@ai-matrx/print 0.8.7 → 0.8.9

package/CHANGELOG.md

@@ -1,5 +1,55 @@
 # Changelog
 
+## 0.8.9
+### Unreleased
+
+- **New subpath `/markdown-html` — THE ONE Markdown→HTML engine** (Matrx
+  Alchemy CONTRACT §1.9, ALC-16). `createMarkdownHtmlEngine({ projectors })`
+  → `parse` (content-ir marks every kind island, each becomes an atom
+  placeholder, the `/document` parse runs once), `fragment` (skins
+  `document` / `article` / `inline`; profiles `generic` / `google-docs` /
+  `word`), `document`, `plainText` (tables as TSV, TeX kept as `$…$`). Every
+  string ends in the allow-list. Optional asset resolver for private images
+  (an unresolved image becomes its description with a notice).
+- **Security (LIST T26):** the display-math SVG the PDF writer draws is
+  allow-list sanitized (`MATH_SVG_SCHEMA`, `sanitizeMathSvg`) — TeX `\href`,
+  `\style`, `\class`, `\cssId`, `\bbox[…]`, `\mmlToken[…]` no longer reach it.
+- `/safe-html`: `engineHtmlSchema`, `sanitizeEngineHtml`.
+- `/document`: `documentToHtml` accepts `srcFor` and `transformBody`; an
+  image `srcFor` result now also passes the URL check (EPUB included).
+
+- **Tables print by THE GFM rule** (`@ai-matrx/content-ir/source`, verify-RC-B4):
+  tables written without edge pipes print as tables; an escaped `\|` stays in
+  its cell by backslash parity and prints as `|`; a line that starts another
+  block (`- x`, `> y`) ends a pipe-less table; rows print at the header's width.
+  The private `splitTableRow` and the pipe-led-only table regex are gone.
+  Requires `@ai-matrx/content-ir` with `/source` table exports.
+
+## 0.8.8
+
+- **Security: every markdown→HTML path is allow-list sanitized.** The regex
+  converter (`markdownToHtml`, `markdownToArticleHtml`, and through them
+  `renderMarkdownDocument`, `markdownToHtmlForSkin`, `markdownToPdfBlob`) used a
+  block-list and let `<iframe>`, `<form>`, `<meta http-equiv=refresh>`,
+  `<object>`/`<embed>`, `<base>`, `<link>`, `<svg>`, `<textarea>`/`<select>` and
+  inline `style` through. Output now goes through `PRINT_HTML_SCHEMA`
+  (hast-util-sanitize's GitHub schema plus the converter's own `.matrx-*` /
+  `print-*` / `code-*` classes, table-cell `text-align`, the flashcard
+  checkbox + label, figures, page-break data/aria attributes). Sync, DOM-free.
+- **Security (`/document`):** TeX `\href`, `\style`, `\class`, `\cssId`,
+  `\bbox[css]` and `\mmlToken[attrs]` no longer write `href`/`style` into the
+  exported MathML (`MATHML_SCHEMA`); a front-matter `header`/`footer` can no
+  longer close the document's `<style>` element.
+- **New subpath `/safe-html`:** `sanitizeHtmlFragment`, `SAFE_HTML_SCHEMA`
+  (plain GitHub allow-list for email/CMS), `PRINT_HTML_SCHEMA`,
+  `sanitizePrintHtml`, `MATHML_SCHEMA`, `sanitizeMathml`.
+- **Visible differences:** author raw-HTML classes outside the print
+  vocabulary are dropped (`<div class="note">` → `<div class="matrx-div">`);
+  author inline styles are dropped; HTML serialization is normalized (`<img …>`
+  not `<img … />`, text `>` and `"` literal instead of `&gt;`/`&quot;`) —
+  same rendered page.
+- Consumer action: none — refresh to latest.
+
 ## 0.8.7
 
 - **Fixed: hidden reasoning merged two paragraphs.** `removeThinkingContent`
@@ -583,3 +633,4 @@ originals die at the adoption swap):
   defaults; the template registry is now immutable; the dev-time geometry
   check no longer reads `process.env` bare (which would throw in browser
   consumers).
+

package/dist/document.cjs

@@ -170,6 +170,249 @@ function contentColumnWidths(b, n, avail, measure, pad = 10) {
 
