Reimplementation of XREF objects registration for propper rearrange of deletions. Deletion of pages objects on page removal. Snapshots ref deletion registration

adnsistemas · adnsistemas · commit 87ac8a904d7b · 2025-07-23T00:09:06.000-03:00
diff --git a/apps/deno/tests/test20.ts b/apps/deno/tests/test20.ts
@@ -14,13 +14,15 @@ import {
 
 /**
  * This test modifies the pdf adding a page and a placeholder for an electronic signature.
+ * It also deletes the first page of the PDF.
  * The file should have an incremental update at the end, and the start of the file be exactly the original file.
  */
 export default async (assets: Assets) => {
   const pdfDoc = await PDFDocument.load(assets.pdfs.simple, {
     forIncrementalUpdate: true,
   });
   const page = pdfDoc.addPage([500, 200]);
+  pdfDoc.removePage(0);
   const font = pdfDoc.embedStandardFont(StandardFonts.Helvetica);
   const fontBold = pdfDoc.embedStandardFont(StandardFonts.HelveticaBoldOblique);
   const advertencia = `-- Esta nota es meramente informativa, no es la firma real del documento --`;
@@ -164,5 +166,5 @@ export default async (assets: Assets) => {
     acroForm.dict.set(PDFName.of('Fields'), fields);
   }
   (fields as PDFArray).push(widgetDictRef);
-  return await pdfDoc.save();
+  return await pdfDoc.save({ useObjectStreams: false });
 };
diff --git a/apps/node/tests/test20.ts b/apps/node/tests/test20.ts
@@ -13,13 +13,15 @@ import {
 
 /**
  * This test modifies the pdf adding a page and a placeholder for an electronic signature.
+ * It also deletes the first page of the PDF.
  * The file should have an incremental update at the end, and the start of the file be exactly the original file.
  */
 export default async (assets: Assets) => {
   const pdfDoc = await PDFDocument.load(assets.pdfs.simple, {
     forIncrementalUpdate: true,
   });
   const page = pdfDoc.addPage([500, 200]);
+  pdfDoc.removePage(0);
   const font = pdfDoc.embedStandardFont(StandardFonts.Helvetica);
   const fontBold = pdfDoc.embedStandardFont(StandardFonts.HelveticaBoldOblique);
   const advertencia = `-- Esta nota es meramente informativa, no es la firma real del documento --`;
@@ -163,5 +165,5 @@ export default async (assets: Assets) => {
     acroForm.dict.set(PDFName.of('Fields'), fields);
   }
   (fields as PDFArray).push(widgetDictRef);
-  return await pdfDoc.save();
+  return await pdfDoc.save({ useObjectStreams: false });
 };
diff --git a/apps/web/test34.html b/apps/web/test34.html
@@ -65,6 +65,7 @@
         forIncrementalUpdate: true,
       });
       const page = pdfDoc.addPage([500, 150]);
+      pdfDoc.removePage(0);
       const font = pdfDoc.embedStandardFont(StandardFonts.Helvetica);
       const fontBold = pdfDoc.embedStandardFont(
         StandardFonts.HelveticaBoldOblique,
diff --git a/package.json b/package.json
@@ -71,7 +71,7 @@
     "scratchpad:flame": "rimraf isolate*.log && node --prof scratchpad/build/scratchpad/index.js && node --prof-process --preprocess -j isolate*.log | flamebearer",
     "apps:node": "tsc --build apps/node/tsconfig.json && node apps/node-build/index.js",
     "apps:deno": "deno run --allow-read --allow-write --allow-run apps/deno/index.ts",
-    "apps:web": "http-server -c-1 .",
+    "apps:web": "http-server -c-1 -o apps/web/test1.html",
     "apps:web:mac": "bash -c 'sleep 1 && open http://localhost:8080/apps/web/test1.html' & yarn apps:web",
     "apps:rn:ios": "cd apps/rn && yarn add ./../.. --force && react-native run-ios",
     "apps:rn:android": "yarn apps:rn:emulator & cd apps/rn && yarn add ./../.. --force && react-native run-android",
diff --git a/src/api/PDFDocument.ts b/src/api/PDFDocument.ts
@@ -688,8 +688,10 @@ export default class PDFDocument {
     const pageCount = this.getPageCount();
     if (this.pageCount === 0) throw new RemovePageFromEmptyDocumentError();
     assertRange(index, 'index', 0, pageCount - 1);
+    const page = this.getPage(index);
     this.catalog.removeLeafNode(index);
     this.pageCount = pageCount - 1;
+    this.context.delete(page.ref);
   }
 
