FENCE.test(line) || |
| 482 | HEADING.test(line) || |
| 483 | HR.test(line) || |
| 484 | QUOTE.test(line) || |
| 485 | COMMENT.test(line) || |
| 486 | /^ {0,3}([*+-]|1[.)])[ \t]+\S/.test(line) || |
| 487 | (line.includes("|") && next !== undefined && tableSep(next)) |
| 488 | ); |
| 489 | } |
| 490 | |
| 491 | function blocks(lines: string[], opts: MarkdownOptions, depth: number, tight = false): string { |
| 492 | if (depth > MAX_DEPTH) return `<p>${escapeHtml(lines.join("\n"))}</p>`; |
| 493 | let i = 0; |
| 494 | const parts: string[] = []; |
| 495 | while (i < lines.length) { |
| 496 | const line = lines[i]!; |
| 497 | if (!line.trim()) { |
| 498 | i++; |
| 499 | continue; |
| 500 | } |
| 501 | // Fenced code. |
| 502 | const fence = FENCE.exec(line); |
| 503 | if (fence) { |
| 504 | const marker = fence[1]!; |
| 505 | const lang = fence[2] ?? ""; |
| 506 | const indent = line.length - line.trimStart().length; |
| 507 | const body: string[] = []; |
| 508 | i++; |
| 509 | while (i < lines.length && !new RegExp(`^ {0,3}${marker[0] === "`" ? "`" : "~"}{${marker.length},}\\s*$`).test(lines[i]!)) { |
| 510 | body.push(lines[i]!.replace(new RegExp(`^ {0,${indent}}`), "")); |
| 511 | i++; |
| 512 | } |
| 513 | i++; |
| 514 | const named = /^[\w+#.-]{1,30}$/.test(lang) ? lang : ""; |
| 515 | const hl = named && opts.highlight ? opts.highlight(body.join("\n"), named) : null; |
| 516 | const cls = named ? ` class="language-${escapeHtml(named)}${hl !== null ? " sh" : ""}"` : ""; |
| 517 | parts.push(`<pre><code${cls}>${hl ?? escapeHtml(body.join("\n"))}</code></pre>`); |
| 518 | continue; |
| 519 | } |
| 520 | // Indented code (4 spaces), only outside lists. |
| 521 | if (/^( {4}|\t)/.test(line) && !tight) { |
| 522 | const body: string[] = []; |
| 523 | while (i < lines.length && (/^( {4}|\t)/.test(lines[i]!) || !lines[i]!.trim())) { |
| 524 | body.push(lines[i]!.replace(/^( {4}|\t)/, "")); |
| 525 | i++; |
| 526 | } |
| 527 | while (body.length && !body[body.length - 1]!.trim()) body.pop(); |
| 528 | parts.push(`<pre><code>${escapeHtml(body.join("\n"))}</code></pre>`); |
| 529 | continue; |
| 530 | } |
| 531 | // HTML comments, hidden as GitHub hides them. A line that starts with "<!--" opens one that |
| 532 | // runs to the line holding "-->"; anything after the close on that line is rendered as usual. |
| 533 | // Left open, it hides the rest of the document (or of the list item or quote it is in). |
| 534 | const comment = COMMENT.exec(line); |
| 535 | if (comment) { |
| 536 | let j = i; |
| 537 | let close = line.indexOf("-->", comment[0].length - 2); |
| 538 | while (close < 0 && ++j < lines.length) close = lines[j]!.indexOf("-->"); |
| 539 | if (close < 0) break; |
| 540 | const rest = lines[j]!.slice(close + 3).trimStart(); |
| 541 | if (rest) lines[j] = rest; |
| 542 | i = rest ? j : j + 1; |
| 543 | continue; |
| 544 | } |
| 545 | const h = HEADING.exec(line); |
| 546 | if (h) { |
| 547 | const level = h[1]!.length; |
| 548 | const text = stripComments(headingText(h[2] ?? "")); |
| 549 | const html = inline(text, opts, depth); |
| 550 | const id = escapeHtml(`${opts.idPrefix ?? "h-"}${slugify(text)}`); |
| 551 | parts.push(`<h${level} id="${id}"><a class="anchor" href="#${id}" aria-hidden="true">#</a>${html}</h${level}>`); |
| 552 | i++; |
| 553 | continue; |
| 554 | } |
| 555 | if (HR.test(line)) { |
| 556 | parts.push("<hr>"); |
| 557 | i++; |
| 558 | continue; |
| 559 | } |
| 560 | if (QUOTE.test(line)) { |
| 561 | const body: string[] = []; |
| 562 | while (i < lines.length && lines[i]!.trim()) { |
| 563 | const m = QUOTE.exec(lines[i]!); |
| 564 | if (m) body.push(m[1]!); |
| 565 | else if (body.length && !startsBlock(lines[i]!, lines[i + 1])) body.push(lines[i]!); |