 // src/document/math.ts
 var import_source = require("@ai-matrx/content-ir/source");
+
+// src/safe-html.ts
+var import_hast_util_from_html = require("hast-util-from-html");
+var import_hast_util_sanitize = require("hast-util-sanitize");
+var import_hast_util_to_html = require("hast-util-to-html");
+var STRIP_WITH_CONTENT = [
+  "script",
+  "style",
+  "noscript",
+  "template",
+  "iframe",
+  "object",
+  "embed",
+  "textarea",
+  "select",
+  "head",
+  "title"
+];
+var FORBIDDEN_TAGS = [
+  ...STRIP_WITH_CONTENT,
+  "form",
+  "meta",
+  "base",
+  "link",
+  "frame",
+  "frameset",
+  "applet",
+  "button",
+  "svg",
+  "math"
+];
+var baseAttributes = import_hast_util_sanitize.defaultSchema.attributes ?? {};
+var SAFE_HTML_SCHEMA = {
+  ...import_hast_util_sanitize.defaultSchema,
+  tagNames: (import_hast_util_sanitize.defaultSchema.tagNames ?? []).filter((t) => !FORBIDDEN_TAGS.includes(t)),
+  strip: STRIP_WITH_CONTENT,
+  // Authored anchors keep their names; the GitHub `user-content-` rewrite
+  // would also break `<label for>` → `<input id>` pairs.
+  clobberPrefix: "",
+  clobber: []
+};
+var PRINT_CLASS = /^(?:(?:matrx|print|code|doc|language)-[\w-]*|inline-code|is-checked)$/;
+function withPrintClasses(defs = []) {
+  let found = false;
+  const out = defs.map((d) => {
+    if (Array.isArray(d) && d[0] === "className") {
+      found = true;
+      return [...d, PRINT_CLASS];
+    }
+    return d;
+  });
+  return found ? out : [...out, ["className", PRINT_CLASS]];
+}
+var printAttributes = Object.fromEntries(
+  Object.entries(baseAttributes).map(([tag, defs]) => [tag, withPrintClasses(defs)])
+);
+var cellAlign = ["style", /^text-align:\s*(?:left|center|right);?$/];
+var PRINT_HTML_SCHEMA = {
+  ...SAFE_HTML_SCHEMA,
+  tagNames: [
+    .../* @__PURE__ */ new Set([...SAFE_HTML_SCHEMA.tagNames ?? [], "section", "figure", "figcaption", "label"])
+  ],
+  attributes: {
+    ...printAttributes,
+    "*": [
+      ...baseAttributes["*"] ?? [],
+      ["className", PRINT_CLASS],
+      "role",
+      "ariaLabel",
+      "ariaHidden",
+      "dataLabel",
+      "dataLanguage",
+      "dataArtifactType"
+    ],
+    // Table-cell alignment is the one inline style the converter writes.
+    td: withPrintClasses([...baseAttributes["td"] ?? [], [...cellAlign]]),
+    th: withPrintClasses([...baseAttributes["th"] ?? [], [...cellAlign]]),
+    // The flashcard flip is a checkbox + label pair (GitHub forces `disabled`).
+    input: withPrintClasses([["type", "checkbox"], "checked"]),
+    label: withPrintClasses(["htmlFor"])
+  },
+  required: { input: { type: "checkbox" } },
+  protocols: {
+    ...import_hast_util_sanitize.defaultSchema.protocols,
+    href: [...import_hast_util_sanitize.defaultSchema.protocols?.href ?? [], "tel"],
+    // Embedded images (`data:image/png;base64,…`) — an <img> never runs script.
+    src: [...import_hast_util_sanitize.defaultSchema.protocols?.src ?? [], "data"]
+  }
+};
+function sanitizeHtmlFragment(html, schema = SAFE_HTML_SCHEMA) {
+  if (!html) return "";
+  const tree = (0, import_hast_util_sanitize.sanitize)((0, import_hast_util_from_html.fromHtml)(html, { fragment: true }), schema);