   /**
diff --git a/src/api/snapshot/DefaultDocumentSnapshot.ts b/src/api/snapshot/DefaultDocumentSnapshot.ts
@@ -4,6 +4,7 @@ import type { DocumentSnapshot } from './DocumentSnapshot';
 export class DefaultDocumentSnapshot implements DocumentSnapshot {
   pdfSize = 0;
   prevStartXRef = 0;
+  deletedCount = 0;
 
   shouldSave(_objectNumber: number): boolean {
     return true;
@@ -24,6 +25,18 @@ export class DefaultDocumentSnapshot implements DocumentSnapshot {
   markObjsForSave(_objs: PDFObject[]): void {
     throw new Error('This method should not be called.');
   }
+
+  markDeletedObj(_obj: PDFObject): void {
+    throw new Error('This method should not be called.');
+  }
+
+  markDeletedRef(_ref: PDFRef): void {
+    throw new Error('This method should not be called.');
+  }
+
+  deletedRef(_index: number): PDFRef | null {
+    throw new Error('This method should not be called.');
+  }
 }
 
 export const defaultDocumentSnapshot = new DefaultDocumentSnapshot();
diff --git a/src/api/snapshot/DocumentSnapshot.ts b/src/api/snapshot/DocumentSnapshot.ts
@@ -3,6 +3,7 @@ import type { PDFObject, PDFRef } from '../../core';
 export interface DocumentSnapshot {
   pdfSize: number;
   prevStartXRef: number;
+  deletedCount: number;
 
   shouldSave: (objectNumber: number) => boolean;
 
@@ -11,4 +12,9 @@ export interface DocumentSnapshot {
 
   markObjForSave: (obj: PDFObject) => void;
   markObjsForSave: (objs: PDFObject[]) => void;
+
+  markDeletedRef: (ref: PDFRef) => void;
+  markDeletedObj: (obj: PDFObject) => void;
+
+  deletedRef: (index: number) => PDFRef | null;
 }
diff --git a/src/api/snapshot/IncrementalDocumentSnapshot.ts b/src/api/snapshot/IncrementalDocumentSnapshot.ts
@@ -4,7 +4,9 @@ import type { DocumentSnapshot } from './DocumentSnapshot';
 export class IncrementalDocumentSnapshot implements DocumentSnapshot {
   pdfSize: number;
   prevStartXRef: number;
+  deletedCount: number = 0;
 
+  private deleted: PDFRef[] = [];
   private lastObjectNumber: number;
   private changedObjects: number[];
 
@@ -56,4 +58,22 @@ export class IncrementalDocumentSnapshot implements DocumentSnapshot {
         .filter((ref) => ref !== undefined) as PDFRef[],
     );
   }
+
+  markDeletedRef(ref: PDFRef): void {
+    if (
+      this.deleted.findIndex((dref) => dref.objectNumber === ref.objectNumber) <
+      0
+    )
+      this.deletedCount = this.deleted.push(ref);
+  }
+
+  markDeletedObj(obj: PDFObject): void {
+    const oref = this.context.getRef(obj);
+    if (oref) this.markDeletedRef(oref);
+  }
+
+  deletedRef(index: number): PDFRef | null {
+    if (index < 0 || index >= this.deleted.length) return null;
+    return this.deleted[index];
+  }
 }
diff --git a/src/core/PDFContext.ts b/src/core/PDFContext.ts
@@ -60,6 +60,7 @@ class PDFContext {
   largestObjectNumber: number;
   header: PDFHeader;
   trailerInfo: {
+    Size?: PDFNumber;
     Root?: PDFObject;
     Encrypt?: PDFObject;
     Info?: PDFObject;
@@ -114,6 +115,7 @@ class PDFContext {
   }
 