| 566 | else break; |
| 567 | i++; |
| 568 | } |
| 569 | // GitHub's alerts: a quote whose first line is only [!NOTE], [!TIP], [!IMPORTANT], [!WARNING] or [!CAUTION]. |
| 570 | const alert = ALERT.exec(body[0] ?? ""); |
| 571 | if (alert && body.length > 1) { |
| 572 | const kind = alert[1]!.toLowerCase(); |
| 573 | parts.push(`<div class="markdown-alert markdown-alert-${kind}"><p class="markdown-alert-title">${kind[0]!.toUpperCase()}${kind.slice(1)}</p>${blocks(body.slice(1), opts, depth + 1)}</div>`); |
| 574 | continue; |
| 575 | } |
| 576 | parts.push(`<blockquote>${blocks(body, opts, depth + 1)}</blockquote>`); |
| 577 | continue; |
| 578 | } |
| 579 | // Tables. |
| 580 | if (line.includes("|") && i + 1 < lines.length && tableSep(lines[i + 1]!)) { |
| 581 | const head = splitRow(line); |
| 582 | const aligns = splitRow(lines[i + 1]!).map((c) => (c.startsWith(":") && c.endsWith(":") ? "center" : c.endsWith(":") ? "right" : c.startsWith(":") ? "left" : "")); |
| 583 | i += 2; |
| 584 | const rows: string[][] = []; |
| 585 | while (i < lines.length && lines[i]!.trim() && lines[i]!.includes("|")) { |
| 586 | rows.push(splitRow(lines[i]!)); |
| 587 | i++; |
| 588 | } |
| 589 | const cell = (tag: string, text: string, idx: number) => `<${tag}${aligns[idx] ? ` style="text-align:${aligns[idx]}"` : ""}>${inline(text, opts, depth)}</${tag}>`; |
| 590 | let t = `<div class="table-wrap"><table><thead><tr>${head.map((c, idx) => cell("th", c, idx)).join("")}</tr></thead>`; |
| 591 | if (rows.length) t += `<tbody>${rows.map((r) => `<tr>${head.map((_, idx) => cell("td", r[idx] ?? "", idx)).join("")}</tr>`).join("")}</tbody>`; |
| 592 | parts.push(`${t}</table></div>`); |
| 593 | continue; |
| 594 | } |
| 595 | // Lists. |
| 596 | const li = LIST_ITEM.exec(line); |
| 597 | if (li) { |
| 598 | const ordered = /\d/.test(li[2]!); |
| 599 | const baseIndent = li[1]!.length; |
| 600 | const start = ordered ? parseInt(li[2]!, 10) : 1; |
| 601 | const items: string[][] = []; |
| 602 | let loose = false; |
| 603 | while (i < lines.length) { |
| 604 | const m = LIST_ITEM.exec(lines[i]!); |
| 605 | if (!m || m[1]!.length > baseIndent + 1 || m[1]!.length < baseIndent || /\d/.test(m[2]!) !== ordered) break; |
| 606 | const contentIndent = m[1]!.length + m[2]!.length + Math.max(1, Math.min(4, (m[3] ?? " ").length - (m[4] ?? "").length)); |
| 607 | const item: string[] = [m[4] ?? ""]; |
| 608 | i++; |
| 609 | let sawBlank = false; |
| 610 | while (i < lines.length) { |
| 611 | const l = lines[i]!; |
| 612 | if (!l.trim()) { |
| 613 | sawBlank = true; |
| 614 | item.push(""); |
| 615 | i++; |
| 616 | continue; |
| 617 | } |
| 618 | const ind = l.length - l.trimStart().length; |
| 619 | if (ind >= contentIndent || (ind > baseIndent && LIST_ITEM.test(l))) { |
| 620 | item.push(l.slice(Math.min(ind, contentIndent))); |
| 621 | sawBlank = false; |
| 622 | i++; |
| 623 | continue; |
| 624 | } |
| 625 | if (!sawBlank && !startsBlock(l, lines[i + 1]) && !LIST_ITEM.test(l)) { |
| 626 | // Lazy continuation of the item's paragraph. |
| 627 | item.push(l.trim()); |
| 628 | i++; |
| 629 | continue; |
| 630 | } |
| 631 | break; |
| 632 | } |
| 633 | while (item.length && !item[item.length - 1]!.trim()) { |
| 634 | item.pop(); |
| 635 | if (i < lines.length && LIST_ITEM.test(lines[i]!)) loose = true; |
| 636 | } |
| 637 | if (item.slice(1).some((l, idx) => !l.trim() && item[idx + 2] !== undefined && !/^\s*([*+-]|\d+[.)])\s/.test(item[idx + 2]!))) loose = true; |
| 638 | items.push(item); |
| 639 | } |
| 640 | const rendered = items.map((item) => { |
| 641 | let first = item[0] ?? ""; |
| 642 | let task = ""; |
| 643 | const tm = /^\[([ xX])\][ \t]+/.exec(first); |
| 644 | if (tm) { |