+  return (0, import_hast_util_to_html.toHtml)(tree, { characterReferences: { useNamedReferences: true } });
+}
+var MATHML_TAGS = [
+  "math",
+  "maction",
+  "annotation",
+  "menclose",
+  "merror",
+  "mfrac",
+  "mglyph",
+  "mi",
+  "mlabeledtr",
+  "mmultiscripts",
+  "mn",
+  "mo",
+  "mover",
+  "mpadded",
+  "mphantom",
+  "mprescripts",
+  "none",
+  "mroot",
+  "mrow",
+  "ms",
+  "mspace",
+  "msqrt",
+  "mstyle",
+  "msub",
+  "msubsup",
+  "msup",
+  "mtable",
+  "mtd",
+  "mtext",
+  "mtr",
+  "munder",
+  "munderover",
+  "semantics"
+];
+var MATHML_ATTRIBUTES = [
+  "xmlns",
+  "display",
+  "displaystyle",
+  "mathvariant",
+  "mathsize",
+  "mathcolor",
+  "mathbackground",
+  "scriptlevel",
+  "dir",
+  "fence",
+  "form",
+  "largeop",
+  "lspace",
+  "rspace",
+  "movablelimits",
+  "separator",
+  "stretchy",
+  "symmetric",
+  "maxsize",
+  "minsize",
+  "accent",
+  "accentunder",
+  "align",
+  "columnalign",
+  "rowalign",
+  "columnspacing",
+  "rowspacing",
+  "columnlines",
+  "rowlines",
+  "frame",
+  "framespacing",
+  "equalrows",
+  "equalcolumns",
+  "linethickness",
+  "numalign",
+  "denomalign",
+  "bevelled",
+  "notation",
+  "width",
+  "height",
+  "depth",
+  "voffset",
+  "lquote",
+  "rquote",
+  "columnspan",
+  "rowspan",
+  "encoding",
+  "open",
+  "close",
+  "separators",
+  "actiontype",
+  "selection",
+  "mslinebreak",
+  "linebreak",
+  "indentalign",
+  "data*"
+];
+var MATHML_SCHEMA = {
+  tagNames: MATHML_TAGS,
+  attributes: { "*": MATHML_ATTRIBUTES },
+  strip: ["script", "style"],
+  clobberPrefix: "",
+  clobber: [],
+  protocols: {}
+};
+function sanitizeMathml(mathml) {
+  return sanitizeHtmlFragment(mathml, MATHML_SCHEMA);
+}
+var MATH_SVG_TAGS = ["svg", "g", "path", "rect", "line", "polyline", "polygon", "circle", "ellipse", "text", "title", "desc"];
+var MATH_SVG_ATTRIBUTES = [
+  "xmlns",
+  "width",
+  "height",
+  "viewBox",
+  "role",
+  "focusable",
+  "ariaHidden",
+  "ariaLabel",
+  "stroke",
+  "fill",
+  "strokeWidth",
+  "strokeLinecap",
+  "strokeLinejoin",
+  "transform",
+  "d",
+  "x",
+  "y",
+  "x1",
+  "y1",
+  "x2",
+  "y2",
+  "points",
+  "cx",
+  "cy",
+  "r",
+  "rx",
+  "ry",
+  "dataMmlNode",
+  "dataC"
+];
+var MATH_SVG_SCHEMA = {
+  tagNames: MATH_SVG_TAGS,
+  attributes: { "*": MATH_SVG_ATTRIBUTES },
+  strip: ["script", "style", "foreignObject"],
+  clobberPrefix: "",
+  clobber: [],
+  protocols: {}
+};
+function sanitizeMathSvg(svg) {
+  return sanitizeHtmlFragment(svg, MATH_SVG_SCHEMA);
+}
+
+// src/document/math.ts
 var MATH_OPEN = "\uE100";
 var MATH_CLOSE = "\uE101";
 var MATH_TOKEN_RE = /\uE100(\d+)\uE101/g;
@@ -236,12 +479,12 @@ async function typesetMathSpans(spans) {
       const mathml = mj.mml(s.tex, s.display);
       const err = /<merror[\s\S]*?<mtext[^>]*>([\s\S]*?)<\/mtext>/.exec(mathml);
       if (err) {
-        out.set(key, { mathml, error: decodeXml(err[1]) });
+        out.set(key, { mathml: sanitizeMathml(mathml), error: decodeXml(err[1]) });
         continue;
       }
-      const entry = { mathml: mathml.replace(/\n\s*/g, "") };