   delete(ref: PDFRef): boolean {
+    if (this.snapshot) this.snapshot.markDeletedRef(ref);
     return this.indirectObjects.delete(ref);
   }
 
diff --git a/src/core/document/PDFCrossRefSection.ts b/src/core/document/PDFCrossRefSection.ts
@@ -10,7 +10,7 @@ export interface Entry {
 
 /**
  * Entries should be added using the [[addEntry]] and [[addDeletedEntry]]
- * methods **in order of ascending object number**.
+ * methods.
  */
 class PDFCrossRefSection {
   static create = () =>
@@ -37,6 +37,22 @@ class PDFCrossRefSection {
   }
 
   addDeletedEntry(ref: PDFRef, nextFreeObjectNumber: number): void {
+    // fix the first entry if required
+    if (!this.subsections.length) {
+      this.subsections = [
+        [
+          {
+            ref: PDFRef.of(0, 65535),
+            offset: ref.objectNumber,
+            deleted: true,
+          },
+        ],
+      ];
+      this.chunkIdx = 0;
+      this.chunkLength = 1;
+    } else if (!this.subsections[0][0].offset) {
+      this.subsections[0][0].offset = ref.objectNumber;
+    }
     this.append({ ref, offset: nextFreeObjectNumber, deleted: true });
   }
 
@@ -159,7 +175,35 @@ class PDFCrossRefSection {
     const chunk = this.subsections[this.chunkIdx];
     const prevEntry = chunk[this.chunkLength - 1];
 
-    if (currEntry.ref.objectNumber - prevEntry.ref.objectNumber > 1) {
+    if (currEntry.ref.objectNumber - prevEntry.ref.objectNumber !== 1) {
+      // the current chunk is not the right chunk, find the right one, or create a new one
+      for (let c = 0; c < this.subsections.length; c++) {
+        const first = this.subsections[c][0];
+        const last = this.subsections[c][this.subsections[c].length - 1];
+        if (first.ref.objectNumber > currEntry.ref.objectNumber) {
+          // goes before this subsection, or at the start of it
+          if (first.ref.objectNumber - currEntry.ref.objectNumber === 1) {
+            // first element of subsection
+            this.subsections[c].unshift(currEntry);
+            if (c === this.chunkIdx) this.chunkLength += 1;
+            return;
+          } else {
+            // create subsection
+            this.subsections.splice(c, 0, [currEntry]);
+            this.chunkIdx++;
+            return;
+          }
+        } else if (last.ref.objectNumber > currEntry.ref.objectNumber) {
+          // goes in this subsection, find its place..
+          const cep = this.subsections[c].findIndex(
+            (ee) => ee.ref.objectNumber > currEntry.ref.objectNumber,
+          );
+          this.subsections[c].splice(cep, 0, currEntry);
+          if (c === this.chunkIdx) this.chunkLength += 1;
+        }
+        // bigger, keep looking
+      }
+      // if got to here, then a new subsection is required
       this.subsections.push([currEntry]);
       this.chunkIdx += 1;
       this.chunkLength = 1;
diff --git a/src/core/parser/PDFParser.ts b/src/core/parser/PDFParser.ts
@@ -24,6 +24,7 @@ import { Keywords } from '../syntax/Keywords';
 import { IsDigit } from '../syntax/Numeric';
 import { waitForTick } from '../../utils';
 import { CipherTransformFactory } from '../crypto';
+import PDFNumber from '../objects/PDFNumber';
 
 class PDFParser extends PDFObjectParser {
   static forBytesWithOptions = (
@@ -307,11 +308,17 @@ class PDFParser extends PDFObjectParser {
 