| 645 | task = tm[1] !== " " ? `<span class="task-box on" role="img" aria-label="Done"></span> ` : `<span class="task-box" role="img" aria-label="To do"></span> `; |
| 646 | first = first.slice(tm[0].length); |
| 647 | } |
| 648 | const body = blocks([first, ...item.slice(1)], opts, depth + 1, !loose); |
| 649 | return `<li${task ? ' class="task"' : ""}>${task}${body}</li>`; |
| 650 | }); |
| 651 | parts.push(ordered ? `<ol${start !== 1 ? ` start="${start}"` : ""}>${rendered.join("")}</ol>` : `<ul>${rendered.join("")}</ul>`); |
| 652 | continue; |
| 653 | } |
| 654 | // Paragraph. |
| 655 | const para: string[] = [line.trimStart()]; |
| 656 | i++; |
| 657 | while (i < lines.length && lines[i]!.trim()) { |
| 658 | // Setext headings (checked before "---" is taken as a thematic break). |
| 659 | if (/^ {0,3}=+\s*$/.test(lines[i]!) || /^ {0,3}-+\s*$/.test(lines[i]!)) { |
| 660 | const level = lines[i]!.includes("=") ? 1 : 2; |
| 661 | const text = stripComments(para.join(" ")); |
| 662 | const id = escapeHtml(`${opts.idPrefix ?? "h-"}${slugify(text)}`); |
| 663 | para.length = 0; |
| 664 | parts.push(`<h${level} id="${id}"><a class="anchor" href="#${id}" aria-hidden="true">#</a>${inline(text, opts, depth)}</h${level}>`); |
| 665 | i++; |
| 666 | break; |
| 667 | } |
| 668 | if (startsBlock(lines[i]!, lines[i + 1])) break; |
| 669 | para.push(lines[i]!.trimStart()); |
| 670 | i++; |
| 671 | } |
| 672 | if (para.length) { |
| 673 | const html = inline(trimBlank(para.join("\n")), opts, depth); |
| 674 | parts.push(tight ? html : `<p>${html}</p>`); |
| 675 | } |
| 676 | } |
| 677 | return parts.join("\n"); |
| 678 | } |
| 679 | |
| 680 | /** Languages named by the document's code fences, to load grammars before rendering. */ |
| 681 | export function fenceLangs(source: string): string[] { |
| 682 | if (source.length > MAX_SOURCE) return []; |
| 683 | const out = new Set<string>(); |
| 684 | for (const line of source.split("\n")) { |
| 685 | const lang = FENCE.exec(line)?.[2]; |
| 686 | if (lang && /^[\w+#.-]{1,30}$/.test(lang)) out.add(lang); |
| 687 | } |
| 688 | return [...out]; |
| 689 | } |
| 690 | |
| 691 | /** |
| 692 | * Markdown without its HTML comments, for plain-text views such as email. Comments inside |
| 693 | * fenced code and code spans stay. |
| 694 | */ |
| 695 | export function stripHtmlComments(source: string): string { |
| 696 | if (!source.includes("<!--") || source.length > MAX_SOURCE) return source; |
| 697 | scanned = 0; |
| 698 | budget = scanBudget(source.length); |
| 699 | const out: string[] = []; |
| 700 | let prose: string[] = []; |
| 701 | let fence: RegExp | null = null; |
| 702 | const flush = () => { |
| 703 | if (prose.length) out.push(stripComments(prose.join("\n"))); |
| 704 | prose = []; |
| 705 | }; |
| 706 | for (const line of source.split("\n")) { |
| 707 | if (fence) { |
| 708 | out.push(line); |
| 709 | if (fence.test(line)) fence = null; |
| 710 | continue; |
| 711 | } |
| 712 | const open = FENCE.exec(line); |
| 713 | if (open) { |
| 714 | flush(); |
| 715 | out.push(line); |
| 716 | fence = new RegExp(`^ {0,3}${open[1]![0] === "`" ? "`" : "~"}{${open[1]!.length},}\\s*$`); |
| 717 | continue; |
| 718 | } |
| 719 | prose.push(line); |
| 720 | } |
| 721 | flush(); |
| 722 | return out.join("\n").replace(/\n[ \t]*(?:\n[ \t]*){2,}/g, "\n\n").trim(); |
| 723 | } |
| 724 | |
| 725 | /** Render Markdown to sanitized HTML. */ |
| 726 | export function renderMarkdown(source: string, opts: MarkdownOptions = {}): string { |
| 727 | if (source.length > MAX_SOURCE) return `<pre>${escapeHtml(source)}</pre>`; |
| 728 | const lines = source.replace(/\r\n?/g, "\n").split("\n"); |
| 729 | scanned = 0; |
| 730 | budget = scanBudget(source.length); |
| 731 | return blocks(lines, opts, 0); |
| 732 | } |