fix(xps): repair scanner XPS packages that MuPDF fails to parse

src/js/utils/xps-repair.ts

@@ -0,0 +1,147 @@
+import JSZip from 'jszip';
+import type { PyMuPDFInstance } from '@/types';
+
+const XML_PART = /\.(rels|fdseq|fdoc|fpage|dict|xml)$/i;
+const PIECE = /^(.*)[/\\]\[(\d+)\](?:\.last)?\.piece$/i;
+const ATTRIBUTE_NAME = /^[A-Za-z0-9._:-]+$/;
+const ATTRIBUTE =
+  /([^\s\p{Cc}=/>"']+)\s*=\s*("[^"]*"|'[^']*'|[^\s"'>]*[^\s"'>/])/gu;
+const VERBATIM: Array<[string, string]> = [
+  ['<!--', '-->'],
+  ['<![CDATA[', ']]>'],
+  ['<?', '?>'],
+  ['<!', '>'],
+];
+
+function decodeXml(bytes: Uint8Array): string {
+  const [a, b, c] = bytes;
+  if (a === 0xff && b === 0xfe)
+    return new TextDecoder('utf-16le').decode(bytes);
+  if (a === 0xfe && b === 0xff)
+    return new TextDecoder('utf-16be').decode(bytes);
+  if (a === 0x3c && b === 0) return new TextDecoder('utf-16le').decode(bytes);
+  if (a === 0 && b === 0x3c) return new TextDecoder('utf-16be').decode(bytes);
+  if (a === 0xef && b === 0xbb && c === 0xbf) {
+    return new TextDecoder().decode(bytes);
+  }
+  const head = new TextDecoder('latin1').decode(bytes.subarray(0, 256));
+  const declared = head.match(
+    /^\s*<\?xml[^>]*?encoding\s*=\s*["']([^"']+)["']/i
+  )?.[1];
+  if (!declared || /^utf-?16/i.test(declared)) {
+    return new TextDecoder().decode(bytes);
+  }
+  try {
+    return new TextDecoder(declared).decode(bytes);
+  } catch {
+    return new TextDecoder().decode(bytes);
+  }
+}
+
+function rebuildTag(inner: string): string {
+  if (inner.startsWith('/')) return `</${inner.slice(1).trim()}>`;
+  const name = inner.match(/^\s*([^\s/>]+)/);
+  if (!name) return `<${inner}>`;
+  const rest = inner.slice(name[0].length);
+  const attributes = Array.from(rest.matchAll(ATTRIBUTE))
+    .filter(([, attribute]) => ATTRIBUTE_NAME.test(attribute))
+    .map(([, attribute, value]) =>
+      /^["']/.test(value) ? `${attribute}=${value}` : `${attribute}="${value}"`
+    );
+  const end = /\/\s*$/.test(rest) ? '/>' : '>';
+  return `<${[name[1], ...attributes].join(' ')}${end}`;
+}
+
+export function repairXml(source: string): string {
+  const text = source.replace(/\0/g, '');
+  let out = '';
+  let i = 0;
+  while (i < text.length) {
+    const start = text.indexOf('<', i);
+    if (start === -1) {
+      out += text.slice(i);
+      break;
+    }
+    out += text.slice(i, start);
+    const verbatim = VERBATIM.find(([open]) => text.startsWith(open, start));
+    if (verbatim) {
+      const close = text.indexOf(verbatim[1], start + verbatim[0].length);
+      const stop = close === -1 ? text.length : close + verbatim[1].length;
+      out += text.slice(start, stop);
+      i = stop;
+      continue;
+    }
+    let j = start + 1;
+    let quote = '';
+    while (j < text.length && (quote || text[j] !== '>')) {
+      if (quote) {
+        if (text[j] === quote) quote = '';
+      } else if (text[j] === '"' || text[j] === "'") {
+        quote = text[j];
+      }
+      j++;
+    }
+    out += rebuildTag(text.slice(start + 1, j));
+    i = j + 1;
+  }
+  return out.replace(
+    /^(\s*<\?xml[^>]*?encoding\s*=\s*)["'][^"']*["']/,
+    '$1"utf-8"'
+  );
+}
+
+function repairPart(name: string, bytes: Uint8Array): Uint8Array {
+  if (!XML_PART.test(name)) return bytes;
+  return new TextEncoder().encode(repairXml(decodeXml(bytes)));
+}
+
+export async function repairXpsPackage(file: Blob): Promise<Blob> {
+  const zip = await JSZip.loadAsync(await file.arrayBuffer());
+  const repaired = new JSZip();
+  const pieces = new Map<string, Array<{ index: number; data: ArrayBuffer }>>();
+
+  for (const entry of Object.values(zip.files)) {
+    if (entry.dir) continue;
+    const piece = entry.name.match(PIECE);
+    if (piece) {
+      const list = pieces.get(piece[1]) ?? [];
+      list.push({
+        index: Number(piece[2]),
+        data: await entry.async('arraybuffer'),
+      });
+      pieces.set(piece[1], list);
+      continue;
+    }
+    repaired.file(
+      entry.name,
+      repairPart(entry.name, await entry.async('uint8array'))
+    );
+  }
+
+  for (const [name, list] of pieces) {
+    list.sort((x, y) => x.index - y.index);
+    const merged = await new Blob(list.map((p) => p.data)).arrayBuffer();
+    repaired.file(name, repairPart(name, new Uint8Array(merged)));
+  }
+
+  return new Blob([
+    await repaired.generateAsync({ type: 'arraybuffer', compression: 'STORE' }),
+  ]);
+}
+
+export async function convertXpsToPdf(
+  pymupdf: PyMuPDFInstance,
+  file: Blob
+): Promise<Blob> {
+  try {
+    return await pymupdf.convertToPdf(file, { filetype: 'xps' });
+  } catch (error) {
+    try {
+      return await pymupdf.convertToPdf(await repairXpsPackage(file), {
+        filetype: 'xps',
+      });
+    } catch {
+      throw error;
+    }
+  }
+}