     const { context } = this;
     context.trailerInfo = {
+      Size:
+        dict.lookupMaybe(PDFName.of('Size'), PDFNumber) ||
+        context.trailerInfo.Size,
       Root: dict.get(PDFName.of('Root')) || context.trailerInfo.Root,
       Encrypt: dict.get(PDFName.of('Encrypt')) || context.trailerInfo.Encrypt,
       Info: dict.get(PDFName.of('Info')) || context.trailerInfo.Info,
       ID: dict.get(PDFName.of('ID')) || context.trailerInfo.ID,
     };
+    // if open for incremental update, then deleted objects need to be preserved, and largestObjectNumber has to be Size-1
+    if (context.trailerInfo.Size && context.pdfFileDetails.originalBytes)
+      context.largestObjectNumber = context.trailerInfo.Size.asNumber() - 1;
   }
 
   private maybeParseTrailer(): PDFTrailer | void {
diff --git a/src/core/parser/PDFXRefStreamParser.ts b/src/core/parser/PDFXRefStreamParser.ts
@@ -66,11 +66,19 @@ class PDFXRefStreamParser {
     this.alreadyParsed = true;
 
     this.context.trailerInfo = {
+      Size: this.dict.lookup(PDFName.of('Size'), PDFNumber),
       Root: this.dict.get(PDFName.of('Root')),
       Encrypt: this.dict.get(PDFName.of('Encrypt')),
       Info: this.dict.get(PDFName.of('Info')),
       ID: this.dict.get(PDFName.of('ID')),
     };
+    // if open for incremental update, make sure next object number doesn't overlap a deleted one
+    if (
+      this.context.trailerInfo.Size &&
+      this.context.pdfFileDetails.originalBytes
+    )
+      this.context.largestObjectNumber =
+        this.context.trailerInfo.Size.asNumber() - 1;
 
     const entries = this.parseEntries();
 
diff --git a/src/core/writers/PDFWriter.ts b/src/core/writers/PDFWriter.ts
@@ -166,6 +166,17 @@ class PDFWriter {
       size += this.computeIndirectObjectSize(indirectObject);
       if (this.shouldWaitForTick(1)) await waitForTick();
     }
+    // deleted objects
+    for (let idx = 0; idx < this.snapshot.deletedCount; idx++) {
+      const dref = this.snapshot.deletedRef(idx);
+      if (!dref) break;
+      const nextdref = this.snapshot.deletedRef(idx + 1);
+      // add 1 to generation number for deleted ref
+      xref.addDeletedEntry(
+        PDFRef.of(dref.objectNumber, dref.generationNumber + 1),
+        nextdref ? nextdref.objectNumber : 0,
+      );
+    }
 
     const xrefOffset = size;
     size += xref.sizeInBytes() + 1; // '\n'
diff --git a/tests/api/PDFDocument.spec.ts b/tests/api/PDFDocument.spec.ts
@@ -164,6 +164,20 @@ describe(`PDFDocument`, () => {
       const pdfDoc = await PDFDocument.load(simplePdfBytes);
       expect(pdfDoc.context.largestObjectNumber).toBe(8);
     });
+
+    it('preserves deleted objects numbers if open for update', async () => {
+      const pdfBytes = fs.readFileSync(
+        './assets/pdfs/with_update_sections.pdf',
+      );
+      const pdfDoc = await PDFDocument.load(pdfBytes, {
+        forIncrementalUpdate: false,
+      });
+      expect(pdfDoc.context.largestObjectNumber).toBeGreaterThanOrEqual(131);
+      const pdfUpdDoc = await PDFDocument.load(pdfBytes, {
+        forIncrementalUpdate: true,
+      });
+      expect(pdfUpdDoc.context.largestObjectNumber).toBe(334);
+    });
   });
 
   describe('embedFont() method', () => {
@@ -595,6 +609,23 @@ describe(`PDFDocument`, () => {
 