+      const entry = { mathml: sanitizeMathml(mathml.replace(/\n\s*/g, "")) };
       if (s.display) {
-        const svg = mj.svg(s.tex, true);
+        const svg = sanitizeMathSvg(mj.svg(s.tex, true));
         entry.svg = svg;
         entry.widthEx = Number(/width="([\d.]+)ex"/.exec(svg)?.[1] ?? 0);
         entry.heightEx = Number(/height="([\d.]+)ex"/.exec(svg)?.[1] ?? 0);
@@ -3351,7 +3594,10 @@ function mathHtml(tex, display, x, inText = false) {
 }
 var href = (id, x) => x.hrefFor ? x.hrefFor(id) : `#${id}`;
 function imageSrc(url, x) {
-  if (x.srcFor) return x.srcFor(url);
+  if (x.srcFor) {
+    const src = x.srcFor(url);
+    return src === null ? null : safeUrl(src);
+  }
   const resolved = x.images?.get(url);
   if (resolved) return resolved.dataUrl;
   return safeUrl(url);
@@ -3437,7 +3683,7 @@ function coverHtml(s, x) {
     s.title ?? ""
   )}</h1>${s.subtitle ? `<p class="doc-cover-subtitle">${(0, import_html_escape.escapeHtml)(s.subtitle)}</p>` : ""}${s.author ? `<p class="doc-cover-author">${(0, import_html_escape.escapeHtml)(s.author)}</p>` : ""}${s.date ? `<p class="doc-cover-date">${(0, import_html_escape.escapeHtml)(s.date)}</p>` : ""}</section>`;
 }
-var cssString = (s) => `"${s.replace(/\\/g, "\\\\").replace(/"/g, '\\"').replace(/\n/g, " ")}"`;
+var cssString = (s) => `"${s.replace(/\\/g, "\\\\").replace(/"/g, '\\"').replace(/\n/g, " ").replace(/</g, "\\3C ").replace(/>/g, "\\3E ")}"`;
 function marginContent(template, s) {
   if (!template) return "none";
   return templateParts(template, s).map((p) => "field" in p ? `counter(${p.field})` : cssString(p.text)).join(" ");
@@ -3532,7 +3778,7 @@ ${[2, 3, 4, 5, 6].map((l) => `.doc-toc .toc-l${l}{padding-left:${(l - 1) * 1.2}e
 }
 function documentToHtml(tree, options = {}) {
   const s = tree.settings;
-  const x = { tree, images: options.images };
+  const x = { tree, images: options.images, ...options.srcFor ? { srcFor: options.srcFor } : {} };
   const sections = tree.sections.map((sec) => {
     const cls = [
       "doc-section",
@@ -3555,8 +3801,8 @@ ${s.author ? `<meta name="author" content="${(0, import_html_escape.escapeHtml)(
 </head>
 <body>
 <main class="doc">
-${s.cover ? coverHtml(s, x) : ""}
-${sections}
+${(options.transformBody ?? ((h) => h))(`${s.cover ? coverHtml(s, x) : ""}
+${sections}`)}
 </main>
 </body>
 </html>`;
@@ -4895,6 +5141,7 @@ var import_micromark_extension_gfm = require("micromark-extension-gfm");
 
 // src/markdown.ts
 var import_html_escape3 = require("@ai-matrx/kit/html-escape");
+var import_source2 = require("@ai-matrx/content-ir/source");
 function removeThinkingContent(content) {
   if (!content) return "";
   const hide = (_match, before, after) => before.length >= 2 || after.length >= 2 ? "\n\n" : "\n";
@@ -4967,6 +5214,7 @@ function looksLikeJson(body) {
     return false;
   }
 }
+var ALLOWED_TAGS = new Set(PRINT_HTML_SCHEMA.tagNames ?? []);
 var LINK_URL = String.raw`((?:[^()\s]|\([^()\s]*\))+)(?:\s+"[^"]*")?`;
 var IMAGE_RE = new RegExp(String.raw`!\[([^\]]*)\]\(` + LINK_URL + String.raw`\)`, "g");
 var LINK_RE = new RegExp(String.raw`\[([^\]]+)\]\(` + LINK_URL + String.raw`\)`, "g");