@hilvana/office 0.1.0 → 0.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/index.d.ts CHANGED
@@ -1,7 +1,7 @@
1
1
  export { wordToPdf, type WordToPdfOptions, type WordToPdfResult, type WordWarning } from './word.js';
2
2
  export { convertWord, type WordEngine } from './convert.js';
3
3
  export { findLibreOffice, libreOfficeToPdf } from './libreoffice.js';
4
- export { DocxError, isDocx, parseDocx } from './parse.js';
4
+ export { DOCX_LIMITS, DocxError, isDocx, parseDocx } from './parse.js';
5
5
  export { fallbackTextFonts, familyFor, nodeFontLoader, type Family, type FontLoader } from './fonts.js';
6
6
  export type * from './model.js';
7
7
  //# sourceMappingURL=index.d.ts.map
package/dist/index.js CHANGED
@@ -1,6 +1,6 @@
1
1
  export { wordToPdf } from './word.js';
2
2
  export { convertWord } from './convert.js';
3
3
  export { findLibreOffice, libreOfficeToPdf } from './libreoffice.js';
4
- export { DocxError, isDocx, parseDocx } from './parse.js';
4
+ export { DOCX_LIMITS, DocxError, isDocx, parseDocx } from './parse.js';
5
5
  export { fallbackTextFonts, familyFor, nodeFontLoader } from './fonts.js';
6
6
  //# sourceMappingURL=index.js.map
package/dist/parse.d.ts CHANGED
@@ -5,9 +5,15 @@ export interface ParsedDocx extends DocModel {
5
5
  headerDistance: number;
6
6
  footerDistance: number;
7
7
  }
8
+ /** Ceilings for what a .docx may expand to in memory. Real documents stay far below them. */
9
+ export declare const DOCX_LIMITS: {
10
+ partBytes: number;
11
+ totalBytes: number;
12
+ entries: number;
13
+ };
8
14
  export declare class DocxError extends Error {
9
- readonly code: 'NOT_DOCX' | 'ENCRYPTED_DOCX' | 'CORRUPT_DOCX';
10
- constructor(code: 'NOT_DOCX' | 'ENCRYPTED_DOCX' | 'CORRUPT_DOCX', message?: string);
15
+ readonly code: 'NOT_DOCX' | 'ENCRYPTED_DOCX' | 'CORRUPT_DOCX' | 'DOCX_TOO_LARGE';
16
+ constructor(code: 'NOT_DOCX' | 'ENCRYPTED_DOCX' | 'CORRUPT_DOCX' | 'DOCX_TOO_LARGE', message?: string);
11
17
  }
12
18
  /** True when the bytes look like a Word .docx package. */
13
19
  export declare function isDocx(bytes: Uint8Array): boolean;
package/dist/parse.js CHANGED
@@ -786,6 +786,8 @@ function readPage(sect) {
786
786
  footer: twip(attr(pgMar, 'footer')) ?? 36,
787
787
  };
788
788
  }
789
+ /** Ceilings for what a .docx may expand to in memory. Real documents stay far below them. */
790
+ export const DOCX_LIMITS = { partBytes: 64 * 1024 * 1024, totalBytes: 256 * 1024 * 1024, entries: 5000 };
789
791
  export class DocxError extends Error {
790
792
  code;
791
793
  constructor(code, message) {
@@ -813,14 +815,68 @@ export function parseDocx(bytes) {
813
815
  if (bytes[0] === 0xd0 && bytes[1] === 0xcf && bytes[2] === 0x11 && bytes[3] === 0xe0)
814
816
  throw new DocxError('ENCRYPTED_DOCX');
815
817
  let files;
818
+ // Guard against zip bombs: a few MB can declare gigabytes of content. Only the parts the
819
+ // converter reads are extracted, and their declared sizes are capped before inflating.
820
+ let total = 0;
821
+ let count = 0;
822
+ let tooLarge;
823
+ const take = (f) => {
824
+ // Deflate never grows data by more than a few bytes per block: a compressed part much larger
825
+ // than its declared size is lying about it, so the cap above would not hold.
826
+ if (f.size > f.originalSize * 1.01 + 1024)
827
+ tooLarge ??= `${f.name} hides its real size`;
828
+ total += f.originalSize;
829
+ if (f.originalSize > DOCX_LIMITS.partBytes)
830
+ tooLarge ??= `${f.name} declares ${f.originalSize} bytes`;
831
+ if (total > DOCX_LIMITS.totalBytes)
832
+ tooLarge ??= `content declares ${total} bytes`;
833
+ return !tooLarge;
834
+ };
816
835
  try {
817
- files = unzipSync(bytes);
836
+ // First the XML parts and relationships, then only the media those relationships point to.
837
+ files = unzipSync(bytes, {
838
+ filter: (f) => {
839
+ if (++count > DOCX_LIMITS.entries)
840
+ tooLarge ??= 'too many parts';
841
+ if (!/^word\/.*\.(xml|rels)$/.test(f.name))
842
+ return false;
843
+ return take(f);
844
+ },
845
+ });
846
+ const referenced = new Set();
847
+ for (const [name, data] of Object.entries(files)) {
848
+ if (!name.endsWith('.rels'))
849
+ continue;
850
+ const dir = name.replace(/_rels\/[^/]+\.rels$/, '');
851
+ for (const m of strFromU8(data).matchAll(/Target="([^"]+)"/g)) {
852
+ const target = m[1];
853
+ if (/^[a-z]+:/i.test(target))
854
+ continue; // external links are never fetched
855
+ referenced.add(normalize(target.startsWith('/') ? target.slice(1) : dir + target));
856
+ }
857
+ }
858
+ if (!tooLarge && referenced.size) {
859
+ Object.assign(files, unzipSync(bytes, { filter: (f) => !(f.name in files) && referenced.has(f.name) && take(f) }));
860
+ }
818
861
  }
819
862
  catch (err) {
820
863
  throw new DocxError('CORRUPT_DOCX', String(err.message));
821
864
  }
865
+ if (tooLarge)
866
+ throw new DocxError('DOCX_TOO_LARGE', tooLarge);
822
867
  if (!files['word/document.xml'])
823
868
  throw new DocxError('NOT_DOCX');
869
+ // Anything that breaks while reading the parts means a damaged document, with a code callers can explain (D21).
870
+ try {
871
+ return readDocx(files);
872
+ }
873
+ catch (err) {
874
+ if (err instanceof DocxError)
875
+ throw err;
876
+ throw new DocxError('CORRUPT_DOCX', String(err?.message ?? err).slice(0, 200));
877
+ }
878
+ }
879
+ function readDocx(files) {
824
880
  const ctx = new Context(files);
825
881
  loadStyles(ctx);
826
882
  const docPart = { path: 'word/document.xml', rels: loadRels(ctx, 'word/document.xml') };
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@hilvana/office",
3
- "version": "0.1.0",
3
+ "version": "0.2.0",
4
4
  "description": "Hilvana office conversions that run locally: Word (.docx) to PDF with real, selectable text.",
5
5
  "keywords": [
6
6
  "docx",
@@ -18,7 +18,7 @@
18
18
  },
19
19
  "repository": {
20
20
  "type": "git",
21
- "url": "https://github.com/gethilvana/hilvana",
21
+ "url": "git+https://github.com/gethilvana/hilvana.git",
22
22
  "directory": "packages/office"
23
23
  },
24
24
  "type": "module",