       await expect(noErrorFunc()).resolves.not.toThrowError();
     });
+
+    it(`saves deleted objects`, async () => {
+      const noErrorFunc = async (pageIndex: number) => {
+        const pdfDoc = await PDFDocument.load(simplePdfBytes);
+        const snapshot = pdfDoc.takeSnapshot();
+        const page = pdfDoc.getPage(pageIndex);
+        snapshot.markDeletedRef(page.ref);
+        const pdfIncrementalBytes = await pdfDoc.saveIncremental(snapshot);
+        expect(pdfIncrementalBytes.byteLength).toBeGreaterThan(0);
+        expect(Buffer.from(pdfIncrementalBytes).toString()).toMatch(
+          `xref\n0 1\n${page.ref.objectNumber.toString().padStart(10, '0')} 65535 f \n${page.ref.objectNumber.toString()} 1\n0000000000 00001 f`,
+        );
+      };
+
+      await expect(noErrorFunc(0)).resolves.not.toThrowError();
+      await expect(noErrorFunc(1)).resolves.not.toThrowError();
+    });
   });
 
   describe(`copy() method`, () => {
@@ -702,6 +733,44 @@ describe(`PDFDocument`, () => {
       await expect(noErrorFunc()).resolves.not.toThrowError();
     });
 
+    it(`registers deleted objects`, async () => {
+      const pdfDoc = await PDFDocument.load(simplePdfBytes, {
+        forIncrementalUpdate: true,
+      });
+      let page = pdfDoc.getPage(0);
+      const delONum = page.ref.objectNumber;
+      const delOGen = (page.ref.generationNumber + 1).toString();
+      pdfDoc.context.delete(page.ref);
+      page = pdfDoc.getPage(1);
+      const timesRomanFont = await pdfDoc.embedFont(StandardFonts.TimesRoman);
+      const fontSize = 30;
+      page.drawText('Incremental saving is also awesome!', {
+        x: 50,
+        y: 4 * fontSize,
+        size: fontSize,
+        font: timesRomanFont,
+      });
+
+      const pdfIncrementalBytes = await pdfDoc.save({
+        useObjectStreams: false,
+      });
+      const pdfRewriteBytes = await pdfDoc.save({
+        rewrite: true,
+        useObjectStreams: false,
+      });
+      expect(pdfIncrementalBytes.byteLength).toBeGreaterThan(
+        simpleStreamsPdfBytes.byteLength,
+      );
+      expect(pdfRewriteBytes.byteLength).toBeGreaterThan(
+        simpleStreamsPdfBytes.byteLength,
+      );
+      // first element in table must point to deleted page, deleted page must have next gen number
+      const rex = new RegExp(
+        `xref[\n|\r\n]0 .*[\n|\r\n]${delONum.toString().padStart(10, '0')} 65535 f[\\s\\S]*0000000000 ${delOGen.padStart(5, '0')} f`,
+      );
+      expect(Buffer.from(pdfIncrementalBytes).toString()).toMatch(rex);
+    });
+
     it(`produces same output than manual incremental update`, async () => {
       const noErrorFunc = async (pageIndex: number) => {
         const pdfDoc = await PDFDocument.load(simplePdfBytes, {
diff --git a/tests/api/PDFPage.spec.ts b/tests/api/PDFPage.spec.ts
diff --git a/tests/core/document/PDFCrossRefSection.spec.ts b/tests/core/document/PDFCrossRefSection.spec.ts

Original file line number	Diff line number	Diff line change
`@@ -60,6 +60,7 @@ class PDFContext {`
`60`	`60`	`largestObjectNumber: number;`
`61`	`61`	`header: PDFHeader;`
`62`	`62`	`trailerInfo: {`
	`63`	`+ Size?: PDFNumber;`
`63`	`64`	`Root?: PDFObject;`
`64`	`65`	`Encrypt?: PDFObject;`
`65`	`66`	`Info?: PDFObject;`
`@@ -114,6 +115,7 @@ class PDFContext {`
`114`	`115`	`}`
`115`	`116`
`116`	`117`	`delete(ref: PDFRef): boolean {`
	`118`	`+ if (this.snapshot) this.snapshot.markDeletedRef(ref);`
`117`	`119`	`return this.indirectObjects.delete(ref);`
`118`	`120`	`}`
`119`	`121`